Report in English with a French summary (KCE reports 176B)
publicité
KCE REPORT 176B DÉPIS STAGE DU CAN NCER DU SEIN N ENTRE 70 ET T 74 ANS 2012 www.kce.fgo ov.be Le Centre fédéral d’ex xpertise des s soins de sa anté Le Ce entre fédéral d’e expertise des so oins de santé es st un parastatal, créé par la loi-p programme (1) du d 24 décem mbre 2002 (articles 259 à 281), sous s tutelle du Ministre M de la San nté publique et d des Affaires sociales. Il est chargé de réalis ser des études éclairant é la décis sion politique da ans le domaine d des soins de san nté et de l’a assurance malad die. Conseil d’a administratio on dent Présid Foncttionnaire dirigeantt de l'INAMI (vice président) Présid dent du SPF Santté publique (vice président) p Présid dent du SPF Sécu urité sociale (vice président) Admin nistrateur général de l'AFMPS Repré ésentants du minis stre de la Santé publique p Repré ésentants du minis stre des Affaires sociales s Repré ésentants du Cons seil des ministres Agenc ce intermutualiste e Organ nisations professio onnelles représen ntatives des médecins m Organ nisations professio onnelles représen ntatives des in nfirmiers Fédérrations hospitalièrres Partenaires sociaux Cham mbre des Représentants Me embres effectifs Pie erre Gillet Jo De Cock Dirrk Cuypers Fra ank Van Massenh hove Xa avier De Cuyper Be ernard Lange Ma arco Schetgen Olivier de Stexhe Ri De Ridder Jea an-Noël Godin Da aniel Devos Mic chiel Callens Pa atrick Verertbrugge en Xa avier Brenez arc Moens Ma Jea an-Pierre Baeyens chel Foulon Mic My yriam Hubinon Joh han Pauwels Jea an-Claude Praet Ritta Thys Pa aul Palsterman Lie eve Wierinck Membres suppléants Benoît Collin Chris D Decoster Jan Be ertels Greet M Musch Franço ois Perl Annickk Poncé Karel V Vermeyen Lambe ert Stamatakis Frédérric Lernoux oghe Bart Oo Frank D De Smet Yoland de Husden Geert M Messiaen d Lemye Roland Rita Cu uypers Meyers Ludo M Olivier Thonon Katrien n Kesteloot Pierre Smiets Leo Ne eels Celien Van Moerkerke Contrôle Comm missaire du Gouv vernement Yv ves Roger Direction Direc cteur Général Direc cteur Général Adjo oint Direc cteurs du program mme d'études Ra af Mertens Je ean-Pierre Closon Ch hristian Léonard Krristel De Gauquierr Contact Centrre Fédéral d’Expe ertise des Soins de e Santé (KCE) e Doorb building (10 étage) Boule evard du Jardin Botanique, 55 B-100 00 Bruxelles Belgique T +32 2 [0]2 287 33 88 F +32 2 [0]2 287 33 85 info@ @kce.fgov.be http:///www.kce.fgov.be e KCE REPORT 176B GOOD CLINICA AL PRACTICE DEPIS STAGE D DU CAN NCER DU D SEIN N ENTRE 70 ET T 74 ANS FRANÇOISE MAMBOURG, M JO R ROBAYS, SOPHIIE GERKENS 2012 www.kce.fgo ov.be COLOPHO ON Titre: d sein entre 70 et e 74 ans Dépistage du cancer du Auteurs: s (KCE), Sophie Gerkens G (KCE) Françoise Mambourg (KCE), Jo Robays Relecture: heer (KCE), Nancy y Thiry (KCE) Frank Hulstaert (KCE)), Pascale Jonckh Experts externe es: Marc c Arbijn (WIV - IS SP), Martine Berliière (UCL Saint-L Luc), Hilde Bosmans (UZ Leuven)), Jean-Benoit Bu urrion (ASB BL Brummammo)), Joëlle Desreux x (CHU Liège), André-Robert A Griv vegnée (Institut JJules Bordet), Pa atrick Neve en (UZ Leuven), Myriam M Provost (S SSMG), Hubert Th hierens (UGent), Reinhilde R Van Ee eckhoudt (WVG), Anne A Vand denbroucke (UCL Saint-Luc), Geertt Villeirs (UZ Gentt). Validateurs exte ernes: Philip ppe Autier (IPRI-L Lyon), Geert Page e (Jan Yperman Ziekenhuis), Z Chanttal Van Ongeval ((KU Leuven) Conflits d’intérê êt: Aucu un conflit déclaré Layout: Ine Verhulst V Disclaimer: • Le es experts exterrnes ont été con nsultés sur une version v (prélimin naire) du rapporrt scientifique. Leurs L re emarques ont été é discutées au cours des réunion ns. Ils ne sont pa as co-auteurs du u rapport scientiffique ett n’étaient pas né écessairement d’accord d avec son contenu. • Une version (fina ale) a ensuite étté soumise aux x validateurs. La a validation du rrapport résulte d’un co onsensus ou d’u un vote majorita aire entre les validateurs. Les va alidateurs ne son nt pas co-auteurrs du ra apport scientifiqu ue et ils n’étaient pas nécessaire ement tous les trrois d’accord ave ec son contenu. • Finalement, ce rap pport a été appro ouvé à l'unanimité par le Conseill d’administration. • Le e KCE reste seu ul responsable des erreurs ou om missions qui pourraient subsiste er de même que e des re ecommandations s faites aux autorités publiques. Date de publica ation: 26 ap pril 2012 Domaine: Good d Clinical Practice e (GCP) MeSH: Breast Neoplasms ; Mammography M ; Mass Screening Classification NLM: N WP 870 8 - Breast - Neo oplasms Langue: franç çais, anglais Format: Adob be® PDF™ (A4) Dépot légal: D/2012/10.273/19 Copyright: Les rapports r KCE son nt publiés sous Lic cence Creative Co ommons «by/nc/nd» http:///kce.fgov.be/fr/co ontent/a-propos-du u-copyright-des-ra apports-kce Comment citer ce rapport? Mam mbourg F, Robays s J, Gerkens S. Dépistage D du can ncer du sein entre e 70 et 74 ans. G Good Clinical Pra actice (GCP P). Bruxelles: Centre Fédérale d’Expertise des Soins de Santté (KCE). 2012. KCE Report 176B. D/2012/10.273/19. Ce document est disponible en télécharrgement sur le site Web du Centre fédéral d’expertisse des soins de sa anté. KCE Report 176 6B PRÉF FACE Dé épistage du cancerr du sein i Faire des choix c en matière de d soins ressemb ble vite à de la dis scrimination, en particulier p lorsque ces choix sont ba asés sur l’âge. Comment C par exe emple justifier le refus r de rembourrsement d’une inttervention coûteusse sur le cœur à une personne, uniquement u sur un critère d’âge, même m si pour le re este cette personn ne est encore en bonne forme? De e tels raisonneme ents conduisent sy ystématiquement à des discussion ns enflammées, no ourries à partir de e systèmes de valleurs parfois diam métralement oppo osés. La présente e étude qui pose la question de sa avoir s’il faut offrirr un dépistage orrganisé du cancerr du sein aux fem mmes âgées de 70 7 à 74 ans, nou us place donc à nouveau sur un sol glissant. Mais s il y a encore d’autres raisons d’être d particulièrement vigilants surr un tel sujet. Com mme dans tout dé épistage organisé é, on s’adresse en n effet à des gens s qui ne présente ent a priori pas de e plainte de santé é et qui n’étaient donc d pas nécessa airement demandeurs d’un tel exam men. L’adage priimum non nocere est donc ici d’auttant plus importan nt. En matière d’argumentation à développer, il y a aussi un défi particulier p à releve er. Le clinicien est plus coutumier de d la logique utilisée pour poser un diagnostic che ez une personne e qui a une plaintte que de celle u utilisée en matière e de screening. Dans le premierr cas, le risque d’un d résultat faux x positif est non seulement plus petit mais est aussi a clairement considéré comme moins importan nt qu’un résultat faux f négatif, qui équivaut à loupe er un diagnostic. Cela explique po ourquoi les inconv vénients d’un dép pistage sont systématiquement sou us évalués. De plu us, le sujet est loin de laisser l’opiinion publique ind différente, un lobb bying intense estt organisé à son propos, et il est (donc) aussi sensible politiqueme ent. Même si on n mobilise toutes les preuves scien ntifiques disponib bles pour fonder un u avis sur la que estion, on ne peutt pas pour autant espérer arrêter la controverse. Nous N osons néan nmoins espérer que ce rapport ap pportera tout ce qu’on q peut attend dre d’un organe d’avis scientifique dans d un tel débat. Jean-Pierre e CLOSON Directeur Général G Adjoint Raf MERTENS Directeur Gén néral ii RÉSU UMÉ Dé épistage du cancerr du sein KCE Report 176B 1 INTRODUC CTION Ce travail fait parttie d’un projet plu C us large ayant pour objet la mise à jour d rapport: «Dépis du stage du cancer du sein», publié en 2005 (rapport KCE n n°11). Il concerne e plus particulièrem ment l’extension du dépistage organisé a aux femmes âgées de 70 à 74 ans qui ne préssentent par ailleu urs ni s symptôme évocate eur, ni facteur de risque particulier. L dépistage du cancer Le c du sein es st un processus ccomplexe qui com mporte d des bénéfices ett des risques. Les L principaux b bénéfices attendu us du d dépistage du cancer du sein sontt la diminution de e la mortalité et de la m morbidité liées à la maladie. La diiminution de morrbidité implique so oit un a allègement des traitements, t soit une diminution d des récidives ou u des s stades métastatiqu ues de la maladie e. L risques principaux liés au dép Les pistage concernen nt la qualité de vie. En e effet, un résultat faussement pos sitif, un diagnosttic excédentaire (surd diagnostic) suivi d’un traitement et l’avance au diagnostic (ou lead time, qui e la durée par la est aquelle le diagnos stic par dépistage e précède le diagn nostic c clinique) ont des conséquences c surr la qualité de vie. L Les résultats faussement positifs ont pour consé équence d’inclure e des fe emmes en bonne santé dans un circuit d’exxamens diagnosttiques a anxiogènes voire invasifs i (biopsies)). L sur-diagnostic peut être défini comme Le c la détection de cas de cancers q n’auraient jamais été perçus clin qui niquement en l’ab bsence de dépista age. Il s s’accroit au fur et à mesure de la diminution de l’e espérance de vie de la p population dépistée. Le sur-traite ement est une cconséquence du surd diagnostic. Vu qu’il est actuellemen nt impossible de p prédire si un canc cer va s développer, la se a très grande ma ajorité des cance ers diagnostiqués sont trraités. E Enfin, le dépistage e met les cancers s en évidence deu ux ou trois ans plus tôt q que ne le ferait un u diagnostic clin nique. Ceci implique que la pers sonne d devient «malade du d cancer» et reço oit des traitementss invasifs plus tôt dans le e décours de sa vie. v KCE Report 176 6B Dé épistage du cancerr du sein QUESTIO ONS POSÉES Ce rapport inv vestigue la question suivante: fau ut-il étendre le dé épistage organisé du can ncer du sein aux ffemmes âgées de e 70 à 74 ans? Si la réponse à cette question est négative, une e question subsid diaire se pose: que répo ondre à la personn ne de cette tranch he d’âge qui demande un dépistage? R RÉSULTAT TS ISSUS DE D LA LITT TÉRATURE E M Mortalité Les résultats des L s différents essais s contrôlés rando omisés permettent de m mettre en évidence les faits suivantts: • MÉTHOD DOLOGIE L’étude des bé énéfices cliniquess du dépistage se e base sur une revue de littérature effecttuée dans OVID M Medline, EMBASE E, CDSR et DAR RE. Cette revue a inclus les articles publiés en Anglais, Allemand, A Néerlandais et Français de jan nvier 2004 à avril 2 2011. L’évaluation du rapport bénéfices-risques de ce dépistage se base sur une revue des étud des de modélisattion recherchées dans Medline, Embase, E NHS EED et Econlit. E Cette revu ue a inclus les arrticles publiés en Anglais, Allemand, Néerrlandais et França ais de janvier 2000 0 à septembre 20 011. Afin de quantifiier le rapport bén néfices-risques da ans le contexte belge, un modèle a été construit dans cce but. La consttruction de ce modèle m a nécessité de rechercher dans Medline, Embase e, HTA EED et Psycinfo P (1950-10/2011)) les études relativves à la qualité de vie pendant et après le dépistage et le e traitement du cancer du sein.. Le modèle con ntient le maximum de do onnées belges utiilisables. Enfin, des reco ommandations de e bonne pratique e ont été rédigée es sur la base des élé éments de pre euve obtenus. Une révision desdites d recommandatio ons a été effectué ée par les expertts externes. Aucu un conflit d’intérêts n’a étté signalé. iii minution de morttalité de 23% sur une Le dépistage entraîne une dim uivi de 13 ans che ez les femmes de e plus de 50 ans ayant période de su bénéficié d’un n dépistage tous le es deux ans. • Cette diminuttion de mortalité se s manifeste entrre 4 et 7 ans aprrès le dépistage. Il convient donc de la mettre en perspective avec l’espérance de vie moyenne de ce groupe d’âg ge qui est de 16 ans a à 70 ans et de 13 1 ans à 74 ans (d données belges d de 2009). D Dans l’interprétatio on des études inte ernationales, il fau ut tenir compte du u petit n nombre de particip pantes âgées de 70 7 à 74 ans; consséquemment, l’effe et sur la a mortalité n’a pu être statistiqueme ent démontré pou ur celles-ci. M Morbidité Outre le gain en années O a de vie, le principal p avantage e attendu du dépistage e de permettre des traitements moins est m agressifs, vvu que le dépista age a p pour objectif de mettre m en éviden nce des petites tumeurs. Les don nnées b belges dont nous s disposons actu uellement ne nou us permettent pa as de v valider cette assertion. Les donnée es les plus récentes (rapport KCE 150) fo ont état de 58% de d chirurgie cons servatrice versus 38% de mastecto omies to otales dans les sttades les moins avancés a (Stades I and II). Près de e 90% d des bénéficiaires de la chirurgie conservatrice reççoivent égalemen nt un trraitement par rad diothérapie, 38% d’entre d elles reço oivent un traiteme ent de c chimiothérapie néo o-adjuvante et 41% un traitement h hormonal. D D’autre part, les essais e contrôlés randomisés r n’ont quantifié ni le tau ux de ré écidives ni l’évolu ution vers les stades métastatique es de la maladie. Il est d donc impossible d’infirmer d ou de co onfirmer l’hypothè èse d’une réductio on de m morbidité sur cette e base. Par contrre, la perte de qu ualité de vie impu utable a métastases es aux st inclue dans le modèle m (voir ci-de essous). iv Dé épistage du cancerr du sein ÉTUDES DE MODÉ ÉLISATION Les principales études de modé élisation ont été ré éalisées au sein du d projet CISNET (Canc cer Intervention a and Surveillance Modeling Network). Ces modèles avaie ent pour objectiif d’évaluer la contribution rela ative du dépistage par mammographie m ett du traitement adjuvant sur la rédu uction de la mortalité due e au cancer du sein observée au ux Etats-Unis de 1975 à 2000, et ils uttilisent les donné ées issues du Breast B Cancer Sc creening Consortium. Les résultats de e ces modèles ind diquent un gain en années de vie allant a de 9 à 22 ans parr 1.000 femmes d dépistées. D’autre es modèles n’utilis sant pas la méthodolog gie CISNET son nt également décrits d dans le rapport scientifique. Ces modèles ne n sont pas adapttables en tant que e tels à la situatio on belge car il est impos ssible d’y inclure les données belg ges. Un nouveau modèle spécifique a donc été construit. UN MODÈ ÈLE DE CO OHORTE POUR P LA BELGIQU UE Méthodologie e Le modèle construit pour ce rap pport est un modè èle de cohorte qu ui évolue par cycles annuels. Il compare deux cohortes th héoriques de fem mmes de plus de 70 ans s, l’une sans inviitation au dépista age (situation actuelle) et l’autre où les fe emmes continuen nt à être invitées au dépistage. Le taux de participation et la répartition dess cancers détectés s par le dépistage e versus les cancers d’in ntervalle sont conssidérés comme étant les mêmes que q dans la tranche d’âge e 50-69 ans. Le dépistage a pour objectif de m mettre en évidenc ce les tumeurs à un u stade précoce (I et II) afin d’évite er l’évolution ve ers le stade IV V (stade métastatique) qui q est incurable. Ce «stage-shift» implique que parmi tous les cancers dép pistés, la proportio on des stades pré écoces (I, II) augm mente en même temps que q la proportion n des stades ava ancés (II et IV) diminue. d D’autre part, nous avons émis l’hypothèse selon n laquelle la surv vie et la qualité de vie dépendent d de l’âg ge de la patiente et e du stade de la tumeur. Elle ne tient pa as compte du fait que le pronostic des d cancers déte ectés par KCE Report 176B 1 le e dépistage est meilleur que cellui des cancers détectés clinique ement (c cancers d’interva alle et cancers survenant che ez les femmes nonp participantes). P Paramètres Ce modèle explloite au maximu C um les données belges, à sa avoir : l’espérance de vie moyenne des s femmes selon leur âge (2009)), les d données du regis stre du cancer (C Communauté Fla amande), les don nnées is ssues du progra amme de dépis stage actuel (50 0-69 ans), le temps n nécessaire pour infirmer un dia agnostic faussem ment positif (Ag gence In ntermutualiste, AIIM/IMA) et les do onnées de survie à cinq ans en fon nction d stade (Registre du e du Cancer). Les données de la Communauté Flam mande o été privilégiées ont s car elles sont pllus complètes et p parce que le dépistage o opportuniste après s 70 ans y est mo oins fréquent que dans le reste du pays. L durée de l’avan La nce au diagnostic et le pourcentage e de sur-diagnosttic ont é estimées au dé été épart de l’analyse e de la littérature. M Mesure de la qualité de vie Les données sur la qualité de vie L e pendant le dépistage et le traite ement p proviennent de la littérature. L’instrrument utilisé pou ur décrire les éta ats de s santé est l’EQ-5 5D (European Quality of Life--5 Dimensions); ces d descriptions ont été é valorisées parr la population gé énérale anglaise («UK ta ariffs»). Nous ne e disposons pas de données rela atives à la population b belge. L variations de la qualité de vie des Les d femmes de plus de 70 ans utillisées d dans le modèles sont s les suivantes: • • La perte de qualité de vie co onsécutive à un résultat de dépistage faussement positif est estimée à 16% pendant 4 45 jours. entes cancéreuses et pendant la première année qu ui suit Pour les patie le diagnostic (quel ( que soit le traitement) t la pertte de qualité de vie est estimée à 16% pour les stade es I, II, III et à 18 8% pour les stade es IV. Pendant les années a suivantes, la perte de qualitté de vie est estim mée à 6% pour les stades I, II, III. Cette C perte demeu ure stationnaire (18%) ( pour les stade es IV. P Plusieurs limitatio ons de cette app proche nous oblig gent à interpréter ces c chiffres avec préca aution KCE Report 176 6B Dé épistage du cancerr du sein Résultats Le scénario de d base montre que le dépistag ge entre 70 et 74 ans permettrait d’év viter 1,3 décès po our 1000 femmes s qui y participentt, ce qui représente une e réduction de 2 21% des décès. Globalement, le nombre d’années de vie e sauvées est estiimé à 13,1 et le gain en QALY à 3,9. Etant donné qu u’il existe une inccertitude importan nte (pour les déta ails, voir les discussions s dans le rapport scientifique) au sujet s des ces estim mations, une analyse de d sensibilité du u modèle a été réalisée. Cette analyse comprend un sc cénario pessimistte et un scénario optimiste. o Le scenario pe essimiste fait l’hyypothèse d’un ex xcédent de diagnostic de 20%, d’un taux de faux positifs d de 10%, entrainan nt une perte de qu ualité de vie de 0,19 perd durant pendant 54 4 jours (temps né écessaire pour infiirmer les résultats). La distribution des cancers dépisté és par stades observée o actuellement da ans le cadre du d dépistage organisé en Flandre (50--69 ans) a été appliqué ée au groupe d dépisté. Ce scén nario pessimiste permet d’estimer un ga ain de 8,7 année es de vie et une perte de 3,1 QA ALY pour 1000 femmes participant au dé épistage. Ceci sig gnifie que dans certaines c circonstances, au demeurant ttout à fait réalisttes, le dépistage e puisse aboutir à une pe erte en terme de q qualité de vie. Le scenario optimiste fait l’hypotthèse d’un excédent de diagnostic c de 3%, d’un taux de faux positifs de 2% %, entrainant une perte p de qualité de d vie de 0,13 perdurant pendant 36 jourss. Ce scénario ap pplique au groupe e dépisté la distribution par stades obsservée actuellem ment dans le ca adre du dépistage orga anisé aux Pays-B Bas (70-74 ans)). Ce scénario optimiste o permet d’estime er un gain de 17,0 années de vie et un gain de 16,,3 QALY pour 1000 fem mmes participant au dépistage. Ceci signifie qu’’il serait nécessaire d’inv viter 67 femmes à participer au dé épistage pendant cinq c ans pour gagner un QALY. v C CONCLUS ION Le dépistage estt organisé dans le but d’améliorrer le bien-être de L d la p population en évitant notamment des décès prématu urés. Il est certain n que p prolonger le dépistage jusqu’à l’âge e de 74 ans devra ait permettre de ga agner q quelques années de vie pour un certain nombre d de femmes. Touttefois, l’influence d’un dé épistage organisé sur la qualité de vie est nettement plus a aléatoire (niveau de d preuve très faible car basé sur u un modèle). Selon des h hypothèses raison nnables, cette inte ervention pourrait même aboutir à une p perte en terme de e qualité de vie. Dans D ces condition ns, il se pourrait que q la b balance bénéfices s-risques de ce dépistage d penche e du côté d’une perte g globale de bien-ê être de la population. Il n’est do onc pas recomm mandé d d’étendre le dépis stage organisé du u cancer du sein a aux femmes âgée es de 7 à 74 ans. 70 vi Dé épistage du cancerr du sein RECO OMMAND DATIONS Sa a KCE Report 176B 1 • L’invittation systématiique des femme es âgées de 70 à 74 ans à parrticiper au dépis stage organisé du cancer du u sein n’est pas recommandée. r • Si une e personne de plus de 70 ans demande une mammographie dans un objecttif de dépisttage, il importe que q le médecin veille v à ce qu’elle e soit bien inform mée des avantag ges et des in nconvénients pottentiels de celle-ci. • Toute mammographie e de dépistage doit répondre aux x exigences euro opéennes en ma atière de qualité, dont notam mment : le contrrôle de la qualité é des installation ns, la double lec cture, l’enreg gistrement et l’optimisation du taux de rappel. C’est pourrquoi, les méde ecins orienteront la personn ne qui demande e un dépistage vers v une structurre qui réponde à ces exigen nces de qualité. • Afin de d minimiser le riisque de perte de qualité de vie lié aux résultats faussement pos sitifs, il impo orte que le taux de rappel après mammographie e soit le plus bas s possible et res ste en desso ous du seuil défin ni par les critères s européens (<5% %). Le KCE re este seul responsab ble des recommandations faites aux au utorités publiques KCE Report 176 6 S Screening Breast Cancer C 1 TABL LE OF CO ONTENT TS LIST OF FIGURES S ................................................................................................................................................. 4 LIST OF TABLES ................................................................................................................................................... 4 LIST OF ABBREVIIATIONS ................................................................................................................................... 5 1. 2. 3. SYNTHÈS SE .............................................................................................................................................. 7 CONTEXT TE .............................................................................................................................................. 7 QUESTIONS POSÉES ............................................................................................................................. 7 DESCRIPT TION DE LA PRO OBLÉMATIQUE ........................................................................................... 8 3.1. APPROCH HE INTUITIVE ........................................................................................................................... 8 3.2. APPROCH HE ÉPIDÉMIOLOGIQUE ......................................................................................................... 8 3.2.1. Objectif O à court terrme ................................................................................................................ 9 3.2.2. Objectif O ultime............................................................................................................................. 9 3.2.3. Faux F positifs et dia agnostics excéden ntaires ............................................................................... 9 MÉTHODO OLOGIE .................................................................................................................................. 11 4. 4.1. ESTIMATIION DES BÉNÉFICES DU DÉPIST TAGE ............................................................................... 11 4.1.1. Diminution D de la mortalité m ........................................................................................................ 11 4.1.2. Amélioration A de la qualité de vie des s patientes ....................................................................... 11 4.2. ESTIMATIION DES INCONV VÉNIENTS DU DÉPISTAGE...................................................................... 11 4.2.1. Diminution D de la qu ualité de vie des participantes p .................................................................... 11 4.3. APPROCH HE PAR MODÉLISATION ...................................................................................................... 12 4.3.1. Mesures M de la qua alité de la vie................................................................................................ 12 4.3.2. Description D du mod dèle ............................................................................................................ 12 4.3.3. Hypothèses H de bas se................................................................................................................ 14 4.3.4. Alimentation A du mo odèle ........................................................................................................... 14 4.3.5. Analyse A de sensibilité .............................................................................................................. 14 RÉSULTA ATS .......................................................................................................................................... 15 DISCUSSION ......................................................................................................................................... 15 5. 6. 2 S Screening Breast Cancer C KCE Reportt 176 6.1. AJOUTER R DES ANNÉES À LA VIE ? ................................................................................................... 15 6.2. 7. AJOUTER R DE LA (QUALITÉ DE) VIE AUX ANNÉES A ? ....................................................................... 16 6.2.1. Traitements T moins s agressifs? ................................................................................................. 16 6.2.2. Faux F positifs ............................................................................................................................. 16 6.2.3. Excès E de diagnosttics et de traiteme ents................................................................................... 16 CONCLUS SIONS ..................................................................................................................................... 16 7.1. FAUT-IL PROLONGER P LE DÉPISTAGE JUS SQU’À L’ÂGE DE 74 ANS ? .......................................... 16 7.2. QUE REPONDRE A LA PE ERSONNE QUI DE EMANDE UN DEP PISTAGE? .......................................... 17 7.3. 8. MESSAGE E CLÉ ...................................................................................................................................... 17 RÉFÉREN NCES ....................................................................................................................................... 18 1. SCIENTIF FIC REPORT ............................................................................................................................ 19 INTRODUCTION .................................................................................................................................... 19 1.1. CONTEXT T OF THIS REPORT .............................................................................................................. 19 1.2. SCOPE OF O THIS REPORT T ................................................................................................................... 19 1.3. BREAST CANCER C SCREE ENING IN BELGIU UM.................................................................................... 19 1.4. CLINICAL QUESTIONS ......................................................................................................................... 20 1.5. 2. SCIENTIF FIC APPROACH ...................................................................................................................... 20 LITERATU URE REVIEWS ....................................................................................................................... 21 2.1. REVIEW OF O CLINICAL STU UDIES......................................................................................................... 21 2.1.1. Methodology M ............................................................................................................................. 21 2.1.2. Description D of scre eening benefit .............................................................................................. 22 2.1.3. Description D of scre eening harms............................................................................................... 25 2.1.4. Screening S conditio ons ............................................................................................................... 27 2.1.5. Key K data ................................................................................................................................... 28 2.1.6. Conclusion C ................................................................................................................................ 29 2.2. REVIEW OF O MODELING STUDIES S ...................................................................................................... 29 2.2.1. Literature L search strategy s ....................................................................................................... 29 2.2.2. Selection S criteria ....................................................................................................................... 29 KCE Report 176 6 S Screening Breast Cancer C 2.2.3. 2.2.4. 2.2.5. 2.3. 3 Quantity of researc Q ch available ................................................................................................. 30 S Selected studies ....................................................................................................................... 30 C Conclusion ................................................................................................................................ 33 3. REVIEW OF O QUALITY OF LIFE STUDIES ........................................................................................... 33 2.3.1. Methods M .................................................................................................................................... 34 2.3.2. Results R ..................................................................................................................................... 35 2.3.3. Discussion D ................................................................................................................................ 43 DECISION N ANALYSIS ........................................................................................................................... 43 3.1. DATA SOU URCES ................................................................................................................................... 44 3.2. MODEL DESCRIPTION D ......................................................................................................................... 44 3.3. DESCRIPT TION OF THE PA ARAMETERS .............................................................................................. 48 3.3.1. Age A specific overa all survival .................................................................................................... 48 3.3.2. Breast B cancer incid dence .......................................................................................................... 48 3.3.3. Participation P rate ...................................................................................................................... 48 3.3.4. Proportion P of scree en detected breas st cancers ........................................................................ 48 3.3.5. Recall R rate ................................................................................................................................ 48 3.3.6. Stage S distribution and a stage shift ............................................................................................ 49 3.3.7. Stage S specific rela ative survival ................................................................................................ 50 3.3.8. QALY Q ........................................................................................................................................ 51 3.4. RESULTS S ............................................................................................................................................... 54 3.5. 4. DISCUSSION ......................................................................................................................................... 58 ANSWER TO CLINICAL QUESTIONS Q ................................................................................................. 60 4.1. BREAST CANCER C RELATE ED MORTALITY ......................................................................................... 60 4.2. DELAY BE ETWEEN THE SC CREENING AND THE T MORTALITY Y REDUCTION ................................... 60 4.3. OVERALL L MORTALITY ......................................................................................................................... 60 4.4. MORBIDIT TY ............................................................................................................................................ 60 4.5. FALSE PO OSITIVE OR FALS SE NEGATIVE RESULTS .......................................................................... 60 4.6. ADDITION NAL DIAGNOSTIC C TESTS ..................................................................................................... 60 4.7. OVER-DIA AGNOSIS AND OVER-TREATMEN O NT .................................................................................... 60 4 S Screening Breast Cancer C 4.8. 5. LIST OF FIGURES F LIST OF TABLES T KCE Reportt 176 WHAT AT TTITUDE SHOULD D BE RECOMMENDED FOR WOM MEN IN CASE OF F SELF REFERRA AL?61 REFEREN NCES ....................................................................................................................................... 62 s for which uttilities are needed d (reflection process) .................................................................... Figure 2.1: Health states Figure 2.2: Percenttage change in utilities .......................................................................................................................... Figure 3.1: Comparrison of the two co ohorts with and without a screening g program.......................................................... Figure 3.2: Comparrtments in the two o cohorts and the transitions t betwee en them ............................................................ Ta able 2.1: Data issu ued from clinical literature review ................................................................................................... Ta able 2.2: Selection n criteria ........................................................................................................................................... Ta able 2.3: Modeling g studies excluded d after full-text ass sessment ................................................................................. Ta able 2.4: results of the different models m in terms of o mortality reduc ction and years o of life gained perr 1000 wo omen screened fo or the different mo odels ..................................................................................................................... Ta able 2.5: Article se election criteria ................................................................................................................................. Ta able 2.6: Health sttates descriptions s for the study of Lidgren L et al. ............................................................................. Ta able 2.7: Description of a “false pos sitive” state (Gerard et al)83.................................................................................. Ta able 2.8: Description of the selected d utilities .............................................................................................................. Ta able 3.1: Stage diistribution among screen detected breast cancers, interval cancers a and cancers amon ng non pa articipants, age 50 0-69, Flemish scre eening program 2001-2006. ................................................................................ Ta able 3.2: Parametters used in the model m ..................................................................................................................... Ta able 3.3 Modeling results: baseline, worst and best case c scenario. .......................................................................... Ta able 3.4 Modeling results: sensitivitty analysis. ........................................................................................................... KCE Report 176 6 LIST OF ABBREVIA A ATIONS S Screening Breast Cancer C AB BBREVIATION DEFINITION CP PG CC CRT CII DC CIS DE ET BC CSC AH HRQ BC CR DN NETB CIISNET IM MA/AIM IN NAMI/RIZIV IC CER KC CE MST M-A NIIS NB BSS NB BCSP NH HS NH HS EED NC CI QA ALY Qo oL RC CT RR R ce Guideline Clinical Practic Cochrane Cen ntral Register of Controlled C Trials Confidence Intterval Ductal Carcino oma in situ Data Extractio on Table Breast Cancerr Surveillance Con nsortium (USA) Agency for He ealth Care Researrch and Quality Belgian Cance er Registry Dutch Nationa al Evaluation Team m for Breast cance er screening Cancer Interve ention and Surveillance Modelling Network N Intermutualistic Agency National Institu ute for Health and d Disability Insuran nce Incremental co ost-effectiveness ratio r Belgian Health hcare Knowledge Centre Mean Sojourn Time Meta-analysis National Institu ute for Statistics Canadian Natiional Breast Canc cer Screening Study Norwegian Bre east Cancer Scree ening Programme es National Healtth Service (UK) NHS Economic Evaluation Data abase National Canc cer Institute (USA)) Quality Adjuste ed Life Year Quality of Life Randomized Controlled C Trial Relative Risk 5 6 S Screening Breast Cancer C SE EER SR R ST T TT TO UK K US SA US SPSTF Surveillance, Epidemiology E and d End Results (US SA) Systematic Re eview Sojourn Time Time-trade-offf United Kingdo om Unites States of America US Preventive e Services Task Force KCE Reportt 176 KCE Report 176 6 SYNT THÈSE Screening Breast Cancer C 7 1 CONTEX 1. XTE Le KCE a déjà publié trois rapports L s sur le dépistage du cancer du sein. Le ra apport de base publié en 2005 (rapport N°11 du KCE) concernait le d dépistage du canc cer du sein en gé énéral, dans la po opulation sans fac cteurs d risque. Le dépistage du cancer du de d sein des femm mes de la tranche d’âge 4 40-49 ans a fait l’o objet d’une mise à jour partielle pub bliée en 2010. Dans ce ra apport (rapport N° 129 du KCE E), le KCE ne recommandait pa as le d dépistage systématique des femm mes de moins de 50 ans. Le trois sième ra apport (rapport 172 1 du KCE), pu ublié en 2012, a posé le problèm me de l’indentification de es femmes expos sées à un risque e accru de cance er du s sein. Le rapport actuel pose la question de l’exxtension du dépistage o organisé du cance er du sein aux fem mmes âgées de 70 0 à 74 ans C Cette question es st régulièrement adressée aux po oliticiens en raiso on de l’augmentation rég gulière de l’espérrance de vie de la population féminine. S la plupart des Si s groupes actifs dans le dépista age demandent cette p prolongation, les autorités publiqu ues font preuve d de moins d’unanimité. S Seuls quatre Etatts membres de l’Union européen nne ciblent la tra anche d d’âge des 70-74 (la France, les Pays-Bas, P l’Espag gne et la Suède)1. Les a autres pays insiste ent sur la nécessité d’informer les ffemmes et de parrtager a avec elles la prise de décision. 2 QUESTIIONS POSÉ 2. ÉES Le dépistage orga L anisé du cancer du sein devrait-il être prolongé jusqu’à l’âge de 74 ans? Si S la réponse à ce ette question est n négative, que répo ondre à la personne qui demande d ce dépistage? L première questtion concerne plus La s spécifiquement les pouvoirs publlics et la a seconde, les pre estataires de soin ns. 8 S Screening Breast Cancer C 3. DESCRIPTION D DE LA PRO OBLÉMATIQ QUE 3.1. Approche intuitive De façon intuittive, le dépistage e du cancer fait sens. Les médias sont généralement enthousiastes e à ll’égard du dépista age. Cette attitud de a été démontrée par Schwartz au dé ébut du 21e siècle e2. Une enquête réalisée onsidéraient que dépister aux Etats-Unis a révélé que 87% des adultes co des personnes intterrogées déclara aient que est une bonne idée. Trois quart d u cancer à un sstade précoce sa auve la vie la plu upart du diagnostiquer un temps. L’enthousiasme des rép pondants était si fort que pour la majorité sion à prendre mais m un d’entre eux le dépistage n’était pas une décis impératif moral3. Cette attitude générale que no ous pouvons rés sumer ainsi “la détection d c sauve de es vies” peut av voir suscité des attentes précoce des cancers irréalistes de la l part des femm mes. Silverman a réalisé des interviews téléphoniques pour p évaluer comment les femmes considèrent le ca ancer du sein et le bénéfice du dépistag ge par mammographie4. La majo orité des c le ccancer du sein comme une maladie répondantes considérait progressive un niforme et croyaitt que tous les ca ancers débutent par une forme curable et e silencieuse. En n résumé, ces fem mmes pensaient que q si le cancer du sein n'est pas détecté é par une mammo ographie et traité de d façon andit, se propage e et tue. Fortes s de ces croyances, les précoce, il gra femmes estima aient que les canccers avancés (et sans s doute la plupart des cancers mortels s) sont liés à un é échec au niveau du dépistage préco oce. Schwartz a sou uligné que 94% de es femmes ne sav vent pas que le dé épistage peut détecter des d cancers qui n ne vont jamais progresser. p De plu us, 92% des répondante es sont persuadé ées du fait que la a mammographie ne peut faire de tort à une personne qui n n'a pas de cancerr du sein5. Le corps médical lui-même n’a appréhende pas toujours le dépis stage de ue de nombreux cliniciens c restent focalisés f manière adéquate. C’est ainsi qu sur le taux de cancers diagnosstiqués (objectif intermédiaire), allors que d dépistage est de diminuer la mortalité. m D’autre part, les l’objectif final du cliniciens parais ssent plus sensib bilisés au risque de d méconnaître un n cancer (faux négatif) qu’aux risques liés aux résultats faussement positifs. KCE Reportt 176 3 3.2. Approche épidémiologiq que Le cancer du sein est le cancer le plus fréquent chez la femme L e. En B Belgique, 10.849 cancers c du sein ont o été diagnostiq qués en 2008. Plu us de trrois quarts des ca ancers du sein so ont diagnostiqués après l’âge de 50 0 ans. L L’âge moyen au moment du diag gnostic est de 62 2 ans. L’incidenc ce du c cancer du sein estt de 370,7/100.00 00 dans le groupe des femmes âgées de 7 à 75 ans6. 70 N Néanmoins, la part relative de la mortalité m due au ccancer du sein da ans la m mortalité totale difffère en fonction de l’âge. En 1999, le cancer du sein n était re esponsable de 18 8% des décès ch hez les femmes â âgées de 50 à 54 4 ans, d 13% dans le groupe de 60 à 64 de 4 ans et de 6% dans le groupe de e 70 à 7 ans (Rapport N°11 du KCE). En 74 E 2006, cette prroportion était de 14% p pour les femmes âgées â de 50 à 54 4, 12% pour le grroupe de 60 à 64 4 ans, 7 pour le groupe 7% e des 70 à 74 ans et 5% pour le gro oupe des 75 à 79 ans6. C Caractéristiques fo ondamentales d’u un dépistage: 1. Le dépistage s’adresse à des personne es en bonne santé Contraireme ent au patient qu ui consulte son m médecin en raiso on d’ une plainte ou d’un symptô ôme, la personn ne qui participe à un dépistage es st présumée inde emne de la malad die recherchée. 2 2. Le dépistage a pour objectif à court terme de cconfirmer l’absenc ce de la maladie. 3 3. Le dépistage a pour objectif ultime u de diminuer la mortalité/morrbidité liée à la maladie. 4 4. Le principe “p primum non nocere” est particulière ement d’applicatio on en ce qui concerrne le dépistage. Rappelons que pour R p mille femme es dépistées enttre 70 et 74 ans, plus d 990 sont indem de mnes du cancer du sein. KCE Report 176 6 S Screening Breast Cancer C 9 3.2.1. Objecttif à court terme 3 3.2.3. Faux pos sitifs et diagnostiics excédentaire es Le dépistage a pour objectif de confirmer l’abs sence de la mala adie. La personne qui participe au dépistage bénéfic cie de la “prés somption n ce qui concerne e le cancer du sein. A l’inverse, la patiente d’innocence” en qui consulte so on médecin parce e qu’elle a une plainte p ou parce qu’elle q a constaté quelq que chose d’inhabituel, devient “suspecte” de maladie. m L’objectif du médecin m et les mo oyens à mettre en e œuvre dans ces c deux situations sont diamétralement o opposés. Dans le e cas d’une mise au point e médecin a le de evoir de tout mettre en œuvre pourr trouver diagnostique, le une étiologie à la plainte ou au symptôme. A l’in nverse, dans le cadre c du m a le devoir de pratiquer uniquement u les examens e dépistage, le médecin indispensables.. Ceci afin de min nimiser les risque es et les inconvén nients du dépistage pourr les 996 femmes (/1.000) qui sontt indemnes du ca ancer du sein. d médecins éttant essentiellement effectuée en n hôpital Le formation des auprès de malades, ce change ement de point de vue est franc chement our un clinicien. contre-intuitif po Avant d’instaurer un A u dépistage orga anisé, il est nécesssaire de s’assure er que la a balance avanta ages/inconvénien nts du dépistage penche du côté é des a avantages. Pour ce faire, l’ample eur de la diminu ution de mortalité é doit c contrebalancer la perte de qualité de d vie consécutive e aux inconvénients et a risques induits aux s par le dépistage e. L Les résultats dits s: “faux-positifs” (suspicion de lésion cancéreus se en d dehors de la prése ence d’un cancer)) sont les effets négatifs indésirable es du d dépistage du canc cer du sein les plus fréquents. Cess résultats fausse ement p positifs créent de e hauts niveaux x d'anxiété et so ont suivis d’examens c complémentaires. P Plus encore que le es faux–positifs, le risque de diagn nostic excédentairre est le e risque majeur du dépistage des femmes âgéess de 70 à 74 ans. Le d diagnostic excéde entaire peut se dé éfinir comme le fa ait de diagnostiquer un c cancer dont l’évo olution est telle qu’il ne se serait jamais manifesté c cliniquement en l’a absence de dépis stage8. Ce risque est d’autant plus élevé q que le cancer es st d’évolution len nte et que l’esp pérance de vie de d la p personne est faib ble. Ce risque est e particulièreme ent méconnu dans la p population. Très peu de femmes savent en effet que certains cancers é évoluent tellemen nt lentement que e même s'ils ne e sont pas traité és ils n n'altèreront pas la santé9. C rapport a pourr objectif de quanttifier les avantage Ce es et les inconvén nients (v voir Figure 1) de ce dépistage afin de pouvoir les m mettre en perspecttive et s s’assurer que les bénéfices l'emporrtent largement su ur les risques de perte d qualité de vie. de 3.2.2. Objecttif ultime Diagnostiquer les cancers à un stade préco oce avant qu’ils ne se étastases) est l’h hypothèse fondatrice du développent ett essaiment (mé dépistage du ca ancer. C’est ainsi que l’on attend du d dépistage qu’il diminue la mortalité spé écifique à la mala adie et conséquem mment la mortalitté totale. Le fait que la technologie t utilisé ée permette de diagnostiquer d des s lésions peu avancées et donc potentiellement curable es ne représente e qu’une étape intermédiaire dans ce proccessus. Il s’agit d’une condition nécessaire p suffisante7. mais qui n’est pas On peut égale ement émettre l’’hypothèse que le dépistage réd duise la morbidité liée à la maladie, en p permettant l’utilisation de traitementts moins invasifs (maste ectomies partielles plutôt que mas stectomies totales s) et en évitant une parttie des évolutions vers les stades métastatiques. m 10 S Screening Breast Cancer C KCE Reportt 176 Figure 1 mise en perspective d des avantages ett des inconvénients potentiels du u dépistage. Faux-négaatif Rééassurance innadéquate Résultat - Rééassurance Diagnostic retardé r Normal Dépisttage par mammo ographie Faux-possitif Exxamens complémeentaires Anormal Résultat + Canncer invaasif Caancer inn situ Traitement précoce Traitement précoce Diminuttion de mortaalité Surtraittement KCE Report 176 6 S Screening Breast Cancer C 4. MÉTHODOLOGIE Nous avons rec cherché des élém ments de réponse aux questions prrécitées, dans la littératu ure clinique, danss les études de modélisation et dans d les données nation nales et internatio onales. Ces reche erches ont été me enées en suivant les procédures en vigue eur au KCE. Elles s sont décrites en n détails e 2 du rapport scie entifique. dans le chapitre 4.1. Estimattion des bénéfices du dépista age 4.1.1. Diminu ution de la morta alité Les principales s données proban ntes relatives au dépistage du ca ancer du sein, sont issue es de huit essaiss contrôlés rando omisés. Sur base e de ces essais, on peutt retenir deux consstats principaux: 1. Le dépista age entraîne une diminution de mortalité m de 23% sur une période de e suivi de 13 ans pour les femmes de plus de 50 an ns ayant bénéficié d’un d dépistage tou us les deux ans. 2. Cette dimin nution de mortalité se manifeste prrincipalement entrre 4 et 7 ans après le dépistage. Il cconvient de la me ettre en perspectiive avec e de vie de la pop pulation-cible. L’es spérance de vie moyenne m l’espérance de ce grou upe d’âge est de 16 ans à 70 ans s et de 13 ans à 74 ans (données belges b de 2009). Les données probantes p issues de ces essais contrôlés c random misés ne peuvent donner une réponse complète à notre qu uestion de base. En E effet, andomisé, l’étude e suédoise dite de es “Two County”, a inclus un seul essai ra des femmes âgées â de 70 à 7 74 ans et le nombre de septuag génaires participant à ce et essai était trop ffaible (10.000 pou ur les deux groupes) pour pouvoir mettre en évidence un effet statistiqu uement significatiff sur la ait entachée de biiais méthodologiques. mortalité. De plus cette étude éta 11 4 4.1.2. Amélioration de la qualité é de vie des patiientes Le dépistage ayant pour objectif de L d mettre en évid dence des tumeurs de p petite taille, un de es avantages atte endus est de perm mettre des traitem ments m moins agressifs. Ni N les données iss sues de essais co ontrôlés randomisés, ni le es données factuelles recueillies en e Belgique, ne pe ermettent de conffirmer c cette attente. L essais contrôlés randomisés n’ont Les n quantifié ni le taux de récidiv ves ni l’évolution vers le es stades métas statiques de la maladie. Il est donc mpossible d’infirm mer ou de conffirmer l’hypothèse e d’une réductio on de im m morbidité sur cette e base. Par contrre, la perte de qu ualité de vie impu utable a métastases es aux st inclue dans le modèle m décrit ci-dessous. L Les données be elges dont nous disposons acctuellement ne nous p permettent pas de e valider cette assertion. Les donn nées les plus récentes (rapport KCE 150)) font état de 58% % de chirurgie con nservatrice versus s 38% d mastectomies totales dans les stades les moinss avancés (Stades I et de III). Près de 90% des d bénéficiaires de la chirurgie cconservatrice reço oivent é également un traittement par radiothérapie, 38% d’entre elles reçoive ent un trraitement de ch himiothérapie né éo-adjuvante et 41% un traite ement h hormonal. 4 4.2. Estimation n des inconvén nients du dépisstage 4 4.2.1. Diminutio on de la qualité de d vie des particcipantes Le dépistage prov L voque une diminuttion de la qualité d de vie d’une partie e des p personnes dépisté ées. Ceci s'expliqu ue par une série d de facteurs: 1. Les résultats faussement pos sitifs du dépistage e sont perçus pa ar les mme de vrais pos sitifs, aussi longtemps que les examens patientes com complémentaires n’ont pas pe ermis de les infirm mer. Ils provoque ent de ar rapport au canc cer du sein et auxx procédures inva asives l'inquiétude pa telles que les ponctions mamm maires. 2 Les diagnostic 2. cs excédentaires et les traitementss qui les suivent (over( diagnosis and d over-treatment,, pour plus de d détails, voir le ra apport scientifique) conduisent c à des inquiétudes grave es et à des traitem ments lourds dont de es amputations mammaires m qui n’o ont pas d’influenc ce sur la survie de la a personne. 12 3. S Screening Breast Cancer C L’avance au a diagnostic peu ut entrainer une perte p de plusieurs années de vie en bonne b santé. Le d dépistage a pour objectif de diagn nostiquer le cancer plus p précocementt que ne le ferait un u diagnostic clin nique. La patiente de evient de ce fait m malade du cancerr plus tôt dans le décours de sa vie. Toutefois, si cette patiente e décède d’une e cause elui-ci n’ait eu le e temps indépendante de son canccer avant que ce alade du cancer” quelques années s trop tôt d’évoluer, elle aura été “ma u diagnostic et au a traitement n’a aient pu sans que cette avance au s espérance de e vie10. influencer son 4.3. Approche par modélissation Les revues de littérature précitée es ne nous ayantt pas permis de quantifier q b et des risques, nous av vons construit un modèle le poids des bénéfices spécifique dans ce but. La construction de ce modèle a nécessité de e des femmes pe endant le rechercher les études relatives à la qualité de vie adie. dépistage et à la qualité de vie des patientes au cours de leur mala 4.3.1. Mesure res de la qualité de la vie Différents instruments sont disp ponibles pour me esurer la qualité de vie. ments sont spécifiquement adaptés à la maladie, comme Certains instrum par exemple, le l questionnaire relatif à la qualité de vie des patientes p atteintes d'un cancer c du sein de l'European Organ nization for Resea arch and Treatment of Cancer C (EORTC).. Ces outils évalu uent l'image du corps, c le fonctionnementt physiologique, la peur de la récidive… Toutefois, il n'est pas possible de prendre en compte ces do onnées de santé é multis dans un modè èle. Elles doiven nt être converties s en un dimensionnelles indice global de d qualité de vie e, à savoir, le Quality-Adjusted Q L Life-Year (QALY). Les QALYs sont le nom mbre d’années de e vie ajustées à la a qualité de vie. d KCE, considèrrent que Les recommandations pharmacco-économiques du ensions) le questionnaire appelé EQ-5D (European Quality of Life-5 Dime eilleurs instrumen nts disponibles pour p évaluer les QALYs. est un des me Avec cet instrument, la qualité d de la vie liée à l'état de santé est mesurée m obilité, l’autonomiie de la en prenant en compte cinq dimensions: la mo antes, la doule eur/la gêne, l’anxiété/la personne, les activités coura our chacune de cces dimensions, plusieurs répons ses sont dépression. Po possibles. Celles-ci reflètent le niveau de sévérité du problème e (aucun KCE Reportt 176 problème, quelqu p ues problèmes, des problèmess modérés, ou des p problèmes graves) Ce questionnairre est soumis à la population conce ernée, s en ce qui conc soit cerne le dépistage, une population n de femmes indemnes d cancer du sein du n et en ce qui co oncerne la malad die, une populatio on de p personnes atteintes de ce cancerr. La revue de lla littérature a permis d d’identifier trois éttudes qui correspondaient à nos ccritères d’inclusion n. Sur b base de ces étude es, les variations de la qualité de vvie des septuagén naires s sont estimées com mme suit: 1. La perte de qualité de vie co onsécutive à un résultat de dépistage p est estimée e à 16% pendant la période néces ssaire faussement positif pour infirmerr ce faux positif.. En Belgique, ccette période durre en moyenne 45 5 jours (minimum m 36, maximum 54 jours) selon n les données AIM (Agence Intermu utualiste) 2 Pour les patie 2. entes cancéreuses et pendant la première année qu ui suit le diagnostic (quel ( que soit le traitement), la pertte de qualité de vie v est estimée à 16% pour les stade es I, II, III et à 18 8% pour les stade es IV. a suivantes, la perte de qualitté de vie est estim mée à Pendant les années 6% pour les stades I,II,III. Ce ette perte demeu ure stationnaire (18%) ( es IV. pour les stade P Plusieurs limitatio ons de cette app proche nous oblig gent à interpréter ces c chiffres avec préc caution. Il s’agit de d résultats provvenant de pays angloa s saxons. Le questio onnaire utilisé, à savoir, l’EQ-5D m mesure les dimen nsions s sanitaires générales et non les dim mensions spécifiqu ues au cancer du sein. L mesures conc Les cernant les patien ntes ne prennent q ent en que sommaireme c compte l’impact à court terme du d diagnostic ett de la chirurgie e. Ce q questionnaire aya ant été utilisé lorrs des consultatio ons ambulatoires s; ses ré ésultats ne reflè ètent pas la qua alité de vie des patientes grave ement m malades ne pouva ant plus se déplac cer. Les particularités de l’étude uttilisée p pourraient expliquer le faible chang gement de qualité é de vie constaté entre le es patientes ayan nt un cancer du sein s et la populattion générale ou entre le es patientes ayan nt développé des métastases m et celles qui n’en ont pa as. 4 4.3.2. Descriptio on du modèle Le modèle compa L are deux cohorte es théoriques. Ce es deux cohortes s sont c constituées de 100.000 femmes do ont l’évolution est suivie jusqu’à la mort. L schéma ci-dess Le sous représente cette c évolution: KCE Report 176 6 S Screening Breast Cancer C (1) cancer invasif diagnostiqué lors du dép pistage Femmes invitées Cohortte A (2) cancer d'intervalle (3 3) cancer invasif diagnostiqué dans le grroupe des non‐ paarticipantes 13 I II III IV I II III IV (5) Décès (toutes causes co onfondues) I II III IV (4)) Cancer canalaire in situ Cohortee B Femmes non invitées (6 6) cancer invasif diagnostiqué dans le grroupe des non‐ in nvitées (7 7) Cancer canalaire in n situ diagnostiqué d dans le groupe des n non‐invitées I II III IV (8) Décès (toutes causes co onfondues) 14 S Screening Breast Cancer C La cohorte A illustre l’hypothèse e d’une prolongatio on du dépistage organisé o s. Elle est constiituée des femme es invitées au dé épistage. jusqu’à 74 ans Parmi celles-ci, certaines participent au dépistag ge (invitées/partic cipantes) c survenantt dans la et d’autres non (invitées/non-parrticipantes). Les cancers épertoriés. Il s’ag git soit de cance ers diagnostiqués lors du cohorte sont ré dépistage (1), soit de cancers d diagnostiqués dans l’intervalle enttre deux épistage (2), soit de cancer diagn nostiqués dans le e groupe sessions de dé des invitées/no on-participantes ((3). Enfin, les ca ancers canalaires s in situ peuvent surven nir dans le group pe des invitées/p participantes tout comme dans le groupe e des invitées/non n-participantes (4)). La très grande majorité des femmes co onstituant cette co ohorte décèdera d’une autre affec ction que le cancer du se ein (5). La cohorte B (c cohorte de contrô ôle) correspond à la situation actue elle. Les membres de ce ette cohorte ne ssont pas invitées au dépistage. Certaines C femmes serontt atteintes d’un cancer invasif (6 6), d’autres d’un n cancer canalaire in situ u (7). La très gran nde majorité des femmes constitua ant cette cohorte décède era d’une autre afffection que le can ncer du sein (8). Le cancer du sein s évolue en qu uatre stades (I, II, III, IV). Le stade e I est le stade le moins avancé. La survie e est d’autant moins bonne et le tra aitement ourd et plus invassif que le stade est avancé au moment du d’autant plus lo diagnostic. 4.3.3. Hypoth hèses de base L’hypothèse de e base est la suiivante: parmi les cancers détectés par le dépistage, la prroportion de stade es peu avancés (II et II) est plus importante que parmi les cancers diagnosstiqués sur base e de la clinique. Tout le épistage provientt des différences s dans la répartittion des bénéfice du dé stades (stage-s shift) consécutive au dépistage. L’autre hypothè èse retenue est que la survie et e la qualité de vie des femmes dépendent uniquementt du stade de la tumeur et de l’âg ge de la ment du diagnosttic, que celui-ci soit s consécutif ou non au femme au mom dépistage. Les cohortes sont s suivies d’ann née en année et évoluent en fonction de paramètres de e transition tels lle nombre de fe emmes atteintes chaque année (incidenc ce) et le taux de ssurvie en fonction du stade du canc cer. KCE Reportt 176 4 4.3.4. Alimentattion du modèle Pour réaliser cet exercice, P e nous av vons autant que ffaire se peut, alim menté n notre modèle avec des données belges. b Ces param mètres sont décriits en d détails dans le cha apitre 3.3. du rapp port. L L’espérance de vie de la population étudiée provien nt des tables de survie s d la population féminine de f belge du u même âge. L’in ncidence du canc cer en fo onction de l’âge et e des stades de la l maladie provien nt du registre belg ge du c cancer (Communa auté flamande). Les données relatiives au dépistage e sont is ssues des progra amme actuels (ffemmes de 50-69 9 ans en Wallon nie, à B Bruxelles et en Co ommunauté Flama ande). U mesure de qualité de vie a été Une é appliquée à cha aque compartime ent du m modèle. Le modèlle contient un cas s de base (base ccase) qui correspo ond à la a situation la plus vraisemblable. “ “Par essence, tous s les modèles son nt faux mais certains sont utiles”a 4 4.3.5. Analyse de d sensibilité Dans notre modè D èle, nous avons émis un certain nombre d’hypoth hèses s simplificatrices, en n raison des données dont nouss disposions et de la n nécessité d’éviter l’utilisation d’un modèle m trop comp plexe. Ce choix co onduit à une incertitude e liée à la struc cture du modèle e, au bon choix x des p paramètres et de e la source des informations. P Pour faire face à ces d différents types d’iincertitude, nous avons a réalisé une e analyse de sens sibilité a approfondie utilisa ant différents scé énarios. Ces diffférents scénarios sont d décrits en détails dans d la table 3.2 du d rapport scientiffique. a citation attribuée au statisticien George G Box. KCE Report 176 6 S Screening Breast Cancer C 5. RÉSULTATS Le scénario de e base montre que la prolongation du dépistage jus squ’à 74 ans permettraitt d’éviter 1,3 décè ès pour 1000 femmes participantes s, ce qui représente une e réduction de 2 21% des décès. Globalement, le nombre d’années de vie e sauvées est estiimé à 13,1 et le gain en QALY à 3,9. L’analyse de se ensibilité du modè èle comprend un scénario s pessimis ste et un scénario optimiste . Le scenario pe essimiste fait l’hyypothèse d’un ex xcédent de diagnostic de 20%, d’un taux de faux positifs d de 10%, entrainan nt une perte de qu ualité de durant pendant 54 4 jours (temps né écessaire pour infiirmer les vie de 0,19 perd résultats). La distribution des cancers dépisté és par stades observée o ans le cadre du d dépistage organisé en Flandre (50--69 ans) actuellement da a été appliqué ée au groupe d dépisté. Ce scén nario pessimiste permet d’estimer un ga ain de 8,7 année es de vie mais une perte de 3,1 QALYs pour 1000 fem mmes participantt au dépistage. Ceci signifie qu ue dans certaines circon nstances, au dem meurant tout à fa ait réalistes, le dé épistage peut aboutir à une u perte en terme e de qualité de vie e. Le scenario optimiste fait l’hypotthèse d’un excédent de diagnostic c de 3%, %, entrainant une perte p de qualité de d vie de d’un taux de faux positifs de 2% pplique au groupe e dépisté 0,13 perdurant pendant 36 jourss. Ce scénario ap ment dans le ca adre du la distribution par stades obsservée actuellem anisé aux Pays-B Bas (70-74 ans)). Ce scénario optimiste o dépistage orga permet d’estime er un gain de 17,0 années de vie et un gain de 16,,2 QALY pour 1000 fem mmes participant au dépistage. Ceci signifie qu’’il serait nécessaire d’inv viter 62 femmes à participer au dé épistage pendant cinq c ans pour gagner un QALY. 15 6 DISCUS 6. SSION Les résultats du modèle L m décrit ci-de essus indiquent que en ce qui conc cerne la a situation de base, le gain en an nnées de vie est de 13 ans pour 1000 fe emmes dépistées s. Ce résultat res ste fiable tout au u long de l’analys se de s sensibilité. A l’inv verse, les QALYs s varient substan ntiellement en fon nction d des hypothèses choisies, c allant d’un d gain relative ement faible à, selon c certaines hypothès ses plausibles, un ne perte en qualité é de vie. 6 6.1. Ajouter de es années à la vie ? L’augmentation de L e l’espérance de vie de la femme est un des argum ments u utilisés pour justifier de poursuivre le dépistage du ccancer du sein ch hez la fe emme âgée de plus de 69 ans. Cet argument présuppose qu ue la p population des septuagénaires s a les mêmes ca aractéristiques qu ue la p population des sexagénaires. Il n’en n est rien en n ce qui concern ne la frréquence et les ca auses de décès. L nombre de dé Le écès observé dan ns la tranche d’âg ge des 70-79 an ns est d deux fois et demi plus élevé que ce elui de la tranche e d’âge des 60-69 9 ans. E fait, la population féminine belge En e perd 4% de sess effectifs entre 50 0 à 59 a ans, 8% entre 60 à 69 ans et 20% % entre 70 à 79 ans (Belgian life table 2 2009). L causes de dé Les écès varient égallement. En Belgiq que, la proportion n des d décès dus au can ncer du sein pass se de 13% entre 60 et 64 ans à 6% 6 de to ous les décès en ntre 70 et 75 ans. A cet âge, la mo ortalité par cance er tout c comme la mortalité cardiovascula aire sont pratique ement équivalentes et re esponsables chac cune d’un peu plu us d’un tiers des décès. Parmi tou us les d décès, la part de décès d consécutifs s au cancer du se ein diminue donc avec l’âge (KCE report 11). 16 S Screening Breast Cancer C KCE Reportt 176 6.2. Ajouter de la (qualité d de) vie aux ann nées ? 7 CONCLU 7. USIONS 6.2.1. Traitem ments moins agrressifs? 7 7.1. Faut-il pro olonger le dépis stage jusqu’à l’âge de 74 ans s? Outre le gain en n années de vie, le principal avanta age attendu du dé épistage est de permettre des traitements moins agre essifs. Toutefois,, ni les s de essais contrô ôlés randomisés, ni les données fa actuelles données issues recueillies en Belgique, B ne perme ettent de confirme er cette attente. 6.2.2. Faux positifs p Dans notre mod dèle, les diagnosttics ”faussement positifs” représentent une source importante de perte de q r qualité de vie. Un taux élevé de résultats ositifs (pouvant a aller jusqu’à 10% %) conjugué à un u délai faussement po d’attente relativ vement élevé (45 5 jours en moye enne) pour les examens e complémentaire es peut amener à un résultat total du dépistage né égatif en termes de QA ALY. Si on parvient à garder ce e taux dans les normes européennes (3 3,5%) comme c’e est le cas dans une région du pays p (en Flandre), le gain en QALY est de e 3 pour 1000 fem mmes. 6.2.3. Excès de diagnostics e et de traitements s Le risque de diagnostic excéden ntaire est le risque e majeur de ce dé épistage agénaires. Si nouss appliquons un ta aux de surdiagnos stic de 3 pour les septua %, on peut s’atttendre à ce que d dans chaque cohorte de 100.000 femmes, f 108 femmes su upplémentaires auront un diagnos stic de cancer et subiront très vraisembla ablement un traitement. Si nous appliquons un taux de surdiagnostic de 10 %, ce nombrre monte à 367. D’autre part, toutes t les femme es dont le canc cer est diagnostiq qué par screening devie ennent malades d du cancer deux ou o trois ans plus tôt t qu’en cas de diagnos stic clinique. Cecci a un impact négatif n sur la qua alité des années de vie qui q leur restent. La conclusion de cette étude est que la réponse à ccette question estt non. L C Cette affirmation est basée, d’une e part, sur les ré ésultats du modè èle et d d’autre part sur le e contexte spécifiq que de cette question. Les résulta ats du m modèle démontre ent un gain de 13 1 années de vie pour 1000 fem mmes d dépistées. Toutefo ois, certaines hyp pothèses qui sontt loin d’être irréallistes, in ndiquent que le résultat net du prolongement d du dépistage po ourrait ré ésulter en une pe erte globale en qu ualité de vie. Ces résultats ne sont donc p décisifs en ta pas ant que tels et do oivent être interprrétés dans le con ntexte p particulier d’un dé épistage organisé é. Le dépistage o organisé s’adresse par d définition à un individu qui n’ex xprime ni plainte e ni demande. Cette s spécificité implique d’être d’autant plus vigilant auxx principes éthiqu ues11. L trois principes Les s éthiques de bas se applicables nottamment au dépistage s sont: les principes s de bienfaisance e ou de non malfa aisance, le princip pe de ju ustice ou d’équité et le principe d’autonomie12. L principes de bienfaisance Les b ou de d non malfaisancce sont définis co omme s suit: “Ne pas faire e de mal (primum m non nocere) esst le premier. Il do oit se d doubler d’un devo oir de bienfaisanc ce qui va de pairr avec une attitud de de b bienveillance”. Le principe de justic ce ou d’équité estt: “cette préoccup pation q fait intervenir la qui l dimension colle ective des problèm mes de santé, da ans le s sens d’une préfére ence pour les plus s faibles, les plus démunis”12. L dépistage estt organisé dans le but d’améliorrer le bien-être de Le d la p population en évittant notamment des d décès préma aturés. Cependan nt, les ré ésultats obtenus par le modèle ne n permettent pa as d’exclure que dans c certaines situation ns, le dépistage puisse p affecter négativement la qualité d vie dans la tran de nche d’âge étudié ée. Dans ces cond ditions, il y a risqu ue de v violation du princip pe de base ”primu um non nocere”(ne e pas faire de mal). KCE Report 176 6 S Screening Breast Cancer C D’autre part, le e dépistage est nettement moins efficace e pour les femmes dont l’espéranc ce de vie est la plus basse. Cette différence d’e efficacité existe certes dans d les autres tranches d’âge mais elle y es st moins prononcée. Le respect du princcipe de justice ou u d’équité se révè èle donc n supplémentaire de répondre par la négative à la question q être une raison posée. 7.2. Que rép pondre à la perrsonne qui dem mande un dépis stage? Le contexte de cette question d diverge de celui de la question pré écédente s: l’individu est de emandeur et le problème p doit être e évalué en deux points sur un plan individuel. Le princip pe d’autonomie s’’applique particulièrement pe est défini comme suit: ”le respe ect de la bien à cette situation. Ce princip l principe de base, le respect de l’autonomie de d cette personne est le personne en découle ; Il s’agit de reconnaître la a capacité de l’in ndividu à autodétermination n et libre choix) et e à régir faire des choix pour lui-même (a our que la personne puisse faire un libre sa conduite (autogestion)”12. Po mée des choix, il imporrte qu’elle soit cclairement et corrrectement inform avantages et des inconvénie ents du dépista age dans sa situation s e droit d’être inforrmé (Article 7) et le droit au conse entement personnelle. Le éclairé sont dé écrits dans la loi belge relative au ux droits des patients. Le consentement éclairé de la patiente ne peut être e obtenu uniquem ment sur ure d’un documen nt d’information. Il s’agit d’un processus qui base de la lectu devrait idéalement inclure un éch hange d’idées ave ec le praticien. Il convient éga alement que le m médecin développe pour sa patie ente qui demande le dé épistage, une stra atégie qui en minimise les inconvénients13. Ainsi, une attitu ude articulée en tro e: ois étapes peut êttre recommandée • Information n spécifique à la trranche d’âge • Prise de décision d en foncction de l’appréc ciation personnelle de la patiente14. • n de la personne qui le souhaite ve ers un dépistage dont les Orientation modalités minimisent m les incconvénients. 17 Les critères défin L nis dans le cadrre du programme e européen prév voient n notamment la su urveillance de la a qualité techniq que des équipem ments u utilisés, la double lecture des mammographies et l’o optimisation du tau ux de ra appel1. En Belgique, les unités de mammographie a agréées répondan nt aux c critères définis dans le cadre du prrogramme europé éen, il est donc logique d d’orienter les femm mes qui demande ent explicitement un dépistage verrs ces s structures. 7 7.3. Message clé Le dépistage estt organisé dans le but d’améliorrer le bien-être de L d la p population en évitant notamment des décès prématu urés. Il est certain n que p prolonger le dépistage jusqu’à l’âge e de 74 ans devra ait permettre de ga agner q quelques années de vie. Toutefo ois, l’influence de e cette mesure sur s la q qualité de vie est e nettement pllus aléatoire. Se elon des hypoth hèses ra aisonnables, cettte intervention po ourrait même ab boutir à une pertte en te erme de qualité de d vie. Dans ces conditions, il se p pourrait que la balance b bénéfices-risques d’une perte globa ale de de ce dépistage penche du côté d b bien-être de la pop pulation. 18 S Screening Breast Cancer C KCE Reportt 176 8. RÉFÉR RENCES 1. 2. 3. 4. 5. 6. 7. Perry N, N Broeders M, de e Wolf C, Tornberg g S, Holland R, vo on Karsa L. Euro opean guideliness for quality assu urance in breastt cancer screeniing and diagnosiss. Fourth edition---summary docume ent. Ann Oncol. 2008;19(4):614-2 22. Schwarrtz LM, Woloshin n S. News media a coverage of sc creening mammo ography for women in their 40s and a tamoxifen for primary preventtion of breast cancer. JAMA. 2002;287(23):3136-42.. Schwarrtz LM, Woloshin S, Fowler FJ, Jr.., Welch HG. Enthusiasm for cancer screening in the United States s. JAMA. 2004;29 91(1):718. man E, Woloshin S, Schwartz LM M, Byram SJ, We elch HG, Silverm Fischho off B. Women's vviews on breast cancer c risk and sc creening mammo ography: a qualittative interview sttudy. Med Decis Making. 2001;21(3):231-40. Schwarrtz LM, Woloshin S, Sox HC, Fisc chhoff B, Welch HG. US women n's attitudes to fa alse positive ma ammography resu ults and detectio on of ductal carcinoma in situ: cros ss sectional surve ey. BMJ. 2000;32 20(7250):1635-40 0. Belgian n Cancer Registrry, editor. Cance er incidence in Belgium, B 2004-2005. Brussels; 20 008. Paulus D, Mambourg F, Bonneux L. [Brea ast cancer screening]. C Practice (G GCP). Brussels: Belgian B Health Care Good Clinical Knowle edge Centre (KCE E); 2005 02/05/200 05. KCE Reports 11 Availab ble from: http://kc ce.fgov.be/index_ _en.aspx?SGREF=5221&CREF=93 348 8 8. 9 9. 10. 11. 12. 13. 14. Mandelbla att JS, Cronin KA A, Bailey S, Berrry DA, de Koning g HJ, Draisma G, et al. Effects of mammogra aphy screening under u dules: model esstimates of pottential different screening sched m appears in Ann n Intern Med. 2010 0 Jan benefits and harms.[Erratum 0):738-47. 19;152(2)::136]. Ann Intern Med. 2009;151(10 Woloshin S, Schwartz LM, Byram SJ, Sox H HC, Fischhoff B, Welch W men's understand ding of the mam mmography screening HG. Wom debate. Arrch Intern Med. 20 000;160(10):1434 4-40. Mandelbla att JS, Silliman R. Hanging in the balance: making decisions about the benefits s and harms of brreast cancer screening e oldest old witho out a safety net o of scientific eviden nce. J among the Clin Oncol. 2009;27(4):487--90. Doumont D, Verstraeten K. Enjeux éth hiques du dépistage unauté Française. 2012(7):3-7. organisé. Santé en Commu D les ca ancers, mais à quelle condition n In: Gallois. Dépister UNAFORM MEC, editor. Médecine. Paris; 2005 5. p. 72-7. USPSTF. Screening for Brreast Cancer: U.S S. Preventive Serrvices dation Statementt Annals of Intternal Task Forrce Recommend Medicine 2011(151):716-26 2 6. Woloshin S, Schwartz LM. The bene efits and harm ms of AMA. mammogrraphy screening: understanding tthe trade-offs. JA 2010;303((2):164-5. KCE Report 176 6 SCIENTIFIC REPORT T S Screening Breast Cancer C 19 1 INTROD 1. DUCTION 1 1.1. Context of o this report This report is a partial update of the clinical practicce guideline (CPG T G) on b breast cancer screening published d in 20051. Therefore, the KCE ex xperts m made a list of clinical c questions s related to brea ast cancer scree ening. R Representatives o stakeholders’ orrganizations were of e then invited to re eview th he choice and the e wording of the questions, to highliight the main prob blems re elated to each question and to score the relevance of cllinical q questions(see KC CE report 172)2. Selected questio ons were then divided o over three KCE re eports. A first KCE E report published d in 2010 is focuse ed on b breast cancer scre eening with mam mmography for wo omen in the age group g o 40-49 years (KC of CE report 129)3. The second is focused on identific cation o women at risk fo of or breast cancer and a technical metthods for breast ca ancer s screening (KCE re eport 172)2. 1 1.2. Scope of this report This report focuse T es on the extensio on of organized brreast cancer screening w with mammography to older wom men. Eligible pop pulation is define ed as w women between 70-74 7 years of age e with average risk of breast cancer. 1 1.3. Breast ca ancer screening g in Belgium The Belgian fed T deral and regional governmentss signed a pro otocol a agreement in 200 01 for an organized screening programme for wo omen a aged 50-69 years s, to be organiz zed by the regional governments s with a appropriate financ cial resources sup pplied by the fede eral government. Since S 2 2001, Flanders, th he Walloon region n and the Brusse els capital region have e each introduced an a organized scre eening programm me within their sp pecific c context of alread dy existing practtices. Indeed, op pportunistic screening re emains quite freq quent in the Wallo oon and Brussels region among wo omen in n the age-group 50-69, 5 but also among a younger (4 40-49 years of ag ge) or o older women (>7 70 years). In Flanders, screening mammographies s are d dominant in the ag ge-group 50-69. In the age-group 7 70-79 overall cove erage d drops, mainly beca ause organized sc creening stops at age 69. The cove erage b means of diagn by nostic mammogra aphy decreases also with 3%, indic cating th hat substitution of screening mamm mography by opp portunistic screeniing at 20 S Screening Breast Cancer C the age of 70 is not frequent in Flanders. At this age, total coverage c ortunistic (including both diagnostic or folllow up mammogrraphies and oppo mains at 18% in F Flanders, 33% in Brussels and 30% % in the screening) rem Walloon region (KCE report 172))2. 1.4. Clinical questions This specific report addresses th he following questiions: 1. What are clinical c benefits off an extension off breast cancer orrganized screening in i women betwee en 70 and 74 years s? 1.1. What is the effect of an n extension (70-74 4 years) of breas st cancer er related mortality y? organized screening on the breast cance ong is the delay b between the scree ening and the associated 1.2. How lo breastt cancer related m mortality reduction? ? 1.3. What is the effect of an n extension (70-74 4 years) of breas st cancer organized screening on the overal mortallity? n extension (70-74 4 years) of breas st cancer 1.4. What is the effect of an organized screening on morbidity? ension of breast cancer 2. What are the specific harms of an exte s in wom men between 70 and 74 years?H Harms in organized screening terms of false positive o or false negative re esults? 2.2. Harms s in terms of additional diagnostic te ests? 2.3. Harms s in terms of over--diagnosis? 2.4. Harms s in terms of overttreatment? 3. What attitu ude should be re ecommended for women in case e of self referral? KCE Reportt 176 1 1.5. Scientific approach For each clinical question, a sys F stematic search of the literature was p performed and dis scussed with the support of extern nal experts chose en for th heir scientific co ompetency in se everal fields: gyynaecology, radio ology, e epidemiology, or health economic cs. For question n 3, we searche ed for m models. To quan ntify what the im mplications of ourr findings are on n the B Belgian situation we applied da ata from the Intermutualistic Ag gency (IMA/AIM), cancerr registry and datta from the literature on the Belgia an life ables and constru ucted a simple tim me dependent Markov chain with annual ta c cycles. T methodology used and the resu The ults are described d in each chapter. KCE Report 176 6 S Screening Breast Cancer C 2. LITER RATURE RE EVIEWS 2.1. Review w of clinical stud dies 2.1.1. Metho odology 2.1.1.1. Sou urces A broad searc ch of the electronic databases OVID O Medline, EM MBASE, CDSR and DAR RE was conducted d in April 2011. Se earch was conduc cted first for systematic reviews r (SR) and meta-analysis (M M-A). 2.1.1.2. Sea arch terms For searching on o Medline database, the following g MeSH terms we ere used in combination with usual langua age: Breast neoplasms (MESH) an nd mass e detection) ((MESH) and mam mmography (MES SH). For screening (or early EMBASE, the following Emtree terms were used: u 'cancer scrreening', phy'. These MESH H and Emtree term ms were 'breast cancer' and 'mammograp entify systematic reviews combined with a standard searcch strategy to ide nalysis (M-A). (SR) or meta-an 2.1.1.3. In- and exclusion crriteria Databases werre searched for S SR and M-A in English, E French, Dutch D or German. This report r is a update e of previous KCE E report1 (search made in 2004), thus we w used a date restriction (2004 4-2011) and a la anguage restriction (English, Dutch, Frencch and German). Inclusion criteria used for d on title, abstra act or full text were: w population (women selection based without breastt cancer and w without particula ar breast cance er risk), intervention (m mammography), o outcome (mortality, morbidity, additional diagnosis tests, over diagnosis and over treatme ent), design (SR or o metaCT), key question (screening), age e of population (> >70 and analysis or RC <75 years), and original publica ation. Relevant pu ublications were selected b 2 reviewers (FM M, JR). independently by 2 2.1.1.4. 21 Additio onal evidence We identified two SR4, 5 as the more W m extensive so ource for the rese earch q question 2. Therrefore the evide ence-identified through those SR R-was u updated by search hing Medline and the Cochrane Database of Systematic R Reviews from the search date of the t two SR’s on (search date Nov v-Dec 2 2008). Additional hand h searching of o reference lists w was also undertak ken to e ensure that no potentially relevant studies were missed. We also sca anned re eference lists off SR and of ou ur previous repo ort on breast ca ancer 3 s screening . T The identified studies were selecte ed based on title e and abstract. For F all e eligible studies, the full-text was s retrieved. In ccase no full-text was a available, the stud dy was not taken in nto account. T The description and results of the literature searche es and flow of sttudies s search are in Appe endix 1.1. 2 2.1.1.5. Quality ty appraisal The methodologic T cal quality of systtematic reviews a and associated riisk of b bias were rated using the check klists of the Dutch Cochrane Centre C (w www.cochrane.nl). The assessmen nt of the risk of b bias in the include ed SR w conducted by was y a team of two rev viewers (FM, JR).. T methodologic The cal quality of selec cted additional evvidence was also rated u using the adeq quate checklists of the Dutch Cochrane Centre C (w www.cochrane.nl). T results of the quality appraisal are The a in Appendix 1.5. 2 2.1.1.6. Identiffied systematic reviews r In n the systematic search for literature reviews, 53 citations on the topic w were identified in n database searrches. The majo ority of citations were e excluded on the basis b of title and abstract; 10 citattions were retriev ved in fu ull and reviewed in more detail. On O the basis of tthe full text, 5 rev views w were included4-8. T The reviews writte en by Götzsche and Nelson4, 5 a are mainly focuse ed on m mortality as outco ome, those from Biesheuvel and JJorgensen6, 7 on overd diagnosis and the review of Virnig8 on ductal carcinoma in situ (DCIS). 22 S Screening Breast Cancer C KCE Reportt 176 As a first step,, a quality appraisal of all the rev views was carried out to 4 5 determine theirr suitability for incclusion. Götzsche e and Nelson SR4, were 5 judged to be off high quality with a low risk of bias s. Nelson review was an update of one e other review performed by Humphrey H for th he U.S. Preventive Tas sk Force9. Humph hrey review was also a judged to be e of high quality and used here as comple ementary information source. The review writtten by Biesheuve el6 was judged to be b of good quality y (quality appraisal of selected trials not ssufficiently describ bed) and those written w by s judged to be off high quality. The e review written by b Virnig Jorgensen7 was was also judged d to be of high quality. For DCIS, this upd F date was carried out in July 2011 identifying 7 citations. A citations were excluded All e on the basis b of title and abstract. F overtreatmentt, this update was For s carried out in Ju uly 2011 identifyin ng 19 c citations on Medlin ne and 7 citations s on the Cochrane e Library. The ma ajority o citations were excluded of e on the basis b of title and abstract; 2 pape ers on o overtreatment werre retrieved in full and reviewed in more detail. On the b basis of the full te ext, we retrieved again a the SR writtten by Götzsche4 and s selected one pub blication presentiing data issued from the UK Breast B S Screening Program mme11. The description and resultss of those update es are in n Appendix 1.4. 2.1.1.7. 2 2.1.1.9. Iden ntified RCT 5 The evidence was w updated using g the key words reported in Nelson n SR by searching Med dline and the Cocchrane Database e of Systematic Reviews R from the search h date of this SR on (search date Nov 2008). The literature search for relev vant RCTs carried out in Medline, EMBASE and CCRT C (in April 2011) iden ntified 432 citation ns. The majority of o citations were excluded e on the basis off title and abstractt; the other paperrs (n=8) were retrrieved in full and review wed in more deta ail. On the basis of the full text, all eight studies were ex xcluded because of the study des sign (not an RCT T) shows the flow of rand domized controlled d trials from selec ction to in-or exclusion. By hand searc ching of referencce lists of Götzs sche and Nelson4, 5, the Swedish RCT’s s were identified. Among those, the e Two County tria als is the only RCT tha at includes wom men aged 70-74 years at the time of randomization. Quality appraisall of this RCT was s carried out to de etermine f inclusion. The Two County trials was judged to be b of fair their suitability for quality by Nelson and of low qu uality by Götzsche e and included fo or further analysis10. 2.1.1.8. Iden ntified additional evidence For diagnostic errors and over--diagnosis, this update was carrie ed out in o citations were excluded e July 2011 identtifying 10 citations. The majority of on the basis of o title and abstrract; 2 papers on n diagnostic erro ors were retrieved in full and reviewed in more detail. On the basis of the full text, ers are discussio ons and those two studies were excluded. Most pape he two main SR4, 5. comments on th Ongoiing clinical trials In n addition to the database search hes, the ClinicalT Trials.gov website e was s searched for clinic cal trials. The sea arch terms ‘breastt neoplasm’ as well w as ‘s screening’ and ‘m mammography’ were w used to search for studies.. The m majority of search h results (n=135 5) were ongoing trials. Two poten ntially re elevant trials (NC CT00963911, NC CT00247442) werre identified but were c considered as out of scope after receiving more iinformation on th he full p protocol. 2 2.1.1.10. Data extraction e Data from system D matic reviews and from trials were e extracted into a data e extraction table (D DET) summarizin ng key design fea atures and resultts. All d data extraction tab ble are in Appendiix 1.6. 2.1.2. Descripttion of screening 2 g benefit 2 2.1.2.1. Sourc ces In n the years 196 60-1980, USA, Sweden, S Canada and United King gdom c conducted random mized controlled trrials of mammogra aphy screening. In n US, th he HIP trial (N = 60 995) started d in 1963. In Sw weden, the Malmö ö trial (phase I and II, N = 60 076) started in 1976 and 1978 8, the Two county y Trial = 133 065) in 19 977-78, the Stock kholm (Kopparberg and Ostergötland, N= n 1980 and finally y the Göteborg tria al (N= 51 611) in 1981. trrial (N= 60 117) in In n Canada, the Na ational Breast Sc creening Trials (N NBSS-1 and 2, N = 89 8 835) were initiated d in 1980. In Unite ed Kingdom, the E Edinburgh, trial sttarted in n 1979 in 1980 (N N=44 268) and th he UK Age Trial iin 1991 (trial limitted to KCE Report 176 6 S Screening Breast Cancer C women aged 40-49 years)12. Numerous publlications and so ome SR ow available. summarizing theirs results are no ased on the SR R (2002) commiss sioned to assist the US This part is ba Preventive Serv vices Task Force e (USPTSF) and its update of 200 095,9 and 4 on the Cochran ne SR . We ana alysed more in de etail one RCT named the Swedish Two-C County trial (Osttergötland) which h was included by both SR10,13,14. Both reviews in ncluded the same e trials in their me eta-analysis: the HIP H trial, Malmö I and II, the Two coun nty trial, the NBS SS trials (1 and 2), the A Trial. The Ed dinburgh Stockholm trial, the Göteborg trrial and the UK Age d as poor quality b by both authors an nd excluded there efore4, 9. study was rated Nelson updated d the meta-analyssis from Humphrey y9 to include new findings about younger women (40-49 yyears of age). Th herefore, we refe er to the blication for morta ality analysis perrformed on wome en aged Humphrey pub from 50 to 74 years. Götzsche e4 performed firstt a meta-analysis s among 4 years of age. T Then he did a sep parate analysis forr women women 39 to 74 younger than 50 0 years and for w women older than 50 5 years. Two-County trrial We analysed the Swedish T Two-County trial in order to fin nd more ears). The Swedis sh Twoinformation on our specific population (70-74 ye he first eight rand domized trials. We W used County trial is the largest of th e publications tha at describe this study. s We used the first therefore three publication of th he initiator10, the p publication of Nys ström13 who was selected s by Nelson and the last publication of Tabar published in July 201114. The wedish National Board B of Two-County trial was commissioned by the Sw W and inclu uded women in two Swedish counties: c Health and Welfare Kopparberg and d Östergötland. In n 1977-78, 134 867 women aged 40 4 to 74 years were cluster-randomized d by geographic area. They we ere also ocioeconomic stattus, urban or rura al residency, and d size of stratified by so cluster. Finally, 78 085 women n were invited to o the screening. Among n aged 70-74 years in the screenin ng group those, they werre 10 568 women and 7 462 in the t control group. At this age, wo omen were invited d to two screening roun nds with a screen ning interval of 33 3 months. The trial t was closed in 1984 after approximate ely 7 years of scre eening10. 23 In n 2002, Nyström performed one review r of the Sw wedish RCT’s including th he Malmö, Oste ergötland, Stockh holm, Göteborg trials. Results of o the K Kopparberg trial were w not available e at this time. Thiss publication asse essed th he age-dependen ncy of the effectt of screening. T The author calcu ulated m mortality relative risks r for consecutive 5-years age g group based on re esults frrom the Ostergöttland trial. The median m follow-up time was 17.9 years. y U Unfortunately, with hout the Kopparbe erg part of the Sw wedish trial, the nu umber o women 70 to 74 of 4 years of aged enrolled e was low ((approximately 50 000 in e each group)13. F Finally, we found one publication summarizing s long g term data (29 years) y o mammographic on c screening effectt on mortality14. T Trials quality and d bias A studies include All ed by Humphrey in i 2002 and later by Nelson were rated a fair5, 9. Götzs as sche assessed th he randomization n quality. This author a d divided his results s on results based d on adequately ra andomized contro ol trial a results based on suboptimally randomized and r contro ol trial4. N Nevertheless, the third meta-analy yses were judged of high quality with w a lo ow risk of bias (se ee Appendix 1.5.1). S Some publications s based on The Swedish Two-Co ounty reported va arying n numbers of women enrolled. To explain this variation n, Nyström replied d that s some studies ana alysed results by y year-of-birth wh hile some others used e exact age at rand domization13. Nev vertheless, Götzscche assessed this s trial a suboptimally ra as andomized and likely to be biase ed. He argued that for O Ostergötland, a pu ublic notary alloca ated the clusters b by tossing a coin while w witnesses were prresent. Breast ca ancer mortality in the control group p was a almost twice as high h in Kopparbe erg compared to Ostergötland (0.0021 v versus 0.0012, p = 0.02). The au utopsy rate was 36% for all the TwoC County trial and ca ause-of-death ass sessments were n not blinded4. Acco ording to o that the validity y of local end poin nt committee data a was criticized, a third c committee (named d consensus committee) reviewed the records conta aining a doubtful cause of o death14. 24 2.1.2.2. S Screening Breast Cancer C Bre east cancer relatted mortality red duction For women age ed 39 to 74 yearss and at approxim mately 13 years of o followup, the Humphrrey meta-analysiss (M-A)9 and the Cochrane C review4 showed a significant re eduction in breasst cancer mortalitty of 16% (Relative Risk (RR) 0.84, 95% % confidence inte erval (CI) 0.77 to 0.91) and 19% (Relative ( Risk (RR) 0.81,, 95% (CI) (0.74, 0 0.87) respectively y. For women ag ged 39 to 74 years, the Review of Swedish rand domized control trial sho owed a significan nt reduction in breast cancer mortality at 15.8 years (me edian follow up) o of 21% (Relative Risk R (RR) 0.79, 95% 9 (CI) 0.70 to 0.89). This T study showed d that the effect off breast cancer sc creening in terms of brreast cancer morrtality reduction varies according to age range13. For women age ed at least 50 y a at randomization, three t trials with adequate a randomization did not show a significant reduction in breast cancer y (Relative Risk (RR) 0.94, 95% % (CI) 0.77 to 1.15). Four mortality at 13 years trials with suboptimal randomizzation showed a significant redu uction in m (RR of 0 0.77 (95% CI 0.67 7 to 0.83)). The RR R for all breast cancer mortality 4 seven trials com mbined was 0.77 ((95% CI 0.69 to 0.86) 0 . The review of o Swedish rand domized control trial, applying a more conservative de etermination of ca ause of death for women w aged at le east 70 y at randomizatio on, did not show w a significant reduction in breastt cancer mortality at 17.4 years ((RR) 1.1 12, 95% (CI) 0.73 3 to 1.72)). Unforttunately, mall (approximate ely 5000 women in each this age group was relatively sm owered13. Consequently, we must conclude c group) and this study is underpo t age group5. together with Nelson that data arre insufficient for this 2.1.2.3. Dellay between scre eening and spec cific mortality red duction Tabar publishe ed in July 2011 the last follow-up p result (29-yearr) of the Swedish Two-C County Trial14. T This publication modulated m breastt cancer mortality reducttion in function of length of follow up. u In this report both b data issued from loc cal end point com mmittees and con nsensus-based da ata were presented. The validity of local e end point committe ee data was criticized, we c data. For women age ed 39 to 74 yea ars, this present here consensus publication sho owed specific morrtality reductions of 20% ((RR) 0.8 80, 95% (CI) 0.62 to 1.0 05), 27% ((RR) 0 0.73, 95% (CI) 0..59 to 0.92), and 27% at respectively 10 0, 15 and 20 to 2 29 years of follow w up. In the sam me time, KCE Reportt 176 deaths from breas d st cancer preventted in the study g group increased along le ength of follow-u up. They were re espectively 50, 9 99, 114, 122 and d 126 d deaths prevented at 10, 15, 20, 25 and 29 years of ffollow-up for all wo omen in ncluded in this study. s Author emp phasized that bre east cancer screening p prevents deaths more m in the mediu um to long term than in the imme ediate fu uture. So most of the breast canc cer deaths would have occurred (iin the a absence of screen ning) more than 10 0 years after rand domization. A Authors did not calculate mortaliity relative risks for each age group g s separately. Resultts presented are based on 133 06 65 women aged 40-74 4 (77 080 in the scre eening group and 55 985 in the con ntrol group), while e they w were 10 568 wome en aged 70-74 ye ears in the screening group and 7 462 4 in th he control group p. In Kopparberrg, cancers diag gnosed after the e two s screening rounds in women aged 70-74 years and breast cancer deaths frrom these cases were w still included d in the results14. A cited on previo As ous point, the grou up of women aged d 70-74 years included in n the Swedish Tw wo-County Trial was w relatively sma all (approximately 5000 w women in each gro oup) and this stud dy is underpowere ed13. 2 2.1.2.4. All-cau use mortality The Cochrane SR T R has reported data on all-cause mortality. For wo omen a aged at least 50 y at randomization, two trials with ad dequate randomiz zation (n=73654) did nott show a significant reduction in all-cause mortality at 13 y years (Relative Risk (RR) 1.00, 95% % (CI) 0.95 to 1.0 04). The two trials s with s suboptimal rando omization (n=982 261) also did no ot show a significant re eduction in all-cause breast cance er mortality (RR off 0.99 (95% CI 0..97 to 1.02))4. U Unfortunately stud dies did not have statistical power to detect an all-c cause m mortality reduction n. According to th hat disease speciific mortality is a small frraction of all-caus se mortality in cancer screening trrials, detect a mo ortality re eduction would re equire inclusion off millions of subjecct. 2 2.1.2.5. Morbid dity reduction We found no data W a related to the cancer c related mo orbidity in our selected s sources. In other words, we do no ot accept or reje ect the hypothesis s that s screening reduces s the morbidity of the t breast cancerr disease. KCE Report 176 6 S Screening Breast Cancer C 2.1.3. Descrription of screening harms 2.1.3.1. Sou urces This part is bas sed on the 5 SR se elected in our main search4-8. As ex xplained in part 2.3.5, we e updated those iin July 2011 startiing from the last literature search date. Se ee more details in appendix 1.4. 25 screening round). Conversely, false s e-negative mamm mography results are a little more commo on among women aged 70 to 79 9 years (1.5 per 1000 w women per screen ning round)5. 2 2.1.3.4. Additio onal diagnostic tests t SR written by Götzsche and N Nelson are descrribed in point 3.1 1.1. The n by Biesheuvel a and Jorgensen6, 7 were focused on o overreviews written diagnosis and subsequently on n overtreatment. Each author us sed very ods to address this issue. Biesheuvel analysed reports different metho issued from the first RCTs whiile Jorgensen an nalysed data issu ued from zed screening prrogrammes. The review written by b Virnig publicly organiz was focused on n ductal carcinoma a in situ (DCIS)8. Rates of additiona R al imaging are rela atively low among women aged 70 to 79 y years (64.03 per 1000 women pe er screening round). Biopsy rates s are h higher among wo omen aged 70 to o 79 years (12.2 per 1000 women per s screening round) than among youn nger women. As expected, the nu umber o screen detected of d cancer is highes st in this age grou up. Results indicatte 6.5 s screen-detected in nvasive cancer and 1.4 screen-de etected DCIS per 1000 w women per screen ning round. The BCSC B results indiccate that for every y case o invasive breast cancer detected by mammograph of hy screening in wo omen a aged 70 to 79 yea ars, 154 women have h additional mammography, 10 have o other imaging test, and 2 have biopsies5. 2.1.3.3. 2 2.1.3.5. 2.1.3.2. Stu udy description Perrformance of ma ammography The sensitivity of first mammog graphy for women n aged 70-74 ye ears was wo County trial. T This includes ove er-diagnosis and may be 81% in the Tw difficult to interpret. This data cannot be applied to individual patients ed for patient fa actors (use of hormone h because they are not adjuste herapy, mammog graphic breast de ensity), technical factors replacement th (quality of mam mmography, number of mammogrraphic views) or provider factors (the ex xperience of radio ologists and their propensity to la abel the results of an examination e abno ormal)9. Provider factors may explain that sensibility may y vary between countries4. In the Two County trial, t the specificity of a single mammogrraphic examinatio on was 95.6% forr women ears. This indicate es that 4% of women w who did not n have aged 40-74 ye cancer underw went further diagn nostic evaluation. The positive predictive value of one-tim me mammograph hy was 12% for abnormal a results requiring r further evaluatiion and from 50% to 75% for ab bnormal results requiring r biopsy. Positive e predictive value increases with age a and ranged fro om 18% to 20% among women 70 years of age or older9. Nelson reporte ed data from the Breast Cancer Surveillance Con nsortium (USA) BCSC fo or regularly screen ned women that are a based on resu ults from a single screen ning round. False e-positive mammography results are less common amon ng women aged 70-79 years (68.8 per 1000 wom men per Over-d diagnosis Over-diagnosis off breast cancer at screening ma O ay be defined as s the d detection with scre eening of cancer that would not ha ave presented clin nically d during the woman n’s lifetime (and th herefore would no ot be diagnosed in i the a absence of screen ning)6. N Nelson reported ra ates of over-diagn nosis varying from m less than 1% to o 30% w with most from 1% 1 to 10%. She e explained varia ations by inclusio on or e exclusion of DCIS S cases, by wheth her cases are inciident or prevalentt, and b age. She conclluded that the stu by udies are too hete erogeneous to com mbine 5 s statistically . G Götzsche reported d that the level of over-diagnosis w was about 30% in i the R RCT’s that did no ot introduce early screening in th he control group, and s somewhat larger in the sub optimally randomized tria als before screening of th he control group. He found also a 40% to 60% inccrease in inciden nce of b breast cancer in observational o stud dies performed in Australia, Europe e and U USA after beginnin ng of the screenin ng4. B Biesheuvel analy ysed publications s issued from tthe first RCTs (New Y York/HIP, Malm III, Two County, Canada C a and b, Stockholm, Göte eborg, E Edinburgh) and fro om four populatio on-based program mme (Sweden, No orway, N Netherlands and Italy). He selected papers that a attempted to estimate o over-detection of invasive breast cancer by mam mmography scree ening. 26 S Screening Breast Cancer C Note that he did not include DCIS. He excluded potentially y biased B were descriibed as: differen nt breast cancerr risk in publications. Bias screened and unscreened u population, low particip pation in screenin ng group and high partic cipation in non-sccreening group, offering o screening g to the control group before b or during fo ollow up, inapprop priate adjustment for lead time. After exc clusion, he seleccted 22 estimates s of over-detection from several (some overlapping) sources. Publicatio ons were categorrized as e-incidence or incidence-rate methods m being based on cumulative t are in appe cluding biased stu udies as endix 1.4.3). Exc (definitions of terms described before, he selected tthe least biased over-detection o es stimates. S cases, over-dettection ranged fro om 7% to 21% forr women Excluding DCIS aged 60–69 yea ars6. Jorgensen ana alysed data issued from public cly organized sc creening programmes. He H selected pape ers that published trends in incid dence of breast cancer before and affter the introduction of mamm mography e that when data were present, DC CIS were included d. If not, screening. Note he estimated that t they would ccontribute to 10% % of the diagnos ses in a screened population. After exclu usion of the impllementation phase of the ast seven years s before screening, he compared data covering at lea y after scree ening in screening with data covering at least seven years n screened age e groups. The most common age-range for screened and non mammography screening progra ammes was 50-69 years. No data specific e increase in incid dence of for women aged 70 to 79 years are available. The elated to the inttroduction of sc creening. breast cancer was closely re ase was compen nsated for by a drop in Surprisingly, litttle of this increa incidence of breast cancer in women older than 70 years. Jo orgensen for invasive cance er was 35%. The e rate of calculated that over-diagnosis fo n this meta-analys sis (95% over-diagnosis including DCIS ccases was 52% in CI 46% to 58%)7. sheuvel or by Jo orgensen Discrepancies between results reported by Bies ot of controversial discussions. have led to a lo The approach Biesheuvel B et al. tto adjusting for lea ad time was conte ested by Zahl, Jorgense en and Götzche (2008), who stated that their estimations were substantia ally downwardly biased, due to over-adjustment, o u use of a hypothetical inc crease in incidencce based on theoretical models and use of KCE Reportt 176 ong term follow w up data that are considerablyy diluted. They also lo considered estima c ation unhelpfully wide. w J Jorgensen & Gö ötzsche used line ear regression tto compare obse erved in ncidence with a in an (hypothettical) population that did not und dergo s screening. They assume a a linear in ncrease extrapola ated from prescreening trrends, following the same pattern as the linear tren nd observed in wo omen to oo young to be sc creened. It is difficult to judge if thiis assumption hollds or n not, the graphs the authors prresent show no on-linear increase es in in ncidences before e screening was introduced in the e UK and Norwa ay for w whom no explanattion was given. 2 2.1.3.6. DCIS Historically, DCIS H S was rare and diagnosed by ssurgical removal of a s suspicious breast mass. Since the wide use of mammograph hy, a in ncreasing numbers of patients werre diagnosed with h DCIS. The prog gnosis o the disease is excellent. Maass of s reported data isssued from the SEER S d database (Surveillance, Epidemiolo ogy and End Re esults database of o the U United States National Cancer Ins stitute). Those da ata showed a 10 0-year s survival rate of 96.6% 9 for cases s between 1978 and 1983, whe en no s screening was pe erformed. The rate e was 98.1% bettween 1984 and 1989, w when screening was w performed3, 15. R Recent changes in i DCIS incidenc ce in USA were e emphasized by Virnig. V T This author perfo ormed a SR on incidence, treatm ment and outcomes of D DCIS in name of Agency for Healthcare Research h and Quality (AH HRQ). S She included 63 publications ad ddressing inciden nce for analysis. She c compared data obtained before th he screening (19 973-1975) with cu urrent c century data collected in US where e screening is com mmon. DCIS incid dence ro ose there from 1.87 per 100 000 in 1973–1975 to o 32.5 per 100 000 0 in 2 2004. Incidence in ncreased most in women older tha an 50 years. Incre eased u use of mammogrraphy may explain some but nott all of this incre eased in ncidence8. KCE Report 176 6 2.1.3.7. S Screening Breast Cancer C Ove ertreatment Götzsche reported that the num mber of mastecto omies and lumpe ectomies T trials with adequate a was significantlly larger in the sccreened groups. Three randomization showed a sign nificant increase e in mastectomiies and ( Risk (RR R) 1.31, 95% (CI)) 1.22 to 1.42). Tw wo trials lumpectomies (Relative with suboptimal randomization sshowed the same increase in interv ventions 5% CI 1.26 to 1.61 1)). The RR for alll five trials combined was (RR of 1.42 (95 1.35 (95% CI 1.26 to 1.44)4. Based on recent data from the UK Breast Scree ening Programme e, Dixon mbers of patients s with DCIS. In 1998/99 emphasized the increasing num proximately 1500 cases, but in 200 07/08 there were close to there were app 3500 cases. Although, A most DCIS cases ma ay be treat by breastconserving surrgery, the percen ntage of patients s being treated with w this method has rem mained constant at 30% during this period. Becaus se of the increasing incid dence of DCIS tre eatments, the abs solute numbers off women having mastecttomies has increased from just und der 500 in 1998/99 9 to over 1 900 in 2007/0811 . 2.1.4. Scree ening conditions The sojourn tim me (ST) is the avverage duration of o the preclinical screendetectable phase. Estimation off sojourn time ca an be performed by from matical estimates or using mic crosimulation tec chniques simple mathem (mainly Markov v Models)12. Sojou urn time provides an absolute uppe er limit to the lead time obtainable. If the sojourn time is long, the maximum m nding long16. A lo onger sojourn time e results attainable lead time is correspon ber of additional breast cancer detected, d more liife-years in higher numb gained and high her number of yea ars with cancer du ue to lead-time17. 2.1.4.1. Lite erature search In a first stage, studies assesssing sojourn tim me were searche ed. Ovid c from 19 948 to October Week W 1 2011. Th he main Medline was consulted search terms (MESH) were: Breast Neoplasm ms/ Mass Screening/ or was included in frree text. The sea arch was Mammography//. Sojourn time w limited to paperrs written in Engliish, Dutch, Frenc ch, or German. Re eference lists of the sele ected studies were e checked for add ditional relevant citations. c See more details in appendix 1.4 4.4. 27 Selection criteria S a A retrieved refe All erences were as ssessed against pre-defined inclusion c criteria (in terms of o population, inte ervention, outcom mes, and design-T Table 1) in a two-step procedure: initial assessment of the title, abstractt and k keywords; followe ed by full-text as ssessment of the e selected refere ences. E Estimation of sojjourn time not based b on data were excluded. After e excluding of 3 du uplicates, 40 unique citations we ere identified from m the d databases. Of this s total of 40 refe erences, 23 did n not meet the inclusion c criteria based on title and abstract evaluation. Am mong the 17 cita ations re etained for full-tex xt assessment, 6 did not fulfill the population criteria18-23 a and 1 did not fulfill the outcom me criteria24. Fina ally, 10 studies were 16, 17, 25-32 re etained O search was based Our b on ST dura ation estimations as search. We found f s several publication ns were ST estim mations were issue ed from others sttudies c cited as references by the author. For F example, Zappa in 200332referrred to 29 d data published by y Tabar in 1995 . Duffy in 200526 referred also to those 29 d data . Therefore, we used origina al publications. If one author published tw wo or more articles based on the t same data, we choose the most 30, 31 a accurate for our study s . Finally, 7 publications arre summarized in n data e extraction table (se ee appendix 1.6.7 7). 2 2.1.4.2. Resultts Sojourn times ca S alculated on RCT T’s data W found 4 studies based on the results of the Tw We wo- County Trial. The T TwoCounty Tria al is described in n chapter 2 (poiint 2.2.1.1.)10, 13. First e estimates of sojou urn time publishe ed by Tabar and Duffy were base ed on a approximately the same data. Both h authors used th he same Markov chain m model, but results s were not the sam me. Shen underlin ned that the differrence 25, 29 in n estimates publis shed by the two authors a may be caused by diffferent s statistical methods s or by discrepanc cy in the data. Shen applied his rec cently d developed statistic cal methods base ed on the maximu um likelihood estim mates to o data from the Two T County Triall. Authors estima ated the sensitivitiies of e early detection modalities m as 0.92 2 (SD, 0.09) and the mean ST as a 4.4 y years (SD, 0.76)27. 28 S Screening Breast Cancer C Sojourn times calculated on sc creening program mmes data Spratt estimate ed the duration of breast cancer be efore detection by dividing prevalence rate es at first screenin ng round by incide ence rates in the following f years. Thereforre, he used data ffrom 10 000 wome en aged 35 to 70y y at start included in the Breast Cance er Detection and d Demonstration Project d that sojourn time e ranged (Louisville). Forr women aged 70-74, he estimated between 2.5 y to t 3.8 y28. Fracheboud co ompared the resu ults of the Dutch breast cancer sc creening programme for women aged 70--75 with the hypotthesis developed by Boer optimistic and pes ssimistic assumpttions for in 1995. Boer had described o SCAN model. Op ptimistic assumpttion assumed no o further use in his MIS increase in pre eclinical duration n of breast cancer after 65years of age although pessim mistic assumption n assumed a furth her increase in prreclinical duration with age a 33. Based on 187 207 screenin ng examinations (women aged 70-74 years), Fracheboud d found that dete ection rates in bo oth initial nt screens increa ased steadily witth age and got close to and subsequen assumption which assume a continuously increas sing sojourn time e beyond b tumours le ead to a the age of 69. This increasing sojourn time of breast e in detection of cancers, but also to t more life- years s in lead strong increase time17. Weedon constrructed one inventive solution for sc creening program mme who do not have fu ull registration off interval cancers s or where oppo ortunistic screening is common. Although h Norwegian reg gistration is of ve ery high nce data from the e first screening round, interval between b quality, inciden screening exam mination or registra ation of interval ca ancer may be insufficient. Therefore, he replaced data la acking by data is ssued from questtionnaire 5 women in the Norwegian Breast B Cancer Sc creening send to 336 533 Programme (NBCSP). This new w approach gave estimation of MS ST to 6.9 en aged 60-69 yea ars, although STS S was estimated to o 60%30. years for wome 2.1.4.3. 2 2.1.5. KCE Reportt 176 Key data a D Data issued from literature l search are a summarized in n table 2.1. T Table 2.1: Data is ssued from clinic cal literature reviiew Question 1: Sho Q ould breast canc cer organized sc creening extende ed in w women between 70 and 74 years? ? P Population Women be etween 70-74 yyears of age without breast canc cer and without p particular risk of breast b cancer. In ntervention Organized screening with mammography C Comparison No organized screening O Outcomes: M Mortality (specific) For women n >50 y at rando omization, the sp pecific mortality re eduction after a fo ollow-up of 13 yea ars is 23% (RR: 0.77, (CI) 0.69 to 0.86). In the Two al, specific mortality reduction rea ach at County tria significant reduction r of 27% (RR: 0.73, (CI) 0.59 to 0.92) at 15 1 years of follo ow up and incre eases afterwards.. M Mortality (all caus se) Studies did d not have statistical power to dete ect an all-cause mortality m reduction. F FP 68.8 per 1000 women age ed 70 to 79 yearrs per r (BCSC-USA A) screening round F FN 1.5 per 10 000 women aged d 70 to 79 years s per screening round r (BCSC-USA A) A Additional imagin ng 64.03 per 1000 women age ed 70 to 79 yearrs per r (BCSC-USA A) screening round B Biopsy 12.2 per 1000 women age ed 70 to 79 yearrs per r (BCSC-USA A) screening round Discussion Most estimates s of sojourn time have been ba ased on Models (mainly Markov chain models). Such m models assume a chronological stepwise s er. Unfortunately, it remains unkno own whether canc cer really growth of cance develop accord ding to a chrono ological sequence e. Estimations of sojourn time must consequently be interp preted with caution. KCE Report 176 6 S Screening Breast Cancer C 29 DCIS 1.4 scre een-detected DCIS S per 1000 wome en aged 70 to 79 9 years per screen ning round (BCSC C-USA) 2 2.2. Review of o modeling stud dies Over-diagnosis Over-de etection (excluding DCIS cases), ranged from (7% % to 21%) to 35 5% (no data spe ecific for women a aged 70 to 79 yea ars are available).. Over-treatmen nt The num mber of mastecto omies and lumpe ectomies was sig gnificantly larger in the screened groups (RR:1.35 5 (95% CI 1.26 to o 1.44). In n a first stage, ran ndomized clinical trials analysing th he impact of screening o morbidity and mortality on m were sea arched (see abovve). Then, becaus se the e effectiveness of sc creening require a lot of information n from a wide ran nge of s sources to corre ectly inform decision makers, modeling studies were 35 s searched . M Medline, Embase,, NHS EED and Econlit databasess were consulted from J January 2000 up to t September 2011 (see appendixx 2.1). The search h was limited to papers written in English h, Dutch, French,, Spanish, or Gerrman. R Reference lists of the selected studies were checked d for additional relevant c citations. T keywords use The ed and the results s are detailed in appendix 2.1. The main s search terms (MES SH) were: 2.1.6. Concllusion At this age gro oup, performance e of mammograp phy is high and rates of additional imag ging are relativelyy low. Breast canc cer screening ach hieves a specific mortality reduction of 23% to 27% according to autho ors. This ear in the first ye ears after screeniing. The mortality reduction did not appe gnificant before 10 1 years specific mortaliity reduction is not statistically sig after screening g ((RR) 0.80, (CI) 0.62 to 1.05). Breast cancer mortality m reduction must be put in perspecctive with life-expe ectancy for this ag ge-group in our country. elated to quality of life raises questions On the other hand, aspects re cussion of the ben nefit and harms of breast cancer sc creening pertinent to disc in this age-grou up. First, over-diag gnosis being an in nevitable consequ uence of cancer screening, the risk of overtreatment pers sists. Secondly, the t lead ough difficult to esstimate, may be crucial for older women. time bias altho Screening diagnosed breast can ncer and consecu utive treatment ma ay mean h condition” some years earlier than n clinical the end of “the life in good health ast cancer34. diagnosed brea 2 2.2.1. Literaturre search strateg gy • Breast Neopla asms; and • Mass Screening or Early Detec ction of Cancer ; a and • hy; and Mammograph 2 2.2.2. Selection n criteria All retrieved refe A erences were as ssessed against pre-defined sele ection c criteria (in terms of o population, inte ervention, outcome es, and design - Table T 2 2.2.) in a two-step p procedure: initia al assessment off the title, abstrac ct and k keywords; followe ed by full-text ass sessment of the sselected referenc ces. It s should be noted that studies assessing screening g techniques (suc ch as d digital mammography) were exclude ed because such topic was investig gated in n KCE report 1722. 30 S Screening Breast Cancer C Table 2.2: Sele ection criteria Population Intervention T Table 2.3: Modeliing studies exclu uded after full-tex xt assessment Inclusion crite eria Ex xclusion criteria E Exclusion criteria a Studies caucasian wom men without breast cancer and without particular risk Otther (e.g. woman n at risk, As sian women, etc.) P Population Messecar 2000; 2 Wen 200560, 61. Screening mam mmography Otther, including ma ammography tec chniques (e.g. digital mammo ography) O Outcome Outcomes Morbidity and d Mortality (e.g. LYG and Q QALYs) Otther outcomes (e e.g. over dia agnosis) Design Modeling studie es Otther designs LYG: life-year ga ained; QALY: Qualityy-adjusted life-year gained 2.2.3. KCE Reportt 176 Quantity of research a available After excluding 195 duplicates, 1058 unique citattions were identiffied from g did not allow us to identify ad dditional the databases.. Hand searching citations. Of this total of 1058 re eferences, 1016 did d not meet the inclusion criteria based on title and absstract evaluation. Among the 42 citations he population critteria and retained for full-text assessmentt, 2 did not fulfill th odeling publicatio ons were 15 did not fulfill the design criterria. Finally, 25 mo erning 6 models d developed by modeling groups inv volved in retained, conce CISNET, 2 app plications of these models on differe ent context and 7 models developed by other groups or authors, as som me models have several 3 . The flow ch hart of this selec ction is presented d in the publications17, 36-59 appendix 2.2. In ntervention D Design Advisory Committee C on B Breast Cancer 2006; 2 Anonymou us 2000; Barratt 2 2002a; Barratt 20 002b; Bonneux 2009; 2 Caplan 2001; Carney 2007 7; De Koning 20 000; Feuer 200 04; Grivegnee 2001; 2 Habbema 2006; Mandelblattt 2003; Prevost 2000; 2 auch 2000, Xu 200 0062-76. Rautenstra 2.2.4. Selected 2 d studies 2 2.2.4.1. The CISNET C Project The Cancer Interrvention and Surrveillance Modelin T ng Network (CIS SNET) (http://cisnet.cance er.gov) is a con nsortium of National Cancer Ins stitute ose focus is mo odeling the impa act of (NCI)-sponsored investigators who c cancer control inte erventions on population trends in incidence and mo ortality fo or breast cancer. These models arre also used to prroject future trends s and to o help determine optimal cance er control strateg gies40. Seven grroups d developed their own o breast cancer models spann ning a wide rang ge of m modeling philosop phies: The Univerrsity of Texas M.. D. Anderson Ca ancer C Center model37, University U of Wisc consin model, Ge eorgetown47, Erasmus 5 44 (MISCAN) model57 , Dana-Farber model m Universityy of Rochester mo odel43 51 a Stanford model ). and T The seven models were first used to assess the relative and abs solute c contributions of screening mammo ography and adju uvant treatment to the re eduction in breast-cancer mortaliity in the United States from 197 75 to 77 2 2000 . Mandelbla att et all48 used 6 of those CISNET models to prrovide e estimates of pote ential benefits an nd harms of mam mmography screening u under different scrreening schedules s. One of the 7 mo odels, the Univers sity of KCE Report 176 6 S Screening Breast Cancer C Texas M. D. Anderson A Cancer Center model37 was not used as s it was purely descriptive. The models we ere developed byy different groups but not indepe endently, they were com mpared, discusse ed and adapted during the deve elopment process, they also a used a comm mon set of variab bles and inputs, based on US datasets BCSC B (Breast Ca ancer Surveillanc ce Consortium), SEER S 9 (Surveillance, epidemiology e and d end results), Co onnecticut Tumorr registry and the Berkele ey mortality Datab base. A detailed dis scussion of each h of this models s can be found d in the publications and d on the CISNET website, we will not n discuss each model m in detail, but sum mmarize the poole ed comparison off Mandelblatt et al. a 48 and discuss the ma ain limitations an nd implications fo or our research question. q The models es stimated a large e number of sce enarios, but we will w only present the re esults of the partt relevant to ourr research questtion, the comparison of a screening pollicy screening ag ge 50-69 to a sc creening g age 50-74. policy screening Feuer et al.40 identifies two d dimensions to ch haracterize the types t of surveillance mo odels used here. The first dimen nsion incorporate es micro simulation mod dels at one end o of the spectrum, where w individuals are run through the model one at a tiime, where at each transition a random dividual life histo ories are genera ated, to number is generated and ind mechanistic orr analytic modells, where a sett of analytically derived equations desc cribe the relation nships between key k health states s and/or tumor growth and a metastasis. T The University off Texas M. D. Anderson A Cancer Centerr, University of Wisconsin, Geo orgetown, and Erasmus E models could be characterized as micro simula ation models; the e Danac be characte erized as analytic c; and the remain ning two Farber model could models (University of Rocheste er and Stanford)) could be descrribed as d dimension off model having some aspects of each. The second n runs from bio ologic, where th he model goes beyond characterization observable qua antities to model tthe underlying dis sease onset, grow wth, and progression of disease, to epid demiologic, wherre only a portion n of the ually the observab ble portion). disease process is modeled (usu art with estimatess of breast cance er incidence and mortality m The models sta trends without screening and ttreatment and th hen look at the effect e of a improvementss in survival assoc ciated with treatm ment. screening use and 31 Breast cancer is assumed to ha B ave a preclinical,, screening-detec ctable p period (sojourn time) and a clinical detection po oint. On the bas sis of m mammography se ensitivity (or thresholds of detection n), screening iden ntifies d disease in the prreclinical screenin ng-detection perio od and results in the id dentification of ea arlier-stage or smaller tumors than might be identifie ed by c clinical detection, resulting in redu uction in breast ccancer mortality. Age, e estrogen receptor status, and tumor size– or stage–sspecific treatment have in ndependent effec cts on mortality. Women W can die o of breast cancer or of o other causes. as mentioned before e, the 6 models use a common set s of a age-specific varia ables for breast cancer incidence e, mammography y test c characteristics, tre eatment algorithm ms and effects, a and non-breast ca ancer c competing causes s of death. On the other hand, unob bserved variables such a preclinical dete as ectable times (sojourn time), lead time, dwell time within w s stages of disease e, were in these models estimated d intermediate ou utputs th hat followed from the model structture and assumpttions concerning tumor t g growth. T The stage distribu utions in unscree ened versus scree ened women in these t m models were also o intermediate outcomes, this in ccontrast to some other m models that use th his observable va ariable as input. A As end output from m the m model reductions in mortality, life ye ears gained were e calculated, no QALYs Q w were used. The ha armful effects fals se positive mamm mograms, unnece essary b biopsies and overr diagnosis follow wed from the model, also here no direct o observed input wa as used, no attempt was made to q quantify those harrms in te erms of QALYs. Morbidity M associa ated with surgery for screening-detected d disease or decre ements in quality y of life associatted with false-po ositive re esults living with h earlier knowle edge of a cance er diagnosis or over d diagnosis was nott considered, whic ch makes the mod dels less useful fo or our p purposes. T Table 2.4 gives the results of the e different modelss in terms of mo ortality re eduction and yea ars of life gained for f the different m models. Gains are fairly limited and there is some variabilitty between mode els, with number years g gained per 1000 women screened d ranging from 9 to 17 and numb ber of d deaths averted ran nging from 4 to 6. T This class of mode els relies heavily on o unobservable vvariables, and as most m models are indiv vidual bases the ey are not alwa ays very transpa arent. In ndependent validation was made difficult because results from trials s and 32 S Screening Breast Cancer C the main US breast cancer rregistries were used u to parameterize or m Model outp puts are similar to o the results from m RCT’s calibrate the model. and some observational studiess, but this does not say much ab bout the dies were partly used to validity of the model as data from those stud odel. calibrate the mo Table 2.4: res sults of the diifferent models in terms of mortality m reduction and years of life ga ained per 1000 women w screened d for the els different mode Model Mortality (specific) re eduction over the whole period in % Screening in agegroup 50-69 Screening in agegroup 50-74 eduction screening in Incremental mortality re agegroup 50-74 compa ared to screening in agegroup 50-69 p 1000 women Years of life Gained per screened Screening in agegroup 50-69 Screening in agegroup 50-74 Incremental years of liffe gained screening in agegroup 50-74 compa ared to screening in agegroup 50-69 Incremental days of liife gained per women screened p 50-74 compared to Screening in agegroup screening in agegroup 50-69 women screened D E G M S W 16 22 23 27 17 21 16 21 15 20 23 28 6 4 4 5 5 5 88 106 107 116 111 128 82 96 99 121 84 95 18 9 17 14 22 11 women screened for a biannual scrreening in the age w e group 50-74, with an in ncremental beneffit for biannual screening s 50- 74 4 of 1.7% in term ms of m mortality and 2 liffe years gained per p 1000 women screened. As au uthors h had no choice th han to use US data for most ke ey variables one e can q question in what degree this can really be called an adaptation to the C Catalan context. Carles C et al, 2011 138 finally used th he results of Rue et al 58 a Vilaprinyo et al and a to do a cost effectiveness e analysis, including QA ALYs. T They found 3990 life years gained d for a cohort of 100 000 women for a b biannual screening g in the age group 50-74, with an incremental bene efit for b biannual screening 50- 74 of 299 life years gained d per 100 000 wo omen. T They found 3891 QALYs gained per 100 000 wo omen screened with w a b biannual screening g in the age group 50-74, with an incremental bene efit for b biannual screening 50- 74 of 277 life years gained d per 100 000 wo omen c compared to a sch hedule 50-69. The ey did not report tthe QALYs gained d with e extending the scrreening to 50-74 from 50 -69, ass it was dominate ed by s screening from 45 5- 69, but reporte ed that 186 QAL LYs per 100 000 were g gained by extendin ng the screening to 45-74 from 45 -69. Interestingly, they d did not incorpora ate the results of Vilaprinyo ett al,200958 into their c calculations, but used u US survivall data. They did not take into account o over diagnosis. 2 2.2.4.2. 6,6 3,3 6,2 5,1 8,0 4,0 Model group abb breviations: D _ Dan na-Farber Cancer In nstitute; E _ Erasmu us Medical Center; G _ Georgetown Un niversity; M _ M.D. Anderson A Cancer Center; C S _Stanford Univerrsity; W _ Universityy of Wisconsin/Harv vard Stout et al 200 0656 used the Wissconsin model to o do a cost effec ctiveness analysis, includ ding the use of QA ALYs, but compa arisons of the age e groups 50-74 with age groups 50 - 69 were not made. Rue et al.55 ad dapted de Dana-F Farber Cancer Institute model of Lee L and Zeelen44 to datta in Catalonia. B Because there wa as insufficient info ormation on Catalan surv vival they combin ned the survival da ata from the SEE ER in the US with Catalan data in a prrevious publicatio on of Vilaprinyo, 200958. milar to the ones Lee & Zeelen originally o Obtained results were very sim ality reduction of 21% and 131 life e years gained per p 1000 found, a morta KCE Reportt 176 Modells not related or not using CISNE ET methodology y 3 78 Carter et al, 200539, C developed a micro simulation model based on tumor t g growth using main nly SEER data. Th he model lacks crredibility though mainly m b because of unrealistic assumptions s concerning stag ge specific surviva al, as th hey assume a fix xed survival of 2 years y for stage 4 and complete cure for s stages 1, 2 and 3. This leads to con nsiderably higher years of life gaine ed for s screening than otther models but is in absolute con ntradiction to wha at we k know about stage specific survival. R Rojnik et al, 2008 produced a time dependent d Marko ov model with 4 stages, D DCIS, local, regio onal and distant. Overall model structure was desc cribed b details on how but w the model was parameterized p are e lacking so we ca annot ju udge how this wa as done or if ass sumptions were rreasonable. They y only re eport ICERs so we w have no inform mation on assumed gains in Life Years Y G Gained and QALY Ys. KCE Report 176 6 S Screening Breast Cancer C Neeser et al developed d a sim mple Markov mod del comparing orrganizes screening with a coverage of 70% with opporttunistic screening g with a 0%. They assumed that the orga anized screening reduces coverage of 20 breast cancer mortality m with 15% % based on the IARC handbook, but it is unclear how th hey come to thiss figure as IARC C postulates a re eduction ranging from 5 to 20%. They ca alculated the yea ars of life gained for a 10 ng beginning at 70 (they evaluatted other schedu ules not years screenin relevant for ou ur research quesstion as well. They found that orrganized screening would save 41 lives p per 100 000 and add 0.008 life ye ears (2.9 men screened for 10 years. The model m is somewha at overly days) per wom simplistic by not n taking into account lead time e but applying assumed a reductions imm mediately. No QA ALYs were used and effects on morbidity m was not taken into account. Rauner et al, 2010 2 developed a an ant colonizatio on optimization model but only evaluated the effect of scrreening amongst women 50-70 and a their ental model is nott useful for our pu urposes. It is also o unclear rather experime how they actually modeled stage e specific survival.. 46 Mahnken et al,, 2008 develope ed a method to adjust a for lead tim me bias, length bias and d over-detection and applied this to SEER data, but provided p only adjusted Hazard H ratio’s. Rijnsburger et al, 200453 use ed the MISCAN micro-simulation n model t Rotterdam57 (see above) to replicate r the data a of the developed by the Canadian CNBSS-2 trial on brea ast cancer screen ning among women aged ur purposes. 50–59, so their findings are not rreally useful for ou 36 2 constructe ed a Markov mo odel for two hypothetical Barratt et al 2005 cohorts, with one o cohort wome en undergoing bie ennial screening and the other not, assuming 100% particcipation. Within th his model, they ev valuated o women over 70 0 years old underg going 10 years of biennial the outcomes of screening. The ey assume a 37% % mortality reduc ction, adjusting the t 25% reduction from for non compliance, and assum me that benefit accrues ximal level over first five years afte er starting screen ning and linearly to max that benefit de eclines linearly to o nothing over five f years after stopping s screening. For women who conttinue screening fo or 10 years after the age wer women per thousand die from breast cancer than in of 70s, two few women who sto op screening (sixx v eight deaths from f breast canc cer). The number of diagnoses of breast ccancer in screened women is about 41 and 33 he number in uns screened women about 26. assum ming a risk reductiion of th 50% brings the nu 5 umber of deaths in the screened g group down 6.2 to o 5.1. T This simple mode el has the advanta age of transparen ncy, but does nott take in nto account the efffects of lead time e and stage-shifts on morbidity. 2 2.2.5. Conclusiion Models described are give useful in M nsights and eleme ents but it is diffic cult to a adapt them to the Belgian situation n as we do not ha ave the necessary y data to o parameterize th hem. The CISNET T models give a m modest gain in ye ear of life between 9 an 22 2 years per 1000 0 women screene ed. 2 2.3. Review of o quality of life studies Because breast cancer B c screening g programs are expected to hav ve an im mpact on the qu uality of life (QoL L) of the patientss, models with a oned dimensional health h-outcome measu ure in terms of su urvival are not en nough in nformative. It is important to take e into account all the multidimens sional h health outcomes in the assessmen nt of breast cance er screening progrrams. T value these multidimensional ou To utcomes into a single measure, qu ualitya adjusted life-year (QALY) must be b used. QALYs permit to adjus st the e expected length of life by the health-related q quality of life. These T a adjustments are made m using utilitie es derived from iindividuals’ preferrence fo or different health h states. D Determination of utility values, needed for the ccalculation of QA ALYs, re equires two steps s: 1. The health state description. According to the e pharmaco-econ nomic ealth Care Know wledge Centre (K KCE), guidelines off the Belgian He s should be des scribed on a sta andardized descrriptive health states system. Ideallly, the descriptio on should be don ne by Belgian pa atients using a gene eric descriptive system, s such as the EQ-5D. If health h states descrriptions from Belgian B patients are not available, descriptions frrom similar patien nts in other countrries may be used7 79. 34 S Screening Breast Cancer C 2. The valuattion of these he ealth states. Acco ording to the pharmacoeconomic guidelines g of the K KCE, health state e values should be e valued on a 0 (=v value for death) to o 1 (=value for perfect p health) sca ale by a representa ative sample of th he general public c. Ideally, they sh hould be valued by the Belgian popu ulation but if no original Belgian data d are m other countrie es can be use ed and collected, valuations from 79. discussed7 In this section, the availability a and the quality of o published utility y values ase due to breas st cancer (screen ning and describing the burden of disea ssessed. treatment) is as 2.3.1. Metho ods 2.3.1.1. Lite erature search sttrategy Electronic data abases were con nsulted for origin nal publications on o utility estimates for different health states associatted with breast cancer matic searches were w carried out up u to the screening and treatment. System er 2011 in the fo ollowing databas ses: Medline (via a OVID), end of Octobe Embase (via Embase.com), HTA A and EED (via CRD NHS) and Psycinfo P (via OVID). g various qualifierrs for “quality of life” were used as Subject Searches using heading or tex xt word. See app pendix 3.1 for an n overview of the e search strategies and terms t used. 2.3.1.2. Sellection criteria Identified refere ences were assesssed against pre--defined selection n criteria (in terms of pop pulation, intervention, outcome and d design –Table 2.5) 2 in a two-step proced dure: initial assesssment of the title e, abstract and ke eywords; followed by full-text assessmen nt of the selecte ed references. When W no a and the e citation was un nclear or ambiguo ous, the abstract was available citation was assessed based on n keywords and fu ull-text. Reference e lists of udies were scrutin nized for additiona al relevant citation ns. the selected stu KCE Reportt 176 T Table 2.5: Article selection criteriia I Inclusion criteria a Exclus sion criteria P Population Screened or treatted S p patients for breast c cancer, with a Caucasian o origin and withoutt high r risk factors diseases, non Other d Caucasian, high risk wo omen In ntervention Any intervention relevant A r t the Belgian setttings to Interve entions not used in n Belgium m O Outcome Unique QoL weights U a allowing to derive QALYs ( (=utilities) Multi-d dimension HRQoL L scoress, DALYs, HYEs, … D Design Direct (TTO, PTO D O, SG) o indirect (EQ-5D or D, SF6 HUI, QWB) va 6D, aluation m methods in primary s studies Letterss, secondary studiies, CUA w with QALYs derive ed from th he literature, … Direct valuations using VAS V (not recommended in the mic KCE pharmaco-econom guidelines)79. QoL: Quality of Life. QALY: Quality adju Q usted life year. HRQ QoL: Health-Related d Q Quality of Life. DALY Y: Disability-Adjuste ed Life-Years. HYE: healthy-yearse equivalent; TTO: Tim me-Trade-Off. PTO:: Person Trade-Off. SG: Standard-Gam mble. H HUI: Health Utility In ndex. QWB: Quality of Well Being scale e. CUA: cost-utility a analysis. VAS: visua al analogue scale KCE Report 176 6 2.3.1.3. S Screening Breast Cancer C Sellection process The flowchart of o the selection p process is presen nted in appendix 3.2. 3 The searches on th he databases retu urned 524 citation ns. After exclusion n of 172 duplicates, 352 2 unique citationss were left (see also a appendix 3.2 2). Hand searching allow wed us to identifyy 3 additional cita ations. Two-hund dred and ninety (290) refferences were discarded based on title and abstract, leaving 65 references for f full-text evaluation. Another 49 references r were excluded e at this stage, mostly m because off the unmet desig gn and population criteria. Overall, we sele ected 16 primary sstudies (see appe endix 3.2). 2.3.2. Resullts A summary of the t selected studiies can be found in appendix 3.3. It I should be noted that this summary on nly report metho ods used to deriv ve utility w not values and their results. If other parameters were measured, they were reported in the summary. o utilities was don ne according to the e following stages s: The selection of • Determinattion of health state es for which utilitie es were needed • Selection of o utilities Selection of o a basecase stud dy Selection of o other studies • Pooling of selected utilities a and calculation of percentage changes 2.3.2.1. Dettermination of he ealth states Health states fo or which utility valu ues are needed are a listed in Figure e 2.1. It should be no oted that this fig gure is a schema atic representation n of the reflection proce ess but not the model itself (de escribed in section 3.2). 35 KCE Reports vo ol S Screening Breast Cancer C 36 Figure 2.1: Hea alth states for wh hich utilities are needed (reflection process) Before e screening Short term m impact of the sc reening After scre eening (First year) y After screening (Folllowing years) Women ag ged ≥ 70 years Negative e results Women aged ≥ 70 years Women aged ≥ 70 years Positive results False positive p Non me etastatic No on metastatic Breast canccer stage I Breastt cancer stage I Breast cance er stage II Breast cancer stage II Breast cance er stage III Breast cancer stage III Metasstatic Metastatic Breast cance er stage IV Breast cancer stage IV True po ositive KCE Report 176 6 2.3.2.2. S Screening Breast Cancer C Sellection of utilitiess To select utility values, we first ttried to find Belgia an data as recom mmended co-economic guid delines of the KCE E79. However, no Belgian by the pharmac data could be fo ound. Then, we tried to find the mosst complete study y which best fit with w our m was to avoid as much as possible the use of multiple model. The aim instruments and d multiple popula ations to derive them. Indeed, acco ording to the pharmaco-e economic guidelin nes of the KCE, it is strongly recom mmended to use the sam me descriptive insstrument and the e same set of va alues for quality of life we eights coming from m different studies s79. However, no sttudy assessing all of the heath sta ates described in n section 2.3.2.1 with the e same design w was found. We th herefore tried to find the study with the greatest g number o of health states co orresponding to ou ur model and to use it at the starting point of the selection process. p Selection of th he base case study We found only y one study haviing assessed utility values for bo oth nonmetastatic and metastatic patien nts, i.e. the study of o Lidgren et al80. We had strument the chance that this study alsso used the bettter available ins he pharmaco-econ nomic guidelines of the KCE, i.e. the EQaccording to th 5D79. This study y was therefore th he starting point of our selection pro ocess. 80 Utility values in i the study of Lidgren et al. were derived frrom two methods, i.e. a direct valuatio on method (i.e. the time-trade-offf (TTO) S patients and an indirect valuation v method using a technique) by Swedish generic instrum ment (i.e. the E EQ-5D instrumen nt). Because pharmacoeconomic guide elines of the KCE E recommend the e use of the EQ-5 5D, only these valuations were retained (i.e. utility values from f EQ-5D and not from ents and TTO). In this sttudy, health statess were described by Swedish patie valued using UK K tariffs (because e no tariffs from the Swedish popula ation are available). Hea alth states descriptions can be fo ound in Table 2.6 6. Utility values for non-metastatic patients in the first year y (i.e. the yea ar of the ars as well as utility values for metastatic treatment) and the following yea scribed in Table 2 2.8. patients are des This study had the following limittations: 37 • Utility values were measuring g during out-patie ent visits at a breast b cancer outpattient clinic (Karolinska University hospital), implying the following limita ations: Utility values for f non-metastatic c patients did not fully take into account the shortt term impact of surgery. s Howeverr, on an annual basis, b this shortt term impact was expected to co over a limited leng gth of time and was therefore nott included in the m model. Utility values for metastatic patients p did not represent patien nts in hat these utility values v palliative care. It was therrefore assumed th ected the quality of life of metasta atic patients durin ng the only refle first year of diagnosis. The short term m impact of diagn nosis is also not ffully taken into account (not meas sured at the mom ment of the diagno osis), even if autho ors of this study y reported that this s impact was expected to be includ ded in the valua ation (measured th he year of diagnossis). • Non-metastattic patients were divided d in only tw wo groups, i.e. the e first year of diagnosis and the follo owing years. It w was therefore assu umed emained constant. This that after the year of treatment, utility values re i supported by an US study where no significant assumption is difference in utility u values (from m EQ-5D using U US tariffs) was fou und at year 5, 10 and d 1581. This US sttudy is described in the appendix 3.3. • It should also be noted that utility estimates for n non metastatic pa atients ast cancer in the first year of diag gnosis) and meta astatic (primary brea patients werre similar (0.696 and 0.685 respectivelly). This inconstistency y may be due to th he following reaso ons: Metastatic patients p only inc clude patients going in out-patient consultations (best cases). Generic instru uments such as the EQ-5D are lesss sensitive to ca apture relevant changes in healtth in a specific d disease than dise easei How wever, diease-spe ecific instruments s can specific instruments. only be used u if validated mapping m functionss to derive utilities s from these instruments are avaiilable, which was not the case79. 38 S Screening Breast Cancer C Table 2.6: Health states descriptions for the study of Lidgren et e al. Primary b breast cancer (year 0-1) Patients who had prima ary diagnosis brreast of within 1 year or le ess prior to answe ering the cancer w question nnaire, no recurre ence and no metastatic disease Recurrence (y year 01) Patients who had at least one recurrenc ce (locoeral) within 1 year or less regional and/or contra-late prior to answering the questionnaire, and no metastattic disease. Primary b breast cancer and recurrence following yearrs d with a primary Patients who had been diagnosed ancer or their last recurrence more e than 1 breast ca year prio or to answering th he questionnaire,, and no metastattic disease. Metastatic patients Patients who had metasta atic disease KCE Reportt 176 As showed in this A s table, the description include the following stage: being in nvited for screen ning, having a brreast screen, waiiting for results, being re ecalled for furth her examinations, having furthe er examinations and o obtention of a diagnosis, i.e. no o evidence of brreast cancer. Fo or the a assessment, only three of the five EQ-5D E dimensions were used, i.e. usual a activity; pain/disco omfort; and anxietty/distress and it was assumed tha at the re emaining two dimensions (i.e. mobility m and ability of self-care) were u unaffected. The quality q of life effec cts associated with true negatives s and fa alse positives las sted 12 months while true positivve and false neg gative w were measured for f the remaining life expectanccy. These values s can th herefore not be used u to measure the t short term im mpact of screening g. We d decided to make th he following assumptions: • True negativ ve patients have e utility values e equal to the ge eneral population. • m impact of positive results at scre eening is measure ed by The short term the percentag ge change betwee en true negative and false positive. Selection of otther studies • For other health states, we trried to find stud dies having used d similar ation. The study of o Lidgren et al.80 allowed instruments for the same popula ng utility values for the general Swedish S us to identify a study assessin population strattified by age and gender using the same instrument (EQ-5D with UK tariffs),, i.e. the study of Burström et al.82 These utility valu ues were therefore used for women aged 7 70 and over (see Table 2.8). For the short term t impact of p positive results affter screening, on ne study using the EQ-5 5D instrument wa as identified (Gerard et al.)83. Th his study assessed utility y values for false positive, true positive, false nega ative and true negative. Health H states were e described by the UK population (and not the Swedish po opulation) but UK K tariffs were used d to valuate these health states (as in the other selectted studies). A description d of the “false s given in Table 2 2.7. positive” state is • This impact is s present until the diagnosis, i.e. on n average 45 days s after screening acc cording to IMA da ata. After, either vvaluations of Burs ström et al.82 (generral population for false positive) orr valuations of Lid dgren et al.80 (non metastatic m or mettastatic disease yyear 1 for true po ostive) were used. U Utility values for fa alse positive and true t negative can be found in Table e 2.8. Itt should be noted that the study of Domeyer et al.84 described in app pendix 3 assessed the short term impact of biopsy. How 3.3 wever, to avoid model m c complexity and be ecause the biopsy y is included in the e description of a “false p positive” in the stu udy of Gerard et al. a 83, we decided tto not take the stu udy of 84 D Domeyer et al. in nto account. C Concerning the ev volution of utility values for patientss with metastatic breast b c cancer in the long term, no study wa as found. KCE Report 176 6 S Screening Breast Cancer C 39 Table 2.7: Description of a “fallse positive” statte (Gerard et al)83 • S She is invited by letter for routine breast b screening. • T The appointment is about 2 weeks from receiving the e invitation. • T The visit at the bre east screening ce entre takes about half an hour, whic ch may include wa aiting time. • A female radiogra apher asks about any a symptoms or history of breast disease and expla ains what will hap ppen. • T To take the X-ray y she is asked to undress to the waist. w Each breast is placed in turn n between two sp pecial X-ray plates s and ccompressed to ge et the best possiblle picture. • S She is asked by le etter to go to the breast b screening centre c the followin ng week. • O Other tests are ne eeded because the breast X-ray res sult is not clear. • T This visit may take e up to half a day. • T The breast X-ray is repeated. • T The doctor examines her breasts. • T The doctor may carry out an ultraso ound examination n. • F Fluid from the affe ected area is take en for laboratory analysis a using a fin ne needle to do th his her breast mayy again be compre essed b between the X-ray y plates. The results of the tests are ready within the week • T The tests show no o evidence of brea ast cancer. Quality of life effects e of routine breast ort term) screening (sho QoL of some wom men is affected by y the experience of routine breast screening and brreast cancer diagnosis. The effects s may The Q continue for some time e. Receiving invitation • M Most women are pleased to receive e the invitation. • S Some women are e made nervous, anxious a or depressed, and are worrried about having breast cancer. • M Most women carry y on with their usu ual activities and interests. i • S Some women are e anxious and dep pressed, unable to o concentrate, sleep badly and are moody and irritab ble. They are unable to ccarry on with theirr usual activities and a interests. • P Personal and sexual relationships may m be affected. Routine breastt screen Further tests the Waiting for th he day of the appointme ent 40 S Screening Breast Cancer C At the breast screening s clinic Waiting for the e results KCE Reportt 176 • M Most women are nervous, but are not n anxious or dep pressed. • M Most women are not embarrassed by the screening procedure. • M Most women are not unduly worried d about breast ca ancer developing. • M Most women find the breast X-ray is i uncomfortable and a slightly painfu ul, but this is shortt lived. • S Some women find d the breast X-ray y very uncomfortab ble and painful. • M Most women carry y on with their usu ual activities and interests. i • S Some women are e anxious and dep pressed, un-able to concentrate, sleep badly and arre moody and irrittable. They are unable tto carry on with th heir usual activities s and interests. • P Personal and sexual relationships of o some women may m be affected. If reccalled for further te ests: Clear results after the test • M Most women are very v anxious at be eing recalled for further tests. • O One of the tests, where w the doctor removes fluid from m the affected are ea, is painful. • M Most women are reassured by the clear results. • S Some women rem main anxious for up u to a year before e they are back to o their usual self. Table 2.8: Description of the se elected utilities Author (year) Instrument P Population for he ealth state d description Lidgren et al. (2007)80 EQ-5D S Sweden patients ((Mean age: see health states) Burström et al. (2001)82 EQ-5D Gerard et al. (1999)83 EQ-5D Po opulation for valu uation Hea alth state Mean valu ue UK K tariffs (general population) p Prim mary breast cance er (Year 0-1); Mea an age: 56 0.696 UK K tariffs (general population) p Brea ast cancer (follow wing years); Mean age: 58 0.779 UK K tariffs (general population) p Meta astatic patients; Mean M age 56 0.685 Wom men aged 50-59 0.833 Wom men aged 70-79 0.792 Wom men aged 80-88 0.740 True e negative 0.940 S Sweden patients ((Mean age: see health states) UK K tariffs (general population) p W Women from UK aged a 40-64 yyears (eligible for screening) K tariffs (general population) p UK KCE Report 176 6 2.3.2.3. S Screening Breast Cancer C Poo oling of selected d studies and callculation of percentage cha anges A summary of selected utilities and of calculation of percentage changes c n Figure 2.2. The e utility values of the study of Burs ström et can be found in al.82 were chosen as the initial values of the mode el (first state of th he model ding to women ag ge. Then, the perrcentage (A)). These values varied accord e to these valuess was applied. It was assumed that these change relative percentage cha anges did not varyy according to the women age (no data). d hort term impact of the screening The next stage e concerns the sh g. It was assumed that utility u values for ttrue negative wom men were equal to utility values in the general g populatio on (A). Then, the percentage decrrease in utilities betwee en true negative women and false positive wom men was calculated, i.e. -16% (B). Initia al values were thus t maintained for true en and decreased d by 16% for wom men with a positiv ve result negative wome after screening (false or true possitive). As mention ned in the section 0, these m for 45 days. utilities will be maintained 41 For the first year of screening, wo F omen without bre east cancer had utility v values equal to th he general popula ation (A). For true e positive, utility values v o the study of Lid of dgren et al. were used.80 To make e the link betwee en the s study of Lidgren et al.80 and the study of Burström m et al.82, percentage c changes between values for the sa ame population w were used, i.e. Sw wedish w women aged 50-5 59 (and UK tariffs s). Utility values w were therefore red duced b 16% ((0.696-0..833)/0.833) for non by n metastatic pa atients (C) and by y 18% ((0.685-0.833)/0.833) for metastatic c patients (E). F the next yearrs, people from the For t general popu ulation who developed n non-metastatic or metastatic breas st cancer had utility values reduce ed by 16% (C) and 18% % (E) respectively y (as calculated a above). Non-meta astatic p patients who stay yed in this stag ge had their utiliity decreased by y 6% c compared to the general g population ((0.779-0.833)/0 0.833) and maintained th he years after (D D). Metastatic patients maintained their utility until death d (G). 42 S Screening Breast Cancer C KCE Reportt 176 Figure 2.2: Perrcentage change e in utilities Before e screening Short term m impact of the screeniing (45 days) After screening A (First year) After screeniing (Following yea ars) Women age ed = 70 years True negative Wome en aged = 70 years Women aged = 7 70 years A A A A N metastatic Non Non metasta atic Brea ast cancer stage I Breast cancer sstage I Brea ast cancer stage III Breast cancer sttage II Brea ast cancer stage III Breast cancer sttage III Positive P results False positive True positive B C C D Metastatic Brea ast cancer stage IV V F E Metastaticc Breast cancer sttage IV F G KCE Report 176 6 2.3.3. S Screening Breast Cancer C Discu ussion To include the quality of life im mpact of screenin ng in the analysiis, utility h health state of tthe model had to be identified. The e aim of values for each this chapter wa as therefore to select these values. The method wa as based on the KCE pha armaco-economicc guidelines79. We e tried to avoid the use of multiple instrum ments and multiple valuations and focused on utility y values derived from the e EQ-5D instrume ent. • The analys sis had the following limitations: • No Belgian n data were availa able and a transfferability analysis was not possible (n no access to prim mary data). Even ifi we expected th hat using UK tariffs instead of Belgia an valuations wo ould not greatly in nfluence be interesting for future models. results, Belgian data would b • t impact of su urgery and of diag gnosis was not ta aken into The short term account be ecause no valid da ata were available e. • Even if the e EQ-5D is one o of the best availa able instrument to o assess these utilitiies (according to the KCE pharma aco-economic guid delines), this instrum ment is less sen nsitive than disea ase specific instrruments. Consequen ntly, it can be exp pected that the im mpact of some co onditions such as a mastectomy (p partial or total) would have bee en more important if a disease sp pecific instrumen nt instead of a generic nsitivity could exp plain the instrument had been used. This lack of sen ntage change be etween patients with breast cancer and low percen women in the general pop pulation or between metastatic and a non q of life from disease metastatic patients. The asssessment of the quality struments was ne evertheless not investigated in this chapter specific ins because th hese instruments d do not permit to derive QALYs. • Finally, the e review of the literature showed d an important variability v between re eported utility estimates for breast cancer health sta ates (see appendix 3.3), 3 revealing a high level of uncertainty u aroun nd these parameters s. Because of th his uncertainty, a sensitivity analysis on these parameters should be e done in the chap pter on model resu ults. 43 3 DECISIO 3. ON ANALY YSIS To quantify whatt the implications T s of our findingss are on the Be elgian s situation we cons structed a decisio on analysis model using two diffferent a approaches. For the first simple approach, we ap pplied data from IMA, c cancer registry an nd data from the literature l on the B Belgian life tables s (see b below). For the se econd, we constrructed a simple ttime dependent cohort c w annual cycles with s. W consider perfo We orming one Belgia an decision analyysis a better apprroach th han trying to ada apt the models discussed d in cha apter 2 to the Be elgian s situation. Indeed, Belgian data nee eded to paramete erize these modells are n available and not d we would merely reproduce tthe already published fiindings of these models, m as we wo ould be obliged to o use the same (m mainly U data. US) W look at the effe We ect of introducing mammography sscreening in addition to th he currently existing situation with the opportunistic screening going on at th he current level. This T has the adva antage that we ca an use Belgian da ata as b baseline without having h to modify then, as this can only be done making u of an addition use nal number of no on verifiable assu umptions. We des scribe h here: • a used in this deciision analysis; Available data • Additional lite erature review foc cused on qualityy of life related to the screening and d to the breast can ncer as such; • The model us sed for this decisio on analysis. 44 S Screening Breast Cancer C 3.1. Data so ources Belgian life tab ble (2009) Overall surviva al was taken fro om the Belgian life table of 200 09 from be.STAT (http:///statbel.fgov.be) Belgian Cance er Registry (BCR R) The Belgian Cancer C Registry F Foundation is an n public institution which collects data concerning c new ccancer cases in Belgium and ma akes up statistics from these data (http://w www.kankerregistter.org/). Belgian organized screening As recommend ded by European Commission, Be elgium started a national organized scree ening programme e. The target age groups as defined by the program are wo omen aged 50 to 69 years. Belgian n breast cancer sc creening programs a are organized d by: Brrumammo (Bruxelles, http://www.brum mammo.be/), Cen ntre Communauta aire de Référence e pour le dépistage des s cancers (CCR Ref: http://www.c ccref.org/) (Comm munauté Française) and d BorstKankerOp psporing (BKO) (Vlaamse ( Gemee enschap: http://www.zorg g-en-gezondheid.b be/). Intermutualistiic Agency (IMA) The Intermutua alistic Agency (IMA A) centralises data coming from all Belgian sickness funds s. IMA compiled and published several reports on the national screen ning program con ntaining data on the t target age grroups as defined by the e program (50-6 69 years). IMA complemented this with information on persons outside e the target age-group, with a particular p t used, delayys between scree ening tests and possible focus on the tests confirmation an nd treatments follo owing testing (http p://www.nic-ima.be e/). Dutch Nation nal Evaluation Team for Bre east cancer sc creening (DNETB)85. The Dutch National Evaluatio on Team for Breast B cancer sc creening eport with their findings covering g the period 1990-2007 published a re containing inforrmation on age sspecific stage disttributions in the screened s population. KCE Reportt 176 SEER database S T Surveillance, Epidemiology, an The nd End Results (S SEER) Program of o the N National Cancer In nstitute works to provide informatio on on cancer stattistics in n an effort to red duce the burden of cancer among the U.S. Population (http://seer.cancerr.gov/). SEER colllects data on canccer cases from va arious lo ocations and sourrces throughout th he United States. Data collection began b in n 1973. As they used an outdated distribution we could not incorp porate th hese in the model. 3 3.2. Model de escription In n a first simple approach a we app plied the 22% redu uction in breast ca ancer m mortality caused by b screening com ming from RCT and its range, res sulting frrom the results of o the meta-anallysis of Gøtzsche e et al, 20084 on the B Belgian life table. We assume here e that the reductio on in women age ed 707 is similar to the 74 t reduction in other age group ps. We also ass sume, fo ollowing Barratt et e al 200536 that benefit accrues linearly to a ma aximal le evel over first five e years after startting screening and that benefit dec clines linearly to nothing g over five years s after stopping screening. Life years s saved can then be derived from th he life table. How wever, effects of harms h a effects on qua and ality of life resultin ng from earlier dia agnosis, over-diag gnosis a stage-shift is more difficult to assess and a in this ap pproach. Therefore e this a approach was onlly used for cross validation by com mparing it with a more c complex approach h that makes use of o the stage-shift caused by screen ning. T The second apprroach makes use e of the Belgian C Cancer Registry (BCR) d data on incidence e of invasive canc cer and DCIS forr the construction n of a tiime dependent state transition cohort model with annual cycles. T model compares 2 cohorts: The • A cohort of women w starting at age 70 where screening is extend ded to the population n in the age group 70-74, where a part of the wo omen participates in n the screening an nd where a part o of the cancers is found f by screening,, depending on participation p rate and sensitivity of o the screening. Th here is a mix of sc creen detected an nd not screen detected cases (interva al cancers and cancers c amongstt unscreend wom mens). The screen detected d cancers will have a diffe erent stage distrib bution than the cancers not detected by b screening. KCE Report 176 6 S Screening Breast Cancer C • A cohort of o women starting at age 70 whe ere the screening is not extended beyond b the age off 69 years. For th his cohort all wom men have the stage distribution d of the non screened. All women are followed to death h. The cumulative number of life ye ears, the horts are number of QALYs and deaths to breast cancer of the two coh erall mortality is no ot compared as in n the end everybod dy dies. compared. Ove We assume tha at: • Survival an nd quality of life of the women depe ends only on the stage of the tumor at the moment tthe tumor is dete ected and the age of the nd not on the pressence or absence of screening; women, an • All benefit of the screening rresults from the stage-shift, the diffferences stribution caused by the screening.. in stage-dis Harm caused by false positivves at the mome ent of the scree ening is s by asssuming 3 screening rounds with a 2 years accounted for separately, interval in the e participation women and apply ying recall rates s at the proportion wom men that are alive and without brea ast cancer at the moment the screening ro ound actually take es place. Figure 3.1 show ws the different ccompartments in the two cohorts and the transitions betw ween them. In the unscreen ned cohort, transiitions between co ompartments from m year to year are determ mined by: • Incidence of o breast cancer; • Stage distrribution of unscree ened cancers; • Stage spec cific survival; and • Age specific overall mortalityy due to other cau uses. On top of that, for f the cohort whe ere screening take es place, transition is also determined by some s aspects of tthe screening: • Lead time as a part of the cancers will be found d earlier; • The propo ortion of cancers found by scree ening and proporrtion not found by sc creening and their respective stage e distributions. 45 As survival and quality of life depends on both age and time since A d diagnosis in the model, m a separate e compartment is made for each age a of d diagnosis and stage, and stage specific survivall is than applied d. As s screening is applie ed during 5 years and there is an a assumed lead time e of 2 y years (or 3 in sen nsitivity analysis) the number of co ompartments remained m manageable. T Transitions betwee en stages are nott included as stage es are assessed at a the m moment the diagn nosis is made folllowed by treatme ent. Even if the ca ancer e evolves after trea atment it does no ot necessarily go through the 4 stages a anymore. 46 S Screening Breast Cancer C KCE Reportt 176 Figure 3.1: Comparison of the two cohorts with h and without a screening s progra am IInvasive cancer d detected by screeniing Healthy women iinterval cancer in s screened women Invasive cancer in unscreened womeen Cohorrt with screeening I II III IV I II III IV All ccause deaths I II III IV DCIS Co ohort w without screeening Invasive cancer in unscreened womeen Healthy women I II III IV DCIS All ccause deaths KCE Report 176 6 S Screening Breast Cancer C Figure 3.2: Compartments in th he two cohorts and a the transition ns between them m DCIS & over‐ d diagnosed Breast cancer stage I Breast cancer stage II Healthy women Death Breast cancer stage III Breast cancer stage IV 47 48 S Screening Breast Cancer C 3.3. Descrip ption of the parrameters 3.3.1. Age specific s overall survival Overall surviva al was taken fro om the Belgian life table of 200 09 from be.STAT (http:///statbel.fgov.be) after adjusting fo or breast cancer specific mortality based on data from the Belgian Cancer register. r 3.3.2. Breas st cancer inciden nce For the baseline e without screenin ng the BCR data on o incidence of DCIS D and the 4 stages for invasive for the age group 70-74 4 of the period 2004-2008 ere is some opporrtunistic screening g in that age grou up. From were used. The the IMA data we w infer that in Flanders the cov verage with at le east one mammography in the past 2 yea ars is 18% (for de etails see the KC CE report ps)2. Given that we can 172 on breastt cancer screening in risk group assume that an n important part o of this is also for diagnostic d and follows up purposes, so we w choose to use e data issued from m Flanders becau use they are less contam minated by opportu unistic screening. For the situatio on where screen ning takes place,, incidence in the 70-74 group will increa ase with a numbe er of cancers coming from two sourrces: • Lead time, cancers that would have appeare ed later but are fou und now o lead time. This will lead to a co ompensatory decrease in because of number of cases in the folllowing years. The e moment and de egree of d on the a assumed lead tim me (see point 3.2 2.2). We this shift depends used 2 yea ars lead time in tthe baseline and 3 years in the se ensitivity analysis. • gnosis invasive cancer, we mod deled the over diagnosis Over diag based on the findings in th he literature as described d in the literature ove under 2.1.3.5.. We assume a ra ange of 2 to 30% for over review abo diagnosis excluding e DCIS. • Over diag gnosis DCIS, we e model the overr diagnosis of DC CIS in a different way: w we use the o observation that in Flanders the in ncidence DCIS per 100 1 000 is twice in the group 60-6 69 where screenin ng takes place com mpared to the age e-groups 70-74 and a 75-79 where e only a limited am mount of opportunistic screening takes place. Th his is in contrast with the Brussels ca apital region and Walloon W region where w the CIS is much less pronounced. So we e take as an estim mation of drop in DC KCE Reportt 176 over diagnosis the difference in i DCIS incidence e in Flanders bettween ps 60-69 and 70--74, augmented b by 1.5 to adjust fo or the the age-group fact that scree ening coverage is only 60% as a prroxy for overdiagn nosed DCIS. This brrings us to an ove er diagnosis of DC CIS of 40 per 100 0 000 women per ye ear. 3 3.3.3. Participa ation rate W used a 70% pa We articipation (plaus sible range 60% to o 80%) as baselin ne. 3 3.3.4. Proportio on of screen dettected breast ca ancers The data of the Be T elgian screening program p show tha at in the age grou up 506 49% of the cas 69, ses are found by screening, and th he rest is either interval c cancer or not partticipating in the sc creening. Among the screened wo omen, 7 75% of the found cancers are scree en-detected and 2 25% is interval ca ancer. W used a proporrtion of cancers fo We ound among the w women participating in s screening of 70% (plausible range 60% 6 to 80%). 3 3.3.5. Recall ra ate We assume a rec W call rate of 3.5% % based on the d data from the Fle emish s screening program m concerning follo ow up rounds (ass the screening would w b an extension of the screening be g among women aged 50-69. Fo or the s sensitivity analysis s we used 2% in an a optimistic scen nario and 5 and 10 0% in th he more pessimistic scenario (10% recall rates are observed fo or the m moment in certain regions). A a baseline we assume a delay of 45 days, base As ed on IMA data, with w a p plausible range fo or the sensitivity analysis a of 36 and d 45 days (subtra acting a adding 20%). and T short term im The mpact of positive results r at screeniing were measure ed by th he percentage ch hange in utility va alues between trrue negative and false p positive results. KCE Report 176 6 3.3.6. S Screening Breast Cancer C Stage e distribution and d stage shift We take estima ations of the stag ge distribution for breast cancer amongst a screened and unscreened u from the BCR data on o the Flanders and a data provided by the e Flemish screenin ng program. For the stage distribution d in the unscreened wom men, we can cons sider the stage distributio on amongst wom men in the group 70-74 7 in Flanders s for the years 2004 - 2008 a good esstimation. The stage shift will be e slightly d as there is som me opportunistic screening in tha at group underestimated going on, see above. a For stage distribution in the scre eened population, the base case es stimation e data from the D Dutch National Ev valuation Team fo or Breast is based on the cancer screenin ng report of 2009 9 (DNETB)85 who o provide data spe ecifically for the age group g 70-74 from 1998-2007. Although A using 2 stage distributions fro om different sourrces is a subopttimal way of mod deling a stage shift we think t this approxim mates best the Be elgian situation, as a we do not have date on screen detectted cancer in this s age group. We assume on of cases amon ng the non screen ned and interval cancer c to stage distributio be the same, ba ased on the data from the Flemish screening program. The Flemish screening s program provided data a on the stages among screen detecte ed cancers, interrval cancers and d cancers amongst non participants, co ollected amongstt women who ga ave their consen nt in the period 2001-2006. Stage distribu ution of interval ca ancers and cancer among s is very similar. non participants 49 Table 3.1: Stage distribution amo T ong screen dete ected breast can ncers, in nterval cancers and cancers among a non partticipants, age 50-69, 5 F Flemish screenin ng program 2001-2006. Screen n cancerrs detected In nterval cancers Cancers amongst non participan nts S Stage n % n % n % I 2586 62.5% 62 24 41.5% 1454 41.8% % III 1306 31.6% 65 56 43.6% 1460 42.0% % IIII 232 5.6% 20 00 13.3% 493 14.2% % IV V 15 0.4% 24 4 1.6% 71 2.0% % T TOTAL 4139 100% 15 504 100% 3478 100% % T This baseline stag ge shift we call Sce enario 1: Stage distribution S n of c cancers not foun nd by s screening B BCR data (Flemis sh p population,70-74y y, 20042 2008) S Stage % % I II III IV 31.6 6% 42.3 3% 16.6 6% 9.5 5% Sta age distribution of o can ncers found by scrreening Da ata of the DNETB B scrreening report 20 009 Sttage % I II III IV 80% % 18.7% 0.8% % 0.5% % 50 S Screening Breast Cancer C Important rem mark: It is importtant to note that this shift concerrns only screen detecte ed cancers and that in the cohort with screening interval cancers and nonparticipants n ke eep the stage distribution of can ncer not found by screening. In most casses this is around 50% but depends on the er values of the sccreening and varie es in time. other paramete For the sensitiv vity analysis we usse 2 supplementary scenario’s: As the stage distribution d from the Dutch National Evaluation Team T for Breast cancer screening s report o of 2009 may be more m favorable th han what can be achieved in the Belgian ccontext, we used as a an alternative scenario s bution for screen detected patients s of the age grou up 50-69 the stage-distrib from the Flemis sh cancer screenin ng program. This we call Sce enario 2: Stage distributtion of cancers not fo ound by screening BCR data (Flem mish population,70--74 years, 2004-2008) Stage distribu ution of cancers found d by screening (Fle emish screening pro ogramme (50-69 years) Stage % Stage % I II III IV 31.6% 3 4 42.3% 1 16.6% 9.5% I II III IV 62 2.5% 31 1.6% 5 5.6% 0 0.4% In a third scena ario we use a sligh htly different mode eling approach. Instead of usin ng stage distributtions amongst sc creened and uns screened women, we ass sume that introducing screening in the group 69-74 will shift the stage distrib bution amongst alll breast cancer cases c in the popu ulation to the stage distribution of the wom men 60-69 in the same period, using data om the Belgian bre east cancer registtry. for Flanders fro KCE Reportt 176 T This we call scena ario 3 Stage distribution S n of c cancers not foun nd by s screening B BCR data (Flemis sh p population,70-74 years, 2 2004-2008) Sta age distribution am mongst all breast can ncers if screenin ng lev vels are similar to o lev vels among 60-69 9 in Fla anders S Stage % Staage % I II III IV 31.6 6% 42.3 3% 16.6 6% 9.5% % II II IIII IV V 45.7% 35.9% 12.5% 5.9% 3 3.3.7. Stage sp pecific relative survival Stage specific su S urvival was taken n from Belgian stage specific annual s survival data (take en from KCE repo ort 150A)86. We o only have data up p to 5 y years. We used data from the Dutch D cancer reg gister taken from m the w website (http://ww ww.cijfersoverkank ker.nl) to supplem ment until 7 years s (see a appendix 4.1). We e assumed that survival conditiona al on stage is similar in s screened and uns screened breast cancer c patients. A As a sensitivity ana alysis w used also: we • d for women above 70 years s per Entirely the Dutch survival data T relative surviv val curve shows a lower relative su urvival stagegroup. The for women above 70 compared with the overrall survival. This s may ct that older wome en support the invvasive treatments s less reflect the fac well but it is also a possible thatt there is undertre eatment of the elderly. Moreover, the e data include pa atients that were treated more tha an 20 years ago, this may also explaiin the lower relativve survival. • val data coming g from breast cancer research h UK British surviv (http://info.can ncerresearchuk.orrg/) They provide 10 years survival data but survival is s considerably low wer than the Dutch h or Belgian data. One of the problem ms with 10 year survival data is tthe fact that it re eflects survival of persons treated at a least 11 yearss ago, given the e fast evolution in brreast cancer treattment this is a long time. KCE Report 176 6 S Screening Breast Cancer C • urvival data supplemented by Fren nch 10 year survival data Belgian su coming fro om87. The probllem of the evo olution in breast cancer treatment apply a here as for tthe British data. We did not use e US SEER data as they use an outdated o staging method, so that survivall curves per stage e are not comparrable to the other sources and difficult to incorporate in the model. The survival curves can be found d in the annex. 3.3.8. • In the study of o Lidgren et al.80, non-metastatic p patients were divid ded in only two groups, i.e. the first ye ear of diagnosis a and the following years. y atients in stage I, II, III (groupe ed as non meta astatic Therefore pa patients) were e assumed to have the same utility. This assumption is supported by the fact that for years y 2001-2006, the treatment wa as the ort on same for patients in stage I, II, III according to the KCE repo ncer86. Note that more recent data a may quality indicattors in breast can change this picture because many cancer found d by screening are e now onservative surge ery. Neverthelesss, data to prove e this treated by co assumption are a not available and we found no o study comparing the impact of partial p versus to otal mastectomyy on quality off life corresponding g to our inclusion criteria. • Utility values for f non-metastatic c patients (after th he year of surgery y) and metastatic pa atients were assu umed to remain cconstant across years. y For non-meta astatic patients, th his assumption iss supported by an US study showin ng no significant differences at yyear 5, 10, and 1581. Nevertheless,, as a sensitivity analysis we applyy a 20% decreme ent in QALYs for tak king into account a variation of utilitty values across years. y QALY Y Number of life years was calcullated for each sta age and a stage and this or the quality of llife (QALYs), bas sed on a literature e search was adjusted fo (see point 2.3). We made some e assumptions: • Utility value es at start of the m model (before scrreening) were stra atified by age but percentage changess relative to these e values were ass sumed to ccording to the ag ge of the women (we did not have data on not vary ac this). For the sensitivity a analysis, we app ply a 20% reduction or increase. • w negative results had utility values equal to the general Patients with population.. • In the asse essment of utility values for true ne egative and false positive results, mo obility and ability o of self-care were assumed a to be un naffected by screenin ng. 51 • At baseline, we w did not discou unt QALYs. For th he sensitivity ana alysis, discount rates s of 1.5%, 3% and d 5% were applied d. P Parameters used in i the model are shown s in table 3.2 2. 52 S Screening Breast Cancer C KCE Reportt 176 Table 3.2: Para ameters used in the model Parame eters No screenin ng Base case c Sensitivity ana alysis 3.3.1 Age spe ecific overall survivval Belgian life-table Belgian n life-table Belgian life-tablle 3.3.2 Breast cancer c incidence BCR data (Flanders population, 20042008) BCR data (Flemish ation, 2004-2008) popula increas sed by lead time. over-diiagnosis invasiv ve cancer of DCIS BCR data (Flem mish population, 2004-2008) 2 increased by lea ad time. over-diag gnosis invasive cancer of DCIS Lead tim me 2 years s 3 years Over-dia agnosis invasive cancer 10.0% range from 3 to 30% Over-dia agnosis DCIS 40/100 0 000 women per year 40/100 000 wom men per year 3.3.3 Participa ation rate 70.0% range from 60% % to 80% 3.3.4 Proportiion of screened detected d cancers 70.0% range from 60% % to 80% 3.3.5 Recall rate 3.5% (Flemish ning program) screen range from 2% to 10% Duration n of period after positive result 45 day ys range from 36 to t 54 days QALYs lost in this period 16.0% estimated between 13% to 19% 3.3.6 Scenario 2 Scenario 3 Stage distribution BCR data (F Flemish population, 7074years, 20042008) Data of the DNETB ning report 2009 screen Stage distributio on of Flemish screening 0-69) programme (50 ge distribution of Stag BCR R (Flemish women n 60-6 69 (screened and not screened) Stage I 31.6% 80.0% 62.5% 45.7 7% Stage II 42.3% 18.7% 31.6% 35.9 9% KCE Report 176 6 S Screening Breast Cancer C 53 Stage III 16.6% 0.8% 5.6% 12.5 5% Stage IV V 9.5% 0.5% 0.4% 5.9% % 3.3.7 Stage specific relative survival Belgian stag ge specific ann nual survival data a supplemented until D 7 years by Dutch data Belgian n stage specific annuall survival data supplemented until 7 b Dutch data years by Dutch survival or o British survival or Belgian/French h survival 3.3.8 QALY Stage II III IV -constant - consttant -20.0% Stage III IV -constant -consta ant -20.0% Stage IV V -constant -consta ant -20.0% Age rela ated QALY -constant -consta ant range from + 20 0% to -20% Discoun nted QALY discounting rate e + 1.5%. 3% and 5% 54 S Screening Breast Cancer C 3.4. Results s In the baseline scenario the mod del predicts that th here would be 130 07 years er 100 000 (13,1 p per 1000) women invited for screen ning and of life saved pe 395 per 100 00 00 (3.9 per 1000) QALYs. The mod del also predicts that 128 deaths would be b averted per 100 000 women screened s (1.3 pe er 1000), being a reductio on of 21% (numbe er needed to be offered o screening: 782). Because of the e considerable un ncertainty surroun nding the parametters and model structure e we did an exten nsive sensitivity analysis. a Most unc certainty is not due to ra andom error but d due to issues rela ating to the right choice c of source of inforrmation on the p parameter. We did not do a prob babilistic sensitivity ana alysis, as it wa as not possible to choose app propriate probability-distrributions in a meaningful way. Table 3.4 show ws the results of th he sensitivity analysis, the plausible e ranges used for this an nalysis was discu ussed in 3.3, desc cription of the parrameters and justification n of chosen valuess. The number off years of life gain ned remains fairly y constant under different T number of QA ALY gained or lostt varies much more under assumptions. The different assum mptions. This is partly due to th he fact the a lott of the uncertain or va ariable parameters have an impa act on the quality of life gained rather th han on mortality, such as high rec call rates, over dia agnosis, apart from the values v accorded tto the QALYs. Assumed degre ee of over-diagno osis has a strong impact on QALYs s gained and under the higher assumed d values of 20% or 30% even im mply that b lost instead o of gained. Years of life gained in ncreases QALY would be slightly, this due e to the fact that an over diagnose ed case cannot be ecome a new case in the e model, one could argue that this is i somewhat of an n artifact but the effect is s very small. Recall rates of 10% also can sh hift the balance to owards a loss of QALYs. ecall rates are actu ually found in som me parts of Belgium m. Ten per cent re Assumptions on n the choice of th he appropriate surrvival curve have both an impact on years of life gained and QALYs gained. The Dutch and the e number of life years y gained but lead to a British survival data increase the val data suppleme ented by loss of QALYs in certain scenarios. Belgian surviv where in between. French 10 yearr survival is somew KCE Reportt 176 In ncreasing the ass sumed lead time to t 3 years has an n impact on both years of life gained and QALYs. o Q T model’s estim The mation of the numb ber of QALYs gain ned or lost depends on th he valuation of th hese QALYs. Dim minishing the age related QALYs, this t is th he decrease in quality q of life due e to old age, deccreases the numb ber of Q QALYs gained, as s could be expecte ed. T estimations co The oming from the Lidgren’s paper are e fairly uniform an nd do n vary much in function of the different stages. W not We introduced a larger l d decrement in qua ality of life due to o increasing stage e at diagnosis, with w 3 s scenarios: (i) decrreasing stage II III and IV with 20% %, (ii) decreasing stage IIII and IV with 20 0% or (iii) decrea asing stage IV with 20%. This ha as the e effect of increasing the number of QALYs gained, b because there are e also g gains in QALYs du ue to the stage sh hift alone outside the effect on morrtality, a persons in stag as ge I have in this scenario a better a assumed quality of o life, in n contrast to the Lidgren L data. A could be expec As cted, introducing discount d rates decreases the numb ber of Q QALYs gained. A a worst case scenario, As s we set th he estimation of o over diagnosis at 20%, re ecall rate at 10% %, loss of QALYs per recall at 0.19 9 during a period of 54 d days and using th he stage distributiion coming from the Flemish screening p program (scenario o 2). This gives a gain of 872 Years of Life but a lo oss of 3 QALYs per 10 307 00 000. A a best case sc As cenario, we set the estimation of over-diagnosis at 3%, re ecall rate at 2%, loss of QALYs per p recall at 0.13 during a period of 36 d days and using th he stage distribution coming from m the Dutch screening p program (scenario o 1). This gives a gain of 1704 Years of Life and a gain of 1626 QALYs per 100 1 000. A Applying the 22% % reduction in mortality m from the meta-analysis from G Götzsche et al. to o the Belgian life table, t as describe ed above, gives a very s similar result, 139 cancers deaths due d to breast can ncer avoided and 1145 y years of life saved d. KCE Report 176 6 S Screening Breast Cancer C Table 3.3 Mode eling results: bas seline, worst and d best case scen nario. Scenario Assump ptions Years of life Per 100 00 00 women Quality ad djusted years of life l Per 100 00 00 women Baseline Over diagnosis: 10% 3.5% 3 recall rate at QALYs per recall 0.16 loss of Q during a period of 45 days stage distribution com ming gram Dutch screening prog o 1) (scenario 1307 gaine ed 395 gained d Worst case Over diagnosis: 20% recall rate at 10% QALYs per recall 0.19 loss of Q during a period of 54 days stage distribution com ming gram Flemish screening prog o 2) (scenario 872 gained d 307 lost Best case Over diagnosis: 3% recall rate at 2% QALYs per recall 0.13 loss of Q during a period of 36 days stage distribution com ming gram Dutch screening prog o 1) (scenario 1704 gaine ed 1626 gaine ed 55 56 S Screening Breast Cancer C KCE Reportt 176 Table 3.4 Mode eling results: sen nsitivity analysis s. Stage eshift scena ario 1 Stageshift scenario 2 Sttageshift sc cenario 3 Years s of life QALYs Years of life QALYs Ye ears of life QA ALYs 1307 395 1014 186 12 246 420 0 0.03 1304 526 1011 317 12 245 551 1 0.05 1305 489 1012 280 12 245 514 4 0.1 1307 395 1014 186 12 246 420 0 0.2 1310 208 1018 0 12 249 232 2 0.3 1314 22 1022 -187 12 251 45 0.02 1307 442 1014 234 12 246 459 9 0.035 1307 395 1014 186 12 246 420 0 0.05 1307 348 1014 139 12 246 380 0 0.1 1307 190 1014 -19 12 246 249 9 36 days 1307 417 1014 208 12 246 438 8 45 days 1307 395 1014 186 12 246 420 0 54 days 1307 373 1014 164 12 246 401 1 QALYs loss pe eriod 13% 1307 416 1014 207 12 246 449 9 QALYs loss pe eriod 16% 1307 395 1014 186 12 246 434 4 QALYs loss pe er period 19% 1307 374 1014 166 12 246 420 0 1120 281 869 102 na a na Baseline Assumed overrdiagnosis Recall rate Period betwee en false positive e and confirmation test: duration Period betwee en false positive and confirmatio on test:% QALYs lost Participation rate r 0.6 KCE Report 176 6 S Screening Breast Cancer C 57 0.7 1307 395 1014 186 na a na 0.8 1493 509 1159 270 na a na 0.6 1120 281 869 102 na a na 0.7 1307 395 1014 186 na a na 0.8 1493 509 1159 270 na a na Dutch survivall 1607 710 1181 310 14 481 505 5 Britisch surviv val 1714 399 1148 -3 15 585 374 4 Belgian surviv val supplemented d by French data a 1460 473 1045 96 13 365 477 7 Assumed lead d time 3 years 1098 118 875 77 11 169 187 7 All QALYs min nus 20% 1307 -948 1014 -1089 12 246 -787 All QALYs plus s 20% 1307 1587 1014 1310 12 246 153 34 Stage II III IV -2 20 1307 903 1014 465 12 246 na Stage III IV -20 0 1307 648 1014 370 12 246 na Stage IV -20 1307 450 1014 241 12 246 na Discount rate 1.5% for QALYs 1307 297 1014 121 12 246 274 4 Discount rate 3% 3 for QALYs 1307 215 1014 67 12 246 193 3 Discount rate 5% 5 for QALYs 1307 138 1014 15 12 246 114 4 Effectiveness screening amon ngst participants Survival curve e by stage from o other sources Discounted QA ALYs 58 S Screening Breast Cancer C 3.5. Discussion Under baseline e assumptions, sscreening in the age group 70-74 4 has a limited impact on o breast cancer d deaths avoided and number of yea ars of life saved, amountiing to 1.4 death a avoided per 1000 women offered sc creening in that period an nd 13 years of life e saved per 1000 women, amountin ng to 4.7 days of life gain ned per women offfered screening. This results falll within the range e that the modele ers of the CISNET T project found as reportted by Mandelblattt et al in 200948, where w years of life e gained ranged from 9 to t 22 per thousan nd women screen ned. This, despite the fact that completely y different data and d model structures were used. Years of life ga ained remained fairly constant in the sensitivity analy ysis. We choose a worstt and best case sscenario, with yea ars of life gained ranging from 872 to 170 04. It correspondss also with the sim mplified estimatio on based on the meta-analysis of Götzsche et al.4 and the Belgian B life tables,, despite ation comes from a completely diffferent source of data d and that this estima estimation meth hod. This indicate es that the estima ations of the years of life gained are fairly y robust and conssistent with other studies. s The gain in quality adjusted life years (QALYs) is i considerably le ess, with Ys per 1000 wom men (1.4 quality adjusted day of life per only 3.9 QALY women) offered d screening and uncertainty is large er. One can prese ent these data in anothe er way by statin ng that 250 wom men need to be offered screening for 5 years to gain one year of life. The sensitivity analysis shows that und der certain assumptions introducing g breast cancer sc creening in this age gro oup would actuallly generate a lo oss of QALYs. Th he most important of these is an assum med recall rate off 10%, as is the case in es should certainly y first be certain parts off Belgium, so thesse high recall rate ore proceeding. addressed befo The worst case e scenario would imply a loss of 3 QALYs per 1000 0 women screened, we made m sure that the assumptions off this worst case scenario s are still reason nable assumption ns and not unduly extreme. A nu umber of elements were not considered in the worst cas se scenario beca ause the aseline estimation n for the effect is sometimes mixed. Bringing down the ba p age-group inccreases or decrea ases the final nu umber of quality of life per QALYs gained depending on the chosen values of the other para ameters. ng induces losses s due to This is due to the fact that intrroducing screenin a over diagnosiis but gains due to o the stage shift. false positives and KCE Reportt 176 Under the best ca U ase scenario one would w gain 16 QA ALYs per 1000 wo omen s screened. T The higher variab bility seen in the estimation of the e QALYs comparred to y years of life lost has a numbe er of reasons. T There is conside erable u uncertainty around key parameterrs that determine e a loss of QALY Ys, in p particular concern ning over-diagnosis. Variability due e to recall rates on o the o other hand rather reflects real und derlying difference es in practice bettween c countries and in Belgium between regions. There e is also conside erable u uncertainty around d the valuation off the quality of liffe, and this is nott only u uncertainty conce erning the quality of life surroun nding different breast b c cancer states but also age specific quality of life in B Belgium. If in the future f c cost-effectiveness analyses are considered, thiss problem should be a addressed if we want w to have mea aningful results. Q Quality of life attrib buted to o stage IV has on nly a limited impact on overall num mbers of QALYs gained o lost as survival in this stage is short and proportio or on of stage IV pa atients is s low. C Carles et al. 201138, in an adaptation of the CISNET model of Lee L & Z Zeelen, found an incremental bene efit for biannual sccreening 50-74 off 2.78 life years gained per p 1000 women compared c to a sch hedule 50-69. The ey did n report the QA not ALYs gained with extending the sccreening to 50-74 from 5 50-69, as it was dominated by screening from 45-- 69, but reported d that 1.86 QALYs per 1 000 were gaine ed by extending tthe screening to 45-74 4 estingly, they did not n incorporate th he results of Vilap prinyo frrom 45-69. Intere e al, 201158 into th et heir calculations, but used US survvival data. T model takes into account over--diagnosis and lea The ad time bias. How wever, itt does not take into account leng gth bias, the factt that screen-detected c cancers would ha ave a slower clinical course and have a better su urvival b because screening tends to pick-u up slow growing ttumors some of which w a not life threate are ening. Follow up studies s of screen detected cancers s and n non screen detectted cancer in the e literature show tthat survival of sc creen d detected cases is better than case es among non parrticipants, indepen ndent o stage, and thatt survival of interv of val cases is somewhat in between n88-91. T This indicates tha at there may ind deed be a lengtth time bias, thrrough s selection of less aggressive a cancers by screening. However, the fac ct that in nterval cancers have h a better su urvival than cance er among women not a attending screenin ng indicates that other factors also o play a role, suc ch as KCE Report 176 6 S Screening Breast Cancer C selection bias (such as the so ocial class or other health related d factors en non attending screening) and re esidual confounding after amongst wome adjustment for stage. s However le ength time has no o direct impact on efficacy and on effectiveness, as dete ection of slow grrowing tumors does not gatively correlate e with the ability to detect potentiially lifenecessarily neg threatening can ncers at an earlierr stage. Length tim me has indeed a negative n impact on effic ciency, as detection of indolent tu umors means mo ore harm and greater cos st for no benefit to o women. We unfortunate ely did not have data on stage spec cific survival for tu umors in unscreened wo omen, tumors fo ound by screenin ng and interval cancers. c Moreover, there e is in general co onsiderable uncerttainty around the survival curves that will apply in the future, as treatment ev volves and actual data on e outdated. survival may be Another majorr source of unccertainty is the right choice of o stage distributions of the diagnosed ccancers and stage e-shift. The Flemish data at the stage disttributions of the interval on stage distribution show tha o do not participatte in the cancers and the cancer amongsst the people who v similar. screening are very We choose a modeling m approacch that is essenttially based on th he stage shift and its co onsequences, in ccontrast to most CISNET models that are essentially tumo or growth modelss. This has the adv vantage that it allo owed us to stay closer to the data and make less use of unobserved va ariables, arameters based on Belgian data, but has the disad dvantage incorporating pa that the model is less flexible an nd has more simp plifications. We model m an o screening on tthe proportion of cancers that are e screen overall effect of detected based d on Belgian data a in the group 50-69. This implies that we can only evalua ate the effect of th he screening sche edules actually in place in Belgium, we ca annot vary the screening interval. We W do not have the data however neede ed to parameterize e the CISNET mo odels and would be b forced to use the same e parameters thatt are already used d in the published models, we would just merely m replicate th hem. In conclusion, there t is considera able structural uncertainty around the right choice of the parameters, p so a lot of caution is needed n when inte erpreting the results. This uncertainty is reflected in the wide range of es stimated g and QALY Ys gained in the end e result. Neverrtheless, Years of Life gained there is evidence that continuing g screening until the age of 74 ye ears has 59 modest effect on the m t number of Life e Years Saved bu ut there is conside erable u uncertainty on the e effect on quality y adjusted life yea ars, and the data show th hat under reason nable assumptions the intervention n may even lead d to a lo oss of quality adju usted life years. Itt is important to b bring the recall rattes to a acceptable level before extending screening. s 60 S Screening Breast Cancer C 4. ANSW WER TO CL LINICAL QU UESTIONS S What are clin nical benefits an nd specific harm ms of an exten nsion of breast cancerr organized scre eening in wome en between 70 and 74 years? 4.1. Breast cancer related d mortality What is the effe ect of an extension (70-74 years) of breast cancer orrganized screening on th he breast cancer rrelated mortality? The continued sc creening for breast canc cer between the ages of 70 and 74 makes it pos ssible to obtain an extra 13 years of life fo or 1,000 women screened. s The mo odel also 28 deaths would be averted per 100 000 women screened s predicts that 12 (1.3 per 1000), being a reduction n of 21%. 4.2. Delay between b the sccreening and th he mortality red duction How long is the delay between the screening and the associated d breast d mortality reducction? The morttality reduction appears cancer related between 4 and 7 years after scre eening 4.3. Overall mortality What is the effe ect of an extension (70-74 years) of breast cancer orrganized screening on the t overall morta ality? The effect of an extension n (70-74 years) of breas st cancer organizzed screening on n the overall mo ortality is unclear. Studie es did not have statistical powerr to detect an all-cause a mortality reducttion. 4.4. Morbidity What is the effe ect of an extension (70-74 years) of breast cancer orrganized screening on morbidity? m We found no data relate ed to the cancer morbidity m in randomized control c trials. In other words, on this basis we do no ot accept or reject the hy ypothesis that scre eening reduces th he morbidity of the breast cancer disease. Aim of screening is to detect minor tumors. Conse equently, b diminish by lesss aggressive treatment. The Belg gian data morbidity may be currently at ourr disposal do not enable us to ratiffy this assertion. Actually, A the most recentt data (KCE reporrt 150)86 show 58% % of the interventtions are conservative surgery s versus 3 38% of total ma astectomies in th he least advanced stage es (C Stage I an nd II). Nearly 90% % of patients und dergoing KCE Reportt 176 conservative surge c ery also receive radiotherapy treatm ment, 38% are giv ven a trreatment of neo o-adjuvant chemo otherapy, and 41% receive hormone trreatment. 4 4.5. False pos sitive or false negative n resultss What are the spe W ecific harms in te erms of false possitive or false neg gative re esults? The Belg gian data currently y at our disposall show a recall ra ate of 3 3,5% in Flanders s and of 10% in Walloon and Brussels region n per s screening round. At A this age group, performance of mammography is s high a rates of false negative results are and a relatively low. For USA, rate off false n negative results are 1.5 per 100 00 women aged 70 to 79 years s per s screening round (B BCSC-USA). 4 4.6. Additiona al diagnostic tes sts What are the spec W cific harms in term ms of additional dia agnostic tests? Tw wenty to o forty additional punctures or biop psies may be expe ected per 1000 wo omen o offered screening (three rounds). 4 4.7. Over-diag gnosis and ove er-treatment What are the sp W pecific harms in n terms of overr-diagnosis and overtrreatment? Based d on selected studies, over-detecction (excluding DCIS c cases), ranged fro om (7% to 21%) to o 35% (no data sp pecific for women aged 7 to 79 years are 70 a available). Götzsche G reported d that the numb ber of m mastectomies and d lumpectomies was w significantly llarger in the scre eened g groups (no data specific s for wome en aged 70 to 79 9 years are availa able). T Three trials with adequate a randomiization showed a significant increa ase in m mastectomies and d lumpectomies (R Relative Risk (RR R) 1.31, 95% (CI)) 1.22 to o 1.42). Two trials with suboptimal randomizatio on showed the same in ncrease in interve entions (RR of 1.4 42 (95% CI 1.26 to o 1.61)). The RR for f all fiive trials combined was 1.35 (95% CI 1.26 to 1.44). KCE Report 176 6 S Screening Breast Cancer C 4.8. What attitude a should be recommend ded for women n in case off self referral? It is advisable that when a pattient asks her doctor for a screen ning, the d a strategyy minimizing the drawbacks d of scre eening92. doctor should develop In this way, an attitude stru uctured around three phases can be recommended: • n specific to the ag ge bracket93 Information • Decision making m according tto the patient pers sonal assessmentt94 • Steering off the person who o so wishes towarrds a screening involving methods th hat minimize the d drawbacks. The criteria deffined in the frame ework of the European Programme e notably make provision for the monitorin ng of the technical quality of the eq quipment ble reading of the e mammographies s, and an optimiz zation of used, the doub the recall rate955. In Belgium, the e approved mamm mography units meet m the criteria laid dow wn in the contexxt of the Europea an Programme, and a it is therefore logica al to steer those w women who explicitly request a sc creening towards these structures. s 61 62 S Screening Breast Cancer C 5. REFER RENCES 1. 2. 3. 4. 5. 6. 7. Paulus D, Mambourg F F, Bonneux L. [B Breast cancer scrreening]. Good Clinical Practice (GCP). Brussells: Belgian Health Care edge Centre (KC CE); 2005 02/05 5/2005. KCE rep ports 11 Knowle Availab ble from: http://kc ce.fgov.be/index_ _en.aspx?SGREF=5221&CREF=93 348 Verleye e L, Desomer A, Gailly J, Robays s j. [Identifying wo omen at risk fo or breast cance er/technical meth hods for breast cancer screeniing]. Good clinical Practice (GCP). Bruxelles: Centre e fédéral d'experrtise des soins d de santé (KCE); 2012. KCE Repo orts 172 Availab ble from: http://w www.kce.fgov.be/p publication/report/identifying-women n-atrisk-for--breast-cancertecchnical-methods-fo or-breast-cancer-s sc Mambo ourg F, Robays j,, Camberlin C, Vlayen J, Gailly J.. [Breast cancer screening with m mammography forr women in the agegroup 49 years ]. Good Clinical Practice (GCP). Brussels: Belgian of 40-4 Health Care Knowledge e Centre (KCE); 2010 07/07/201 10. KCE 129 9 Available reports from: ce.fgov.be/index_ _en.aspx?SGREF=14851&CREF=1 16581 http://kc Gotzsche PC, Nielsen M. Screening for breast canc cer with ography. Cochra ane Database of o Systematic Reviews. R mammo 2011(1). Nelson HD, Tyne K, Naik A, Bougatsos C, C Chan BK, Humphrey L. ancer: an update e for the U.S. Pre eventive Screening for breast ca es Task Force. Ann Intern Me ed. 2009;151(10):727-37, Service W237-4 42. Biesheuvel C, Barratt A, Howard K, Houss sami N, Irwig L. Effects E of m and biasses on estimates of invasive breas st cancer study methods overdettection with mammography screen ning: a systematic c review. Lancet Oncol. 2007;8(12 2):1129-38. Jorgens sen KJ, Götzsche e PC. Overdiagno osis in publicly orrganised mammo ography screening programmes: systematic review of incidence trends. Bmj. 20 009;339. 8 8. 9 9. 10. 11. 12. 13. 14. 15. 16. 17. KCE Reportt 176 Virnig BA,, Tuttle TM, Sham mliyan T, Kane RL L. Ductal carcinom ma in situ of the breast: a system matic review of inccidence, treatmentt, and nst. 2010;102(3):170-8. outcomes. J Natl Cancer In Humphrey y LL, Helfand M, Chan BK, Woo olf SH. Breast ca ancer screening: a summary of the evidence for the U.S. Preve entive T Force. Ann Intern Med. 2002;137(5 Part 1):347 7-60. Services Task Tabar L, Fagerberg CJ, Gad A, Baldeto orp L, Holmberg g LH, O et al. Reductio on in mortality fro om breast cancerr after Grontoft O, mass scre eening with mam mmography. Rand domised trial from m the Breast Ca ancer Screening Working W Group off the Swedish National Board of Health H and Welfarre. Lancet. 1985;1 1(8433):829-32. Dixon JM M. Breast screening has increa ased the numbe er of mastectom mies. Breast Canc cer Res. 2009;11 Suppl 3:S19. Smith RA, Duffy SW, Gabe R, Tabar L, Ye en AM, Chen TH. The ed trials of brea ast cancer scree ening: what have e we randomize learned? Radiol R Clin North Am. 2004;42(5):7 793-806, v. Nystrom L, L Andersson I, Bjurstam N, Frise ell J, Nordenskjo old B, Rutqvist LE. Long-term effects of mam mmography scree ening: omised trials. La ancet. updated overview of the Swedish rando 2002;359((9310):909-19. Tabar L, Vitak B, Chen TH, T Yen AM, Co ohen A, Tot T, et al. T Triall: Impact of Mam mmographic Screening Swedish Two-County on Breast Cancer Mortality during 3 Decadess. Radiology. 2011. Maass N, Alkasi O, Baue er M, Jonat W, S Souchon R, Mein nholdment of ductal ca arcinoma in situ of o the Heerlein I. Actual managem ch Gynecol Obste et. 2009;280(5):69 99-705. breast. Arc Duffy SW,, Chen HH, Tabarr L, Day NE. Estim mation of mean so ojourn time in brreast cancer scre eening using a M Markov chain mod del of both entry y to and exit from m the preclinical d detectable phase. Stat Med. 1995 5;14(14):1531-43.. Frachebou ud J, Groenewoud d JH, Boer R, Dra aisma G, de Bruijn AE, Verbeek AL, A et al. Seventy y-five years is an appropriate uppe er age limit for po opulation-based mammography m sccreening. Int J Ca ancer. 2006;118((8):2020-5. KCE Report 176 6 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. S Screening Breast Cancer C Fett MJ J. Computer mod delling of the Sw wedish two county y trial of mammo ographic screenin ng and trade offs between b participa ation and screeniing interval. Journ nal of Medical Scre eening. 2001;8(1)):39-45. Shen Y, Zelen M. R Robust modeling g in screening studies: estimattion of sensitivityy and preclinical sojourn time disttribution. Biostatistics. 2005;6(4):6 604-14. Wu JC, Hakama M, Antttila A, Yen AM, Malila M N, Sarkeala T, et al. o breast cancer based b on Estimattion of natural history parameters of non-ran ndomized organizzed screening data: subsidiary ana alysis of effects of inter-screening g interval, sensitivity, and attendance rate earch & on reduction of advancced cancer. Breast Cancer Rese Treatment. 2010;122(2):553-66. OF, Myles JP, Lynge E, Dufffy SW. Olsen AH, Agbaje O agnosis, sojourn time, and sensittivity in the Cope enhagen Overdia mammo ography scree ening program m. Breast Journal. 2006;12 2(4):338-42. Frisell J, Eklund G, He ellstrom L, Somell A. Analysis of interval eening trial in Sto ockholm. breast carcinomas in a randomized scre h & Treatment. 19 987;9(3):219-25. Breast Cancer Research E Duffy SW. M Modelling the analysis of breast cancer Paci E, screeniing programmes: sensitivity, lead time t and predictiv ve value in the Florence Districct Programme (1 1975-1986). Interrnational Journal of Epidemiology. 1991;20(4):852-8. R de Koning H HJ, van der Maa as PJ. A longerr breast Boer R, carcino oma screening interval for women age older than 65 5 years? Cancerr. 1999;86(8):1506 6-10. Duffy SW, S Day NE, Tabar L, Chen HH, Smith S TC. Markov v models of brea ast tumor progresssion: some age-s specific results. Jo ournal of the Nattional Cancer Insttitute. Monographs s. 1997;22:93-7. Duffy SW, S Gabe R. Wha at should the dete ection rates of can ncers be in brea ast screening programmes? Brittish Journal of Cancer. 2005;92 2(3):597-600. Shen Y, Y Zelen M. Scre eening sensitivity y and sojourn tim me from breast cancer early de etection clinical trrials: mammogra ams and 2 28. 2 29. 3 30. 3 31. 3 32. 3 33. 3 34. 3 35. 3 36. 3 37. 63 physical examinations. Journal of Clinical Onco ology. 2001;19(15):3490-9. Spratt JS, Greenberg RA, Heuser H LS. Geom metry, growth rates s, and o cancer and ca arcinoma in situ of the breast before b duration of detection by screening. Can ncer Research. 19 986;46(2):970-4. Tabar L, Fagerberg F G, Che en HH, Duffy SW, Smart CR, Gad A, et al. Efficac cy of breast cance er screening by a age. New results from the Swedish Two-County Trial. Cancer. 1995 5;75(10):2507-17.. Weedon-F Fekjaer H, Lindqv vist BH, Vatten LJ, Aalen OO, Tre etli S. Estimating g mean sojourn time and scree ening sensitivity using questionna aire data on time e since previous screening. Journ nal of Medical Screening. 2008;15 5(2):83-90. Weedon-F Fekjaer H, Vatten n LJ, Aalen OO, Lindqvist B, Tre etli S. Estimating g mean sojourn time and screen ning test sensitiv vity in breast cancer mammograp phy screening: new results. Journ nal of 2(4):172-8. Medical Screening. 2005;12 Zappa M, Visioli CB, Ciatto o S. Mammograph hy screening in elderly e e and cost--effectiveness. Re eview 16 refs. Critical C women: efficacy Reviews in n Oncology Hema atology. 2003;46(3 3):235-9. Boer R, de Koning HJ, van n Oortmarssen GJJ, van der Maas PJ. P In ge limit for breast cancer screening g. Eur search of the best upper ag J Cancer. 1995;31A(12):2040-3. Mandelbla att JS, Silliman R. Hanging in the balance: making decisions about the benefits s and harms of brreast cancer screening e oldest old witho out a safety net o of scientific eviden nce. J among the Clin Oncol. 2009;27(4):487--90. Karnon J, Goyder E, Tappe enden P, McPhie S, Towers I, Braz zier J, ng in prioritising and et al. A review and crittique of modellin ogrammes. Heallth Technol As ssess. designing screening pro 52):iii-iv, ix-xi, 1-14 45. 2007;11(5 Barratt A, Howard K, Irwig g L, Salkeld G, H Houssami N. Mod del of m in nformation to su upport outcomes of screening mammography: c BMJ. 200 05;330(7497):936 6. informed choices. Berry DA,, Inoue L, Shen Y, Y Venier J, Cohen D, Bondy M, et al. Modeling the impact of tre eatment and scre eening on U.S. breast b 64 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. S Screening Breast Cancer C cancer mortality: a Baye esian approach. J Natl Cancer Inst Monogr. Monographs.(36):3 30-6. 2006;M Carles M, Vilaprinyo E, Cots F, Gregori A, A Pla R, Roman R, et al. arly detection of breast b cancer in Catalonia C Cost-efffectiveness of ea (Spain)). BMC Cancer. 20 011;11. Carter KJ, Castro F, Ke essler E, Erickson n BA. Simulation of begin nd ages for mammography scre eening. J Health hc Qual. and en 2005;27 7(1):40-7. Feuer EJ. E Modeling the impact of adjuva ant therapy and sc creening mammo ography on U.S. breast cancer mo ortality between 1975 and 2000: introduction to th he problem. J Natl Cancer Inst Monogr. Monographs.(36):2 2-6. 2006;M Fryback DG, Stout N NK, Rosenberg MA, Trentham-D Dietz A, ngton PL. The Wisconsin W Breast Cancer Kuruchittham V, Remin miology Simulatio on Model. J Na atl Cancer Inst Monogr. Epidem 2006;M Monographs.(36):3 37-47. Gyrd-H Hansen D. Cost-be enefit analysis of mammography sc creening in Denm mark based on diiscrete ranking da ata. Int J Technoll Assess Health Care. 2000;16(3):811-21. Hanin LG, Miller A, Zo orin AV, Yakovlev AY. The Unive ersity of ster model of bre east cancer detec ction and survival. J Natl Roches Cancerr Inst Monogr. 200 06;Monographs.(3 36):66-78. Lee S, Zelen M. A stoch hastic model for predicting p the mo ortality of cancer. J Natl Can ncer Inst breast Monogr. Monographs.(36):7 79-86. 2006;M Lee SJ J, Zelen M. Mode elling the early de etection of breast cancer. Ann On ncol. 2003;14(8):1199-202. Mahnke en JD, Chan W, Freeman DH, Jr., Freeman JL. Reducing R the effe ects of lead-time e bias, length bia as and over-dete ection in evaluatting screening m mammography: a censored bivaria ate data approach. Stat Methods Med Res. 2008;17(6):643-63. Mandelblatt J, Schechte er CB, Lawrence W, Yi B, Cullen n J. The TRUM population n model of the impact of screen ning and SPECT treatme ent on U.S. bre east cancer trend ds from 1975 to o 2000: 4 48. 4 49. 5 50. 5 51. 5 52. 5 53. 5 54. 5 55. KCE Reportt 176 principles and practice of the t model method ds. J Natl Cance er Inst 2 .(36):47-55. Monogr. 2006;Monographs Mandelbla att JS, Cronin KA A, Bailey S, Berrry DA, de Koning g HJ, Draisma G, et al. Effects of mammogra aphy screening under u dules: model esstimates of pottential different screening sched m appears in Ann n Intern Med. 2010 0 Jan benefits and harms.[Erratum 0):738-47. 19;152(2)::136]. Ann Intern Med. 2009;151(10 Mandelbla att JS, Schechter CB, Yabroff KR, Lawrence W, Dig gnam J, Exterm mann M, et al. To oward optimal sccreening strategie es for older wom men: Costs, benefits, and harm ms of breast ca ancer screening by age, biology, and health statuss. J. Gen. Intern. Med. 6):487-96. 2005;20(6 Neeser K,, Szucs T, Bulliard d JL, Bachmann G, Schramm W. Costeffectivene ess analysis off a quality-conttrolled mammogrraphy screening program from m the Swiss statutory health h-care ve: quantitative as ssessment of the most influential fa actors perspectiv (Structured abstract). 2007;;10(1):42-53. Plevritis SK, S Sigal BM, Salzman S P, Rose enberg J, Glynn P. A stochastic c simulation mode el of U.S. breast ccancer mortality trrends from 19 975 to 2000. J Natl Cancer Inst Mo onogr. 2006;Monographs.(36):86-9 95. Rauner MS, M Gutjahr WJ, Heidenberger K, Wagner J, Pas sia J. Dynamic Policy Modeling for Chronic Disseases: Metaheuristicdentification of Pareto-Optimal P S Screening Strate egies. Based Id Operations Research. 2010 0;58(5):1269-86. Rijnsburge er AJ, van Oortm marssen GJ, Boerr R, Draisma G, To T T, Miller AB, et al. Mammogrraphy benefit in tthe Canadian National creening Study-2 2: a model evaluation. Int J Ca ancer. Breast Sc 2004;110((5):756-62. Rojnik K, Naversnik K, Mateovic-Rojnikk T, Primiczakelj M. ness modeling of different breast ca ancer Probabilistic cost-effectiven nia. Value Health. 2008;11(2):139-4 48. screening policies in Sloven Rue M, Vilaprinyo V E, Lee e S, Martinez-Allonso M, Carles MD, Marcos-Gragera R, et al. Efffectiveness of ea arly detection on breast b KCE Report 176 6 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. S Screening Breast Cancer C cancer mortality reducttion in Catalonia a (Spain). BMC Cancer. 2009;9(326). N Rosenberg MA A, Trentham-Diettz A, Smith MA, Robinson R Stout NK, SM, Frryback DG. Retrrospective cost-e effectiveness ana alysis of screeniing mammography. J Natl Cancer Inst. I 2006;98(11):774-82. Tan SY Y, van Oortmarsssen GJ, de Konin ng HJ, Boer R, Habbema H JD. The e MISCAN-Fadia continuous tumor growth model fo or breast cancer.. J Natl Cancer Inst Monogr. 2006;M Monographs.(36):56-65. Vilaprin nyo E, Rue M, Marcos-Gragera R, Martinez-Alo onso M. Estimattion of age- an nd stage-specific Catalan breast cancer surviva al functions using g US and Cata alan survival data a. BMC Cancerr. 2009;9(98). Wang H, Karesen R, Hervik A, Thore esen SA. Mamm mography esults from the firs st screening round in four screeniing in Norway: Re countie es and cost-efffectiveness of a modeled nationwide screeniing. Cancer Causes Control. 2001;12(1):39-45. Messec car DC. Mammog graphy screening for f older women with w and withoutt cognitive impairrment. J Gerontoll Nurs. 2000;26(4 4):14-24; quiz 52 2-3. Wen YP P, Sheu ML. A co ost-benefit analysis of preventive ca are: The case of o breast cancer screening. Taiwain J. Public Health. 2005;24 4(6):519-28. Advisorry Committee on n Breast Cancer S. Screening fo or breast cancer in England: past and future. J Me ed Screen. 2006;1 13(2):5961. Anonym mous. Landelijk bevolkingsonderrzoek naar bors stkanker volledig g ingevoerd; resultaten van de implementatiefase e 19901997. Landelijk L Evaluattie Team voor be evolkingsonderzo oek naar Borstka anker. Ned Tijdsch hr Geneeskd. 200 00;144(23):1124-9 9. Barratt A, Irwig L, Glasziiou P, Salkeld G, Houssami N, Kerllikowske y reduces with age. Evid.K, et al. Relative benefit of mammography 4):156-7. Based Healthc. 2002;6(4 M, Glasziou PP, Salkeld S GP, Hous ssami N. Barratt AL, Les Irwig M m in n women Benefits, harms and cossts of screening mammography 6 66. 6 67. 6 68. 6 69. 7 70. 7 71. 7 72. 7 73. 7 74. 7 75. 65 70 years s and over: a systematic review. Med J Aust. 2002;176((6):266-71. Bonneux L. De voor- en nadelen n van borsstkankerscreening g: tijd dence-based info ormatie. Nederla ands Tijdschrift voor voor evid Geneesku unde. 2009;153. Caplan LS S. To screen or not to screen: the issue of breast ca ancer screening in older women.. Public Health R Rev. 2001;29(2-4)):23140. Carney PA A, Abraham LA, Miglioretti DL, Ya abroff KR, Sickles s EA, Buist DSM M, et al. Factors associated a with im maging and proce edural events used u to detec ct breast canccer after screening mammogrraphy. Am. J. Roe entgenol. 2007;188(2):385-92. De Koning g HJ. Breast canc cer screening; cosst-effective in prac ctice? Eur J Radiol. 2000;33(1):32 2-7. Feuer EJ, Etzioni R, Cronin n KA, Mariotto A. T The use of modeling to S. mortality: exam mples understand the impact of screening on U.S mmography and PSA testing. Sta at Methods Med Res. from mam 2004;13(6 6):421-42. Grivegnee e AR, Autier P. Approche A econom mique du depistag ge du cancer du sein en Belgique e. Rev Med Brux. 2 2001;22(4):A277--81. Habbema JD, Tan SY, Crronin KA. Impact of mammograph hy on st cancer mortality y, 1975-2000: are e intermediate outcome U.S. breas measures informative? J Natl Can ncer Inst Mo onogr. 5-11. 2006;Monographs.(36):105 Mandelbla att J, Saha S, Teu utsch S, Hoerger T, Siu AL, Atkins D, et al. The co ost-effectiveness of o screening mammography beyond d age 65 years: a systematic re eview for the U.S S. Preventive Serrvices ce. Ann Intern Med d. 2003;139(10):8 835-42. Task Forc Prevost TC, T Abrams KR R, Jones DR. Hierarchical mode els in generalize ed synthesis of ev vidence: an exam mple based on sttudies of breast cancer c screening. Stat Med. 2000;1 19(24):3359-76. Rautenstra auch J. Is mam mmography screening only a poin ntless waste of money? m MMW-Fortschr. Med. 2000 0;142(12):4-10. 66 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. S Screening Breast Cancer C Xu W, Vnenchak P, S Smucny J. Scre eening mammogrraphy in n aged 70 to 79 ye ears. J. 2000;49(3 3):266-7. women Berry DA, D Cronin KA, Ple evritis SK, Frybac ck DG, Clarke L, Zelen Z M, et al. Effect E of screenin ng and adjuvant therapy t on mortality from breast cancer. c N Engl J Med. 2005;353(17 7):1784-92. Carter KJ, Castro F, Kesssler E, Erickson B. A computer model m for dy of breast cance er. Comput Biol Med. M 2003;33(4):345-60. the stud Cleemp put I, Van Wilder P, Vrijens F, Huyb brechts M, Ramae ekers D. Recommandations pour les évaluations pharmacoéconomi p iques en ue. Health technology Assessment (HTA). Bruxelles s: Centre Belgiqu fédéral d'expertise des ssoins de santé (K KCE); 2008. KCE Reports 78B (D//2008/10.273/24) Lidgren n M, Wilking N, Jonsson B, Reh hnberg C. Health h related quality of life in differen nt states of breas st cancer. Qual Life L Res. 6(6):1073-81. 2007;16 Freedm man GM, Li T, An nderson PR, Nicolaou N, Konski A. A Health states of o women after co onservative surgerry and radiation fo or breast cancer.. Breast Cancer R Res Treat. 2010;12 21(2):519-26. Burstro om K, Johannesso on M, Diderichsen n F. Health-related d quality of life by disease and d socio-economic c group in the general Health Policy. 2001 1;55(1):51-69. populattion in Sweden. H Gerard K, Johnston K, Brown J. The ro ole of a pre-score ed multie health classifiication measure in validating co onditionattribute specific c health state desccriptions. Health Econ. E 1999;8(8):6 685-99. Domey yer PJ, Sergentan nis TN, Zagouri F, Zografos GC. Healthrelated quality of life in vvacuum-assisted breast biopsy: sh hort-term y of Life effects,, long-term effectts and predictors. Health & Quality Outcom mes. 2010;8(11):2 2010. Borstka anker LETvb bn. Landelijke e evaluatie van bevolkingsonderzoek na aar borstkanker in Nederland 199 90-2007. 2010. ur S, Vrijens F, Beirens K, Vlay yen J, Devriese S, Van Stordeu Eycken n E. Quality indiccators in oncolog gy: breast bance er. Good Clinicall Practice (GCP). Brussels: Belgian n Health Care Knowledge Centre (KCE); 2010. KCE reports 15 50C (D/2010/10.2 273/101) 8 87. 8 88. 8 89. 9 90. 9 91. 9 92. 9 93. 9 94. 9 95. KCE Reportt 176 Available from: 211&CREF=1884 47 http://kce.ffgov.be/index_en.aspx?SGREF=52 INC. Surv vie attendue des patients p atteints d de cancers en Fra ance : état des lie eux. 2010. Mook S, Van V 't Veer LJ, Ru utgers EJ, Ravdin PM, van de Velde e AO, van Leeuw wen FE, et al. In ndependent progn nostic value of sc creen detection in invasive brreast cancer. J Natl Cancer Inst. 2011;103((7):585-97. Cortesi L, Chiuri VE, Rusce elli S, Bellelli V, N Negri R, Rashid I, et al. s of screen-dete ected breast ca ancers: results of a Prognosis population n based study. BM MC Cancer. 2006;6:17. Joensuu H, H Lehtimaki T, Ho olli K, Elomaa L, T Turpeenniemi-Hujjanen T, Kataja V, et al. Risk fo or distant recurre ence of breast ca ancer hy screening or o other methods. Jama. J detected by mammograph 2004;292((9):1064-73. Olsson A, Borgquist S, Buttt S, Zackrisson S S, Landberg G, Manjer M a prognosis in b breast cancer detected J. Tumourr-related factors and by screening. Br J Surg. 20 012;99(1):78-87. Physicians s AAoF. Summa ary of Recomme endations for Cllinical Preventive e Services. In: AA AFP Policy Action AAFP; 2010. Woloshin S, Schwartz LM. The bene efits and harm ms of AMA. mammogrraphy screening: understanding tthe trade-offs. JA 2010;303((2):164-5. Jorgensen n KJ, Gotzsche PC. Content of invitations for pu ublicly funded screening mammog graphy. BMJ. 2006 6;332(7540):538-4 41. Perry N, Broeders B M, de Wolf W C, Tornberg S, Holland R, von Karsa K L. Europe ean guidelines fo or quality assura ance in breast ca ancer screening and diagnosis. Fourth F edition--sum mmary documentt. Ann 08;19(4):614-22. Oncol. 200 KCE Report 176 6 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. S Screening Breast Cancer C cancer mortality reducttion in Catalonia a (Spain). BMC Cancer. 2009;9(326). Stout NK, N Rosenberg MA A, Trentham-Diettz A, Smith MA, Robinson R SM, Frryback DG. Retrrospective cost-e effectiveness ana alysis of screeniing mammography. J Natl Cancer Inst. I 2006;98(11):774-82. Tan SY Y, van Oortmarsssen GJ, de Konin ng HJ, Boer R, Habbema H JD. The e MISCAN-Fadia continuous tumor growth model fo or breast cancer.. J Natl Cancer Inst Monogr. 2006;M Monographs.(36):56-65. Vilaprin nyo E, Rue M, Marcos-Gragera R, Martinez-Alo onso M. Estimattion of age- an nd stage-specific Catalan breast cancer surviva al functions using g US and Cata alan survival data a. BMC Cancerr. 2009;9(98). Wang H, Karesen R, Hervik A, Thore esen SA. Mamm mography screeniing in Norway: Re esults from the firs st screening round in four countie es and cost-efffectiveness of a modeled nationwide screeniing. Cancer Causes Control. 2001;12(1):39-45. Messec car DC. Mammog graphy screening for f older women with w and withoutt cognitive impairrment. J Gerontoll Nurs. 2000;26(4 4):14-24; quiz 52 2-3. Wen YP P, Sheu ML. A co ost-benefit analysis of preventive ca are: The case of o breast cancer screening. Taiwain J. Public Health. 2005;24 4(6):519-28. Advisorry Committee on n Breast Cancer S. Screening fo or breast cancer in England: past and future. J Me ed Screen. 2006;1 13(2):5961. Anonym mous. Landelijk bevolkingsonderrzoek naar bors stkanker volledig g ingevoerd; resultaten van de implementatiefase e 19901997. Landelijk L Evaluattie Team voor be evolkingsonderzo oek naar Borstka anker. Ned Tijdsch hr Geneeskd. 200 00;144(23):1124-9 9. Barratt A, Irwig L, Glasziiou P, Salkeld G, Houssami N, Kerllikowske K, et al. Relative benefit of mammography y reduces with age. Evid.Based Healthc. 2002;6(4 4):156-7. Barratt AL, Les Irwig M M, Glasziou PP, Salkeld S GP, Hous ssami N. Benefits, harms and cossts of screening mammography m in n women 6 66. 6 67. 6 68. 6 69. 7 70. 7 71. 7 72. 7 73. 7 74. 7 75. 67 70 years s and over: a systematic review. Med J Aust. 2002;176((6):266-71. Bonneux L. De voor- en nadelen n van borsstkankerscreening g: tijd voor evid dence-based info ormatie. Nederla ands Tijdschrift voor Geneesku unde. 2009;153. Caplan LS S. To screen or not to screen: the issue of breast ca ancer screening in older women.. Public Health R Rev. 2001;29(2-4)):23140. Carney PA A, Abraham LA, Miglioretti DL, Ya abroff KR, Sickles s EA, Buist DSM M, et al. Factors associated a with im maging and proce edural events used u to detec ct breast canccer after screening mammogrraphy. Am. J. Roe entgenol. 2007;188(2):385-92. De Koning g HJ. Breast canc cer screening; cosst-effective in prac ctice? Eur J Radiol. 2000;33(1):32 2-7. Feuer EJ, Etzioni R, Cronin n KA, Mariotto A. T The use of modeling to understand the impact of screening on U.S S. mortality: exam mples from mam mmography and PSA testing. Sta at Methods Med Res. 2004;13(6 6):421-42. Grivegnee e AR, Autier P. Approche A econom mique du depistag ge du cancer du sein en Belgique e. Rev Med Brux. 2 2001;22(4):A277--81. Habbema JD, Tan SY, Crronin KA. Impact of mammograph hy on U.S. breas st cancer mortality y, 1975-2000: are e intermediate outcome measures informative? J Natl Can ncer Inst Mo onogr. 2006;Monographs.(36):105 5-11. Mandelbla att J, Saha S, Teu utsch S, Hoerger T, Siu AL, Atkins D, et al. The co ost-effectiveness of o screening mammography beyond d age 65 years: a systematic re eview for the U.S S. Preventive Serrvices Task Forc ce. Ann Intern Med d. 2003;139(10):8 835-42. Prevost TC, T Abrams KR R, Jones DR. Hierarchical mode els in generalize ed synthesis of ev vidence: an exam mple based on sttudies of breast cancer c screening. Stat Med. 2000;1 19(24):3359-76. Rautenstra auch J. Is mam mmography screening only a poin ntless waste of money? m MMW-Fortschr. Med. 2000 0;142(12):4-10. 68 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. S Screening Breast Cancer C Xu W, Vnenchak P, S Smucny J. Scre eening mammogrraphy in women n aged 70 to 79 ye ears. J. 2000;49(3 3):266-7. Berry DA, D Cronin KA, Ple evritis SK, Frybac ck DG, Clarke L, Zelen Z M, et al. Effect E of screenin ng and adjuvant therapy t on mortality from breast cancer. c N Engl J Med. 2005;353(17 7):1784-92. Carter KJ, Castro F, Kesssler E, Erickson B. A computer model m for the stud dy of breast cance er. Comput Biol Med. M 2003;33(4):345-60. Cleemp put I, Van Wilder P, Vrijens F, Huyb brechts M, Ramae ekers D. Recommandations pour les évaluations pharmacoéconomi p iques en Belgiqu ue. Health technology Assessment (HTA). Bruxelles s: Centre fédéral d'expertise des ssoins de santé (K KCE); 2008. KCE Reports 78B (D//2008/10.273/24) Lidgren n M, Wilking N, Jonsson B, Reh hnberg C. Health h related quality of life in differen nt states of breas st cancer. Qual Life L Res. 2007;16 6(6):1073-81. Freedm man GM, Li T, An nderson PR, Nicolaou N, Konski A. A Health states of o women after co onservative surgerry and radiation fo or breast cancer.. Breast Cancer R Res Treat. 2010;12 21(2):519-26. Burstro om K, Johannesso on M, Diderichsen n F. Health-related d quality of life by disease and d socio-economic c group in the general populattion in Sweden. H Health Policy. 2001 1;55(1):51-69. Gerard K, Johnston K, Brown J. The ro ole of a pre-score ed multiattribute e health classifiication measure in validating co onditionspecific c health state desccriptions. Health Econ. E 1999;8(8):6 685-99. Domey yer PJ, Sergentan nis TN, Zagouri F, Zografos GC. Healthrelated quality of life in vvacuum-assisted breast biopsy: sh hort-term effects,, long-term effectts and predictors. Health & Quality y of Life Outcom mes. 2010;8(11):2 2010. van Borstka anker LETvb bn. Landelijke e evaluatie bevolkingsonderzoek na aar borstkanker in Nederland 199 90-2007. 2010. Stordeu ur S, Vrijens F, Beirens K, Vlay yen J, Devriese S, Van Eycken n E. Quality indiccators in oncolog gy: breast bance er. Good Clinicall Practice (GCP). Brussels: Belgian n Health Care Knowledge Centre (KCE); 2010. KCE reports 15 50C (D/2010/10.2 273/101) 8 87. 8 88. 8 89. 9 90. 9 91. 9 92. 9 93. 9 94. 9 95. KCE Reportt 176 Available from: http://kce.ffgov.be/index_en.aspx?SGREF=52 211&CREF=1884 47 INC. Surv vie attendue des patients p atteints d de cancers en Fra ance : état des lie eux. 2010. Mook S, Van V 't Veer LJ, Ru utgers EJ, Ravdin PM, van de Velde e AO, van Leeuw wen FE, et al. In ndependent progn nostic value of sc creen detection in invasive brreast cancer. J Natl Cancer Inst. 2011;103((7):585-97. Cortesi L, Chiuri VE, Rusce elli S, Bellelli V, N Negri R, Rashid I, et al. Prognosis s of screen-dete ected breast ca ancers: results of a population n based study. BM MC Cancer. 2006;6:17. Joensuu H, H Lehtimaki T, Ho olli K, Elomaa L, T Turpeenniemi-Hujjanen T, Kataja V, et al. Risk fo or distant recurre ence of breast ca ancer detected by mammograph hy screening or o other methods. Jama. J 2004;292((9):1064-73. Olsson A, Borgquist S, Buttt S, Zackrisson S S, Landberg G, Manjer M J. Tumourr-related factors and a prognosis in b breast cancer detected by screening. Br J Surg. 20 012;99(1):78-87. Physicians s AAoF. Summa ary of Recomme endations for Cllinical Preventive e Services. In: AA AFP Policy Action AAFP; 2010. Woloshin S, Schwartz LM. The bene efits and harm ms of mammogrraphy screening: understanding tthe trade-offs. JA AMA. 2010;303((2):164-5. Jorgensen n KJ, Gotzsche PC. Content of invitations for pu ublicly funded screening mammog graphy. BMJ. 2006 6;332(7540):538-4 41. Perry N, Broeders B M, de Wolf W C, Tornberg S, Holland R, von Karsa K L. Europe ean guidelines fo or quality assura ance in breast ca ancer screening and diagnosis. Fourth F edition--sum mmary documentt. Ann Oncol. 200 08;19(4):614-22.
