IJC International Journal of Cancer

Transcription

1 IJC International Journal of Cancer Unbiased estimates of long-term net survival of solid cancers in France Valerie Jooste 1,2,3, Pascale Grosclaude 1,4, Laurent Remontet 5,6,7,8, Guy Launoy 1,9, Isabelle Baldi 1,10, Florence Molinie 1,11, Patrick Arveux 1,12, Nadine Bossard 5,6,7,8, Anne-Marie Bouvier 1,2,3, Marc Colonna 1,1,3 and the French Network of Cancer Registries (FRANCIM) 1 FRANCIM (French Network of Cancer Registries), Toulouse, France 2 Registre des cancers digestifs, INSERM U866, Universite de Bourgogne, Dijon, France 3 Centre Hospitalier Universitaire de Dijon, Dijon, France 4 Registre des cancers du Tarn, Institut Claudius Regaud, Toulouse, France 5 Hospices Civils de Lyon, Service de Biostatistique, Lyon, France 6 Universite de Lyon, Lyon, France 7 Universite Lyon 1, Villeurbanne, France 8 CNRS, UMR5558, Laboratoire de Biometrie et Biologie Evolutive, Equipe Biotatistique-Sante, Villeurbanne France 9 Registre des tumeurs digestives du Calvados, INSERM U1086, Caen, France 10 Registres des tumeurs cerebrales, Institut de Sante Publique, d Epidemiologie et de Developpement, Bordeaux, France 11 Registre des cancers de Loire-Atlantique-Vendee, CHU, Nantes, France 12 Registre des Cancers du Sein de Côte d Or, Centre Georges François Leclerc, Dijon, France 13 Registre du Cancer de l Isère, Grenoble, France In cancer studies, net survival (observed if cancer was the only cause of death) is a useful indicator but survival estimation at 5 years is insufficient for planning healthcare needs. We estimated the net survivals at 5 and 10 years in a cohort of 387,961 patients who had solid tumors between 1989 and 2004 and were followed-up until January 1, The cases were actively followed-up. Net survival was estimated with the unbiased Pohar-Perme method. The standardized net survival used the international cancer survival standard weights. In men, the standardized net survivals ranged from 92% at 5 years and 89% at 10 years (testis) to 6% at 5 years and 5% at 10 years (pancreas). In women, it ranged from 91% at 5 years and 88% at 10 years (thyroid) to 10% at 5 years and 7% at 10 years (pancreas). The most frequent cancers had the highest net survivals: 84% at 5 years and 71% at 10 years for prostate and 84% at 5 years and 74% at 10 years for breast cancer. Advanced age was associated with poorer prognosis. In most cancers, the net survivals at 5 and 10 years increased over periods of diagnosis. Net cancer survival is unaffected by mortalities due to other causes. It is the only indicator suitable for comparisons between countries or periods of diagnosis within a given country. The 10-year net survival confirmed the persistent unfavorable role of age in prognosis and the general improvement of cancer management over the last decade. Key words: cancer, registries, population-based study, net survival, survival analysis Conflict of interest: All authors declare they have no financial or personal relationships with other people or organizations that could inappropriately influence (bias) their work. Grant sponsor: Institut National du Cancer (France); Grant sponsor: Institut de Veille Sanitaire (France); Grant sponsor: Agence Nationale de la Recherche (France); Grant number: ANR- 09-BLAN-0357 DOI: /ijc History: Received 11 Apr 2012; Accepted 3 Sep 2012; Online 24 Sep 2012 Correspondence to: Bouvier AM, CHU Dijon, F-21079; Registre Bourguignon des Cancers Digestifs Inserm U866, Faculte de Medecine, BP 87900, Dijon Cedex France, Tel.: [ ), Fax: +[ ), anne-marie.bouvier@u-bourgogne.fr Up to now, the available epidemiological indicators of cancer survival in France were 5-year relative or overall survivals. 1 These survival estimates have shown that the most frequent cancer localizations were amongst the high survival group (prostate and breast) or the moderate survival group (colon and rectum). This highlights the importance of longer-term cancer survival studies, say up to 10 years after diagnosis, because long-term survival estimates are objective key indicators for quantifying and planning healthcare needs. Among various indicators, net survival defined as the survival that would be observed if cancer was the only possible cause of death is of major importance in populationbased studies because comparisons of net survivals are not affected by differences in other-cause mortalities; this allows comparisons between countries or periods. Indeed, relative survival has been long assimilated to net survival, but it has been demonstrated that, because of their inability account for

2 Jooste et al What s new? The relative survival rates reported for many types of cancer are subject to bias. Net survival, on the other hand, allows for comparisons between populations, because it isn t affected by mortalities due to other causes. In this study, the authors used the new, unbiased Pohar-Perme method to evaluate net survival in a large cohort of patients with solid tumors, and extended the standard follow-up period to 10 years. Their results confirm that prognosis for most types of cancer worsens with age, and that cancer management has improved survival rates over the last decade in France. informative censoring due to other-cause mortality, existing relative survival methods generated biased net survival estimates. Opportunely, among advances recently made in the field of net survival methodology, 2,3 a new interesting estimator of net survival was proposed by Perme et al. 2 : this Pohar-Perme estimator is unbiased and should be considered as the reference method. The French network of cancer registries (FRANCIM) and the Service de Biostatistique (Hospices Civils de Lyon, France) are currently maintaining a common database that counts more than 900,000 patients diagnosed between 1975 and For the first time in the French population, Pohar- Perme estimator is used on part of FRANCIM data to estimate 5- and 10-year net survivals. This study focuses on the changes of net survival over periods of diagnosis and according to age regarding 33 cancer sites. The dynamics of the excess mortality rate over the time elapsed since diagnosis is also provided for selected cancers sites. Material and Methods Data The study included all solid neoplasms (except non-melanoma skin cancers) diagnosed between January 1, 1989 and December 31, 2004 by the French registries (Table 1). The quality and the completeness of the participant registries are currently assessed every 4 years via an audit by the Comite National des Registres. Quality controls were performed at registry and whole database levels. These controls used tools provided by the International Agency for Research on Cancer. All cancers diagnosed in patients over 15 were included in the analysis. The 33 cancer sites considered were defined according to the codes of the International Classification of Diseases-Oncology 3 (ICD-O-3; Table 1). An active search for the vital status at January 1, 2008 was carried out for 370,773 cases diagnosed between January 1, 1989 and December 31, 2004 using a single standardized procedure. The information was collected at first line via birthplace public services or an electronic request to the Repertoire National d Identification des Personnes Physiques. Both procedures required the knowledge of the birthplace. Whenever the birthplace was missing, other sources of information on the vital status were used such as the medical records or the public services of the places of residence. Nevertheless, priority was given to the first line standardized strategy. The general principle was to minimize the number of lost to follow-up patients (alive at some date before January 1, 2008) without compromising the quality of the information or introducing bias. Finally, the proportions of lost to follow-up patients were 2.4% at 5 years and 2.9% at 10 years (Table 1). Estimation methods The net survivals were obtained using the new Pohar-Perme estimator of the net cumulative rate. 2 The expected mortality rates (necessary for this estimator) were available by age, sex, year of diagnosis and Departement of residence (French administrative area). The dynamics of the excess (net) mortality rates over time elapsed since diagnosis was obtained by smoothing and deriving the net cumulative rate estimate. Age-standardized net survival estimates were calculated using the international cancer survival standard weights. 4 Results Table 1 presents the 5- and 10-year age-standardized net survivals by cancer site. Concerning the three most frequent incident cancers in France, the age-standardized 10-year survivals were somewhat similar for breast and prostate cancers (74 and 71%, respectively) but lower for colon-rectum cancer (49% in males and 54% in females). Testicular cancer and lip cancer had the highest 5- and 10-year survivals. The 10-year net survival was over 80% for thyroid cancer and skin melanoma. It was under 25% for lung, esophagus, stomach, liver, biliary tract, pancreas, brain, mesothelioma, oropharynx and hypopharynx cancers. For most other sites, the 10-year net survival ranged between 40 and 60%. For all cancer sites that affect men and women, the 5- and 10-year net survivals were higher in women than in men, except bladder and biliary tract cancers. Tables 2 and 3 show the net survivals at 5 and 10 years after diagnosis by age groups for the most common cancer sites. For bladder, stomach, liver and ovarian cancers, 5- and 10-year net survivals decreased with increasing age. In men, these net survivals remained relatively constant in the age groups for esophagus cancer and skin melanoma and in the age groups for colorectal, corpus uteri and thyroid cancers. In men, the net survival for lung cancer was constant up to 55 years then decreased regularly in older age groups. For breast and prostate cancers, 5- and 10-year survivals were higher in the age groups than in the youngest age groups. Table 4 shows the trends in 5- and 10-year standardized net survivals by cancer sites and by 5-year periods from 1989

3 2372 Unbiased estimates of net survival of solid cancers Table 1. Five- and ten-year age-standardized net survivals (95% confidence intervals) in men and women for various cancer sites Lost to follow-up at 10 years % Dead within 10 years 5 Year std net survival 10 Year std net survival Cancer site ICD-O-3 code 1 Cases Men Women Men Women Breast C50 66, % 28.2% 84 (83 84) 74 (73 75) Colon-rectum C18 C21 64, % 61.1% 57 (56 57) 60 (59 60) 49 (48 50) 54 (53 55) Prostate C61 54, % 41.1% 84 (83 85) 71 (70 73) Lung C33 C34 39, % 89.7% 13 (12 13) 17 (16 18) 8 (8 9) 12 (11 13) Bladder C67 13, % 66.3% 56 (55 58) 50 (47 52) 47 (45 49) 44 (40 47) Stomach C16 13, % 82.9% 25 (23 26) 31 (30 33) 20 (19 22) 25 (24 28) Liver C22 9, % 92.6% 11 (10 12) 14 (12 16) 5 (5 6) 9 (7 11) Pancreas C25 9, % 94.9% 6 (5 7) 10 (9 11) 5 (4 5) 7 (5 8) Kidney C64 9, % 50.0% 64 (62 66) 65 (64 67) 53 (50 56) 58 (55 61) Esophagus C15 9, % 91.8% 11 (10 12) 16 (14 19) 7 (6 8) 10 (8 13) Skin melanoma C44 9, % 26.9% 83 (81 85) 88 (87 89) 77 (73 81) 84 (81 86) Corpus uteri C54 8, % 39.1% 72 (71 73) 67 (65 69) Ovary C569 C574 7, % 66.8% 37 (36 39) 30 (28 31) Thyroid gland C739 6, % 12.6% 84 (81 87) 91 (90 92) 75 (70 80) 88 (85 91) Larynx C32 5, % 62.4% 51 (50 53) 58 (52 64) 41 (38 44) 52 (42 64) Brain C70 C72 5, % 83.8% 20 (19 22) 24 (22 26) 13 (12 15) 18 (17 20) Cervix uteri C53 4, % 41.7% 65 (64 67) 58 (57 60) Oropharynx C09 C10 4, % 76.6% 30 (28 33) 44 (39 50) 17 (14 20) 33 (27 40) Oral cavity C03 C06 4, % 71.1% 36 (34 38) 51 (47 56) 22 (19 26) 39 (33 47) Hypopharynx C12 C13 4, % 83.1% 23 (22 25) 23 (18 30) 13 (11 15) 13 (9 20) Biliary tract C23 C24 3, % 87.4% 22 (20 25) 17 (15 20) 17 (14 21) 16 (13 19) Tongue C01 C02 3, % 74.9% 32 (29 35) 47 (42 52) 18 (14 23) 33 (27 40) Testis C62 2, % 8.7% 92 (89 95) 89 (84 95) Soft tissues 2, % 54.2% 59 (56 63) 59 (56 63) 54 (49 58) 52 (48 56) C381,C382,C383, C47,C480,C49, C696,C76,C809 Small intestine C17 1, % 63.5% 47 (43 51) 48 (44 53) 43 (36 50) 37 (32 42) Vagina and C51 C52 1, % 66.2% 53 (49 57) 42 (37 47) vulva Pleural C384 1, % 95.9% 4 (3 6) 10 (7 16) 1 (1 4) 6 (2 15) mesothelioma 2 Nasal cavities C30 C % 64.9% 47 (43 52) 42 (35 51) 38 (30 47) 31 (23 40) Lip 2 C % 39.8% 91 (86 96) 93 (82 100) 88 (78 100) 89 (58 100) Bone 2 C40 C % 52.7% 54 (49 60) 58 (52 64) 40 (34 47) 54 (47 61) Salivary glands C07 C % 52.6% 52 (47 59) 69 (63 76) 43 (35 52) 62 (53 71) Penis C % 54.8% 68 (62 75) 58 (49 68) Nasopharynx C % 58.3% 50 (44 56) 58 (49 68) 39 (32 47) 52 (42 63) 1 For skin melanoma, only morphology codes are included. For ovary, ( 84423, 84513, 84613, 84623, 84723, } are excluded. For pleural mesothelioma only are included. 2 Nonstandardized 5- and 10-year net survivals. For soft tissues, only 89103, 89123, 89203, 89913, 88103, 88113, 88213, 88233, 88203, 88223, 91503, 91603, 94913, 95803, 91403, 85873, 88063, 88363, 88703, 88803, 88813, 89213, 89823, 89903, 91353, 91363, 91413, 91423, 91613, 92313, 92513, 92523, 93733, 95813, 88303, 89633, 91803, 92103, 92203, 92403, 92603, 93643, or or or or or or or or or or or or or or or or

4 Jooste et al Table 2. Five-year net survivals for the most common cancer sites according to age-classes in men and women Men Women Cancer site þ þ Breast 86 (85 87) 90 (89 90) 88 (88 89) 86 (85 87) 76 (74 77) Colon-rectum 64 (61 67) 63 (61 65) 60 (59 61) 57 (56 58) 49 (48 51) 68 (65 71) 65 (63 67) 65 (64 67) 61 (59 62) 50 (49 51) Prostate 1 84 (81 86) 90 (89 91) 89 (88 90) 79 (78 80) 58 (55 62) Lung 17 (15 20) 16 (15 17) 16 (15 17) 14 (13 14) 8 (7 8) 30 (26 34) 23 (20 26) 21 (19 23) 17 (15 19) 10 (8 11) Bladder 80 (74 86) 69 (65 72) 64 (62 67) 55 (53 57) 41 (38 43) 55 (42 72) 60 (51 70) 57 (51 63) 52 (48 57) 36 (32 40) Stomach 35 (30 41) 31 (28 35) 28 (26 31) 24 (22 26) 17 (15 19) 38 (32 46) 35 (30 41) 37 (33 42) 34 (31 37) 21 (19 23) Liver 20 (15 28) 16 (14 20) 14 (13 16) 10 (8 11) 6 (5 8) 32 (23 44) 20 (14 30) 20 (15 26) 8 (6 11) 7 (5 10) Pancreas 15 (10 22) 7 (5 10) 6 (4 7) 7 (6 8) 3 (2 4) 7 (6 8) 35 (27 46) 15 (11 20) 11 (9 14) 7 (5 8) Kidney 75 (71 80) 71 (68 74) 68 (65 70) 65 (62 67) 54 (50 59) 81 (76 86) 78 (74 82) 73 (69 76) 66 (63 69) 50 (46 55) Esophagus 15 (11 20) 14 (12 16) 13 (12 14) 12 (11 14) 6 (5 8) 28 (16 49) 18 (13 26) 24 (19 30) 14 (11 19) 9 (6 12) Skin melanoma 86 (84 89) 85 (82 88) 87 (84 90) 82 (78 85) 70 (64 77) 94 (93 96) 90 (88 93) 90 (88 93) 85 (82 88) 75(70 81) Corpus 85 (80 90) 86 (84 89) 82 (80 84) 71 (69 73) 56 (53 59) Ovary 71 (68 75) 55 (52 58) 43 (40 45) 34 (32 37) 21 (19 24) Thyroid 99 (97 100) 94 (91 97) 86 (81 91) 75 (67 83) 55 (42 72) 99 (99 100) 99 (99 100) 97 (96 99) 88 (85 92) 58 (51 66) 1 Due to its repartition, corresponding age classes for prostate were 15 54, 55 64, 65 74, and 84þ. Table 3. Ten-year net survivals according to age classes in men and women for the most common cancer sites Men Women Cancer site þþ þþ Breast 75 (74 76) 83 (82 83) 80 (79 81) 76 (75 77) 65 (62 68) Colon rectum 58 (55 62) 55 (53 57) 52 (50 53) 49 (48 51) 43 (40 46) 61 (58 65) 58 (56 61) 59 (57 61) 54 (53 56) 46 (44 50) Prostate 1 72 (67 76) 83 (81 84) 79 (77 80) 61 (59 64) 32 (24 43) Lung 13 (11 16) 12 (11 13) 10 (9 10) 8 (7 9) 5 (4 6) 25 (21 29) 17 (14 20) 16 (13 18) 11 (9 13) 4 (3 7) Bladder 75 (69 83) 60 (56 64) 54 (52 57) 42 (40 45) 32 ((28 38) 47 (34 65) 56 (47 67) 48 (41 55) 43 (38 49) 35 (29 44) Stomach 29 (24 35) 27 (24 31) 24 (21 27) 20 (18 22) 13 (10 16) 32 (25 40) 33 (28 40) 32 (28 38) 28 (25 31) 13 (10 17) Liver 16 (10 24) 11 (8 14) 6 (5 8) 3 (2 5) 2 (1 5) 29 (20 41) 12 (6 24) 15 (10 21) 4 (2 7) 2 (1 8) Pancreas 15 (10 23) 5 (4 8) 4 (3 5) 4 (3 6) 3 (1 5) 24 (16 35) 10 (6 16) 7 (5 11) 5 (4 6) 2 (1 5) Kidney 69 (64 74) 64 (60 68) 57 (54 61) 52 (48 56) 43 (35 53) 78 (73 84) 72 (67 78) 66 (62 70) 57 (53 62) 41 (34 51) Esophagus 10 (7 14) 8 (6 10) 8 (7 10) 7 (6 8) 3 (2 7) 28 (16 49) 12 (7 21) 13 (9 20) 9 (5 14) 5 (2 9) Skin Melanoma 81 (78 84) 80 (76 84) 78 (73 83) 76 (70 83) 67 (47 96) 91 (89 93) 88 (85 91) 87 (83 90) 80 (75 84) 66 (51 84) Corpus 82 (77 88) 81 (78 84) 78 (76 80) 65 (63 68) 51 (45 58) Ovary 65 (61 69) 43 (40 46) 34 (32 37) 25 (23 28) 16 (12 21) Thyroid 99 (97 100) 90 (84 95) 77 (69 87) 59 (47 73) 26 (10 68) 99 (99 100) 98 (96 100) 96 (93 98) 85 (80 90) 43 (27 67) 1 Due to its repartition, corresponding age classes for prostate were 15 54, 55 64, 65 74, and 84þ.

5 2374 Unbiased estimates of net survival of solid cancers Table 4. Age-standardized net survivals by periods of diagnosis for the most common cancer sites 5-year 10-year (1989; 1991) (1992; 1994) (1995; 1997) (1998; 2000) (2001; 2004) (1989; 1991) (1992; 1994) (1995; 1997) (1998; 2000) (2001; 2004) Cancer site Breast 80 (79 81) 80 (79 82) 83 (82 84) 84 (83 86) 87 (86 88) 71 (6 74) 70 (68 73) 74 (71 76) ND ND Colon-rectum 54 (53 56) 57 (56 58) 57 (56 58) 59 (58 60) 59 (58 60) 48 (46 49) 50 (48 52) 51 (49 52) ND ND Prostate 71 (68 74) 76 (73 78) 78 (76 80) 83 (82 85) 90 (89 91) 55 (51 59) 63 (60 66) 67 (64 70) ND ND Lung 13 (12 14) 14 (13 15) 12 (11 13) 14 (13 15) 15 (14 16) 8 (7 9) 9 (8 10) 8 (7 9) ND ND Bladder 59 (56 62) 55 (52 58) 58 (56 61) 55 (52 58) 54 {51 57) 50 (47 54) 46 (42 50) 50 (46 53) ND ND Stomach 28 (25 30) 27 (25 29) 25 (23 25) 26 (24 29) 27 (26 28) 22 (19 24) 21 (19 24) 20 (17 22) ND ND Liver 7 (6 10) 10 (8 12) 9 (8 12) 12 (10 14) 13 (12 15) 4 (3 6) 6 (4 8) 5 (3 7) ND ND Pancreas 9 (7 11) 5 (4 7) 7 (6 9) 7 (5 8) 8 (7 10) 7 (5 9) 4 (3 5) 5 (4 7) ND ND Kidney 59 (55 63) 62 (58 66) 62 (59 66) 64 (61 67) 69 (67 72) 48 (43 54) 54 (49 59) 51 (46 55) ND ND Esophagus 9 (8 11) 11 (9 13) 11 (9 13) 11 (9 13) 13 (11 15) 5 {4 7) 7 (5 9) 6 (5 8) ND ND Skin melanoma 84 (81 87) 83 (80 86) 85 (82 88) 86 (83 88) 88 (86 90) 79 (73 85) 76 (72 80) 80 (75 84) ND ND Corpus uteri 69 (65 72) 69 (66 73) 72 (69 75) 72 (69 76) 72 (70 75) 63 (58 67) 65 (60 69) 69 (65 74) ND ND Ovary 36 (33 40) 35 (32 39) 37 (34 40) 35 (33 39) 43 (40 46) 29 (39 34) 29 (25 33) 29 (26 32) ND ND Thyroid 82 (78 86) 86 (82 89) 91 (87 94) 89 (86 92) 91 (88 93) 75 (71 80) 81 {77 85) 86 (80 93) ND Brain 24 (21 28) 25 (21 28) 19 (17 23) 20 (17 23) 22 (20 25) 18 (16 22) 16 (13 20) 13 (11 16) ND Cervix uteri 68 (65 72) 65 (61 69) 66 (62 70) 65 (61 70) 63 (59 67) 63 (59 67) 56 {52 60) 61 (57 65) ND ND to The greatest increase was seen in prostate cancer; the 5-year net survival increased from 71% (in ) to 90% (in ) and the 10-year net survival from 55% (in ) to 67% (in ). For breast cancer, the 5- and 10-year net survivals increased smoothly from 80% (in ) to 87% (in ) and from 71% (in ) to 74% (in ), respectively. A similar trend was seen for thyroid and kidney cancers. There was only a minor increase in survival for colon-rectum cancer, ovarian cancer and melanoma. The 5- and 10-year net survivals decreased over time for cervix uteri cancer (from 68% at 5 years in to 63% in ) and for bladder and brain cancers. They remained constant for the other cancer sites. Figure 1 presents the dynamics of the excess mortality rate over time elapsed since diagnosis by age groups for four major cancer sites. Very different excess rate levels were reached depending on the cancer site (see different scales in Fig. 1). These dynamics varied with age for all cancer sites: the initial excess rates were consistently higher in elderly than in young patients. For lung cancer, a high excess risk of death remained during the second year after diagnosis whatever the age. The excess rate decreased over time elapse since diagnosis for colorectal cancer. Such a decreasing trend was not observed in breast and prostate cancers, which exhibited a maximal excess rate at 2 years and then a low but non decreasing trend. Discussion Previous French studies on cancer survival provided only 5- year survival estimates. 1,5 This study provides the first estimations of 10-year survival for solid cancers in France. Besides, it is the first to use an unbiased estimator of net survival. Indeed, it has been shown, theoretically, 2 then on simulated data, 3 that Pohar-Perme estimator is the only one to estimate really net survival (without bias); it provides thus information that does not depend on the general population mortality. Nonhomogeneous potential follow-ups (e.g., dependent on age) may lead to bias in any estimator of the net survival, including the Pohar-Perme estimator. 2 However, Danieli et al. have shown that, even when the potential follow-up are highly dependent on age, the Pohar-Perme estimator still shows a high performance at 10 years with low bias. 3 This estimator is therefore suitable for comparison of country-specific cancer survivals or for the analysis of survival trends, which is not the case of other traditional methods 2,3 usually called relative survival methods. The benefit from using the Pohar Perme estimator (compared to the traditional relative survival methods) is particularly important when reporting nonage standardized survival and is less when reporting age standardized survival: a full discussion of this point is available in the article of Roche et al. 6 International publications of estimates that use this new method are still scarce; this makes it difficult to carry out comparisons of net survivals between countries. However, as

6 Jooste et al Figure 1. Dynamics of the excess mortality rates for colon-rectum, lung, breast and prostate cancer. pointed by Roche et al., 6 in most cancers, 5-year age standardized net estimates are reasonably comparable to relative survival estimated but 10-year estimates may not. This study was carried out on non-selected French population-based cancer registries covering 16 administrative areas. The vital status was updated using a standardized procedure combining administrative data and medical files collected by registry researchers. The quality of the information relative to patients born in France was better than that of patients born abroad. This could hypothetically bias the estimations. However, the latter patients accounted for only 9.5% of the censored patients and the overall percentages of lost to follow-up patients were only 2.4% at 5 years and 2.9% at 10 years. This enhances somewhat the overall reliability of the estimations. The high disparities in net survival between cancer sites observed in France persisted throughout the study period. The 10-year age-standardized net survival varied between 5% (liver) and 89% (testis) in men and between 7% (pancreas) and 88% (thyroid) in women. It was 70% for prostate cancer, 76% for breast cancer, 48% for colorectal cancer in men and 52% for colorectal cancer in women. The latter three cancers represented 41% of all incident cases. Except for bladder cancers and some ORL cancers whose prognoses were worse in women than in men, the 5-year net survivals were generally higher in women than in men and, at 10 years, this difference tended to increase. Survival decreased in the older age classes for all cancer sites, except breast and prostate cancers, which exhibited higher survivals in middle-aged groups than in young patients. Otherwise, survival worsened considerably after 75 years, especially for gynaecologic, prostate, kidney and thyroid cancers; survival decreased by at least 10% between cancers diagnosed at ages and over 75 years. Similarly, to 5-year net survivals, the 10-year net survivals decreased with increasing age classes. Generally, the excess mortality rates change with age. The initial rates are always higher in old than in young patients. In young patients, the risk of death from cancer is often the highest during the second year after diagnosis. This reflects not only the ability of the treatments to delay death but also the fact that advanced tumors are more frequently diagnosed in old than in young people. The latter fact does not hold for colon-rectum cancers; thus, the high early mortality rates may be explained by operative mortality because surgical procedures are currently opted for in almost 90% of colon-rectum cancers. For some cancer sites, the excess rates decreased to low values by 5 years and even before whereas, for others, the excess rates decreased regularly over 10 years then flattened at non-null values. Ongoing analyses taking into account the histological type, the stage at diagnosis and the therapeutic management should be able to explore such trends more clearly.

7 2376 Unbiased estimates of net survival of solid cancers In France, the group of good prognosis cancers (lip, testis and thyroid cancers) had nearly a 90% 10-year net survival but accounted for only 7% of all cancers. In contrast, the group of bad prognosis cancers (lung, esophagus, stomach, liver, pancreas, biliary tract and brain) accounted for 19% of all cancers. Among the most common cancers, breast and prostate cancers had 74% and 71% 10-year net survival, respectively, whereas the intermediate-prognosis colorectal cancer had an intermediate 50% 10-year survival. For most cancer sites, women had higher survivals than men. This was already reported in previous studies. It is usually hypothesized that biological phenomena could explain such disparities, that women take better care of their health and that their cancers are diagnosed earlier. Differences in comorbidities and risk factors may partly explain the differences in net survival. Another partial explanation is that specific cancer risk factors, such as tobacco and/or alcohol consumption, are also risk factors for other severe diseases; this changes the expected death rate, then net survival. This is the case of larynx and hypopharynx cancers. 7 For most cancer sites, the highest 10-year net survivals were found in the youngest age group (15 44 years). However, this was not the case for breast and prostate cancers where the highest survivals were seen in age group years. For breast cancer, this fact is already well-known North Eastern European registries have reported similar results but attributed them to the screening programmes established for women over 50 years. Indeed, breast cancers diagnosed in the eighties (i.e., before the start of the programmes) had better prognoses in young than in elderly women. 8 A study on SEER data has shown that, all breast cancer stages combined, the survivals increased over age groups up to then decreased. These survival differences were not due to differences in stage distribution because they persisted even when the analysis was stratified by cancer stages. 9 In France, differences in net survival according to age were observed before the start of the screening programme in women aged years. 1 The survival trends in prostate cancer should be interpreted with caution because of the increasing use of biopsies and PSA detection. The steady increase may be partly attributed to the effects of the lead-time bias and the increased detection of less aggressive tumors. Early diagnosis and curative treatment may also have had some impact in improving the survival of patients with aggressive tumors. Nordic countries registries have reported bad prognoses in the youngest patient categories. 10 Prostate cancer is rare in people <50 but its forms are frequently more aggressive. The 10-year net survival was found decreasing more or less steadily with advancing age. This decrease was also reported over successive 10-year age groups for lung, bladder, stomach, liver, ovary and kidney cancers. 11 Ten-year net survivals remained constant up to 65-yearsold in colorectal and corpus uteri cancers in men and women and up to 75 years in esophagus cancer and skin melanoma in men only. In all sites, survival decreased sharply in patients 75 and over. The relationship between age and survival is more marked for many cancer sites in the USA (SEER program 12 ). Elderly patients may be less rapidly referred for diagnosis and, when referred, receive less aggressive medical or interventional management. However, these managements vary with the cancer site. For instance, it has been shown in elderly patients with stage III colon cancer that adjuvant chemotherapy was much more rapidly implemented in the USA than in Europe. 13 Achieving better outcomes in elderly cancer patients is also a major challenge for Europe. Ten-year net survival remained constant or increased slightly over the period suggesting that little or no treatment advances have been achieved over this period. Mass screening for breast or colorectal cancer were not yet implemented. Survival after cervix cancer decreased slightly. A study by Eurocare has shown a slight decrease of survival between the late eighties and the late 90s. 14 This decrease was not found in all European countries; it was obvious in France ( 2.1%), England and Wales ( 2.7%) and Austria ( 3.1%) but not in Northern Europe. 15 This trend may be the consequence of screening; in France, over the last 20 years, screening and early treatment of preneoplastic or in situ tumors have decreased the incidence of invasive cancers. This generated or selected advanced-stage or bad-prognosis cancers. The role of the histological subtype in survival differences is still unclear. A Finnish study has shown that the incidence of squamous cell carcinoma of the cervix declined after introduction of the screening programme while the incidence of adenocarcinoma did not change. 16 Some studies have reported a higher survival for squamous cell carcinoma while others have reported no difference in prognosis. Conclusion This study provided the first long-term net survival estimations for solid tumors in France. Our results confirmed the persistent unfavorable role of age in prognosis and general improvements of the net survivals over time. Because unaffected by the mortality due to other causes, these estimations will allow reliable comparisons between periods of diagnosis within a given country or between countries as soon as registries shall adopt the Pohar-Perme estimator. Acknowledgements The research was carried out within the context of a four-institute research-program partnership that involved the Institut National du Cancer (INCa), the Institut de Veille Sanitaire (InVS), FRANCIM and Hospices Civils de Lyon. The authors thank the French Assurance Maladie and all participant cancer centers, hospitals, private clinics and anatomical pathologists for the identification of the cases and the data collection.

8 Jooste et al References 1. Bossard N, Velten M, Remontet L, et al. Survival of cancer patients in France: a population-based study from The Association of the French Cancer Registries (FRANCIM). Eur J Cancer 2007;43: Perme MP, Stare J, Esteve J. On estimation in relative survival. Biometrics 2012;68: Danieli C, Remontet L, Bossard N, et al. Estimating net survival: the importance of allowing for informative censoring. Stat Med 2012;31: Corazziari I, Quinn M, Capocaccia R. Standard cancer patient population for age standardising survival ratios. Eur J Cancer 2004;40: Bouvier AM, Remontet L, Hedelin G, et al. Conditional relative survival of cancer patients and conditional probability of death: a French National Database analysis. Cancer 2009;115: Roche L, Danieli C, Belot A, et al. Cancer net survival on registry data: use of the new unbiased Pohar-Perme estimator and magnitude of the bias with the classical methods. Int J Cancer, submitted. 7. Berrino F, Richiardi L, Boffetta P, et al. Occupation and larynx and hypopharynx cancer: a job-exposure matrix approach in an international case-control study in France, Italy, Spain and Switzerland. Cancer Causes Control 2003;14: Tryggvadottir L, Gislum M, Bray F, et al. Trends in the survival of patients diagnosed with breast cancer in the Nordic countries followed up to the end of Acta Oncol 2010; 49: Ries L, Young J, Keel G, Eisner M, Lin Y, M-J H, editors. SEER Survival Monograph: Cancer Survival Among Adults: U.S. SEER Program, , Patient and Tumor Characteristics. Bethesda, MD: National Cancer Institute, SEER Program, NIH Pub. No ; Bray F, Klint A, Gislum M, et al. Trends in survival of patients diagnosed with male genital cancers in the Nordic countries followed up until the end of Acta Oncol 2010;49: Colonna M, Bossard N, Remontet L, et al. Changes in the risk of death from cancer up to five years after diagnosis in elderly patients: a study of five common cancers. Int J Cancer 2010; 127: Surveillance,, and End Results (SEER) Program Populations ( ) ( National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, released January Jessup JM, Stewart A, Greene FL, et al. Adjuvant chemotherapy for stage III colon cancer: implications of race/ethnicity, age, and differentiation. JAMA 2005;294: Bielska-Lasota M, Inghelmann R, van de Poll- Franse L, et al. Trends in cervical cancer survival in Europe, : a population-based study. Gynecol Oncol 2007;105: Klint A, Tryggvadottir L, Bray F, et al. Trends in the survival of patients diagnosed with cancer in female genital organs in the Nordic countries followed up to the end of Acta Oncol 2010;49: Pukkala E, Malila N, Hakama M. Socioeconomic differences in incidence of cervical cancer in Finland by cell type. Acta Oncol 2010;49:180 4.