Long-term population-based survival data detailed by cancer subtype are important to measure the overall outcomes of malignancy managements. We provide net survival estimates at 1, 3, 5 and 10-year postdiagnosis on 37,549 hematological malignancy (HM) patients whose ages were >15 years, diagnosed between 1989 and 2004 and actively followed until 2008 by French population-based cancer registries. These are, to our knowledge, the first unbiased estimates of 10-year net survival in HMs detailed by subtypes. HMs were classified according to the International Classification of Diseases-Oncology 3. Net survival was estimated with the unbiased Pohar-Perme method. The results are reported by sex and age classes. The changes of these indicators by periods of diagnosis were tabulated and the trends of the net mortality rates over time since diagnosis graphed. In all, 5- and 10-year age-standardized net survivals after HMs varied widely from 81 and 76% for classical Hodgkin lymphoma (CHL) to 18 and 14% for acute myeloid leukemia (AML). Even in HMs with the most favorable prognoses, the net survival decreased between 5- and 10-year postdiagnosis. Women had better prognoses than men and age at diagnosis was an unfavorable prognostic factor for most HMs. In patients <55 years old, the net mortality rate decreased to null values 5-year postdiagnosis in AML and 10-year postdiagnosis in CHL, precursor non-HL, chronic myelogenous leukemia, diffuse large B-cell lymphoma and follicular lymphoma. The prognoses improved for various HMs over the study period. The obtained unbiased indicators are important to evaluate national cancer plans.
Hematological malignancies (HMs) form the fourth most frequently diagnosed cancers in both males and females in the economically developed regions of the world.1–3 Numerous disease subtypes have been identified and these differ widely in clinical presentation, treatment requirements and prognosis.4 Population-based survival data detailed by subtypes are important to measure the overall outcomes of HM managements.5 Several studies in Europe and the USA have reported estimates of 5-year relative survivals for the main subtypes of HMs but few of them provided longer-term survival estimates although the latter are more relevant in some good-prognosis HMs such as Hodgkin lymphoma (HL).2, 6–11
In France, the first population-based relative survivals of HM patients were published in 2007.12 Since then, significant advances have been made in net survival methodology.13, 14 In cancer studies, net survival is the survival that would be observed if cancer was the only possible cause of death. It constitutes a very interesting indicator within the context of population studies. Although relative survival has been assimilated to net survival, it was demonstrated that relative survival provides biased net survival estimates because it is unable to take into account informative censoring related to other-cause mortality.13, 14 Net survival is not affected by changes in other-cause mortality. This property allows net survival comparisons between countries or time periods. Recently, Perme et al.13 proposed a new unbiased estimator that should be considered as the reference method. Using this new estimator, we estimated 5- and 10-year net survivals of HM patients by sex, age classes and ten major HM subtypes. Age-standardized net survival (ASNS) estimates are also given by sex and different periods of diagnosis. Changes of the net mortality rate over time since diagnosis are also given by age classes.
AML: acute myeloid leukemia; ASNS: age-standardized net survival; CHL: classical Hodgkin lymphoma; CLL/SLL: chronic lymphocytic leukemia/small lymphocytic lymphoma; CML: chronic myelogenous leukemia; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; HM: hematological malignancy; ICD-O-3: International Classification of Diseases-Oncology 3; LPL/Waldenström: lymphoplasmacytic lymphoma/Waldenström macroglobulinemia; MDS: myelodysplastic syndrome; NHL: non-Hodgkin lymphoma; PCN: plasma cell neoplasm; WHO: World Health Organization
Material and Methods
The study included all HMs in patients whose ages were >15 years, diagnosed between 01/01/1989 and 31/12/2004 and registered by FRANCIM, the French network of cancer registries; this represented 37,549 cases (Table 1). The quality and completeness of the registries were audited every 4 years by the Comité National des Registres. In addition, more than 3 years, a specific program funded by the Institut National du Cancer was dedicated to coding and registering HMs in the registries who contributed data for this analysis.
Table 1. Distribution, deaths, 5- and 10-year age-standardized net survival (ASNSs) according to the ICD-O-3 (37,549 hematological malignancies (HMs), diagnosed [1989–2004] in patients aged 15–99 years, collected by 16 French cancer registries, and followed up to January 1, 2008)
This effort allowed us to define five HM groups and ten major subtypes, instead of seven groups previously.12 These groups and subtypes were classified according to the International Classification of Diseases-Oncology 3 (ICD-O-3)15 using the groupings published by Morton et al.16 in 2007 for the Lymphoid neoplasms (Table 1). Nodular lymphocyte predominant HL, mast-cell tumors, histiocytosis and chronic myelomonocytic leukemia were excluded because rare (n = 737; i.e., 2%) and difficult to cluster with other groups.
An active search for the vital status at 01/01/2008 of all cases was carried out; first via the birthplace public services or via an electronic request to the Répertoire National d'Identification des Personnes Physiques. When the birthplace was unknown, other sources of information were used (medical records or public services of the places of residence). The principle was to minimize the number of lost to follow-up patients without compromising the quality of the information or introducing biases. A patient was considered lost to follow-up when he was followed up for <10 years and when the date of last contact was before 01/01/2008. Finally, the average loss to follow-up rate was 2.8% (range, 0.6–4.6%) (Table 1).
The net survivals were obtained using the new Pohar-Perme estimator of the net cumulative rates.13 The expected mortality rates (necessary for that estimator) were available by sex, year of diagnosis, age and Département of residence (Roche et al., 2012). The dynamics of the net mortality rates over time elapsed since diagnosis was obtained by smoothing and deriving the net cumulative rate estimate. The ASNS estimates were calculated using the International Cancer Survival Standard weights.17
Table 1 lists the number of cases of each HM according to the ICD-O-3 codes, the follow-up characteristics and the 5- and 10-year ASNS by sex. The median follow-up in alive-patients was 7.1 years. Within the first 10 years, 22,206 deaths were registered. Nonspecific codes for non-Hodgkin lymphoma (NHL) not otherwise specified (NOS) category (i.e., 9590/3, 9591/3, 9675/3) represented 9.9% (2,762/27,931) of all lymphomas (data not shown).
Classical Hodgkin lymphoma
The 5- and 10-year ASNSs for classical Hodgkin lymphoma (CHLs) were 81% (95% confidence interval [CI]: 80–83) and 76% (95% CI: 73–78), respectively. Ten-year ASNS was higher in women than in men (80 vs. 73%) (Table 1). Age at diagnosis was a prognostic factor in both sexes, especially after 65 years old (Tables 2 and 3). The 10-year net survival was 94% in women aged 15–45 years and 17% in women aged 75 years or more. The 5-year net survival remained constant (81%) during the last four diagnosis periods (1992–2004), whereas it was higher (87%) during the first diagnosis period (1989–1991) (Table 4). The net mortality rate was low 2 years postdiagnosis but did not reach zero 10 years after diagnosis, especially in patients >55 years old at diagnosis (Fig. 1).
Table 2. Five-year net survivals (95% CIs of HMs according to sex and age categories (cases diagnosed [1989–2004] in patients aged 15–99 years, followed up to January 1, 2008 in the 16 French cancer registries)
Table 3. Ten-year net survivals (95% CIs) of HMs according to sex and age categories (cases diagnosed [1989–2004] in patients aged 15–99 years, followed up to January 1, 2008 in the 16 French cancer registries)
Table 4. Age-standardized 5- and 10-year net survival according to the period of diagnosis (HMs diagnosed in patients aged 15–99 years, followed up to January 1, 2008 in the 16 French cancer registries)
Here, according to the World Health Organization (WHO) and the InterLymph nested classification,4, 16 “all NHLs” considered all malignant neoplasms of lymphoid origin, except HL. The 5- and 10-year ASNSs in all NHLs were 60% (95% CI: 59–60) and 44% (95% CI: 43–45), respectively. The 10-year prognosis was better in women than in men (46 vs. 42%) (Table 1). In both sexes, the net survival decreased with increasing age at diagnosis (Tables 2 and 3). The 10-year ASNS was 39% (95% CI: 32–46) in the lymphoma NOS category (data not shown). Overall, a slight improvement of the 5-year ASNS (6 points) was observed over the study period (1989–2004) (Table 4). The results relative to the main NHL subtypes are discussed in the following sections.
The 5- and 10-year ASNSs in chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL) were 78% (95% CI: 77–80) and 57% (95% CI: 54–60), respectively. The differences between 5- and 10-year net survivals were among the highest of all NHLs (minus 23% in men and minus 19% in women). The 10-year ASNSs were higher in women than in men (62 vs. 53%) (Table 1). An impact of age on prognosis was seen after 55 years old and, 10 years after the diagnosis, it was more marked in men >75 years old (Tables 2 and 3). The 5-year ASNS remained rather constant over the study period, around 76%, whereas the 10-year net survival was more variable (Table 4). The net mortality rate was relatively low shortly after diagnosis but increased steadily over time, in all age groups, mostly after 4 years of follow-up (Fig. 1).
Follicular lymphoma (FL) had one of the best prognoses of all NHL subtypes; the 5- and 10-year ASNSs were 70% (95% CI: 67–73) and 57% (95% CI: 52–61), respectively. ASNSs were similar in both sexes (Table 1). Age at diagnosis had a prognostic impact on 10-year survival, especially in patients >65 years old (Tables 2 and 3). The 10-year prognosis of FL improved from 48 to 56% between 1989–1991 and 1995–1997 (Table 4). The net mortality rate was low shortly after diagnosis and decreased regularly between 5 and 10 years after diagnosis in adults >55 years old (Supplementary data).
Diffuse large B-cell lymphoma
The 5- and 10-year ASNSs were 47% (95% CI: 45–49) and 39% (95% CI: 36-41), respectively and similar in both sexes (Table 1). Age at diagnosis was a major prognostic factor; 43- and 40-point decreases were found in 5- and 10-year net survivals, respectively, between young men (15–45 years old) and men >75 years old (Tables 2 and 3). These figures were 46 and 51 points in women, respectively. Regarding the trends of survival over the periods of diagnosis, there was a 10-point improvement of the 5-year ASNS in the most recent period (2001–2004) (Table 4). The net mortality rate decreased markedly 2 years after diagnosis but did not reach zero up to 10 years after diagnosis, except in patients <65 years old, showing no excess risk 2 years after diagnosis (Fig. 1).
The 5- and 10-year ASNSs were 73% (95% CI: 70–76) and 51% (95% CI: 47–55), respectively. The difference between the 5- and 10-year ASNSs was −21% in men and −24% in women. Ten years after diagnosis, men and women prognoses were comparable (Table 1). The impact of age on the 10-year prognosis was very important after 75 years old (Table 3). The trends of survival over the periods of diagnosis showed constant rates 5 years after diagnosis, whereas the 10-year prognosis was more variable (Table 4). Although low, early after diagnosis, the net mortality rate increased regularly with the time elapsed since diagnosis in all age groups, especially after 6 years of follow-up (Supplementary data).
Plasma cell neoplasms
Plasma cell neoplasm (PCN), including Multiple myeloma, had the most unfavorable long-term prognoses of all mature lymphoid malignancies; the 5- and 10-year ASNSs were 45% (95% CI: 43–46) and 24% (95% CI: 22–26), respectively. The prognoses were comparable in both sexes (Table 1) and remained constant between 1989–1991 and 2001–2004 (Table 4). The prognostic impact of age was clear in patients >65 years old (Tables 2 and 3). After a decrease during the first year after diagnosis, the net mortality rate remained low but non-null through the 10 years of follow-up and high in patients >65 years old even 10 years after diagnosis (Fig. 1).
The 5- and 10-year ASNS estimates were close; 35% (95% CI: 31–39) and 31% (95% CI: 27–35), respectively, and similar in both sexes (Table 1). The net survival decreased clearly with increasing ages at diagnosis (Tables 2 and 3). The prognoses at 5 years improved significantly and constantly from 22 to 46% between 1989 and 2004 (Table 4). High net mortality rates were observed shortly after diagnosis, especially in the elderly. These rates decreased rapidly over the first 2 years then slowlier to reach a non-null value 10 years after diagnosis, except in young adults (Supplementary data).
Acute myelogenous leukemia
Acute myeloid leukemia (AML) had the severest prognosis of all HMs with 5- and 10-year ASNSs of 18% (95% CI: 16–19) and 14% (95% CI: 13–16), respectively. The 5- and 10-year net survivals were comparable by sex except in the youngest (Table 1). The great impact of age at diagnosis on prognosis was clear since the first year after the diagnosis and remained high after 5 and 10 years of follow-up (Tables 2 and 3). Besides, the 5-year net survival increased by 8 points between 1989–1991 and 2001–2004 (Table 4). The net mortality rate was high shortly after diagnosis and a slow decrease after 5 years toward null values in patients <65 years old (Fig. 1).
Chronic myelogenous leukemia
The 5- and 10-year ASNSs were 53% (95% CI: 49–56) and 31% (95% CI: 26–36), respectively. Ten years after diagnosis, there was no more difference between men and women (Table 1). The impact of age at diagnosis on the 10-year net survival was important in patients >65 years old (Table 3). Besides, the 5-year ASNS increased markedly from 47 to 69% during the most recent period of diagnosis (i.e., 2001–2004), which corresponds to the advent of tyrosine kinase inhibitor therapy (Table 4). The net mortality rate showed a two-fold decrease during the first year after diagnosis before increasing by 50% during the second year (Fig. 1). From the fifth to the tenth year after diagnosis, the net mortality rate remained high in the elderly.
With a 10-year ASNS of 26% (95% CI: 22–31), the 10-year prognosis of the myelodysplastic syndrome (MDS) was as severe as that of PCN. The net survival was higher in women than in men (32 vs. 21%) (Table 1). Prognosis was affected by age at diagnosis (Tables 2 and 3). There were constant decreases of 5- and 10-year nonstandardized net survivals over the study period (Table 4). The net mortality rate decreased regularly over the first 3 years and stabilized at a relatively high value during the last years of follow-up (Supplementary data).
To our knowledge, these are the first estimations of long-term net survival using an unbiased estimator of net survival on population-based registry data with HM typing according to the WHO classification and to the grouping proposals of the InterLymph consortium. Indeed, it was shown, first theoretically,13 then on simulated data,14 that the Pohar-Perme estimator is the only estimator able to provide, simply, unbiased estimates of net survival 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.13 However, Danieli et al.14 have shown that, even when the potential follow-ups are highly dependent on age, the Pohar-Perme estimator still shows a high performance at 10 years with low bias. This estimator is thus suitable for international comparisons of cancer survivals or for the analyses of survival trends in a given country, which is not the case of the traditional “relative survival” methods (Roche et al., 2012).13, 14 Nevertheless, it should be noted that 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. (2012) The net survival estimates are particularly interesting within the context of HMs (wide ranges of diseases, prognoses and ages at diagnosis).
Our results showed that, even in HMs with the most favorable prognoses (CHL, CLL/SLL, FL or CMD), the net survival decreased between 5 and 10 years postdiagnosis by 5 to 25% points depending on the subtype. This effect was mainly observed in the elderly, there was no plateau in net survival (i.e., no cure) except in the youngest patients. Our observations highlighted also the important effect of age at diagnosis (in all HMs) on the net survival and on the pattern of net mortality rates. In the oldest patients (>75 years), short after diagnosis, the net mortality rate was three to ten times higher than in the youngest ones (15–45 years) and never reached zero over the first 10 years of follow-up, except for precursor NHL that represented a small proportion of cases. In patients 15–45 years old, we reported null net mortality rates during the first 10 years of follow-up not only in good-prognosis HMs (CHL, FL or CML), but also in poor-prognosis ones (AML, DLBCL or precursor NHL). Finally, we observed a continuous increase of the 5-year net survival during all the study period for precursor NHL, whereas improvements of prognosis were found only in the more recent period (i.e., 2001–2004) in CML, FL and DLBCL.
In the present report, we have added several histological HMs subtypes to our former one12: NHLs (FL, DLBCL, LPL/Waldentröm and Mature T-cell NHL) and MDS. All the subtypes had the highest reliability of diagnosis over the time period.18 Using the Interlymph consortium,16 grouping of lymphoma cases allowed a standardization essential for comparing subtype-specific reports. Although the reliability of the diagnosis may have been optimal for new subtypes of NHL reported in this analysis (because they corresponded to well-defined entities), it seems that this may not have been the case for MDS whose diagnosis and case definition are based on both a differential diagnosis with other syndromes such as B12 or folate deficiencies and the ability of cytologists to conclude firmly on the basis of bone marrow smears. However, a centralized pathological revision would have been too expensive. Besides, the registration procedures might have minimized the impact of such a lack of reliability in the most recent period of diagnosis. Hence, the improvement of case selection may have affected the net survival estimates in MDS and the trends we found may have resulted from a better case classification rather from real trends in survival. Finally, we reported <10% of NOS cases of lymphoma, which is a good indicator of data quality versus other publications on the same period.
We reported 10-year net survivals ranking from 76% for CHL to 14% for AML, and, except for FL, these were always higher in women than in men. In chronic HMs such as CLL/SLL, PCN, LPL/Waldenström, FL, CML or MDS, the net survival decreased importantly between 5 and 10 years of follow-up (>20%). This observation may be the consequence of disease progressions or relapses during the follow-up after 5 years. Besides, late toxicity might likely explain the same observation in acute HMs (such as AML or precursor NHL) or in CHL although the rates of net survival decreased by <6%. Recently, Verdecchia et al.19 mentioned this excess of mortality after 5 years in CML and NHL but not in CHL. The non-null probability of dying from CHL even 10 years after diagnosis should be explored with the same methods by stages and histological subtypes.
Interestingly, our results also showed improvements in net survival in various subtypes that could be interpreted as changes in treatment patterns. The recent improvement of CML prognosis is the most striking example and may be attributed to the introduction of tyrosine kinase inhibitors by the end of 2002, but the long-term outcome of this treatment cannot be assessed yet.20–22 The recent better net survivals reported in FL and DLBCL may also be related to the improvement of treatment efficacy (without added toxicity) since the advent of anti-CD20 treatment.23–25 Improvements have also been reported in the youngest age groups in severe-prognosis HMs such as precursor NHL. This may reflect a progress in initial treatment and better side-effect management in childhood lymphoblastic leukemia or AML.26 However, contrarily to the recent results from Europe and the USA,2, 9, 19 we found no improvement of CHL prognosis within the study period. Yet, our results are consistent with those published in 2007 on cases diagnosed between 1987 and 1997 and, despite distinct methods, our estimations of the 5-year net survival rates are close to those of others.2, 8, 11, 19 One explanation could be that the long-term risk-to- benefit ratios obtained with old standard treatments used in most HL cases were not challenged by new treatments even if they are achieving better short-term results.27–29
In conclusion, our study provides unbiased estimates of long-term net survival in all HMs and ten major subtypes, allowing comparisons between countries. These indicators are important to evaluate national cancer plans and the efforts to control the burden of the disease.
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 also thank all the sources of information: Hospitals, Cancer Centers, Private clinics, General Practitioners, Laboratories (Pathology, Haematology), the French Assurance Maladie and the Centre d'épidémiologie sur les causes médicales de décès (CépiDc).