Multidimensional penalized splines for survival models: illustration for net survival trend analyses.

BACKGROUND: In descriptive epidemiology, there are strong similarities between incidence and survival analyses. Because of the success of multidimensional penalized splines (MPSs) in incidence analysis, we propose in this pedagogical paper to show that MPSs are also very suitable for survival or net survival studies. METHODS: The use of MPSs is illustrated in cancer epidemiology in the context of survival trends studies that require specific statistical modelling. We focus on two examples (cervical and colon cancers) using survival data from the French cancer registries (cases 1990-2015). The dynamic of the excess mortality hazard according to time since diagnosis was modelled using an MPS of time since diagnosis, age at diagnosis and year of diagnosis. Multidimensional splines bring the flexibility necessary to capture any trend patterns while penalization ensures selecting only the complexities necessary to describe the data. RESULTS: For cervical cancer, the dynamic of the excess mortality hazard changed with the year of diagnosis in opposite ways according to age: this led to a net survival that improved in young women and worsened in older women. For colon cancer, regardless of age, excess mortality decreases with the year of diagnosis but this only concerns mortality at the start of follow-up. CONCLUSIONS: MPSs make it possible to describe the dynamic of the mortality hazard and how this dynamic changes with the year of diagnosis, or more generally with any covariates of interest: this gives essential epidemiological insights for interpreting results. We use the R package survPen to do this type of analysis.

How to produce sound predictions of incidence at a district level using either health care or mortality data in the absence of a national registry: the example of cancer in France.

BACKGROUND: In many countries, epidemiological surveillance of chronic diseases is monitored by local registries (LR) which do not necessarily cover the whole national territory. This gap has fostered interest in using non-registry databases (e.g., health care or mortality databases) available for the whole territory as proxies for incidence at the local level. However, direct counts from these databases do not provide reliable incidence measures. Accordingly, specific methods are needed to correct proxies and assess their epidemiological usefulness. METHODS: This study's objective was to implement a three-stage turnkey methodology using national non-registry data to predict incidence in geographical areas without an LR as follows: constructing a calibration model to make predictions including accurate prediction intervals; accuracy assessment of predictions and rationale for the criteria to assess which predictions were epidemiologically useful; mapping after spatial smoothing of the latter predictions. The methodology was applied to a real-world setting, whereby we aimed to predict cancer incidence, by gender, at the district level in France over the 2007-15 period for 24 different cancer sites, using several health care indicators and mortality. In the present paper, the spatial smoothing performed on predicted incidence of epidemiological interest is illustrated for two examples. RESULTS: Predicted incidence of epidemiological interest was possible for 27/34 solid site-gender combinations and for only 2/8 haematological malignancies-gender combinations. Mapping of smoothed predicted incidence provided a clear picture of the main contrasts in incidence between districts. CONCLUSIONS: The methodology implemented provides a comprehensive framework to produce valuable predictions of incidence at a district level, using proxy measures and existing LR.

Multidimensional penalized splines for incidence and mortality-trend analyses and validation of national cancer-incidence estimates.

BACKGROUND: Cancer-incidence and mortality-trend analyses require appropriate statistical modelling. In countries without a nationwide cancer registry, an additional issue is estimating national incidence from local-registry data. The objectives of this study were to (i) promote the use of multidimensional penalized splines (MPS) for trend analyses; (ii) estimate the national cancer-incidence trends, using MPS, from only local-registry data; and (iii) propose a validation process of these estimates. METHODS: We used an MPS model of age and year for trend analyses in France over 1990-2015 with a projection up to 2018. Validation was performed for 22 cancer sites and relied essentially on comparison with reference estimates that used the incidence/health-care ratio over the period 2011-2015. Alternative estimates that used the incidence/mortality ratio were also used to validate the trends. RESULTS: In the validation assessment, the relative differences of the incidence estimates (2011-2015) with the reference estimates were <5% except for testis cancer in men and < 7% except for larynx cancer in women. Trends could be correctly derived since 1990 despite incomplete histories in some registries. The proposed method was applied to estimate the incidence and mortality trends of female lung cancer and prostate cancer in France. CONCLUSIONS: The validation process confirmed the validity of the national French estimates; it may be applied in other countries to help in choosing the most appropriate national estimation method according to country-specific contexts. MPS form a powerful statistical tool for trend analyses; they allow trends to vary smoothly with age and are suitable for modelling simple as well as complex trends thanks to penalization. Detailed trend analyses of lung and prostate cancers illustrated the suitability of MPS and the epidemiological interest of such analyses.

Trends in probabilities of death owing to cancer and owing to other causes in patients with colon cancer.

BACKGROUND: It is of interest to both the clinicians and patients to estimate the probability of death owing to cancer in the presence of other causes as time elapses since diagnosis. The objective of this study was to depict for patients diagnosed with colon cancer between 1990 and 2010 in France, the probability of surviving up to 10 years after diagnosis and to disentangle the probability of death owing to cancer from that of death owing to other causes. PATIENTS AND METHODS: Individuals with cancer were described, up to 10 years after diagnosis, as belonging to one of three categories: those who died owing to a cause related to cancer, those who died owing to another cause and those who survived. Net survival, crude probabilities of death related to colon cancer, death related to another cause and survival were estimated by modeling excess mortality hazard. RESULTS: In women of all ages, 5 and 10-year net survival improved over calendar time. The 10-year probability of survival decreased when age increased in both sexes. It was higher in women than in men, and this difference increased with age. Crude probabilities of death related to colon cancer decreased between 1990 and 2010 for men and women, although this was not observed in the eldest men. CONCLUSION: Crude probability of death related to colon cancer is an important indicator for patients and health policy makers. Results of cancer screening should be faced to trends in probability of death related to colorectal cancer.

Flexible and structured survival model for a simultaneous estimation of non-linear and non-proportional effects and complex interactions between continuous variables: Performance of this multidimensional penalized spline approach in net survival trend analysis.

Cancer survival trend analyses are essential to describe accurately the way medical practices impact patients' survival according to the year of diagnosis. To this end, survival models should be able to account simultaneously for non-linear and non-proportional effects and for complex interactions between continuous variables. However, in the statistical literature, there is no consensus yet on how to build such models that should be flexible but still provide smooth estimates of survival. In this article, we tackle this challenge by smoothing the complex hypersurface (time since diagnosis, age at diagnosis, year of diagnosis, and mortality hazard) using a multidimensional penalized spline built from the tensor product of the marginal bases of time, age, and year. Considering this penalized survival model as a Poisson model, we assess the performance of this approach in estimating the net survival with a comprehensive simulation study that reflects simple and complex realistic survival trends. The bias was generally small and the root mean squared error was good and often similar to that of the true model that generated the data. This parametric approach offers many advantages and interesting prospects (such as forecasting) that make it an attractive and efficient tool for survival trend analyses.

For a sound use of health care data in epidemiology: evaluation of a calibration model for count data with application to prediction of cancer incidence in areas without cancer registry.

There is a growing interest in using health care (HC) data to produce epidemiological surveillance indicators such as incidence. Typically, in the field of cancer, incidence is provided by local cancer registries which, in many countries, do not cover the whole territory; using proxy measures from available nationwide HC databases would appear to be a suitable approach to fill this gap. However, in most cases, direct counts from these databases do not provide reliable measures of incidence. To obtain accurate incidence estimations and prediction intervals, these databases need to be calibrated using a registry-based gold standard measure of incidence. This article presents a calibration model for count data developed to predict cancer incidence from HC data in geographical areas without cancer registries. First, the ratio between the proxy measure and incidence is modeled in areas with registries using a Poisson mixed model that allows for heterogeneity between areas (calibration stage). This ratio is then inverted to predict incidence from the proxy measure in areas without registries. Prediction error admits closed-form expression which accounts for heterogeneity in the ratio between areas. A simulation study shows the accuracy of our method in terms of prediction and coverage probability. The method is further applied to predict the incidence of two cancers in France using hospital data as the proxy measure. We hope this approach will encourage sound use of the usually imperfect information extracted from HC data.

Focus on an unusual rise in pancreatic cancer incidence in France.

Background: Pancreatic cancer is one of the most lethal. Most countries have exhibited a stable or decreasing incidence over time. The aim of this study was to provide updated French temporal trends in pancreatic cancer incidence and mortality over the past three decades. Methods: Incidence was estimated using the French National Network of Cancer Registries (FRANCIM) and mortality using the French Mortality Statistics Office. World age-standardized incidence and mortality were modelled by age-period-cohort models. The net cumulative risk of developing pancreatic cancer by birth cohort was calculated, as were annual percentage changes (APCs) in incidence and mortality. Results: Between 1982 and 2012, age-standardized incidence increased from 4.8 in 1980 to 9.6 per 100 000 in men and from 2.3 to 6.8 in women. The mean APC was 2.3% (2.1-2.6) and 3.6% (3.3-3.9), respectively. The cumulative risk of developing pancreatic cancer before age 75 rose from 0.62% for males born around 1920 to 1.17% for those born around 1950. It was respectively 0.31% and 0.86% for women. Mortality did not vary in men (8.1 per 100 000). It slightly increased in women from 4.0 in 1982 to 5.4 in 2012. Conclusion: Pancreatic cancer incidence and mortality exhibited diverging trends. Incidence increased over the last 30 years in France whereas mortality did not vary in men and moderately increased in women. Incidence remained lower than mortality up to 2002. One cannot exclude the possibility that a similar trend may appear in other countries. Etiological studies are required to further explain this increase.

Survival of solid cancer patients in France, 1989-2013: a population-based study.

This study provides updates of net survival (NS) estimates at 5, 10, and 15 years as well as survival trends for 35 solid cancers in France using data from 19 population-based cancer registries. The study considered all cases of solid cancer diagnosed between 1989 and 2010 in patients older than 15 years of age who were actively followed up until 30 June 2013. NS was estimated using the Pohar-Perme method. The age-standardized NS used the international cancer survival standard weights. The 5-year age-standardized NSs ranged from 4% (pleural mesothelioma) to 93% (prostate) in men and from 10% (pancreas) to 97% (thyroid) in women. The 10-year age-standardized NSs ranged from 2% (pleural mesothelioma) in both sexes to 95% (testis) in men and 91% (thyroid) in women. The most frequent cancers (namely, breast and prostate cancers) had the highest NSs: 87 and 93% at 5 years and 78 and 84% at 10 years, respectively. Several cancers (especially lung, pancreas, and liver cancer) had very poor prognoses (5-year NSs under 20%). Fifteen-year NSs remained high for testis cancer. In most cancers, 5- and 10-year age-standardized NSs increased between 1989 and 2010. Advanced age was associated with a poor prognosis and little improvement in survival. The increases in cancer survival are probably related to earlier diagnosis and therapeutic advances over the last decade. However, poor prognoses are still found in some alcohol-related and tobacco-related cancers and in elderly patients, highlighting the need for more prevention, diagnosis, and treatment efforts.

Performance of two formal tests based on martingales residuals to check the proportional hazard assumption and the functional form of the prognostic factors in flexible parametric excess hazard models.

Net survival, the one that would be observed if the disease under study was the only cause of death, is an important, useful, and increasingly used indicator in public health, especially in population-based studies. Estimates of net survival and effects of prognostic factor can be obtained by excess hazard regression modeling. Whereas various diagnostic tools were developed for overall survival analysis, few methods are available to check the assumptions of excess hazard models. We propose here two formal tests to check the proportional hazard assumption and the validity of the functional form of the covariate effects in the context of flexible parametric excess hazard modeling. These tests were adapted from martingale residual-based tests for parametric modeling of overall survival to allow adding to the model a necessary element for net survival analysis: the population mortality hazard. We studied the size and the power of these tests through an extensive simulation study based on complex but realistic data. The new tests showed sizes close to the nominal values and satisfactory powers. The power of the proportionality test was similar or greater than that of other tests already available in the field of net survival. We illustrate the use of these tests with real data from French cancer registries.

Incidence trends for potentially human papillomavirus-related and -unrelated head and neck cancers in France using population-based cancer registries data: 1980-2012.

Human papillomavirus (HPV) has been recently recognised as a carcinogenic factor for a subset of head and neck cancers (HNC). In Europe, France has one of the highest incidence rates of HNC. The aim of this study is to explore changes in HNC incidence in France, potentially in relation with infection by HPV. HNC were classified into two anatomical groups: potentially HPV-related and HPV-unrelated. Trends over the period 1980-2012 were analysed by an age-period-cohort model based on data from eleven French cancer registries. Among men, the age-standardised incidence rate (ASR) of HNC decreased in both groups, but less so for HPV-related sites as compared to unrelated sites, especially in recent years (annual percentage change [APC] over the period 2005-2012: -3.5% vs. -5.4%). Among women, the ASR increased in both groups, but more rapidly for HPV-related as compared to unrelated sites (APC over the period 2005-2012: +1.9% vs. -0.4%). This preferential growth of HPV-related versus unrelated HNC was observed in the cohorts born from 1930 to 1935. The differences in trends between possible HPV-related and HPV-unrelated sites suggest an increasing incidence of HNC due to HPV infection. The difference was less marked in men as compared to women, most likely because of a higher contamination in the HPV-related group by cancers due to tobacco or alcohol consumption. The pattern observed is consistent with observations made in other countries, with studies of HPV prevalence in HNC and the evolution of sexual behaviour in France.

Trends in net survival from 15 cancers in six European Latin countries: the SUDCAN population-based study material.

The aim of the SUDCAN collaborative study was to compare the net survival from 15 cancers diagnosed in 2000-2004 in six European Latin countries and provide trends in net survival and dynamics of excess mortality rates up to 5 years after diagnosis from 1992 to 2004 in France, Italy, Spain, and Switzerland, and from 2000 to 2004 in Belgium and Portugal. This paper presents a detailed description of the data analyzed and quality indicators. Incident cases from Belgium, France, Italy, Portugal, Spain, and Switzerland were retrieved from 56 general or specialized population-based cancer registries that participated in the EUROCARE-5 database. Fifteen cancer sites were analyzed. The data were checked according to the EUROCARE protocol. The percentages of excluded cases, cases based on death-certificate only, cases lost to follow-up at 5 years after diagnosis, and the proportions of microscopically verified cases were evaluated across countries and cancer sites. Data exclusions for major flaws were negligible. Cases based on death-certificate only were quite rare, except for some poor-prognosis cancers in some countries. The site-specific proportions of microscopically verified cases were generally high, but slightly lower in Italy than elsewhere. The percentage of cases lost to follow-up at 5 years after diagnosis was generally low. The net survival analyses in 2000-2004 included 873 314 tumors, whereas trend analyses included 1 426 004 tumors. The quality of the data analyzed was generally good. In fact, the analyzed data have been already checked and accepted for EUROCARE-5. However, slight differences in quality indexes, for some cancers, should be kept in mind in the interpretation of survival comparisons across countries.

New insights into survival trend analyses in cancer population-based studies: the SUDCAN methodology.

The main objective of the SUDCAN study was to compare, for 15 cancer sites, the trends in net survival and excess mortality rates from cancer 5 years after diagnosis between six European Latin countries (Belgium, France, Italy, Portugal, Spain and Switzerland). The data were extracted from the EUROCARE-5 database. The study period ranged from 6 (Portugal, 2000-2005) to 18 years (Switzerland, 1989-2007). Trend analyses were carried out separately for each country and cancer site; the number of cases ranged from 1500 to 104 000 cases. We developed an original flexible excess rate modelling strategy that accounts for the continuous effects of age, year of diagnosis, time since diagnosis and their interactions. Nineteen models were constructed; they differed in the modelling of the effect of the year of diagnosis in terms of linearity, proportionality and interaction with age. The final model was chosen according to the Akaike Information Criterion. The fit was assessed graphically by comparing model estimates versus nonparametric (Pohar-Perme) net survival estimates. Out of the 90 analyses carried out, the effect of the year of diagnosis on the excess mortality rate depended on age in 61 and was nonproportional in 64; it was nonlinear in 27 out of the 75 analyses where this effect was considered. The model fit was overall satisfactory. We analysed successfully 15 cancer sites in six countries. The refined methodology proved necessary for detailed trend analyses. It is hoped that three-dimensional parametric modelling will be used more widely in net survival trend studies as it has major advantages over stratified analyses.

Trends in net survival from head and neck cancer in six European Latin countries: results from the SUDCAN population-based study.

The aim of the SUDCAN collaborative study was to compare the trends in 1- and 5-year net survival and the trends in the dynamics of the excess mortality rates in head and neck cancers between six European Latin countries (Belgium, France, Italy, Portugal, Spain and Switzerland). The data were extracted from the EUROCARE-5 database. First, the net survival was studied over the 2000-2004 period using the Pohar-Perme estimator. For trend analyses, the study period was specific to each country. The results are reported from 1992 to 2004 in France, Italy, Spain and Switzerland and from 2000 to 2004 in Belgium and Portugal. The analyses were carried out using a flexible excess rate modelling. There were significant differences between countries in 5-year age-standardized net survivals over the 2000-2004 period, ranging from 33 to 34% in France and Portugal from 42 to 44% in Switzerland and Italy, respectively. The age-standardized net survival improved considerably from 1992 to 2004 in Italy, Spain and Switzerland, but not in France because of lack of improvement in the elderly. The increase in net survival was linked to a decrease in the excess mortality rate up to 3-4 years after diagnosis. The net survival from head and neck cancers improved over the study period, but significant differences were still observed in 2004. Differences in sex ratio and anatomical distribution contributed only partially towards these disparities. Differences in stage at diagnosis, time to treatment and/or proportion of human papillomavirus-related cases are also probably involved in the survival disparities observed. Overall, the prognosis of these tumours remains poor.

Trends in net survival lung cancer in six European Latin countries: results from the SUDCAN population-based study.

Survival is a key measure of the effectiveness of a healthcare system. European Latin countries have some similarities in their health systems; it is thus interesting to examine their differences in survival from cancer, here, lung cancer. The aim of the SUDCAN collaborative study was to compare the trends in the 1- and 5-year net survival from lung cancer and the trends in the excess mortality rates between six European Latin countries (Belgium, France, Italy, Portugal, Spain and Switzerland). The data were extracted from the EUROCARE-5 database. First, the net survival was studied over the 2000-2004 period using Pohar-Perme estimator. For trend analyses, the study period was specific to each country. The results are reported from 1992 to 2004 in France, Italy, Spain and Switzerland and from 2000 to 2004 in Belgium and Portugal. The analyses were carried out using a flexible excess rate modelling. Overall, the 1-year net survival from lung cancer ranged between 36 (Spain) and 43% (Belgium and Switzerland) and the 5-year net survival ranged between 11 (Spain) and 15% (Belgium and Switzerland). Between 1992 and 2004, the age-standardized survival increased considerably at 1 year, but increased less at 5 years after diagnosis. This increase was observed at ages 60 and 70, but was less obvious at age 80. There was little difference in net survival from lung cancer between European Latin countries, particularly in the more recent years. However, survival was slightly lower in Spain and Portugal than in France, Italy, Belgium and Switzerland. High-resolution studies with data on treatment, stage at diagnosis and comorbidities are needed to understand the reasons for these differences.

Trends in net survival from skin malignant melanoma in six European Latin countries: results from the SUDCAN population-based study.

In Europe as a whole, survival from skin malignant melanoma (SMM) has increased constantly since the 1980s. The aim of the SUDCAN collaborative study was to compare the trends in the 5-year net survival from SMM and in related excess mortality rate between six European Latin countries (Belgium, France, Italy, Portugal, Spain, and Switzerland). The data were extracted from the EUROCARE-5 database (end of follow-up: 01 January 2009). First, the net survival was studied over the 2000-2004 period using the Pohar-Perme estimator. For trend analyses, the study period was specific to each country. The results are reported from 1992 to 2004 in France, Italy, Spain, and Switzerland and from 2000 to 2004 in Belgium and Portugal. The analyses were carried out using a flexible excess rate modeling. Over the 2000-2004 period, the 5-year net survival from SMM ranged from 79 (Portugal) to 90% (Switzerland). In all countries, net survival was higher in women versus men and in young versus old age groups. From 1992 to 2004, the 5-year net survival increased the most in the countries with the lowest survivals in 1992 (+9% in Italy and Spain vs. +2% in Switzerland or +4% in France). The differences between countries decreased between 1992 and 2004. Although survival increased to a lower or higher extent in all countries during the period studied, significant differences in net survival from SMM persisted among the six countries studied. Health policies should mainly enhance early diagnosis by increasing public awareness and with screening campaigns. Furthermore, new immunotherapies, which will be approved soon hopefully, should also be used to improve the outcomes of SMM treatment.

Trends in net survival from ovarian cancer in six European Latin countries: results from the SUDCAN population-based study.

European Latin countries have some similarities in their health systems. It is thus interesting to look at their differences in cancer survival (here, ovarian cancer) through monitoring of specific indicators of quality care. The aim of this SUDCAN collaborative study was to compare the trends in 1 and 5-year net survival from ovarian cancer and the trends in the excess mortality rates between six European Latin countries (Belgium, France, Italy, Portugal, Spain and Switzerland). The data were extracted from the EUROCARE-5 database. First, the net survival was studied over the 2000-2004 period using the Pohar-Perme estimator. For trend analyses, the study period was specific to each country. The results are reported from 1992 to 2004 in France, Italy, Spain and Switzerland, and from 2000 to 2004 in Belgium and Portugal. The analyses were carried out using a flexible excess rate modelling. Over the period 2000-2004, there were slight differences in the 5-year age-standardized net survivals from ovarian cancer; they ranged from 36% in Spain to 42% in Belgium. Net survival was much higher in young than in old age groups, but this difference was more marked in Spain and less marked in France. Between 1992 and 2004, the net survival increased in all countries, mainly in young and middle-aged women. However, the differences in 5-year net survival between these countries were larger in 2004 than in 1992. Slight differences were observed in survival from ovarian cancer between the six European Latin countries. A considerable improvement in survival was observed in all countries, especially in young and middle-aged women. This study highlights the need for further monitoring of ovarian cancer outcomes.

Trends of incidence and survival in squamous-cell carcinoma of the anal canal in France: a population-based study.

Data on anal cancer epidemiology are rare. The aim of this study was to report on trends of incidence and survival for anal cancer in France before the implementation of the human papilloma virus vaccine. This analysis was carried out on 1150 squamous-cell carcinomas of the anal canal diagnosed from 1989 to 2004 in a population of 5.7 million people covered by eight population-based cancer registries. Time trends in incidence were modeled using an age-period-cohort model. Net survival rates were obtained using the recently validated unbiased Pohar-Perme estimator. The incidence of squamous-cell carcinoma of the anal canal increased from 0.2 to 0.5/100 000 person-years among men and from 0.7 to 1.3/100 000 person-years among women from 1982 to 2012. Among women, the increase peaked after 2005, with an annual percentage change of +3.4% between 2005 and 2012, as compared with +2.6% among men. The net survival was 56% (95% confidence interval, 49-64) at 5 years and 48% (33-70) at 10 years among men. It was higher among women, at 65% (61-69) and 56% (50-63) at 5 and 10 years, respectively. The prognosis improved between 1989-1997 and 1998-2004. This improvement was slightly greater for men than for women, thus progressively reducing the gap between sexes. The incidence of squamous-cell anal canal cancer increased slightly among both sexes, but the increase was more marked among women than among men. The potential benefit of prophylactic female human papilloma virus vaccination against cervical cancer in France should be further evaluated.

Ovarian cancer in France: trends in incidence, mortality and survival, 1980-2012.

OBJECTIVE: The aim of this epidemiological study was to describe the incidence, mortality and survival of ovarian cancer (OC) in France, according to age, period of diagnosis, and histological type. METHODS: Incidence and mortality were estimated from 1980 to 2012 based on data in French cancer registries and from the Centre for Epidemiology of Causes of Death (CépiDc-Inserm) up to 2009. Net survival was estimated from registry data using the Pohar-Perme method, on cases diagnosed between 1989 and 2010, with date of last follow-up set at 30 June 2013. RESULTS: In 2012, 4615 cases of OC were diagnosed in France, and 3140 women died from OC. World population age-standardized incidence and mortality rates declined by respectively 0.6% and 1.2% per year between 1980 and 2012. Net survival at 5years increased slightly, from 40% for the period 1989-1993 to 45% for the period 2005-2010. Net survival varied considerably according to histological type. Germ cell tumors had better net survival at 10years (81%) compared to epithelial tumors (32%), sex cord-stromal tumors (40%) and tumors without biopsy (8%). CONCLUSIONS: Our study shows a decline in incidence and mortality rates from ovarian cancer in France between 1980 and 2012, but net survival remains poor overall, and improved only slightly over the whole study period.

Cancer incidence estimation at a district level without a national registry: a validation study for 24 cancer sites using French health insurance and registry data.

BACKGROUND: District-level cancer incidence estimation is an important issue in countries without a national cancer registry. This study aims to both evaluate the validity of district-level estimations in France for 24 cancer sites, using health insurance data (ALD demands--Affection de Longue Durée) and to provide estimations when considered valid. Incidence is estimated at a district-level by applying the ratio between the number of first ALD demands and incident cases (ALD/I ratio), observed in those districts with cancer registries, to the number of first ALD demands available in all districts. These district-level estimations are valid if the ratio does not vary greatly across the districts or if variations remain moderate compared with variations in incidence rates. METHODS: Validation was performed in the districts covered by cancer registries over the period 2000-2005. The district variability of the ALD/I ratio was studied, adjusted for age (mixed-effects Poisson model), and compared with the district variability in incidence rate. The epidemiological context is also considered in addition to statistical analyses. RESULTS: District-level estimation using the ALD/I ratio was considered valid for eight cancer sites out of the 24 studied (lip-oral cavity-pharynx, oesophagus, stomach, colon-rectum, lung, breast, ovary and testis) and incidence maps were provided for these cancer sites. CONCLUSION: Estimating cancer incidence at a sub-national level remains a difficult task without a national registry and there are few studies on this topic. Our validation approach may be applied in other countries, using health insurance or hospital discharge data as correlate of incidence.

A Suitable Approach to Estimate Cancer Incidence in Area without Cancer Registry.

Objective. Use of cancer cases from registries and PMSI claims database to estimate Département-specific incidence of four major cancers. Methods. Case extraction used principal diagnosis then surgery codes. PMSI cases/registry cases ratios for 2004 were modelled then Département-specific incidence for 2007 estimated using these ratios and 2007 PMSI cases. Results. For 2007, only colon-rectum and breast cancer estimations were satisfactorily validated for infranational incidence not ovary and kidney cancers. For breast, the estimated national incidence was 50,578 cases and the incidence rate 98.6 cases per 100,000 person per year. For colon-rectum, incidence was 21,172 in men versus 18,327 in women and the incidence rate 38 per 100,000 versus 24.8. For ovary, the estimated incidence was 4,637 and the rate 8.6 per 100,000. For kidney, incidence was 6,775 in men versus 3,273 in women and the rate 13.3 per 100.000 versus 5.2. Conclusion. Incidence estimation using PMSI patient identifiers proved encouraging though still dependent on the assumption of uniform cancer treatments and coding.