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

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.

Impact of socio-economic environment on incidence of primary liver cancer in France between 2006 and 2016

BACKGROUND AND AIMS

To measure the impact of socio-economic environment on the incidence of hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (iCCA).

METHOD

The study used data from the French Network of Cancer Registries (FRANCIM) between 2006 and 2016. Classification of patients into HCC and iCCA was performed according to the topographical and morphological codes of the 3rd edition of the International Classification of Diseases for Oncology. Patient addresses were geolocalized and assigned to an IRIS, the smallest French geographic unit. Socio-economic environment was assessed by the European Deprivation Index (EDI). Sex- and age-standardized incidence rates with 95% confidence intervals (CI) were estimated per 100 000 inhabitants, by national quintiles, for each IRIS, sex and age group. Quintile 1 (Q1) characterized the most affluent areas. A Poisson regression was performed to model the impact of deprivation.

RESULTS

We included 22 249 cases (79.64% HCC, 16.97% iCCA). Incidence rates were 11.46 and 2.39 per 100 000 person-years for HCC and iCCA, respectively. There was an over-incidence of HCC in quintiles 2, 3, 4 and 5 compared to quintile 1: Q1 10.28 [9.9-10.66] per 100 000 person-years, Q2 11.43 [10.48-12.47] (p < .0001), Q3 11.81 [10.82-12.89] (p < .0001), Q4 12.26 [11.25-13.37] (p < .001) and Q5 11.53 [10.57-12.57] (p < .0001). By contrast, there was no difference for iCCa. Deprivation was significantly associated with HCC in men (p = .0018) and women (p = .0009), but not with iCCA (p = .7407).

CONCLUSION

The incidence of HCC is related to socio-economic environment, unlike iCCA. HCC and iCCA should be studied separately in epidemiological studies.

Trends in incidence of invasive vaginal cancer in France from 1990 to 2018 and survival of recently diagnosed women - A population-based study

OBJECTIVE

The aim of this study was to analyze trends in the incidence of vaginal cancer in France over a 28-year period and to present survival for recently-diagnosed women.

METHODS

French cancer registries provided data on invasive vaginal cancers diagnosed from 1990 to 2015 and followed up through June 2018. Trends in incidence were analyzed using a Poisson model with a bidimensional penalized spline of age and year at diagnosis. Net survival analysis was restricted to recently-diagnosed cases (2010-2015) and used a novel approach based on a bidimensional penalized spline of age and time-since-diagnosis to model excess mortality hazard.

RESULTS

With 162 new cases estimated in France in 2018, vaginal cancer represented 0.9 % of genital cancers in French women. In 2018, the world population age-standardized incidence rate was 0.2 per 100,000 person-years, median age at diagnosis was 75 years. The standardized incidence rate decreased significantly by 3 % per year (95 % CI, -3.8; -2.2) between 1990 and 2018 (0.4 cases per 100,000 person-year in 1990, vs 0.2 in 2018). Age-standardized net survival at 1 and 5 years after diagnosis was respectively 74 % and 45 %.

CONCLUSIONS

This study confirms that vaginal cancer is still a rare malignancy in France with 5-year net survival that remains low. We observed a consistent decrease in the incidence rate between 1990 and 2018. It may be too early to attribute these trends to a positive impact of vaccination campaigns against hrHPV infection, since vaginal cancer mainly affects older women and HPV vaccination has only been available since the early 2000s, and only targets young girls.

Evolution of the Incidence of Oral Cavity Cancers in the Elderly from 1990 to 2018

OBJECTIVES

To describe the evolution of the incidence of oral cavity cancers (OCC) among elderly patients in France between 1990 and 2018 and to compare it to the incidence of other cancers sharing the same main risk factors.

MATERIAL AND METHODS

The incidence of cancers in mainland France from 1990 to 2018 was estimated from incidence data observed in every cancer registry of the Francim network. Incidence was modeled by a 2-dimensional penalized spline of age and year of diagnosis, associated with a random effect corresponding to the registry. The elderly population was divided into two groups: 70-79 years old and ≥80 years old.

RESULTS

There was a 72% increase in the number of OCC cases in women over 70 years of age between the periods 1990-1999 and 2010-2018. As for men, there was a stabilization in the number of cases (+2%). Over the same period, for laryngeal and hypopharyngeal cancers, there was a decrease in incidence in elderly men and an increase in elderly women, although less marked than for OCC.

CONCLUSIONS

Since the 1990s, the incidence of OCC has been increasing in elderly subjects in France, particularly in women. Population aging and growth or alcohol and tobacco consumption alone do not seem to explain this increase, which is not observed in the same proportions for other upper aerodigestive tract cancer subsites sharing the same main risk factors.

Estimation of Oncologic Surgery Case Volume Before and After the COVID-19 Pandemic in France

IMPORTANCE

COVID-19 has had a major effect on health care activities, especially surgery. At first, comparisons were proposed using 2019 activities as the highest standard. However, while such an approach might have been suitable during the first months of the pandemic, this might no longer be the case for a longer period.

OBJECTIVE

To examine approaches that may better assess the use of cancer surgeries.

DESIGN, SETTING, AND PARTICIPANTS

In a cross-sectional design, the nationwide French hospital facility data (Medicalised Information System Program) were used to assess cancer surgery for 6 cancer site categories in adults from January 1, 2010, to December 31, 2021.

EXPOSURE

Estimated cancer surgery activity during the COVID-19 pandemic.

MAIN OUTCOMES AND MEASURES

Three models were proposed to assess the expected number of surgical procedures between 2020 and 2021 and make a comparison with those observed in earlier years.

RESULTS

In France, cancer removal surgeries account for approximately 7000 hospitalizations per year for liver cancer; 4000 for pancreatic cancer; 7700 for ovarian cancer; 1300 for esophagus cancer; 23 000 for ear, nose, and throat (ENT) cancer; 78 000 for breast cancer; and 16 600 for thoracic cancers. For most cancer sites, the number of surgical procedures increased from 2010 to 2019: liver, 14%; pancreas, 38%; ovary, 14%; esophagus, 18%; breast, 8%; and thoracic, 29%. Assuming stability, these values underestimate the gap in activity observed in 2020-2021. For other procedures, a decrease was observed: stomach, -10%, and ENT, -6%. Assuming stability, these values overestimate the gap in activity observed in 2020-2021. At the end of 2021, according to the model, the gap in activity observed in 2020-2021 was estimated at between -1.4% and 1.7% for breast, -6.6% and -7.3% for thoracic, -3.1% and -2.5% for ovarian, -4.2% and -1.7% for pancreas, -6.7% and 5.9% for stomach, and -13.0% and -13.9% for esophageal cancers. For ENT, liver, and urologic cancers, because the trend was different before and after 2015, it was necessary to opt for modeling using only the most recent period. The cumulative gap in activity observed in 2020-2021 was estimated at -1.0% for ENT cancers, -5.3% for liver cancers, and -2.9% for urologic cancers.

CONCLUSIONS AND RELEVANCE

The findings of this study suggest that short- and medium-term trends must be considered to estimate COVID-19 cancer surgery activities. Breast cancer is the site for which the activity showed the smallest decrease during the pandemic, with almost full recovery in 2021.

How Can the EU Beating Cancer Plan Help in Tackling Lung Cancer, Colorectal Cancer, Breast Cancer and Melanoma?

Cancer is the second leading cause of mortality in EU countries, and the needs to tackle cancer are obvious. New scientific understanding, techniques and methodologies are opening up horizons for significant improvements in diagnosis and care. However, take-up is uneven, research needs and potential outstrip currently available resources, manifestly beneficial practices—such as population-level screening for lung cancer—are still not generalised, and the quality of life of patients and survivors is only beginning to be given attention it merits. This paper, mainly based on a series of multistakeholder expert workshops organised by the European Alliance for Personalised Medicine (EAPM), looks at some of those specifics in the interest of planning a way forward. Part of this exercise also involves taking account of the specific nature of Europe and its constituent countries, where the complexities of planning a way forward are redoubled by the wide variations in national and regional approaches to cancer, local epidemiology and the wide disparities in health systems. Despite all the differences between cancers and national and regional resources and approaches to cancer care, there is a common objective in pursuing broader and more equal access to the best available care for all European citizens.

Diverging incidence trends of oral tongue cancer compared to other head and neck cancers in young adults in France

While head and neck cancer incidence decreased worldwide due to reduced tobacco and alcohol consumption, oral tongue cancer (OTC) incidence has been reported to be increasing in several countries. Our study examines the incidence trends of OTC in France from 1990 to 2018, globally and by age; and compares the incidence trends with the evolution of the incidence of other human papilloma virus-unrelated head and neck squamous cell carcinoma, that is, cancers of the remaining subsites of the oral cavity (RSOCC) and laryngeal cancers for the period 1990 to 2018. World age-standardized incidence rates of oral tongue cancers (C02), cancers of the remaining subsites of the oral cavity (RSOCC, C03-06) and laryngeal cancers (C32) were estimated using the French National Network of Cancer Registries for the period 1990 to 2018. Trends in national incidence rates were estimated from a mixed-effect Poisson model including age and year effects using penalized splines and a district-random effect. In women aged 30 and 40, a significant increase in OTC incidence was observed, while ROSCC showed a nonsignificant incidence decrease. In young men aged 25, a marginally significant increase of OTC incidence years was observed, while incidence rates of RSOCC significantly declined. The results suggest a tendency towards diverging incidence trends for OTC compared to RSOCC and laryngeal cancer in young adults. The observed trends may reflect changes in underlying exposures or emerging exposures not yet identified, and stress the need to further investigate the etiology of oral tongue cancers.

Temporal and geographical variations of thyroid cancer incidence and mortality in France during 1986-2015: The impact of overdiagnosis

BACKGROUND

France is among the countries showing fastest growth of thyroid cancer (TC) incidence and highest incidence rates in Europe. This study aimed to clarify the temporal and geographical variations of TC in France and to quantify the impact of overdiagnosis.

METHODS

We obtained TC incidence data in 1986-2015, and mortality data in 1976-2015, for eight French departments covering 8% of the national population, and calculated the age-standardised rates (ASR). We estimated the average annual percent changes (AAPC) of TC incidence, overall and by department and histological subtype. Numbers and proportions of TC cases attributable to overdiagnosis were estimated by department and period, based on the comparison between the shape of the age-specific curves with that observed prior to changes in diagnostic practice.

RESULTS

During 1986-2015, there were 13,557 TC cases aged 15-84 years. Large variations of TC incidence were observed across departments, with the highest ASR and the fastest increase in Isère. Papillary subtype accounted for 82.8% of the cases, and presented an AAPC of 7.0% and 7.6% in women and men, respectively. Anaplastic TC incidence decreased annually 3.0% in women and 0.8% in men. Mortality rates declined consistently for all departments. The absolute number (and proportion) of TC cases attributable to overdiagnosis grew from 1074 (66%) in 1986-1995 to 3830 (72%) in 2006-2015 in women, and varied substantially across departments.

CONCLUSIONS

Overdiagnosis plays an important role in the temporal and regional variations of TC incidence in France. Monitoring the time trends and regulating the regional healthcare practice are needed to reduce its impact.

Cancer incidence and mortality trends in France over 1990-2018 for solid tumors: the sex gap is narrowing

Objective

To analyze trends in cancer incidence and mortality (France, 1990–2018), with a focus on men-women disparities.

Methods

Incidence data stemmed from cancer registries (FRANCIM) and mortality data from national statistics (CépiDc). Incidence and mortality rates were modelled using bidimensional penalized splines of age and year (at diagnosis and at death, respectively). Trends in age-standardized rates were summarized by the average annual percent changes (AAPC) for all-cancers combined, 19 solid tumors, and 8 subsites. Sex gaps were indicated using male-to-female rate ratios (relative difference) and male-to-female rate differences (absolute difference) in 1990 and 2018, for incidence and mortality, respectively.

Results

For all-cancers, the sex gap narrowed over 1990–2018 in incidence (1.6 to 1.2) and mortality (2.3 to 1.7). The largest decreases of the male-to-female incidence rate ratio were for cancers of the lung (9.5 to 2.2), lip - oral cavity - pharynx (10.9 to 3.1), esophagus (12.6 to 4.5) and larynx (17.1 to 7.1). Mixed trends emerged in lung and oesophageal cancers, probably explained by differing risk factors for the two main histological subtypes. Sex incidence gaps narrowed due to increasing trends in men and women for skin melanoma (0.7 to 1, due to initially higher rates in women), cancers of the liver (7.4 to 4.4) and pancreas (2.0 to 1.4). Sex incidence gaps narrowed for colon-rectum (1.7 to 1.4), urinary bladder (6.9 to 6.1) and stomach (2.7 to 2.4) driven by decreasing trends among men. Other cancers showed similar increasing incidence trends in both sexes leading to stable sex gaps: thyroid gland (0.3 to 0.3), kidney (2.2 to 2.4) and central nervous system (1.4 to 1.5).

Conclusion

In France in 2018, while men still had higher risks of developing or dying from most cancers, the sex gap was narrowing. Efforts should focus on avoiding risk factors (e.g., smoking) and developing etiological studies to understand currently unexplained increasing trends.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-021-08261-1.

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.