Association between cancer and suicide causes of death in death certificates, France, 2000–2013

Abstract

Introduction Cancer, by its impact on health, adverse events of treatments, or stigma, can impact quality of life significantly, and promote the onset of suicidal ideations and suicide deaths. The aim of the study was to compare the co-occurrence of suicide and cancer in death certificates, according to cancer localization, and to evaluate, with simulation analyses, to what extent these associations can be extrapolated to the general population. Methods The study was conducted on the French medical causes of death database. All deaths that occurred in France between 2000 and 2013 were included in the study, except deaths from unknown cause (International Statistical Classification of Diseases ICD-10 code R99), and deaths in individuals aged 14 years and younger. Suicide deaths were identified according to the underlying cause of death diagnosis (ICD-10 codes X60-X84 and Y87.0). Cancers were identified according to the underlying and other causes of death diagnoses (ICD-10 codes C00-C97), taking into account the order in which causes were declared in the death certificate, and then classified in different cancer localizations. Risks of suicide according to cancer localization were compared by logistic regression, adjusting for age and area of death, and stratified by sex. A sensitivity analysis was conducted identifying cancer only from other causes of death (i.e. excluding the underlying cause of death), in order to maximize the comparability between suicides and non-suicidal deaths. Simulation analyses were conducted using French cancer and suicide attempt incidence data to measure the impact of the selection effect, when working on deceased individuals only, on the association between suicide and cancer. Results In total, 7,375,283 deaths were included in the study, of which 148,095 suicides (109,267 in men, 38,828 in women). Compared to lung cancer, the highest adjusted odds ratios (ORs) for suicide in men were those of prostate cancer (OR = 8.1, 95% confidence interval (95% CI): 7.1–9.0), bladder cancer (OR = 3.5, 95% CI: 2.9–4.2), and larynx cancer (OR = 3.3, 95% CI: 2.6–4.0); the lowest was for central nervous system cancer (OR = 0.3, 95% CI: 0.2–0.5). In women, the highest adjusted ORs for suicide were those of thyroid cancer (OR = 7.1, 95% CI: 2.3–18.1), bladder cancer (OR = 4.7, 95% CI: 2.3–9.4), and larynx cancer (OR = 4.4, 95% CI: 1.6–12.4). When restraining cancer definition to other causes of death (sensitivity analysis), cancers with the highest association with suicide in men were pancreas (OR = 2.1), stomach (OR = 1.8) and prostate cancer (OR = 1.5); liver cancer was associated with the lowest risk of suicide (OR = 0.4). In women, cancers with the highest association with suicide were bladder (OR = 2.2), breast (OR = 2.0), and pancreas cancer (OR = 1.9); liver cancer was associated with the lowest risk of suicide (OR = 0.2). Simulation analyses are in the process of being implemented, but the selection effect seems to bias the associations towards underestimation, the bias increasing with the lethality of the localization concerned. Conclusions Results were highly sensitive to the source of information used for cancer definition. The analysis based on both underlying and other causes of death for cancer definition was held as the main analysis since it used all the information available. This study was conducted on dead individuals only, and thus, results of this study have to be interpreted in light of this limitation. Biases involved in analyses of multiple causes of death association, as measured in this study, are likely to be generalizable to other studies using this design.