Abstract
BACKGROUND: The reported inverse association between cancer and subsequent Alzheimer's disease and related dementias (ADRD) remains uncertain. OBJECTIVES: To investigate the association between these common conditions of old age and explore possible causal factors. DESIGN, SETTING, PARTICIPANTS AND MEASUREMENTS: We conducted a large population-based cohort analysis using data from 3,021,508 individuals aged 60 and over in the UK Clinical Practice Research Datalink (CPRD), over a period up to 30 years (1988-2018). Cox proportional hazards models were fitted to estimate hazard ratios (HR) for risk of dementia associated with previous cancer diagnosis. Competing risk models were employed to account for competing risk of death. Two-sample Mendelian Randomization analysis based on meta-analysis data from large-scale GWAS studies was also conducted. RESULTS: In the CPRD cohort, 412,903 participants had cancer diagnosis and 230,558 were subsequently diagnosed with dementia over a median follow-up period of 7.9 years. Cancer survivors had a 25% lower risk of developing dementia (HR=0.75, 95% CI:0.74-0.76) after adjustment for potential confounders. Accounting for competing risk of death provided a sub-distribution HR of 0.56 (95% CI:0.55-0.56). Results were consistent for prevalent and incident cancer and different common cancer types. Two-sample Mendelian Randomization analysis, using 357 cancer-related instrumental single-nucleotide polymorphisms (SNPs) revealed evidence of vertical pleiotropy between genetically predicted cancer and reduced risk of Alzheimer's disease (OR=0.97,95% CI:0.95-0.99). CONCLUSION: Our results provide strong epidemiological evidence of the inverse association between cancer and risk of ADRD and support the potential causal nature of this association via genetic instruments. Further investigations into the precise underlying biological mechanisms may reveal valuable information for new therapeutic approaches.