Androgen Deprivation Therapy and Dementia: New Opportunities and Challenges in the Big-Data Era
Menée aux Etats-Unis à partir de données portant sur plus d'un million de bénéficiaires du système Medicare atteints d'un cancer de la prostate entre 2001 et 2014 et âgés de plus de 67 ans, cette étude analyse l'association entre une thérapie anti-androgénique et le risque de développer la maladie d'Alzheimer
Androgen deprivation therapy (ADT) extends life for appropriately selected patients diagnosed with prostate cancer.1 Level-1 evidence supports this survival benefit in a wide breadth of clinical scenarios, which leads to high rates of ADT use globally.2 A complete understanding of the adverse effects of this therapy is paramount to permit informed decision making and patient selection. As a part of this effort, a body of literature developed that supported a potential association between ADT and adverse cognitive function.3,4 Subsequently, two research articles were published that used an electronic medical record (EMR)–based informatics approach to support a link between ADT and dementia in large, multi-institutional analyses.5,6 However, not all studies have supported this association.7 Given the widespread use and effectiveness of ADT, investigations that replicate and assess causality of an association between ADT and dementia are critical before changes in patient care are considered. In the article that accompanies this editorial, Baik et al8 examine the risk of dementia after ADT among 1.2 million fee-for-service Medicare beneficiaries with prostate cancer. In the cohort, 35% of men with prostate cancer received ADT, and 19% of the full cohort developed dementia during a mean follow-up time of 5.5 years. In propensity score–matched competing risk regression, there was a small, statistically significant increased risk of all-cause dementia (standardized hazard ratio, 1.01; 95% CI, 1.01 to 1.02) and a small decreased risk of Alzheimer’s disease (standardized hazard ratio, 0.98; 95% CI, 0.97 to 0.99). Although these results reached statistical significance, the analyses as presented support no meaningful effect of ADT on dementia risk. This is certainly reasonable, particularly given that even minor contributions from factors such as unmeasured confounding, exposure/outcome misclassification, or unrecorded competing risks could result in statistical significance because of the exceptional power of the study. Therefore, small-magnitude effects should be interpreted with great caution. This study has a number of notable strengths. Baik et al8 used the Centers for Medicare and Medicaid Services Virtual Research Data Center, which allowed them to leverage a big-data research approach. To my knowledge, the observational cohort of more than a million patients with prostate cancer represents the most well-powered study to date on the association of ADT and dementia. The large cohort allowed the authors to adjust analyses for a wide variety of relevant clinical factors, which might not have been possible in smaller data sets. The analyses used robust statistical methods, including propensity score–matched analyses and competing risk regression. In addition, broad access to data about Medicare beneficiaries allowed exposures, outcomes, and model covariates that occurred outside the index health care system of a patient to be captured; this was not possible in many previous studies,9 in which analyses were limited to a specific site or sites.