Choosing questions before methods in dementia research with competing events and causal goals

Abstract

Several of the hypothesized or studied exposures that may affect dementia risk are known to increase the risk of death. This may explain counterintuitive results, where exposures that are known to be harmful for mortality risk sometimes seem protective for the risk of dementia. Authors have attempted to explain these counterintuitive results as biased, but the bias associated with a particular analytic method cannot be defined or assessed if the causal question is not explicitly specified. Indeed, we can consider several causal questions when competing events like death, which cannot be prevented by design, are present. Current dementia research guidelines have not explicitly considered what constitutes a meaningful causal question in this setting or, more generally, how this choice justifies and should drive particular analytic decisions. To contextualize current practices, we first perform a systematic review of the conduct and interpretation of longitudinal studies focused on dementia outcomes where death is a competing event. We then describe and demonstrate how to address different causal questions (referred here as "the total effect" and "the controlled direct effect") with traditional analytic approaches under explicit assumptions. Our application focuses on smoking cessation in late-midlife. To illustrate core concepts, we discuss this example both in terms of a hypothetical randomized trial and with an emulation of such a trial using observational data from the Rotterdam Study.

Socioeconomic Deprivation, Genetic Risk, and Incident Dementia

“… 22 Sixth, estimates may be biased because the competing risk of death can precede dementia diagnosis. 51 , 52 Seventh, analyses were restricted to those aged ≥60 years, limiting cases, and to European ancestry, limiting generalizability. Finally, some associations were not replicated in complete-case data, likely owing to disproportionally missing data in higher deprivation ( Appendix Table 18 , available online).…”

Preexisting Diabetes Mellitus and All-Cause Mortality in Adult Patients With Sepsis: A Population-Based Cohort Study

“…We have thus chosen to present both crude (to give an extent of the overall risk that can be expected in this population) and adjusted estimates (to further describe the specific role of diabetes while “controlling” for other comorbid and general conditions). Third, our analysis of organ support measures is also subject to misclassification and the competing event of all-cause in-hospital death (48, 49). Furthermore, our definition focused on invasive mechanical ventilation, and we did not consider noninvasive support strategies; hence, it remains possible that patients with diabetes have a higher degree of acute respiratory failure subject to noninvasive support.…”

Towards a Clearer Causal Question Underlying the Association Between Cancer and Dementia

“…35 However, in the context of cancer diagnosis as the proxy for Pin1 overexpression, we cannot rule out the effect of cancer on death, represented as the arrow between falseRt+1 and falseDt+k. Since the total effect of falseAt+1 in falseYt+k includes the indirect causal pathway mediated by the effect of cancer diagnosis on mortality, this may translate into an inverse association 36 if we consider the effect of cancer itself, not the treatment after cancer diagnosis.…”