The importance of using disease causal models in studies of preventive interventions: Learning from preeclampsia research.

Abstract

Interventions aimed at preventing disease are commonly studied as strategies of primary or secondary prevention. Unfortunately, this dichotomy can be misleading, and studies might unknowingly exclude people at high risk of the disease that could benefit from the intervention. Here I use the example of aspirin for prevention of preeclampsia to illustrate this problem. I use directed acyclic graphs to represent several causal models of aspirin and preeclampsia, each making different assumptions regarding the causal relation between previous preeclampsia, aspirin, and subsequent preeclampsia. Afterwards, I discuss the implications of each model. Aspirin started being recommended to pregnant women that had presented preeclampsia in previous pregnancies, but not to women at high risk due to other factors. Studies started evaluating aspirin in women at high risk due to these other causes and found it also reduced the risk of preeclampsia in them. Thanks to a shift towards risk-based interventions, guidelines started recommending aspirin to all women considered at high risk of preeclampsia. Furthermore, recent studies have begun using blood markers in women without classic risk factors to identify additional women that might benefit from aspirin. With such advances, performing "secondary prevention" once the first event occurred will increasingly represent a failure to intervene on time. Explicitly illustrating disease causal models helps to identify those individuals that are most likely to benefit from risk reduction, regardless of whether they were previously afflicted by the disease. This is beneficial when designing studies and when implementing preventive interventions.