Reverse Causality in Cardiovascular Epidemiological Research: More Common Than Imagined?

Abstract

It is now well established that observational studies seeking evidence that a risk factor causes a disease are susceptible to various biases.Among the most important are residual confounding, whereby unmeasured or imprecisely measured confounders prevent causal inferences being drawn from associations between any 2 parameters, and reverse causality.Although the former is typically acknowledged in reports of such studies, the latter concept appears to be less well understood and, therefore, is more often overlooked as a potential explanation for apparent, often unexpected, associations between risk factors and adverse outcomes.With the potential for reverse causality in mind, Ravindrarajah and colleagues 1 in this issue of Circulation, sought to determine whether the association of lower systolic blood pressure (SBP) values with higher mortality in observational studies in the elderly (>75-80 years of age) could be afflicted by this same phenomenon.They did so because recent trial data, in particular the SPRINT study (Systolic Blood Pressure Intervention Trial), 2 demonstrated that lowering SBP to a target of <120 mm Hg led to a 33% reduction in all-cause mortality in those >75 years of age, findings in almost complete contrast with such observational data.To test their hypotheses, the authors analyzed data from the Clinical Practice Research Datalink, a primary care cohort collected in the United Kingdom that, importantly, includes serial measurements of many risk factors.Using this resource, they not only addressed the association between SBP and mortality, but they did so according to (1) frailty category and (2) antihypertensive treatment.They also asked whether SBP levels declined before death and, if so, how did this decline compare with the pattern of SBP in those who survived, and whether these patterns differed in those on or off blood pressure-lowering treatments.The researchers were able to confirm that men and women aged >80 and with SBP levels <120 mm Hg did indeed have higher risks of mortality than those with SBP in the range of 120 to 139 mm Hg.This J-curve phenomenon has been shown in many previous epidemiological studies which has raised concerns about the safety of intensive blood pressure (BP) lowering in the elderly population.Interestingly, this association was clearly observed across all categories of frailty.This latter finding, they note, is somewhat at odds with prior findings in NHANES (National Health and Nutrition Examination Survey), 3 where associations of low SBP with mortality were most evident in frailer subjects, a finding somewhat supportive of reverse causality.That said, different frailty indices (which appear complex in their computation) were used in each case and could have led to differential findings and, so, these frailtyassociated differences need to be considered with a degree of caution.More importantly, by exploiting their access to serial SBP measurements the authors were able to demonstrate a greater decline in SBP in the 5 years before death than was observed in those who remained alive, with a particularly marked decline in the 2 years before death.This pattern was the same whether or not individuals Reverse Causality in Cardiovascular Epidemiological Research More Common Than Imagined?