To the Editor: We read with interest the study by Zhang et al.,1 examining the association between inpatient use of angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs) and mortality in patients hospitalized with coronavirus disease 2019 (COVID-19), published in Circulation Research. The use of ACEIs or ARBs was associated with an important 58% risk reduction in all-cause mortality (hazard ratio, 0.42; 95% confidence interval, 0.19, 0.92). The association was even stronger with the use of propensity score models, resulting in a risk reduction of almost 70%. We believe that these remarkable effects are mainly driven by immortal time bias. In this study, patients were deemed exposed to ACEIs or ARBs if they had received a prescription at any time during the hospitalization period. In contrast, patients who did not receive these drugs during the hospitalization period were categorized as nonusers. This method of identifying users and nonusers can introduce immortal time bias.2 Indeed, by design, ACEI or ARB users had to survive from the date of admission to the date of an ACEI or ARB prescription to be categorized as a user. This time period was both misclassified as exposed (instead of unexposed) and immortal, as it was impossible for patients to die during this time span. In contrast, it was possible for nonusers to die at any point during the hospitalization period. This is particularly evident in Zhang et al.’s1 Figures 2A and 2B, where there were no deaths in the first 7 days of follow-up for the ACEIs/ARBs group, whereas deaths occurred as early as the second day after admission in the non-user group. Given the maximum 28-day follow-up, even a small amount of misclassified exposed person–time could lead to important bias, as was shown previously with β-blockers for patients hospitalized for chronic obstructive pulmonary disease.3 The authors also compared the risk of death among users of ACEIs/ARBs versus users of other antihypertensive drugs. In this analysis, the authors first identified ACEI/ARB users and then users of other antihypertensive drugs. This hierarchical exposure definition introduced immortal time bias,4 as the time between hospital admission and first ACEI/ARB prescription was misclassified as exposed and immortal because no events could have occurred during that period. In summary, although residual confounding is always a concern in observational studies, immortal time bias can introduce severely skewed results, though it can be avoided using proper methods of analysis.2–4 It would be informative to repeat the analyses by modeling ACEI or ARB exposure using time-dependent methods.
Read full abstract