Protective association of angiotensin blockade with influenza: a result of immortal time bias?
Daniel Mølager Christensen, Thomas Alexander Gerds, Gunnar Gislason, Christian Tobias Torp-Pedersen · European Heart Journal - Cardiovascular Pharmacotherapy · 2020
Chung et al.1 reported an association between exposure to angiotensin-converting enzyme inhibitors (ACE-i) and angiotensin receptor blockers (ARBs) for more than 1.5 years and lower susceptibility to influenza in a large UK cohort. ARB treatment for >10 years was remarkably associated with lower incidence of influenza with a hazard ratio (HR) of 0.11. Such findings could direct the search for effective therapies against COVID-19, as influenza viruses have been shown to use the ACE2 receptor to mediate lung damage,2 which is the receptor SARS-CoV-2 binds to in order to facilitate entry into the human cell.3 However, their analyses may have been biased as exposure was defined by number of prescriptions claimed during the study period. This introduces an immortal time bias as patients who start the treatment survive long enough to do so. This bias may lead to results that wrongly indicate a favourable effect of treatment, especially for long durations.4 In addition, it does not seem biologically plausible that angiotensin blockade should be associated with protective effects against influenza infection with effect sizes of the reported magnitude. Nor was ACE-i and ARB treatment associated with risk of COVID-19 infection in a different study.5 Thus, we aimed to determine if exposure to angiotensin blocking drugs was associated with a lower risk of influenza compared to no treatment. We used nationwide registries holding data about demographics, medication, and in-hospital and out-hospital diagnoses to identify all Danish residents aged 30 years and above during 2005–2017. To ensure patient anonymity, data from Danish registries is not available to be shared with the public. We investigated ACE-i, ARB, and ACE-i or ARB exposure compared to no exposure and the associated hazard rate of influenza using a Cox regression model fitted with a nested case–control design. Influenza cases were matched on date of influenza diagnosis (case date), age, and sex with up to four non-influenza population controls using incidence density matching. Exposure to treatment was evaluated as at least one prescription during three years prior to case date. We adjusted for heart failure, ischaemic heart disease, hypertension, dementia, diabetes, chronic obstructive pulmonary disease, chronic kidney disease, cancer, and stroke at the beginning of the exposure window (time zero for follow-up), i.e. 3 years prior to case date (Table 1). Characteristics of cases and controls Characteristics of cases and controls We included 20 232 influenza cases and 80 464 controls. At the case date, 4734 (23.4%) cases and 14 846 (18.5%) controls were exposed to ACE-i or ARB. In fully adjusted analyses, no association was observed between ACE-i or ARB treatment compared to no treatment (Figure 1): HR 1.03, 95% confidence interval 0.99–1.09. We performed an additional analysis evaluating exposure only in the one year prior to case date and found similar results. Association between angiotensin-converting enzyme inhibitor or angiotensin receptor blocker exposure and rate of influenza. In our analyses, ACE-i, ARB, and ACI-i or ARB exposure were not associated with a lower hazard of influenza. Our analyses were not affected by immortal time bias as cohort entry was not determined by exposure status. However, our study had several limitations. Firstly, we cannot rule out bias due to unmeasured confounding, e.g. we did not have access to information about seasonal influenza vaccination. Secondly, our exposure definition defers from that of Chung et al. (any prescription vs. cumulative prescriptions) which should be considered when comparing the studies. Lastly, an active comparator design would be more appropriate to mitigate unmeasured bias in observational studies of drug therapy.6 In spite of these limitations, our findings do not support an effort to investigate a potential protective effect of angiotensin-blockade with regards to COVID-19. In conclusion, the present study found no protective association between angiotensin-blockade and influenza susceptibility. Conflicts of interest: C.T.-P. reports grants from Bayer, grants from Novo Nordisk outside the submitted work. All other authors reported no conflicts of interest.