Association of renin–angiotensin–aldosterone system inhibition with Covid‐19 hospitalization and all‐cause mortality in the UK biobank

With growing evidence on the protective effect of angiotensin‐converting enzyme inhibitors (ACEIs) and angiotensin receptor blockers (ARBs) in coronavirus disease 2019 (Covid‐19), we aimed to thoroughly investigate the association between the use of major classes of antihypertensive medications and Covid‐19 outcomes in comparison with the use of ACEIs and ARBs.


| INTRODUCTION
There has been a debate over the role of renin-angiotensinaldosterone system (RAAS) and RAAS inhibition in coronavirus disease 2019 . Angiotensin converting enzyme 2 (ACE2) is a transmembrane enzyme that functions as receptor for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). 1 After SARS-CoV-2 binds to ACE2 receptor, endocytosis of the viral complex results in ACE2 downregulation and accumulation of angiotensin II (AngII) with proinflammatory, vasoconstrictive and profibrotic effects. ACE2 is present in different organs including heart, kidney and lungs, the target organ for SARS-CoV-2. ACE2 also counteracts the activation of RAAS via degrading AngII to angiotensins 1-7 that exert their vasodilatory, anti-inflammatory and antiproliferative effects through mitochondrial assembly receptor. [2][3][4] The controversy about the use of ACE inhibitors (ACEIs) and angiotensin receptor blockers (ARBs) in patients infected with SARS-CoV-2 started with findings that showed higher prevalence and mortality of patients with cardiovascular diseases such as hypertension among patients with Covid-19. 5 Moreover, in an animal study, increased expression of ACE2 messenger RNA (mRNA) with the use of RAAS inhibitors was observed, suggesting higher susceptibility to SARS-CoV-2 among the users of these medications and hence it was hypothesized that their use might be related to Covid-19 severity and mortality. 6,7 However, current findings do not support this association [8][9][10][11][12] and evidence regarding beneficial effects of RAAS inhibition is growing, illustrating a potential protective effect of RAAS inhibitors in relation to severe clinical outcomes of Covid-19, particularly in patients with hypertension. 13 The favourable effect of RAAS inhibitors in Covid-19 needs investigation in a large cohort study considering the inadequate and conflicting evidence at hand. Therefore, we aimed to investigate the association of ACEIs and ARBs with adverse Covid-19 outcomes in comparison with other antihypertensive drugs in the large UK Biobank.

| Study population
The UK Biobank is a large population-based prospective cohort with about 500 000 participants living in the UK aged 40-69 years when recruited in 2006-2010. The collection of data involved a selfcompleted touch-screen questionnaire, a computer-assisted interview, physical and functional measures, and the collection of biological samples, as previously described in detail. 14 The data are also linked to electronic health-related records, including death, cancer, hospital admissions and primary care records. The UK Biobank study has obtained ethical approval from regulatory authorities and all participants provided signed electronic informed consent.
The UK Biobank has released Covid-19 data for its participants starting from March 2020. The data comprises of diagnostic Covid-19 test data, primary care data provided directly by the system suppliers, hospital inpatient, critical care and death data that are being updated regularly. 15,16 Currently, the primary care data are only available for England and study participants from Scotland and Wales needed to be excluded. A part of the primary care data are the prescription data of general practitioners (GPs).
Participants from England with no recorded GP prescription data and those who died before March 2020 were further excluded from the analyses. Moreover, to account for the confounding by indication bias, only patients with pharmaceutically treated hypertension were included. Finally, among 149 962 English study participants with recent use of antihypertensive medications, a total of 124 143 (82.8%) had diagnosed hypertension and could be included in the analyses.

| Covariates
Sociodemographic, lifestyle and health-related data were taken from the touchscreen interview conducted at the baseline examination of the UK Biobank. Age at onset of the pandemic was calculated by adding the years passed between date of attending an assessment centre for baseline examination and 1 March 2020 to the baseline age. The ethnic background was categorized as white, black and other (all other ethnic groups combined), education according to years (≤9, 10-12, ≥13), and smoking status by never, former or current smoker. The Townsend deprivation index was calculated based on the participants living areas defined by the corresponding postal codes. 20 The intensity of physical activity (low, moderate, high) was based on the International Physical Activity Questionnaire. The amount of ethanol consumed was estimated using the amount and type of beverages used and classified in to the WHO drinking categories as follows: abstainers, category I (mild) including women with an alcohol consumption of <20 g/d or men with <40 g/d, and category II (moderate) including women with an alcohol consumption of 20-39.99 g/d or men with 40-59.99 g/d and category III (heavy) including women with an alcohol consumption of ≥40 g/d or men with ≥60 g/d.
Blood samples were donated, and height, weight and systolic blood pressure (SBP) measures were taken as part of the health assessments during the baseline examination in the recruiting centres.
In addition to self-reported chronic diseases and major cardiovascular events in the touchscreen interview at baseline, GP diagnosis data were used to complete diagnoses as good as possible up to the baseline date for this analysis, which was 16 March 2020 (first recorded positive SARS-CoV-2 test in the UK Biobank study population). The comorbid conditions assessed were chronic obstructive pulmonary disease, diabetes mellitus, heart failure, coronary heart disease (CHD), history of myocardial infarction (MI) and history of stroke.
To specify the use of low-dose aspirin, lipid-lowering drugs and number of drugs concurrently used in participants only the GP data were used. The time window to find users of low-dose aspirin and lipid-lowering drugs was defined as 6 months prior to March 2020.
For the total number of drugs used by each participant, a shorter, 3-month period prior to March 2020 was considered. Prescriptions with the same ATC code in this interval were only counted once.

| Statistical analysis
Multivariable logistic regression models were fitted and odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were estimated for the risk of Covid-19 hospitalization and 28-day all-cause mortality associated with drug exposures of interest. Covid-19 hospitalization was addressed in the total population of patients with hypertension and 28-day all-cause mortality was evaluated in 2 subpopulations of hypertensive patients: (i) those who tested positive for SARS-CoV-2 and (ii) those who were hospitalized due to Covid-19.
In a first round of analyses, ARB, CCB, BB and diuretic users were directly compared to ACEI users and in a second round of analyses the latter 3 were directly compared to ARB users as the reference group. To increase the statistical power, in the third round of analyses, we combined users of either CCBs, BBs and diuretics as 1 exposure group and compared it first to ACEI users and second to ARB users.
Moreover, unadjusted Kaplan-Meier curves were generated for the third round of analyses and log-rank tests were applied to test for different survival probabilities between the patients who received different classes of antihypertensives.
Patients receiving combinations of other antihypertensive drug classes and ACEIs or ARBs were only assigned to the ACEIs or ARBs users group, respectively. In sensitivity analysis, we removed patients that used drug combinations with ACEIs or ARBs from the respective analyses using 1 of these drug groups as the reference.
All models were first adjusted for age, sex and ethnic background only (simple model), and second for all potential confounders available in the UK Biobank (full model), which included age, sex, ethnic background, socioeconomic deprivation, smoking status, physical activity, alcohol consumption, body mass index (BMI), SBP, eGFR, chronic obstructive pulmonary disease, diabetes mellitus, heart failure, CHD, history of MI, history of stroke, use of low-dose aspirin, use of lipidlowering drugs and number of drugs concurrently used.
The linearity assumption of the continuous co-variables age, BMI, SBP, socioeconomic deprivation index and number of drugs was checked for the outcome "Covid-19 hospitalization in the total population" by modelling restricted cubic splines, 22 and the respective curves are shown in Figures TABLE S1-S5. As the linearity assumption was violated for all variables, these were modelled with the categories shown in Table 1 Table TABLE S1). Analyses of the 5 imputed datasets were combined using the SAS procedure PROC MIANALYZE. Two-sided P values <.05 were considered significant.

| ACEIs and Covid-19 outcomes
In the total population, 1015 (0.8%) were hospitalized due to Covid-   As summarized in Table 2 analyses are shown in Figure 1 and log-rank tests came the same conclusion as the logistic regression model: A significant association with 28-day mortality among all Covid-19 patients (log-rank P = .0002) but not among hospitalized Covid-19 patients (log-rank P = .15).
Stratified by SARS-Cov2 wave, associations were weaker for the first wave (Table S2) and stronger for the second wave (  Table 3 shows the same analyses as As already seen for ACEIs, associations were also weaker for SARS-Cov2 first wave (Table S4) and stronger for second wave (Table S5) in analyses with ARBs as the reference. The results for second wave were comparable to those reported for first and second waves combined, and the same associations were statistically significant.

| Interaction of ARB use and CHD
The only interaction detected with statistical significance after correction for multiple testing was between ARB use (with reference to CCB, BB or diuretics use) and CHD with respect to the outcome Covid-19 hospitalization adjusted for the variables of the full model (P = .0002). An analysis stratified by CHD showed that users of CCB, BB or diuretics with a diagnosis of CHD have a $3-fold higher Covid-19 hospitalization risk than ARB users with CHD, whereas the correspoding risk of users of CCB, BB or diuretics without CHD is only 32% higher than that of ARB users without CHD (Table S8). This difference by CHD diagnosis was not observed for ACEI use and Covid-19 hospitalization risk (Table S8).

| Sensitivity analyses
After removal of patients who used drug combinations with ACEIs or ARBs from the respective analyses using 1 of these drug groups as the reference, the results remained similar to the main analyses and showed unchanged conclusions in every case (Tables S6 and S7