Clinical Outcomes From COVID-19 Following Use of Angiotensin-Converting Enzyme Inhibitors or Angiotensin-Receptor Blockers Among Patients with Hypertension in South Korea: A nationwide study

There is ongoing debate as to whether angiotensin-converting enzyme inhibitors (ACEIs) or angiotensin-receptor blockers (ARBs) use is associated with poor prognosis of coronavirus disease-2019 (COVID-19). We sought to investigate the association between ACEI/ARB use and risk of poor clinical outcomes from COVID-19. We identified 1,290 patients with hypertension, of which 682 had recorded ACEI/ARB use and 608 without the use during 30 days preceding the date of COVID-19 diagnosis in completely enumerated COVID-19 cohort in South Korea. Our primary endpoint was the clinical outcomes comprised of all-cause mortality, use of mechanical ventilation, intensive care unit (ICU) admission, and sepsis. We used inverse probability of treatment weighting (IPTW) to mitigate selection bias, and Poisson regression model to estimate the relative risks (RR) and 95% confidence intervals (CI) to compare outcomes in ACEI/ARB users with non-users. Compared to non-use, ACEI/ARB use was associated with lower clinical outcomes (IPTW adjusted RR, 0.60; 95% CI, 0.42-0.85; p=0.0046). When assessed by individual outcomes, ACEI/ARB use was not associated with all-cause mortality (IPTW adjusted RR, 0.62; 95% CI, 0.35-1.09; p=0.0973) and respiratory events (IPTW adjusted RR, 0.99; 95% CI, 0.84-1.17; p=0.9043). Subgroup analysis showed a trend toward protective role of ACEIs and ARBs against overall outcomes in men (IPTW adjusted RR, 0.84; 95% CI, 0.69-1.03; p-for-interaction=0.008) and with pre-existing respiratory disease (IPTW adjusted RR, 0.74; 95% CI, 0.60-0.92; p-for-interaction=0.0023). We present clinical evidence to support continuing ACE/ARB use in completely enumerated hypertensive COVID-19 cohort in South Korea.


Introduction
The rapid spread of coronavirus disease 2019 (COVID-19), caused by severe acute 64 respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing pandemic, having infected 65 >2,800,000 people over 213 countries as of April 28, 2020 1 . Concerns have been raised 66 regarding poor prognosis of COVID-19 associated with the use of angiotensin-converting 67 enzyme inhibitors (ACEI) and angiotensin II receptor blockers (ARBs) 2 . ACEIs and ARBs 68 were shown to upregulate ACE2 expression and activity in several experimental studies [3][4][5] . 69 Given that the binding of ACE2 and viral spike protein of SARS-CoV-2 allows coronavirus 70 entry into host cells, it had been hypothesized that potential upregulation of ACE2 owing to 71 use of ARBs leads to an increased risk of severe and fatal health outcomes 2,6 . 72 In theory, patients with hypertension being managed with ACEIs or ARBs could be 73 at increased risk of COVID-19 since increased ACE2 by these drug classes may increase 74 viral entry into cells. Alternatively, increased ACE2 had been recognized to counterbalance 75 the pro-inflammatory and vasoconstrictive effect of ACE, mainly through conversion of 76 angiotensin II (Ang II) to Ang-(1-7), a peptide with potential protective anti-inflammatory 77 properties that counterbalances pro-inflammatory activity of Ang II [7][8][9] . 78 In view of these two opposing mechanistic hypotheses, demands for clinical research 79 remain very high. To date, several observational studies claimed that the use of ACEI/ARB 80 was not associated with increased all-cause mortality 10-14 . However, these studies were 81 limited due to methodological issues in study design, with the former lacking causal 82 relationship assessment and latter suffering from immortal time bias arising from 83 misclassification of exposure period. Given the lack of robust population-based study 84 assessing the association between the use of ACEI/ARB and sequelae of COVID-19, we 85 analyzed completely enumerated hypertensive COVID-19 cohort in South Korea to assess 86 COVID-19. 88 Methods 89 We retrieved healthcare database from the Health Insurance Review & Assessment 90 Service of South Korea, which covers over 50 million entire South Korean population, from 91 January 1, 2015 to April 8, 2020. We used completely enumerated database of 69,793 92 subjects who underwent COVID-19 screening test in South Korea. The database contains 93 both inpatient and outpatient prescriptions, demographic (age, sex, and insurance type) and 94 clinical information on visit dates for hospitalization and ambulatory cares, procedures, and 95 diagnosis records coded using Korea Standard Classification of Diseases, 7 th revision (K 96 CD-7), which is based on International Classification of Diseases, 10 th revision (ICD-10). 97 The overall agreement of diagnostic records of hypertension, stroke, and heart disease was 98 93.73%, 98.80%, and 97.93%, respectively 15 . 99 We conducted a retrospective cohort study of ACEI/ARB use and adverse outcomes 100 from COVID-19 among patients with hypertension. We identified patients with laboratory 101 confirmed diagnosis of COVID-19 between December 1, 2019 and April 8, 2020. Diagnoses 102 were made based on the diagnostic results from on reverse transcription polymerase chain 103 reaction (RT-PCR) method targeting the RNA-dependent RNA polymerase (RdRP), N, and E 104 genes as recommended by the interim guidance of World Health Organization 16 . Cohort 105 entry was defined as the date of incident diagnosis of COVID-19. We required patients to 106 have a recorded diagnosis of hypertension within a 5-year lookback period from the cohort 107 entry. 108 Exposure to ACEIs and ARBs was ascertained within 30 days preceding the cohort 109 entry. Our exposure of interest was patients ever prescribed with ACEIs or ARBs either as a 110 monotherapy or combination therapy. Non-users were those who had no prescription record 111 of either ACEIs or ARBs during the exposure ascertainment period.

112
. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 1, 2020. . clinical outcomes indicative of poor prognosis for   113   COVID-19, comprised of all-cause mortality, use of mechanical ventilation, intensive care   114 unit (ICU) admission, or sepsis. We also assessed all-cause mortality and respiratory events 115 (acute respiratory distress syndrome, interstitial lung disease, pneumonia, and respiratory 116 failure) individually as the secondary endpoints. Each patient was followed until the 117 occurrence of outcome of interest or data-censoring date. 118 We assessed baseline characteristics within 1 year before cohort entry.  (including antibiotics, antivirals, antimalarials (chloroquine and hydroxychloroquine), 128 corticosteroids, and intravenous immunoglobulin) were assessed but not included in the 129 propensity score model as they act as the intermediates between exposure and outcome of our 130 study. 131 We estimated propensity scores for receiving ACEIs or ARBs by fitting a 132 multivariable logistic regression model using all pre-defined covariates assessed 1 year 133 before cohort entry. We used inverse probability of treatment weighting (IPTW) based on the 134 propensity scores to mitigate selection bias by the characteristics between ACEI/ARB users 135 and non-users. IPTW creates a pseudo population, where the weighted exposure and 136 comparator groups are representative of the patient characteristics in the overall population, 137 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

Our study outcomes included
The copyright holder for this preprint this version posted August 1, 2020. . and thus generates the population average treatment effect 17 . We summarized baseline 138 characteristics of the study cohort using counts and proportions or mean for categorical or 139 continuous variables, respectively. We used Poisson regression model to estimate relative 140 risk (RR) and corresponding 95% confidence intervals (CI) of each outcome in ACEI/ARB 141 users compared to non-users in patients with COVID-19. The unweighted model was 142 adjusted for pre-defined covariates including age, sex, CHA DS -VASc score, diabetes, 143 CVD, and baseline respiratory diseases for parsimony. These covariates were also used in the 144 IPTW weighted models for doubly robust estimation of causal effect. We chose adjusted 145 IPTW weighted model as a main model to report the RRs and corresponding 95% CIs for the 146 clinical outcomes in ACEI/ARB users, compared with non-user. 147 Given that ACEIs and ARBs, apart from hypertension, are primarily prescribed for 148 patients with diabetes and CVD, we repeated analysis with the restricted cohort of patients 149 with these health conditions to exclude confounding by indication. To evaluate whether the 150 association differed by underlying patients' conditions, we conducted additional subgroup 151 analysis using the interaction terms by age group, sex, CHA DS -VASc score, pre-existing 152 respiratory disease, and hospitalization after diagnosis of COVID-19. In subgroup analysis, 153 we used overall outcomes which include all-cause mortality, use of mechanical ventilation, 154 ICU admission, sepsis, or the occurrence of respiratory events to increase the statistical 155 power. Additionally, we conducted sensitivity analysis where we redefined exposure 156 assessment window as within 180 days preceding the cohort entry to address potential 157 exposure misclassification. 158 Our study complies with the Declaration of Helsinki and the study protocol was 159 approved by the Institutional Review Board (IRB) of Sungkyunkwan University (SKKU 160 2020-03-021) and obtaining informed consent was waived by the IRB.

162
. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint patients with past medical history of hypertension, of which 682 had recorded use of 164 ACEI/ARB and 608 without the use during 30 days preceding cohort entry (Figure 1). The 165 characteristics of the ACEI/ARB users compared with non-users are described in Table 1 compared with non-use. 181 We conducted sensitivity analysis where we redefined the exposure risk window as 182 180 days preceding the cohort entry to account for potential exposure misclassification 183 ( Table 3). There were 28 (3.7%) and 23 (4.4%) cases of adverse outcomes in ACEI/ARB 184 users and non-users, respectively. The results from sensitivity analysis were generally 185 consistent with the main analysis; ACEI/ARB use was associated with lower clinical 186 outcomes (IPTW adjusted RR, 0.65; 95% CI, 0.46-0.90; p=0.0094), all-cause mortality 187 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . consistent with the findings from main analysis. 213 While the underlying pathogenic link between hypertension and COVID-19 remains 214 to be elucidated, concerns have been raised that ACEIs and ARBs, mainstay of therapy for 215 hypertension and diabetes, may contribute to the adverse outcomes observed in COVID-19 216 patients 2 . Indeed, interaction between SARS-CoV-2 and ACE2 was proposed a potential 217 mechanism for COVID-19's cell entry 18 13 . The most recent single tertiary center-based study 236 in the United States also reported null association between ACEI/ARB use and poor 237 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint outcomes from COVID-19 among 2,573 hypertensive COVID-19 patients, with the median 238 difference in percentage points between ACEI/ARB users and non-users of -0.5% (95% CI, - is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint underdiagnosis of COVID-19 is likely to be minimal. Third, our study results were consistent 264 in the subgroup analysis by pre-existing diabetes or CVD, which suggests the robustness of 265 our results from confounding by indication given that ACEIs and ARBs are primarily 266 prescribed for patients with these coexisting comorbidities. 267 Our study also has some limitations. First, potential misclassification of diagnosis- In conclusion, our study findings did not identify increased risk of adverse outcomes 285 with the use of ACEIs or ARBs among COVID-19 patients with hypertension. We present 286 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint clinical evidence to support current medical societies recommendation on continuing ACEIs 287 or ARBs as prescribed in COVID-19 patients.

288
. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. .  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. .  . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020.    . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 1, 2020. . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint Figure 2. Relative risks (95% confidence intervals) of overall clinical outcomes* in ACEI/ARB users compared to non-users in selected population subgroups Abbreviations: ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin II receptor blockers; IPTW, inverse probability of treatment weighting; CI, confidence interval *Overall clinical outcomes include all-cause mortality, use of mechanical ventilation, admission to intensive care unit, sepsis, or the occurrence of respiratory events. † Relative risks were adjusted for age, sex, CHA2DS2-VASc score, diabetes, cardiovascular . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 1, 2020. . https://doi.org/10.1101/2020.07.29.20164822 doi: medRxiv preprint disease, and respiratory disease. ‡ To make IPTW, age at cohort entry, sex, income level, CHA DS -VASc score, medical history (including diabetes, cardiovascular disease (CVD), stroke, other cerebrovascular disease, hyperlipidemia, respiratory disease, chronic kidney disease, cancer, thromboembolism, and dementia), comedications (including other antihypertensives (calcium channel blockers (CCBs), diuretics, . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted August 1, 2020. .