key: cord-0907084-cqblm8j1
authors: Sandhu, Alexander T; Kohsaka, Shun; Lin, Shoutzu; Woo, Christopher Y; Goldstein, Mary K.; Heidenreich, Paul A
title: Renin–Angiotensin–Aldosterone System Inhibitors and SARS-CoV-2 Infection: An Analysis from the Veteran's Affairs Healthcare System: Sandhu. ACEI, ARB, and Association with COVID.
date: 2021-06-12
journal: Am Heart J
DOI: 10.1016/j.ahj.2021.06.004
sha: 446d341b5591b87ab74f22dff64a4dc4d15d5e74
doc_id: 907084
cord_uid: cqblm8j1

Background: Angiotensin-converting enzyme inhibitors (ACEI) and angiotensin receptor blockers (ARB) are known to impact the functional receptor for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The association between chronic therapy with these medications and infection risk remains unclear. Objectives: The objective was to determine the association between prior ACEI or ARB therapy and SARS-CoV-2 infection among patients with hypertension in the U.S. Veteran's Affairs health system. Methods: We compared the odds of SARS-CoV-2 infection among three groups: patients treated with ACEI, treated with ARB, or treated with alternate first-line anti-hypertensives without ACEI/ARB. We excluded patients with alternate indications for ACEI or ARB therapy. We performed an augmented inverse propensity weighted analysis with adjustment for demographics, region, comorbidities, vitals, and laboratory values. Results: Among 1,724,723 patients with treated hypertension, 659,180 were treated with ACEI, 310,651 with ARB, and 754,892 with neither. Before weighting, patients treated with ACEI or ARB were more likely to be diabetic and use more anti-hypertensives. There were 13,278 SARS-CoV-2 infections (0.8%) between February 12, 2020 and August 19, 2020. Patients treated with ACEI had lower odds of SARS-CoV-2 infection (odds ratio [OR] 0.93; 95% CI: 0.89-0.97) while those treated with ARB had similar odds (OR 1.02; 95% CI: 0.96-1.07) compared with patients treated with alternate first-line anti-hypertensives without ACEI/ARB. In falsification analyses, patients on ACEI did not have a difference in their odds of unrelated outcomes. Conclusions: Our results suggest the safety of continuing ACEI and ARB therapy. The association between ACEI therapy and lower odds of SARS-CoV-2 infection requires further investigation.

Angiotensin-converting enzyme inhibitors (ACEI) and angiotensin receptor blockers (ARBs), two of the 65 most commonly prescribed chronic therapies for hypertension, have garnered widespread interest as they 66 may impact the incidence or severity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) 67

infection. [1] [2] [3] [4] Angiotensin-converting enzyme 2 receptor (ACE2) is the cellular receptor for the SARS-68

CoV-2 spike protein and is present in lungs. 1 Select animal models have shown ACEI and ARB therapy 69

are associated with increased tissue ACE2 levels, although this has not been demonstrated in lung tissue 70 or in humans. 5 However, this has sparked concerns that ACEI or ARB therapy may increase the risk of 71

developing SARS-CoV-2 infection. 6,7 72 73 To date, multiple observational studies have found mixed results regarding the association between 74 chronic ACEI/ARB utilization and coronavirus disease 2019 (COVID-19) outcomes. [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] Early studies, 75

including reports from Wuhan, China, demonstrated that hypertension was common among patients with 76

SARS-CoV-2 infection and associated with worse outcomes, 4,22 leading to speculation that the increased 77 risk may be related to ACEI or ARB treatment. Others have hypothesized that ACEI or ARB treatment 78 may prevent the lung injury of SARS-CoV-2 given prior data that ACEI or ARB therapy may reduce lung 79 injury with acute respiratory infection. 23, 24 However, for most studies, the relatively modest sample sizes 80

have yielded inadequate precision to detect small but potentially meaningful effects. Additionally, most 81 studies included patients without hypertension or with alternate indications for ACEI/ARB therapy, which 82 may introduce selection bias if the conditions treated by ACEI/ARB impact SARS-CoV-2 infection risk . 83 Finally, studies that evaluated patients being tested or those infectedas opposed to a population-based 84 cohort -may inaccurately estimate the effect on outcomes if ACEI/ARB therapy also impact the risk of 85

contracting the disease or developing symptoms. Therefore, there remains limited population-based US 86 data evaluating the association between ACEI/ARB therapy and SARS-CoV-2 infection. 87

With over 10% of the adult population over 40 taking ACEI therapy, understanding the impact on SARS-89

CoV-2 infection is critical. 25 We performed an observational cohort analysis among Veterans Affairs 90

(VA) patients with a diagnosis of hypertension on at least one of the five most common anti-hypertensive 91 therapy classes (ACEI, ARB, calcium channel blocker [CCB], thiazide diuretic, or beta-blocker To be included, we required a medication fill for one of the five most common anti-hypertensives -ACEI, 118

ARB, thiazide diuretics, CCBs, or beta-blockers -within the 6 months before February 12, 2020. The 119

first-line drug requirement identified a more homogeneous population receiving standard treatment to 120

improve the comparability of ACEI/ARB users and non-users. We excluded patients without therapy to 121 reduce differences in hypertension severity or healthcare utilization. 122

We excluded patients with alternate first-line indications for ACEI/ARB therapy: heart failure with 124 reduced ejection fraction (HF-rEF), diabetic nephropathy, renal artery stenosis, or non-diabetic advanced 125 chronic kidney disease (estimated glomerular filtration rate below 60 ml/min) (diagnosis codes listed in 126 Supplement We adjusted for medical comorbidities potentially associated with the risk of COVID-19 or likelihood of 166 ACEI/ARB therapy using vital signs, diagnoses, medications, and lab values. Vital signs included most 167 recent heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, and oxygen saturation 168

within 1 year before the index date. Body mass index (BMI) was calculated from height and weight. Non-169 physiologic vital signs were dropped (detailed in Supplement Table 1 ). We adjusted for comorbidities 170 using diagnostic codes within the prior 2 years: chronic kidney disease, chronic obstructive pulmonary 171 disease, diabetes, heart failure, ischemic heart disease, liver disease, other pulmonary disease, 172 malignancy, and prior stroke (codes listed in Supplement Table 2 ). In addition, we used pharmacy records 173

to measure insulin fills within 6 months before the index date. We included common laboratory tests 174 previously found to be associated with disease morbidity across conditions including potentially COVID- We assumed missing vital signs, labs, and BMI were missing at random and used multiple imputation 189

using chained equations with linear regression to construct 40 imputed datasets. 44 We included all model 190

variables in the imputation model. We added indicator variables for missing values given missingness 191 may be informative. 45 We performed multiple sensitivity analyses around this approach. First, we dropped 192 vitals, labs, and BMI from the analysis. Second, we performed a complete case analysis. Third, we 193

repeated the multiple imputation but assumed missing variables were more extreme than predicted (1 194 standard deviation higher or lower risk than predicted The primary analysis was a comparison of the adjusted odds of SARS-CoV-2 infection between patients 201 treated with ACEI, ARB, or neither. We used an augmented inverse-propensity weighted logistic 202 regression model adjusted for patient demographics, geography, social risk, comorbidities, vital signs, and 203 laboratory values. 46 We calculated stabilized propensity weights using a multinomial logistic regression 204 that evaluated the propensity to be in each arm based on the same characteristics listed above. 47 We used 205 robust standard errors. We then repeated the analysis for a second model that directly compared ACEI 206

and ARB therapy. 207 208 As a secondary outcome, we evaluated the association between ACEI, ARB, or neither therapy and 209

SARS-CoV-2-associated hospitalization. We used the hospitalization definition described above. 210

We performed stratified subgroup analyses based on age, race/ethnicity, AHRQ SES Index, diabetes, and 212

ischemic heart disease. We divided continuous characteristics based on the median value. For 213 race/ethnicity, we only included Black, Hispanic, and White given the small sample size and limited 214 interpretability of the "Other" category. We re-estimated the propensity weights within each subgroup to 215 reduce potential bias. 48 With two subgroups, we tested the heterogeneity of odds ratios by calculating log 216 odds ratios and determining the Z statistic given we ran two independent models. This analysis raised the concern that risk may differ across the therapies in our control arm. To evaluate 229 this, we performed an exploratory analysis with the following exposure arms: ACEI without ARB/CCB, 230

ARB without ACEI/CCB, CCB without ACEI/ARB, and thiazide diuretic or beta-blocker without 231 ACEI/ARB/CCB. We recalculated our propensity weights with four potential outcomes. 232 233 Third, we stratified our exposure into two groups: chronic versus only recent exposure. We defined 234 chronic exposure as a medication fill both within 6 months before the index data and between 6-18 235 months while those with only recent exposure did not have a medication fill between 6-18 months. We 236 excluded patients who switched exposures between the recent and prior exposure periods. We repeated 237 our analysis in each subgroup. The average patient age was 67.4 (SD 12.0 years) ( Table 1 ). The ACEI and ARB cohorts 260 received more anti-hypertensive medication classes than the control cohort with neither. The largest 261 source of imbalance between cohorts was diabetes status (Figure 2 ). Both ACEI and ARB cohorts had 262

higher prevalence of diabetes and insulin utilization. We evaluated the association between ACEI and ARB therapy with alternate outcomes (Supplement 299 Figure 2 ). We find no significant association between ACEI therapy and diagnoses of influenza, fractures, 300

or UTIs. However, we did find a significant association between ARBs and lower odds of both Finally, we repeated our primary analysis with multiple sensitivity analyses around our approach to 311 missing data. We found no substantial difference in our results across these analyses (results listed in 312

Supplement Table 6 ). 313

Among VA patients with medically treated hypertension, neither ACEI nor ARB therapy were associated 316

with higher odds of SARS-CoV-2 infection. ACEI therapy was associated with lower odds of SARS-317

CoV-2 infection compared with alternate anti-hypertensives. In an exploratory analysis in which we 318 separated our control population into those on CCB versus other anti-hypertensives, CCBs were 319 associated with higher odds of SARS-CoV-2 infection compared with patients on ACEI/ARB/other anti-320

hypertensives. Overall, our results support existing evidence and society guidelines that neither ACEI nor 321 ARB therapy should be discontinued due to concerns of SARS-CoV-2 infection. 52,53 322 323 Multiple studies have evaluated the association between ACEI/ARB therapy and COVID-19 diagnosis.

Most studies have found no significant association between either therapy and COVID-19 diagnoses. 325

However, there are important differences both across existing studies and with ours with regards to the 326 study design and results. Several studies evaluated test positivity among a population being tested or 327 evaluated severe adverse outcomes among those hospitalized for SARS-CoV-2. These designs introduce 328 potential collider bias if the exposure -ACEI or ARBinfluence the likelihood of having symptomatic 329 infection leading to testing or hospitalization. Therefore, we evaluated a broader population. However, 330

our approach may also introduce collider bias if the likelihood of being tested is associated with anti-331 hypertensive therapy independent of symptoms (e.g. patients on ACEI being more likely to be tested 332

given the initial concerns about ACEIs ACEI/ARB therapy to non-users will lead to an imbalance in hypertension severity, which is important 349

given the potential relationship between hypertension and COVID-19 related outcomes. This imbalance is 350 also present in our study, which is why we adjusted for the number of anti-hypertensives and prior blood 351 pressure. 352

Two large studies found non-significant associations between ACEI/ARB therapy and COVID-19 related 354 outcomes but had substantially different point estimates across ACEIs and ARBs, potentially suggestive 355 of heterogeneity across these therapies. De Abajo and colleagues compared hospitalized COVID-19 356 patients with hypertension with matched controls without COVID-19 in Madrid, Spain. 11 Compared with 357 alternate anti-hypertensives, the adjusted odds ratio for COVID-19 hospitalization was 0.80 (95% CI: 358 0.64-1.00) for ACEIs compared with 1.10 (95% CI: 0.88-1.37) for ARBs. Fosbøl and colleagues 359 evaluated severe COVID-19 outcomes among hypertension patients without heart failure or chronic 360 kidney disease. 8 They found a hazard ratio of 0.85 (95% CI: 0.70-1.01) for ACEI therapy compared with 361

1.15 (95% CI: 0.96-1.37) for ARBs. We found a significant association between ACEI therapy and lower 362

odds of SARS-CoV-2 infection but not for ARB therapy. Interpreting our ARB results are challenging. 363

While the falsification endpoints were consistently negative for ACEI, we also found lower odds of 364 vertebral/hip fracture and UTI with ARB therapy. This is suggestive of residual confounding and 365 potential downward bias in our ARB estimates. 366

While most raised concerns that ACEI/ARB therapy may increase COVID-19 risk due to upregulated 368 ACE2 levels, there is also a biologic mechanism by which these therapies may reduce risk. 55 Angiotensin 369 II drives lung injury and inflammation. ACEIs decrease production of Angiotensin II and ARBs blocks its 370

effect. ACEIs and ARBs may have different effects on SARS-CoV-2 given they act via different 371 mechanisms and have differential effects on ACE2 levels. In animal models, losartan led to greater 372 increases in ACE2 levels compared with lisinopril, although this is not consistently observed with renal 373 ACE2 levels. 5 There is limited data regarding the impact on pulmonary ACE2 levels, the impact in human 374

tissue, or how changes in ACE2 levels impact SARS-CoV-2 infection and severity. 55 There are multiple important strengths of our analysis. First, this national US study includes a racially gi 397 population, which is a potential contributor to differences in outcomes, including those related to 398 comorbidities and socioeconomic characteristics. Second, we adjust important characteristics that are 399 often absent in administrative data: laboratory values and vital signs. We also adjusted for social risk, an 400 important factor in this pandemic. Third, our large sample allowed us to restrict our analysis to a 401 population with hypertension without other indications for ACEI/ARB therapy while still providing 402 precise estimates. is still potential confounding. Our ARB falsification analysis findings accentuate this concern. We 410 selected these falsification endpoints given an expectation that ACEI/ARB therapy would not affect 411

outcomes. An alternate explanation would be an improper selection of a priori falsification endpoints 412

given limited studies suggesting lower risk of fracture with ARB than ACEI therapy. 58 However, this has 413 not been seen with UTIs. 414

The second limitation is potential exposure measurement error given we classified exposure based on 416 medication fills within 6 months. This potentially classified patients as ACEI or ARB users despite being 417 non-adherent or having stopped therapy. However, most ACEI/ARB treatment is chronic as seen in our 418

analysis. Additionally, such bias would generally attenuate finding towards the null. We also have 419 potential measurement error in identifying infections given not all patients were tested for SARS-CoV-2 420 and the VA dataset may miss some infections outside of the VA. 421 422 Third, there is potential time-related bias because our study does not account for patients with less time at 423 risk for developing infection due to death without SARS-Cov2 infection. Finally, the risk of infection has 424 marked spatiotemporal variation. Small differences in regional utilization of ACEI/ARB may have 425 important effects. We controlled for patient region and geographic characteristics, but additional analyses 426

with more granular geographic data will be important. 427

We found ACEI therapy was associated with lower odds of SARS-CoV-2 infection in a VA population 431

with medically-treated hypertension. Given the current evidence and the clinical benefits of ACEI/ARB 432 therapy for multiple conditions, neither ongoing ACEI nor ARB therapy should be stopped due to risks of 433 developing COVID-19 infection. In an exploratory analysis, we found the odds of SARS-CoV-2 infection 434

were similar between ACEI, ARB, and beta-blocker/thiazide users but were higher among CCB users. 435

Further analyses to evaluate potential protective effects of ACEI therapy or an increase in risk with CCB 436 therapy will be important. propensity weighting. One imputed dataset was randomly selected among the 40 datasets to calculate 706 standardized mean differences after imputation and inverse propensity weighting. Multiple datasets were 707 compared with minimal difference in the standardized mean differences. Characteristics across each 708 cohort after inverse propensity weighting are detailed in Supplement Table 4 . Standardized mean 709 differences between ACEI and ARB cohorts before and after IP weighting are listed in Supplement Table  710 5. 

Rapid response: RE: Preventing a covid-19 pandemic: ACE inhibitors as a 452 potential risk factor for fatal Covid-19

455 Effect of angiotensin-converting enzyme inhibition and angiotensin II receptor blockers on 456 cardiac angiotensin-converting enzyme 2

Urinary angiotensin-converting enzyme 2 in hypertensive patients may be increased by 461 olmesartan, an angiotensin II receptor blocker

Risk Factors Associated With Acute Respiratory Distress Syndrome and Death in 466 Patients With Coronavirus Disease

COVID-19, ACE2, and the cardiovascular consequences

Can angiotensin receptor-blocking drugs perhaps be harmful in the COVID-19 472 pandemic?

System Inhibitors in Patients with Covid-19

Association of Angiotensin-Converting Enzyme 478 Inhibitor or Angiotensin Receptor Blocker Use With COVID-19 Diagnosis and Mortality

Risk of severe COVID-19 disease with ACE inhibitors and 481 angiotensin receptor blockers: cohort study including 8.3 million people. Heart

Aldosterone System Inhibitors and Risk of Covid-19

Use of renin-angiotensin-aldosterone system 491 inhibitors and risk of COVID-19 requiring admission to hospital: a case-population study

Renin-Angiotensin-Aldosterone System 494 Blockers and the Risk of Covid-19

Association of Use of Angiotensin-Converting Enzyme Inhibitors and Angiotensin II Receptor 499

COVID-19)

Association of Renin-angiotensin-aldosterone System 502 Inhibitors With Coronavirus Disease 2019 (COVID-19)-Related Outcomes in Korea: A 503 Nationwide Population-based Cohort Study

Angiotensin-converting enzyme inhibitors and 506 angiotensin-receptor blockers are not associated with increased risk of SARS-CoV-2 infection

Renin-angiotensin-aldosterone system 510 inhibitors and COVID-19 infection or hospitalization: a cohort study

Renin-Angiotensin-Aldosterone System Inhibitors and Risk of Death in Patients 515 Hospitalised with COVID-19: A Retrospective Italian Cohort Study of 43,000 Patients. Drug Saf

Association of Angiotensin-Converting Enzyme Inhibitors and Angiotensin 519

Receptor Blockers with the Risk of Hospitalization and Death in Hypertensive Patients with 520 Coronavirus Disease-19

Angiotensin-Converting Enzyme Inhibitors 524 or Angiotensin Receptor Blockers Use and COVID-19 Infection Among

Hypertension From a US Integrated Healthcare System

Covid-19 Testing, Hospital Admission, and Intensive Care Among 2,026,227 United States 530 Veterans Aged

Angiotensin II receptor blocker 536 or angiotensin-converting enzyme inhibitor use and COVID-19-related outcomes among US 537 Veterans

Clinical course and risk factors for mortality of 540 adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

COVID-19 and the cardiovascular system

Effects of 545 spironolactone and eprosartan on cardiac remodeling and angiotensin-converting enzyme 546 isoforms in rats with experimental heart failure

Prescription Drug Use Among Adults Aged 40-79 in the 549 United States and Canada

Department of Veterans Affairs, Office of Research and Development. COVID-19 Shared Data 551 Resource

COVID-19 Response Team

Evidence for Limited Early Spread of 557 COVID-19 Within the United States

ACCF/AHA guideline for the management of heart failure: a 564 report of the American College of Cardiology Foundation/American Heart Association Task 565 Force on Practice Guidelines

Management of High Blood Pressure in Adults: Executive 572 Summary: A Report of the American College of Cardiology/American Heart Association Task 573 Force on Clinical Practice Guidelines. Circulation

National trends in the ambulatory treatment of 576 hypertension in the United States

Use of Antihypertensive Agents and 579

Association With Risk of Adverse Outcomes in Chronic Kidney Disease: Focus on Angiotensin-580

Converting Enzyme Inhibitors and Angiotensin Receptor Blockers

Medical 584 Therapy for Heart Failure With Reduced Ejection Fraction: The CHAMP-HF Registry

Trends in Angiotensin-Converting Enzyme Inhibitor and 587

Angiotensin II Receptor Blocker Use among Those with Impaired Kidney Function in the United 588 States

Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus 591 pneumonia in Wuhan, China: a descriptive study

Hospitalization Rates and Characteristics of 598 Patients Hospitalized with Laboratory-Confirmed Coronavirus Disease 2019 -COVID-NET, 14 599 States

Factors associated with 605 COVID-19-related death using OpenSAFELY

Accuracy of Administrative Coding to Identify Reduced 608 and Preserved Left Ventricular Ejection Fraction

Creation of New Race-Ethnicity Codes and 613 Socioeconomic Status (SES) Indicators for Medicare Beneficiaries: Final Report

Concentration of Potentially 617 Preventable Spending Among High-Cost Medicare Subpopulations: An Observational Study

Haemoglobin A1c is a 620 predictor of COVID-19 severity in patients with diabetes

Anemia is 623 associated with severe illness in COVID-19: A retrospective cohort study

Association of 626 Diagnostic Coding-Based Frailty and Outcomes in Patients With Heart Failure: A Report From 627 the Veterans Affairs Health System

Multiple imputation using chained equations: Issues and guidance 630 for practice

Multiple imputation with missing indicators as proxies for unmeasured 632 variables: simulation study

Estimation of causal effects using propensity score weighting: an 635 application to data on right heart catheterization

Use of stabilized inverse propensity 638 scores as weights to directly estimate relative risk and its confidence intervals. Value Health

641 Comparison of propensity score methods for pre-specified subgroup analysis with survival data

Agency for Healthcare Research and Quality

Summary of common statistical approaches to test for heterogeneity

Antihypertensive Drugs and COVID-19 Risk: A Cohort Study 650 of 2 Million Hypertensive Patients. Hypertension

HFSA/ACC/AHA statement addresses concerns re: using RAAS 655 antagonists in COVID-19

Risks and 659 Impact of Angiotensin-Converting Enzyme Inhibitors or Angiotensin-Receptor Blockers on 660 SARS-CoV-2 Infection in Adults: A Living Systematic Review

Risk Factors of Fatal Outcome in Hospitalized 666 Subjects With Coronavirus Disease 2019 From a Nationwide Analysis in China

Controversies of renin-angiotensin 669 system inhibition during the COVID-19 pandemic

Effect 677 of Discontinuing vs Continuing Angiotensin-Converting Enzyme Inhibitors and Angiotensin II 678

Receptor Blockers on Days Alive and Out of the Hospital in Patients Admitted With COVID-19: 679 A Randomized Clinical Trial

Continuation versus 687 discontinuation of renin-angiotensin system inhibitors in patients admitted to hospital with 688 COVID-19: a prospective, randomised, open-label trial

Osteoporotic Fractures in Men (MrOS) Research Group. ARB 691 users exhibit a lower fracture incidence than ACE inhibitor users among older hypertensive men

Abbreviations: ACEI: angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blocker

Agency for Healthcare Research and Quality Socioeconomic Status; IHD: ischemic heart 717 disease

Only 721 recent exposure includes only patients with a medication fill within the prior 6 months without earlier 722 medication fills. Propensity weights were recalculated within each subgroup. We list the p-values for 723 significant differences in the odds ratios across subgroups. For characteristics with two subgroups, the 724 statistical significance of differences in odds ratios between the two groups were tested by calculating z-725 statistics for the difference in log(odds ratios). The odds ratios for race -which we divided

Abbreviations: ACEI: angiotensin-converting enzyme inhibitor; AHRQ SES: Agency for Healthcare 733 Research and Quality Socioeconomic Status; ARB: angiotensin receptor blocker; BMI: body mass index; 734 bpm: beats per minute; ICD: International Classification of Diseases; rpm: respirations per minute

* Categorical variables presented as percentages; continuous variables presented as mean (standard 737 deviation). Proportion of missing variables for vital signs, BMI, and laboratory values detailed in 738

† Defined based on ICD diagnosis codes as opposed to estimated glomerular filtration rate below

‡ Defined as the presence of at least two ICD diagnosis codes indicating potential frailty