Summary

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) uses the angiotensin-converting enzyme 2 (ACE2) as the receptor for its spike protein to invade cells and replicate.1 ACE2 presents a high homology with ACE, a key enzyme in the regulation of blood pressure.2 In some animal studies, renin–angiotensin–aldosterone system (RAAS) inhibitors (a category that includes ACE inhibitors and angiotensin-receptor blockers) have been reported to increase expression of ACE2.3, 4, 5 These findings have led some researchers to postulate that the use of these drugs might enhance the access of SARS-CoV-2 into cells, predisposing patients to infection or increasing severity of COVID-19.6, 7, 8 This hypothesis was fuelled by results from the first case series that was published in which age, hypertension, diabetes, and coronary heart disease—conditions associated with the use of RAAS inhibitors—were identified as potential risk factors for severe cases and in-hospital deaths.9, 10, 11, 12 By contrast, other authors have proposed use of angiotensin-receptor blockers as a preventive measure, or even a therapy, for COVID-19 because of their potential to reduce lung injury caused by angiotensin II.13 RAAS inhibitors are among the most widely used drugs globally for indications such as hypertension, heart failure, kidney complications of diabetes, and myocardial infarction; hence their discontinuation because of COVID-19 could cause patients harm.14 Scientific societies and drug regulatory agencies alike have advised against their discontinuation until sound evidence is available.15

To provide a solution to this urgent issue, we designed a pharmacoepidemiological study with the aim of assessing whether the odds of exposure to RAAS inhibitors relative to other antihypertensive drugs was higher among patients with COVID-19 admitted to hospital than in the general population, adjusted for age, sex, and cardiovascular comorbidities and risk factors.

Methods

Study design and participants

In this case-population study, we used an epidemiological approach, conceived as a surveillance method to assess adverse drug effects,16 to provide a rapid answer to our study question. Briefly, this method uses data from a series of patients with an illness from a specific region (cases) and data from patients randomly sampled from a primary health-care database in the same region (population controls). Assuming that the primary health-care database represents the source population of the cases, a random sample of controls from that database would provide a valid estimate of the prevalence of the exposure and covariates in the source population, approaching the primary base paradigm of case-control studies.17

Here, we collected data on patients (cases) aged 18 years or older with a PCR-confirmed diagnosis of COVID-19 who were admitted to seven hospitals in Madrid, Spain, between March 1 and March 24, 2020. During the study period, the Madrid region was in the upward part of the outbreak, with an average of 465 inpatient admissions, 48 intensive care unit (ICU) admissions, and 76 deaths per day associated with COVID-19 (official data are available online). The seven participating hospitals belong to the National Health System (NHS) and are distributed all over the Autonomous Community of Madrid (appendix p 2). Cases were selected consecutively according to the date of admission to hospital until the planned sample size was met in a competitive manner. For our population controls, we used a random sample of ten individuals per case, individually matched to our cases by sex, age (exact), region (ie, Madrid), and date (day and month), extracted from the last available year (2018) from the primary health-care database Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria (BIFAP; figure 1 ).

Open in a separate window

Figure 1

Case-population study design

COVID-19 cases requiring admission to hospital were selected consecutively from seven hospitals in Madrid, Spain. Data were collected for ten individuals per case who were matched for age, sex, and index date (day and month) of hospital admission of cases (matched controls) from the 2018 Madrid region database of BIFAP, a national primary health-care database. Drug exposure and comorbidities before the index date (2020 for cases and 2018 for controls) were collected from primary health-care records of the NHS in Madrid: for cases through HORUS (an online platform to access primary-care clinical records from any health-care centre of the NHS in Madrid) and for controls through BIFAP. BIFAP=Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria. NHS=National Health System.

The Ethics Research Committee of the University Hospital Príncipe de Asturias (the coordinating centre) assessed the study protocol and granted a favourable opinion on March 18, 2020, including a waiver for the informed consent of patients taking part in the study. The data extracted were fully anonymised and no attempt was made to interview patients or their relatives. The study complied with the provisions of the Spanish legislation and the Declaration of Helsinki 2013. The Scientific Committee of BIFAP granted access to pseudonymised data for controls on March 26, 2020. The study protocol is registered in the EU electronic Register of Post-Authorisation Studies, EUPAS34437, and is available online.

Results

We collected data on 1139 cases and 11 390 matched controls. Despite being matched by age and sex, cases had a greater prevalence of diverse comorbidities than did controls (table 1 ).

The prevalence of use of different subgroups of antihypertensive drugs was higher in cases than in controls, which yielded positive associations with risk of COVID-19 requiring admission to hospital in both crude and fully adjusted analyses compared with non-use of any antihypertensive drug (adjusted OR for RAAS inhibitors 1·71, 95% CI 1·46–2·01; and for other antihypertensive drugs 1·82, 1·47–2·26; appendix p 9). When use of other antihypertensive drugs was set as the reference category, the adjusted OR associated with the current use of RAAS inhibitors was 0·94 (0·77–1·15; table 2 ). No significant increase in the risk of COVID-19 requiring admission to hospital was observed with either ACE inhibitors or angiotensin-receptor blockers, nor did we find any difference when these drugs were used in monotherapy or in combination with other drugs (table 2).

The adjusted ORs associated with other antihypertensive drugs compared with non-use by different pharmacological subgroups were 1·96 (95% CI 1·43–2·69) for calcium-channel blockers, 1·79 (1·32–2·43) for diuretics, 1·68 (1·16–2·43) for β-blockers, and 1·96 (0·57–6·71) for α-blockers (appendix p 9). When the current use of calcium-channel blockers was set as the reference category, the adjusted OR was 0·87 (0·65–1·18) for RAAS inhibitors, 0·74 (0·54–1·02) for ACE inhibitors, and 1·02 (0·74–1·40) for angiotensin-receptor blockers (appendix p 10).

Compared with long-term (longer than 1 year) users of other antihypertensive drugs, long-term users of RAAS inhibitors had an adjusted OR of 0·96 (95% CI 0·76–1·21) for risk of COVID-19 requiring admission to hospital. Likewise, compared with short-term users (up to 1 year) of other antihypertensive drugs, short-term users of RAAS inhibitors had an adjusted OR of 1·39 (95% CI 0·92–2·10).

The results of the assessment of the potential effect modification of RAAS inhibitors by age (<70 and ≥70 years), sex, diabetes, hypertension, and background cardiovascular risk are shown in figure 2 . No significant interaction was observed with any variable, except for diabetes, for which a significantly reduced risk of COVID-19 requiring admission to hospital associated with RAAS inhibitors was observed (adjusted OR 0·53; 0·34–0·80; test of interaction, p=0·004; still significant after applying the Bonferroni correction for multiple testing [corrected significance level fixed at 0·008]).

Open in a separate window

Figure 2

Association between current use of RAAS inhibitors and risk of COVID-19 requiring admission to hospital compared with current use of other antihypertensive drugs, stratified by different variables

Number (%) of cases and controls given is the number exposed to RAAS inhibitors in each stratum and the resulting odds ratio, adjusted for the matching variables and other comorbidities different from the one examined. RAAS=renin–angiotensin–aldosterone system.

393 (34·5%) of 1139 cases had the most severe form of COVID-19: 283 (24·8%) died in hospital and 110 (9·7%) survived but required admission to an ICU. The mean duration of hospital stay was 13 days (SD 8) among survivors and 11 days (SD 7) among non-survivors. Broadly, compared with patients with less severe COVID-19 in our study, patients with the most severe COVID-19 were older (mean age 75·3 years [SD 12·3] vs 65·8 years [SD 15·9]; p<0·0001) and a higher proportion were male (264 [67·2%] of 393 with severe disease vs 431 [57·8%] of 746 with less severe disease; p=0·001; table 3 ). Compared with current use of other antihypertensive drugs, the adjusted OR associated with the current use of RAAS inhibitors was 1·08 (95% CI 0·80–1·47) among the most severe cases and 0·86 (0·66–1·11) among the less severe cases (table 3). In both severity groups, the 95% CIs for the adjusted ORs of ACE inhibitors and angiotensin-receptor blockers compared with other antihypertensive drugs overlapped with each other, indicating no significant differences (table 3).

In our first sensitivity analysis, the adjusted OR of current use of RAAS inhibitors compared with current use of other antihypertensive drugs corrected for secular trends was 0·90 (95% CI 0·74–1·11), and 0·76 (0·61–0·95) for ACE inhibitors and 1·09 (0·87–1·36) for angiotensin-receptor blockers. In our second sensitivity analysis, up to March 16, the adjusted OR associated with current use of RAAS inhibitors compared with current use of other antihypertensive drugs was 0·98 (95% CI 0·77–1·25) and for March 17 onwards was 0·88 (0·62–1·24; figure 2). In our third sensitivity analysis, the exclusion of aldosterone antagonists and renin inhibitors from the RAAS inhibitor group hardly had any effect (overall adjusted OR 0·92, 95% CI 0·76–1·12; among the most severe cases, 1·06, 0·78–1·44; and among the less severe cases, 0·84, 0·65–1·09; appendix pp 11–12). And in our final sensitivity analysis, the inclusion of hypertension in the multivariable model had little effect on the adjusted ORs for RAAS inhibitors (1·00, 0·82–1·23), ACE inhibitors (0·85, 0·68–1·06), or angiotensin-receptor blockers (1·17, 0·93–1·47).

Discussion

Here we show that the current use of RAAS inhibitors is not associated with an increased risk of COVID-19 requiring admission to hospital (including fatal cases and those admitted to an ICU) compared with other antihypertensive drugs. No substantial difference was observed between ACE inhibitors and angiotensin-receptor blockers, nor among short-term and long-term users. Sex, age, and background cardiovascular risk did not significantly affect the results, although use of RAAS inhibitors was associated with a reduced risk of COVID-19 requiring admission to hospital in patients with diabetes.

Our results do not support the hypothesis that previous intake of RAAS inhibitors facilitates or increases the severity of COVID-19. This hypothesis emerged when two facts were linked: the finding that RAAS inhibitors upregulate the expression of ACE2,3, 4, 5, 19 and the observation that hypertension, diabetes, and ischaemic heart disease were highly prevalent among patients with severe COVID-19,9, 10, 11, 12 conditions for which RAAS inhibitors are widely used. However, these facts are only two pieces of a more complex puzzle. Evidence to support the idea that ACE2 might have a dual role in COVID-19 is increasing.20 On the one hand, ACE2 overexpression might increase the susceptibility of cells to SARS-CoV-2, but, on the other hand, its downregulation associated with older age, male sex, and cardiovascular comorbidities, and further heightened by SARS-CoV-2 binding and internalisation21, 22 could increase the unopposed action of angiotensin II and have a key role in the subsequent organ injury.23 ACE2 counteracts the deleterious effect of RAAS axis. In patients with COVID-19, Liu and colleagues24 reported that serum concentrations of angiotensin II were significantly higher in infected individuals than non-infected individuals and was linearly associated with viral load and lung damage. ACE2 inactivates angiotensin II and increases the generation of angiotensin 1–7, a peptide that, acting on the Mas receptor, exerts a vasodilatory effect and anti-inflammatory and antioxidative actions.20 In patients with acute respiratory distress syndrome (ARDS), Reddy and colleagues25 observed a higher ratio of angiotensin 1–7:angiotensin I among survivors than non-survivors, which adds evidence to the idea that the counter-regulation exerted by the axis of ACE2–angiotensin 1–7–Mas receptor benefits patients with ARDS. In line with this theory, recombinant soluble ACE2 has been shown in animal models of ARDS to protect subjects from lung injury,26, 27, 28 and clinical trials are underway in patients with COVID-19 (NCT04287686). Finally, to our knowledge, the case series of patients with COVID-19 published to date has not been adjusted for important potential confounding factors, such as sex, age, and cardiovascular comorbidities.29 In our study, we used controls who were matched to cases for sex and age, and we also adjusted for a number of comorbidities.

To our knowledge, two epidemiological studies have been published to date that aimed to explore the association between RAAS inhibitors and COVID-19 comparing severe cases with less severe cases among inpatients.30, 31 Neither of these studies found an increased risk of severe outcomes associated with these drugs, and in one study30 a substantially reduced risk of death and transfer to a critical care unit within 7 days of admission to hospital was observed among users of ACE inhibitors (OR 0·29, 95% CI 0·10–0·75). Notably, in both studies, the authors considered not only exposure to ACE inhibitors before (7 days) or at admission to hospital, but also during hospital stay. In our study we used population controls, not just a case series, and only considered drugs used before admission.

Several cardiovascular comorbidities are clearly linked to severe COVID-19, and RAAS inhibitors are often used to treat such cardiovascular conditions; thus substantial confounding by indication can be expected.32 In our view, this bias explains the positive association we found between current use of any antihypertensive drug and risk of COVID-19 requiring admission to hospital compared with non-use of such drugs, which did not disappear after full adjustment for comorbidities. This problem can be mitigated if current users of RAAS inhibitors are compared with current users of other drugs that share, at least partially, their indications, but not their safety issues.32 For instance, some researchers have recommended calcium-channel blockers as a suitable alternative treatment to RAAS inhibitors, because calcium-channel blockers have not been reported to increase the expression of ACE2.8 Following this reasoning, other antihypertensive drugs in general, and calcium-channel blockers in particular, can be considered a more valid comparator group than non-use of such drugs. We followed this approach and reassuringly we found no increased risk when using either of these drug groups as the reference group. Nevertheless, RAAS inhibitors do not share all indications with other antihypertensive drugs; hence, the comorbidity pattern associated with use of these drugs might not overlap and some residual confounding by indication could remain. To explore the extent of this possibility, we examined the comorbidity pattern associated with various types of antihypertensive drugs among controls and found much fewer differences among the types of drug than when they were compared with non-users, which reinforces the validity of our approach (appendix p 13).

The risk reduction associated with RAAS inhibitors among individuals with diabetes who had been admitted to hospital for COVID-19 (not explained by chance) deserves a comment. Diabetes has been reported as a risk factor for severe COVID-19.33 The biological underpinning for this clinical observation is not yet known, but experimental models of diabetes in mice have shown that ACE activity is high in the lungs.34 If this activity also occurs in humans, patients with diabetes who have COVID-19 might present a great imbalance in the ACE:ACE2 ratio (ACE activity would be high due to diabetes and ACE2 would be low due to downregulation induced by SARS-CoV-2) that could explain the increased severity of COVID-19 among patients with diabetes and also the protective effect of RAAS inhibitors suggested by our data.

The lack of an increased risk of COVID-19 requiring admission to hospital associated with RAAS inhibitors (compared with other antihypertensive drugs) was found in both patients with most severe and less severe disease, suggesting that the outpatient use of these drugs neither facilitates nor aggravates the infection.

Our study has several limitations. First, we used different data sources to extract information from cases (hospital records and primary health-care records through HORUS, consulted case by case by local researchers) and controls (BIFAP database, which automatically extracted data from primary health-care records); however, both HORUS and BIFAP access the same primary health-care data, which was the main source for determining comorbidity and drug exposure in our study, making major information bias unlikely. Second, cases were recorded as of March, 2020, whereas controls were sourced from records in March, 2018, which might have affected the results due to secular trends in use of antihypertensive drugs; however, when we corrected for secular trends we saw little change in the estimated adjusted OR. Third, we did not collect data on smoking and other lifestyle habits (eg, exercise, diet, alcohol intake), which might have affected the association between RAAS inhibitors and COVID-19 requiring admission to hospital; however, these habits are correlated with several comorbidities that were included in the regression model. Fourth, we obtained information on drug prescriptions, but adherence to treatment cannot be guaranteed. Hence, when we have used the term drug use, we should say drug prescribed. The effect of non-adherence in our results is difficult to measure but would most probably cause a non-differential misclassification of exposure that, if relevant, would distort the estimates towards the null value. Fifth, we did not collect information on clinical and analytical covariates during patients' stay in hospital, nor did we collect information on in-hospital treatment (eg, if RAAS inhibitors were continued or withheld in hospital), which might have been interesting for the assessment of the association of antihypertensive drugs with severity of and fatality due to COVID-19. Sixth, we did not analyse any dose effects. Finally, as in any other observational study, residual confounding due to unmeasured or unknown confounders cannot be ruled out.

In summary, the results of the present study suggest that the outpatient use of RAAS inhibitors does not increase the risk of COVID-19 requiring admission to hospital, including its most severe forms. This finding should be confirmed using other data sources, study designs, and populations. Meanwhile, the data available, along with the important role of ACE inhibitors and angiotensin-receptor blockers in the management of several cardiovascular diseases, do not support their discontinuation as a preventive measure against COVID-19.

Data sharing

After publication, study data will be made available on reasonable request to the corresponding author. A proposal with a detailed description of study objectives and statistical analysis plan will be needed for assessment of requests. Additional materials might also be required during the process of assessment. De-identified participant data will be provided after approval from the principal researchers of the participating hospitals.

Acknowledgments

Contributors

MED-ACE2-COVID19 study group

Declaration of interests

Contributor Information

MED-ACE2-COVID19 study group^†:

References