Cardiovascular drugs and COVID‐19 clinical outcomes: a systematic review and meta‐analysis of randomized controlled trials

Aims: To update our previously reported systematic review and meta‐analysis of observational studies on cardiovascular drug exposure and COVID‐19 clinical outcomes by focusing on newly published randomized controlled trials (RCTs). Methods: More than 500 databases were searched between 1 November 2020 and 2 October 2021 to identify RCTs that were published after our baseline review. One reviewer extracted data with other reviewers verifying the extracted data for accuracy and completeness. Results: After screening 22 414 records, we included 24 and 21 RCTs in the qualitative and quantitative syntheses, respectively. The most investigated drug classes were angiotensin‐converting enzyme inhibitors (ACEIs)/angiotensin receptor blocker (ARBs) and anticoagulants, investigated by 10 and 11 studies respectively. In meta‐analyses, ACEI/ARBs did not affect hospitalization length (mean difference −0.42, 95% confidence interval [CI] −1.83; 0.98 d, n = 1183), COVID‐19 severity (risk ratio/RR 0.90, 95% CI 0.71; 1.15, n = 1661) or mortality (risk ratio [RR] 0.92, 95% CI 0.58; 1.47, n = 1646). Therapeutic anticoagulation also had no effect (hospitalization length mean difference −0.29, 95% CI −1.13 to 0.56 d, n = 1449; severity RR 0.86, 95% CI 0.70; 1.04, n = 2696; and, mortality RR 0.93, 95% CI 0.77; 1.13, n = 5689). Other investigated drug classes were antiplatelets (aspirin, 2 trials), antithrombotics (sulodexide, 1 trial), calcium channel blockers (amlodipine, 1 trial) and lipid‐modifying drugs (atorvastatin, 1 trial). Conclusion: Moderate‐ to high‐certainty RCT evidence suggests that cardiovascular drugs such as ACEIs/ARBs are not associated with poor COVID‐19 outcomes, and should therefore not be discontinued. These cardiovascular drugs should also not be initiated to treat or prevent COVID‐19 unless they are needed for an underlying currently approved therapeutic indication.


| INTRODUCTION
Cardiovascular diseases, mainly ischaemic heart disease, stroke and heart failure, were the leading causes of global mortality in 2017, accounting for approximately 17.8 million deaths. 1 By contrast, the coronavirus disease 2019 (COVID-19) pandemic 2,3 has killed >5.5 million people (out of approximately 335 million infected) as of 19 January 2022. 4 Due to the possible bidirectional interaction between COVID-19 and cardiovascular disease, 5,6 we conducted a baseline systematic review and meta-analysis 7 to evaluate the available evidence on the association between cardiovascular drug exposure and COVID-19 clinical outcomes. Additionally, and because COVID-19 related evidence is rapidly evolving, we planned for periodic updating for up to 2 years to incorporate any novel evidence.
In the baseline review (search date 1 November 2020), we included 429 and 390 studies in the qualitative and quantitative syntheses, respectively, with the majority of these being observational studies (only 2 randomized control trials [RCTs]). 7 In the adjusted estimates, angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin receptor blockers (ARBs), the most commonly reported drug classes, were not associated with COVID-19 infectivity (odds ratio [OR] 0.92, 95% confidence interval [CI] 0.71 to 1.19), hospitalization (OR 0.93, 95% CI 0.70 to 1.24), severity (OR 1.05, 95% CI 0.81 to 1.38) or mortality (OR 0.84, 95% CI 0.70 to 1.00). However, and even though adjustment may account for some confounders present in observational studies, it does not account for nonmeasured confounders and pooling of adjusted estimates may be problematic. 8,9 For more reliable evidence, therefore, a decision was made a priori to focus on RCTs, the current gold standard of evidence, 10 in this update.
In this update, we report RCT evidence for 7 cardiovascular drug classes/subclasses, namely: angiotensin-converting enzyme inhibitors (ACEIs), angiotensin receptor blockers (ARBs), anticoagulants, antiplatelets, antithrombotics, calcium channel blockers (CCBs) and lipid-modifying drugs (LMDs). Although evidence is still emerging, all these drug classes/subclasses have been implicated in modulating the outcomes of COVID-19. Specifically, it has been suggested that ACEIs and ARBs, through their respective modes of action, increase the expression of angiotensin-converting enzyme 2 (ACE2). 11,12 ACE2 converts angiotensin (Ang) II into the vasodilator and antitrophic heptapeptide, Ang-(1-7), which exerts protective effects on the lung and cardiac vasculature making it potentially beneficial in COVID-19 patients. 5,12,13 Although both ACEIs and ARBs increase ACE2 expression, Ang II levels are decreased with ACEIs, 12 which means there will be less substrate for ACE2 to convert into Ang- (1)(2)(3)(4)(5)(6)(7). Consequently, ARBs may be better than ACEIs at attenuating inflammation and acute lung injury in COVID-19 patients. 14 In contrast to these protective effects, SARS-CoV-2, the virus that causes COVID-19, uses ACE2 to enter target cells, 15 which led to the hypothesis that ACEIs/ARBs could modulate COVID-19 disease outcomes. COVID-19 patients have increased haemostatic and thrombotic risks-these have been postulated to originate from the development of a cytokine storm after SARS-CoV-2 infection, followed by hyperinflammation, endothelial disruption, platelet activation and coagulopathy among other mechanisms. 16,17 By helping avert haemostatic and thrombotic complications, anticoagulants, antiplatelets and other antithrombotic agents could therefore positively impact COVID-19 clinical outcomes such as hospitalization, severity and mortality. By contrast, CCBs may interfere with SARS-CoV-2 replication by reducing intracellular calcium levels, and the resulting anti-inflammatory (reduction of COVID-19 related inflammation), anticoagulatory (reduction of microvascular coagulation) and vasodilatory (improvement of local vasoconstriction) effects, may decrease COVID-19 severity and associated mortality. 18 Lastly, LMDs have also been reported to possess antiviral, immunomodulatory, anti-inflammatory and antithrombotic properties, 19,20 which may all contribute to better COVID-19 clinical outcomes.

| METHODS
We followed a predefined protocol (PROSPERO: CRD42020191283 21 ) as previously reported. 7 This manuscript follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines 22 (Table S1).

| Identification of studies
As previously detailed, 7,21,23 we searched >500 databases (including MEDLINE and Scopus) through the University of Liverpool's DIS-COVER platform, preprint servers, and COVID-19 specific databases/ registries, but this time focused on RCTs published/posted between 1 November 2020 and 2 October 2021. To the DISCOVER search strategy (previously only included medical subject headings and text words related to "cardiovascular drugs" and "COVID-19"), we added additional terms (random* OR RCT* OR [clinical AND trial*]) to further limit the search results. The DISCOVER search results were uploaded in EndNote (version X9) 24 and studies were de-duplicated based on title, author, year of publication and reference type information. We also hand-searched the lists of references from the identified studies and previous systematic reviews, although no additional eligible articles were identified. Any additional studies published after 2 October 2021 and referred to us (e.g. by experts) before data analysis were also included.

| Selection criteria
We included RCTs that investigated the association between cardiovascular drug exposure (key drug classes were derived from Chapter 2 ["Cardiovascular system"] of the British National Formulary 25 as previously outlined 7 ) and the COVID-19 clinical outcomes listed below.
We excluded non-English studies but did not exclude any studies based on where and when they were published.

| Assessment of study quality
I.G.A. used the revised Cochrane risk-of-bias tool for randomized trials 26 to assess the quality of each included study estimate, with S.P./ R.M.T. verifying accuracy, and disagreements being resolved by consensus.

| Data synthesis
The R meta package 27 (R version 3.6.1) was used to pool count data (preferred to summary estimates such as the odds ratios, as it is more accurate 8 ) in cases where 2 or more studies reported on the same exposure-outcome combination. The pooling was done using random-effects meta-analysis with the inverse-variance method for effect size and the DerSimonian-Laird estimator for variance. 27 For the dichotomous outcomes (infectivity, hospitalization, severity and mortality), we generated risk ratios (RR; with 95% confidence intervals [CIs]) as odds ratios are not easily interpretable 28 and case control studies were not included (in which RRs, unlike odds ratios, are not applicable). Mean differences (with 95% confidence intervals) were generated for hospitalization length, the only continuous outcome. Where median values and ranges/ interquartile ranges of hospitalization length were reported, we used them to estimate the means and standard deviations. 29,30 Means and standard deviations could also be combined using formulae available in the Cochrane Handbook. 30 Lastly, we prepared Forest plots for each exposure-outcome combination and narratively reported/tabulated the studies that singly reported on an exposure-outcome combination, as part of the qualitative synthesis.

| Heterogeneity measures
As previously reported, 7 the magnitude of the inconsistency in the study results was assessed by visually examining forest plots and considering the I 2 statistic (arbitrarily defined heterogeneity extent categories were: I 2 < 30%, low; I 2 = 30-70%, moderate; and I 2 > 70%, high). Estimates with high heterogeneity were deemed to be inconsistent and would result in a downgrading of the strength of evidence. 31

| Publication bias
We did not assess publication bias since there were fewer than 10 studies for each of the reported exposure-outcome combinations.

| Subgroup and sensitivity analyses
We conducted subgroup analyses based on drug subclasses, study quality (only studies with low risk of bias included) and the hypertension comorbidity, which was guided by our earlier findings. 23 Where a study reported different estimates for the same outcome domain (for example due to different follow-up periods), we conducted sensitivity analyses to determine the impact of the inclusion of 1 estimate, instead of the other(s).

| Confidence in cumulative evidence
We used the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) 31

| Study selection and characteristics
After screening 22 414 titles and/or abstracts, 35 full-text records were assessed for eligibility, of which 24 and 21 were included in the qualitative and quantitative syntheses, respectively ( Figure 1). Table 1 shows the characteristics of the included studies. Of the 24 26 all study estimates were rated as having a low risk of bias except for estimates from 6 studies 33,34,38,41,42,51 (Table S2). Table 1 summarises the main results for all 24 studies, while   Table 2 provides summary results for the studies included in the quantitative synthesis/meta-analysis. Table 2 also includes the GRADE strength of evidence rating, 31 which ranges from moderate to high since all meta-analyses included at least 1 large properly conducted RCT.
Hospitalization was also investigated by only 1 study (117 patients), 37 and compared to placebo, newly-initiated losartan (an ARB, 25 mg once or twice daily, depending on the estimated glomerular filtration rate) given for 10 days, increased the number of hospitalizations (5.2 vs. 1.7%, absolute difference of 3.5%) although this was not statistically significant (95% CI À4.8 to 13.2%, P = .320).
Seven studies 34,36,38,[40][41][42]57 investigated whether ACEI/ARB exposure (both treatment-experienced and -naïve patients) affected length of hospitalization. One study, 38 reported only medians and a hazard ratio and could therefore not be included in the primary metaanalysis. For 6 studies 34,36,[40][41][42]57 (n = 1183 patients), ACEIs/ARBs did not influence the duration of hospitalization (mean difference À0.42, 95% CI À1.83; 0.98 d, I 2 = 51%, Figure 2 moderate, blood oxygen saturation <94%, or lung infiltrates >50%, or ratio of partial pressure of arterial oxygen to fraction of inspired oxygen <300; and severe, invasive mechanical ventilation or haemodynamic instability or multiple organ dysfunction or failure. d Mild disease includes cases not meeting the criteria for classification as moderate or severe disease; moderate disease was characterised by an oxygen saturation <94%, pulmonary infiltrates >50%, or a partial pressure of oxygen to fractional concentration of oxygen in inspired air ratio <300; and severe disease was defined as respiratory failure, haemodynamic instability, or multiple organ dysfunction. e  moderate, blood oxygen saturation <94%, or lung infiltrates >50%, or ratio of partial pressure of arterial oxygen to fraction of inspired oxygen <300; and severe, invasive mechanical ventilation or haemodynamic instability or multiple organ dysfunction or failure. d Mild disease includes cases not meeting the criteria for classification as moderate or severe disease; moderate disease was characterised by an oxygen saturation <94%, pulmonary infiltrates >50%, or a partial pressure of oxygen to fractional concentration of oxygen in inspired air ratio <300; and severe disease was defined as respiratory failure, haemodynamic instability, or multiple organ dysfunction. e Based on the GRADE rating. 31 Evidence from RCTs starts at a high rating and can be downgraded to moderate, low or very low based on risk of bias, imprecision, inconsistency (I 2 > 70% threshold used), indirectness and publication bias. None of the estimates were downgraded due to including studies with high risk of bias since conclusions remained unchanged when these studies were excluded. F I G U R E 2 Forest plots for associations between angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin receptor blockers (ARBs) and COVID-19 outcomes. a Total admission days, reported as means (standard deviations), used. When the 'length of hospital stay', means (standard deviations) estimated from the reported medians (interquartile range), were used, the subgroup pooled mean difference was 0.41 (95% confidence interval [CI] À1.02 to 1.84, I 2 = 65%) days while the overall pooled mean difference was À0.06 (95% CI À1.36 to 1.24, I 2 = 57%) days. b Used intubations to define severity as this was more consistent with the rest of the studies. When severity was defined as the requirement for supplemental oxygen, the subgroup and overall pooled risk ratios were 0.74 (95% CI = 0.46 to 1.18, I 2 = 44%) and 0.91 (95% CI 0.78 to 1.07, I 2 = 23%) respectively. c Used admission to the intensive care unit and/or the requirement for mechanical ventilation to define severity as this was more consistent with the rest of the studies. When severity was defined based on the World Health Organisation COVID-19 ordinal severity scale, the subgroup and overall pooled risk ratios were 0.93 (0.79 to 1.10, I 2 = 2%) and 0.93 (95% CI 0.75 to 1.14, I 2 = 24%) respectively. d Follow up of 30 days, which was preferred as this was more consistent with the rest of the studies. When 90-day follow up was used, the subgroup and overall pooled risk ratios were 0.50 (95% CI 0.20 to 1.29, I 2 = 48%) and 0.89 (95% CI 0.57 to 1.40, I 2 = 31%) respectively. e 14-day follow up. Risk of bias domains: 1 = risk of bias arising from the randomization process; 2 = risk of bias due to deviations from the intended interventions (effect of adhering to intervention); 3 = risk of bias due to missing outcome data; 4 = risk of bias in measurement of the outcome; 5 = risk of bias in selection of the reported result; 6 = overall risk of bias. Colour codes: green = low risk; yellow = some concerns; red = high risk 6 studies (111 patients) that investigated new initiation of ARBs and compared to amlodipine (a CCB) 34 or standard of care, 36 did not show an effect on hospitalization duration (mean difference À2.32, 95% CI À4.81 to 0.16 d, I 2 = 0%, Figure 2). Conversely, 4 [40][41][42]57 out of the 6 studies (1072 patients) investigated whether discontinuation of ACEI/ARB would influence COVID-19 outcomes. Again, there was no effect of continuation vs. discontinuation of ACEI/ARBs (mean difference in duration of hospitalisation 0.07, 95% CI À1.59 to 1.73 d,  Nine trials 34,[36][37][38][39][40][41][42]57 (1646 patients) investigated the association between ACEI/ARB exposure (both treatment-experienced and -naïve patients) and mortality, although 1 study had 0 events. 37 The results were not statistically significant (pooled RR 0.92, 95% CI 0.58 to 1.47, Figure 2)

| Antiplatelet agents
Aspirin, used as an antiplatelet drug, was investigated by 2 studies, 48

| Antithrombotics
In addition to anticoagulants and antiplatelets discussed above separately, sulodexide, an antithrombotic with both anticoagulant and antiplatelet activity (given for 21 d) was investigated and it was protective in terms of the risk of hospitalization (RR 0.60, 95% CI 0.37 to 0.96, P = .03), although the trial was small with only 243 patients included. 56 Additionally, it had no effect on hospitalization length (6.3

| DISCUSSION
To update the previously reported associations between cardiovascular drug exposure and COVID-19 clinical outcomes, 7  Other investigated drugs included anticoagulants, aspirin (an antiplatelet), sulodexide (an antithrombotic with both anticoagulant and antiplatelet activity), amlodipine (a CCB) and atorvastatin (an LMD), although only anticoagulants could be included in the primary meta-analyses, with therapeutic anticoagulation not affecting hospitalization, hospitalization length, severity and mortality outcomes. Although the number of studies included in the meta-analyses were small (2, 4, 2 and 9 for hospitalization, hospitalization length, severity and mortality respectively), many of these studies were platform-based, meaning they could rapidly recruit many patients and the meta-analyses were therefore well-sized (805, 1449, 2696 and 5689 for the respective outcomes). Additionally, most of these trials were rated to have a low risk of bias which led to a high strength of evidence for all the outcomes, except hospitalization, which was ranked as moderate. However, and despite this ranking, it is important to emphasize that we did not explore subgroup analyses based on patient characteristics with the exception of the hypertension comorbidity. We also limited our review to the 5 prespecified outcomes, meaning it is possible that therapeutic anticoagulation is beneficial for specific patients based on different clinical endpoints. This was demonstrated by the ATTACC, ACTIV-4A and REMAP-CAP Investigators 50,52 who showed that therapeutic anticoagulation improved the primary outcome of organ support-free days in noncritically ill

| Limitations of this review
The main limitation of this review is the inclusion of relatively few RCTs, despite our comprehensive search strategy, which meant that many drug classes (including β-blockers, CCBs, diuretics and LMDs) could not be quantitatively synthesized. We were also unable to assess publication bias as we required a minimum of 10 RCTs for each exposure-outcome combination. Nevertheless, this is a recently emerging field and many more RCTs are expected, which will be included in future updates. We did not search trial registries (such as  61 We also relied on single-reviewer extraction. However, 2 other reviewers verified the accuracy and completeness of the extracted data. In addition to not assessing some efficacy endpoints that are being reported in the literature, we did not assess some safety outcomes (e.g. treatment discontinuation due to bleeding with anticoagulation); however, these were outside the scope of this review. Lastly, and despite the randomization of participants to treatment, residual confounding may exist especially if sample sizes are small 28 or responseadaptive randomization is used, 52 which would require the pooling of adjusted estimates. However, most of the studies included in the meta-analyses were relatively large and well-balanced, a reason we rated the strength of evidence as moderate to high.

| CONCLUSIONS
Moderate-to high-certainty RCT evidence suggests that ACEIs/ARBs and therapeutic anticoagulation are not associated with poor COVID-19 clinical outcomes. However, the routine use of therapeutic anticoagulation is questionable as it offers limited benefits over placebo or standard/prophylactic anticoagulation. There are currently many ongoing RCTs that are expected to be completed/published in the coming months and will be incorporated in our next update, which will be conducted within 6 months of this update. As we wait for more evidence, we suggest that patients with COVID-19 on cardiovascular drugs should not discontinue taking them as it is very unlikely that these drugs, specifically ACEIs/ARBs, are harmful.