All-cause and in-hospital mortality after aspirin use in patients hospitalized with COVID-19: a systematic review and meta-analysis

Abstract Background With the results of the largest randomized controlled trial (RECOVERY) and the most extensive retrospective cohort study on coronavirus disease 2019 (COVID-19) recently published, we performed a meta-analysis on the association of aspirin with mortality of COVID-19. We aimed to investigate the role of aspirin in COVID-19 hospitalizations. Materials and Methods We searched PubMed, EMBASE and Cochrane databases for studies from 1 January 2020 until 20 July 2022, that compared aspirin versus non-aspirin use in hospitalized COVID-19 patients. We excluded case reports, review articles and studies on non-hospitalized COVID-19 infections. We used the inverse variance method and random effects model to pool the individual studies. Results Ten observational studies and one randomized controlled trial met the criteria for inclusion. There were 136 695 total patients, of which 27 168 were in the aspirin group and 109 527 were in the non-aspirin group. Aspirin use was associated with a 14% decrease in all-cause mortality compared with non-aspirin use in patients hospitalized with COVID-19 [relative risk (RR) 0.86, confidence interval (95% CI) 0.76–0.97; P = 0.002; I2 =64%]. Among subgroups of studies reporting in-hospital mortality in COVID-19 hospitalizations, aspirin use was associated with a 16% decrease in in-hospital mortality compared with non-aspirin use (RR 0.84, 95% CI 0.71–0.99; P = 0.007; I2 =64%). Conclusion Our study shows that aspirin decreases in-hospital mortality in patients hospitalized with COVID-19. Further studies are needed to assess which COVID-19 patient populations benefit most, such as patients on aspirin for primary versus secondary prevention of atherosclerotic disease. In addition, significant bleeding also needs to be considered when assessing the risk–benefit of aspirin use.


Introduction
The novel coronavirus disease 2019 (COVID-19) since first reported in November 2019 has emerged into a pandemic and resulted in more than one million deaths in the USA alone as of 20 July 2022 [1]. Although most patients with COVID-19 have mild symptoms, the mortality rate for hospitalized patients remains high [1][2][3][4][5]. Since the emergence of COVID-19, systemic corticosteroids for 7-10 days in patients with severe and critical COVID-19 (requiring mechanical ventilation or oxygen support), with a conditional recommendation not to use corticosteroid therapy in patients with non-severe COVID-19 (not requiring respiratory support or oxygen), were provided by WHO in September 2020 [1,6].
Due to severe inflammatory response and hypercoagulability, the risk of thromboembolic events in COVID-19 is reported to be higher when compared to other acute medical illnesses or viral respiratory infections and is associated with a worse prognosis [7]. Immune dysregulation with systemic inflammation (especially interleukin-6 along with complement activation) and thrombosis has been proposed for the pathogenesis of severe COVID-19 [8,9]. With platelet activation and sequestration in critical illnesses, the benefits of antiplatelet therapy secondary to the inhibition of platelet activation and accumulation have been studied extensively. Aspirin due to its antiplatelet and antiviral effects has demonstrated a reduction in replication, propagation and infectivity of many RNA viruses such as MERS-CoV and CoV-229 E in both in-vitro and experimental models [4,[9][10][11][12].
Although aspirin was not among the guideline-recommended treatment for COVID-19, several observational studies along with one large randomized controlled trial (RCT) RECOVERY studied the beneficial effects of aspirin use on mortality in hospitalized patients with COVID-19 [3,4,[11][12][13][14][15]. The first PILOT study, the Collaborative Registry to Understand the Sequelae of Harm in COVID-19 revealed that a combined exposure of pre-hospital and in-hospital aspirin use within the first 24 h of admission led to a decrease in in-hospital mortality [4]. In the Randomized Evaluation of COVID-19 Therapy in a RECOVERY trial, aspirin was not found to be associated with a reduction in 28-day mortality in patients hospitalized with COVID-19. Still, there was a small increase in the rate of being discharged alive within 28 days [3]. Following the publication of the RECOVERY trial [3] and a large observational cohort study [11], we conducted a meta-analysis to assess further the association between aspirin and all-cause mortality with the subgroup of in-hospital mortality in hospitalized patients with COVID-19.

Materials and methods
We included RCTs, quasi-experimental and retrospective cohort studies that reported hazard ratio (HR), odds ratio (OR) or relative risk (RR) of the effect of aspirin on all-cause mortality in patients hospitalized with COVID-19. We independently screened the manuscripts/full papers, abstracts or titles of the studies from the electronic search to identify all potentially eligible studies and extracted data from PubMed/MEDLINE, Web of Science, Embase and Google Scholar from 1 January 2020 until 20 July 2022, that fulfilled the eligibility criteria with no language restrictions, using the search terms ('aspirin' or 'acetylsalicylic acid') and ('COVID-19' or 'Novel Corona Virus Disease 2019' or 'SARS COVID-19 Infection' or 'Coronavirus Disease 2019 Virus' or 'SARS-CoV-2 Infection'). The results from Google scholar are combined with the results of PubMed/MEDLINE because all Google scholar articles were accessed using Pubmed.gov.
Eligible studies compared the use of aspirin versus no aspirin in patients hospitalized with COVID-19 and reported all-cause mortality. We excluded case reports, case series, review articles and studies on non-hospitalized COVID-19 infections. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for reporting systematic reviews as recommended by the Cochrane Collaboration was followed [16]. Search results were saved in EndNote version X9 (Developer: Clarivate analysis) files. We extracted the data manually through a full-text review. Two reviewers (N.B. and P.S.) independently performed the title, abstract and full-text screening. Conflicts were resolved through consensus; if not, the third author (M.A.) resolved the dispute.
We used the Newcastle-Ottawa Scale to assess the quality of observational studies [17]. This scale assigns a maximum of nine points. We scored four for selecting and evaluating exposure, two for comparability and three for assessing the outcome. If a study receives a score of six or higher, then it is considered a highquality publication with a low risk of bias [17]. The RCT was assessed for quality using the Cochrane risk of bias tool [18]. This tool covers six domains of bias: selection bias, performance bias, detection bias, attrition bias, reporting bias and other biases.
For the individual biases in the tool, the assessment of each bias is based on two parts-Support for Judgement and Review of authors' judgement. The Support for judgement provides a summary of characteristics of the trial based on which the risk of each bias is determined and thus the transparency of the judgement is maintained. The second part of the tool involves assigning a judgement of high, low or unclear risk of material bias for each item [18]. The outcome of interest was all-cause mortality. We used the HR or OR or RR, depending on the studies, for the effect measure to generate the pooled risk ratio. We used adjusted HR or OR whenever they were reported. As mentioned in the previous studies, we directly incorporated HR as RRs while creating the forest plot [19][20][21]. We have used the formula RR ¼ OR/ f(1ÀP0) þ (P0 Â OR)g, to transform ORs into RRs. In the formula, P0 is the incidence of the outcome of interest in the non-exposed group [19][20][21]. To further calculate the upper and lower confidence interval (CI), we used the formula standard of error logarithmic (SElog) (RR) ¼ SElog (OR) Â log (RR)/log (OR) [19][20][21]. The analyses were performed using Review Manager 5.4 statistical software (Cochrane Collaboration, Oxford, UK) with an inverse variance method. We assessed the pooled RR and 95% CI using the random effect model. In studies like systematic review and meta-analysis, a treatment effect across various studies is investigated and the effects of treatments will not be the same across all populations [22]. This variation in the effectiveness of treatments is referred to as treatment effect heterogeneity. The I 2 statistic is used as a measure to assess the amount of treatment effect heterogeneity [22]. Sensitivity analysis was performed with the exclusion of individual studies to look for changes in the outcome. We used a statistical significance threshold of P <0.05. All studies reported all-cause mortality as in-hospital, in-hospital 28-day mortality, 30-day mortality, 60-day mortality or overall mortality. The included studies' baseline characteristics are listed in Table 1.

Study selection
Ten observational studies and one RCT met the criteria for inclusion. There were 136 695 total patients, of which 27 168 were in the aspirin group and 109 527 were in the non-aspirin group. Aspirin use was associated with a 14% decrease in all-cause mortality compared with non-aspirin use in patients hospitalized with COVID-19 (RR 0.86, 95% CI 0.76-0.97; P ¼ 0.002; I 2 ¼64%) ( Figure 2).
Among subgroups of studies reporting in-hospital mortality in COVID-19 hospitalizations, aspirin use was associated with a 16% decrease in in-hospital mortality compared with non-aspirin use (RR 0.84, 95% CI 0.71-0.99; P ¼ 0.007; I 2 ¼64%). However, aspirin was not associated with a statistically significant decrease in mortality compared with non-aspirin in a subgroup of studies that included out-of-hospital mortality after hospitalization for COVID-19 (RR 0.56, 95% CI 0.22-1.41; P ¼ 0.05, I 2 ¼67%).

Sensitivity analysis
The sensitivity analysis showed no change in statistical significance of ORs on aspirin's role in preventing all-cause mortality in COVID-19 hospitalized patients with the exclusion of any individual studies. However, with the combined exclusion of Chow et al. [4], Aghajani et al. [25] and Meizlish et al. [13] studies, the OR was statistically non-significant. However, there was no change in the overall beneficial role of aspirin with the exclusion of other studies. We used the exclusion method of individual studies and calculated the pooled RR.

Publication bias and heterogeneity
The funnel plot in our meta-analysis shows asymmetry in the distribution of the included studies due to the high heterogeneity of the included studies. The blue dotted line represents the OR with 95% CI (Figure 3).

Quality of included studies
The quality of included observational studies is shown in Table 2, and the quality of included RCT is shown in Figure 4.
After using the Newcastle-Ottawa Scale, we found that all the included studies truly had aspirin as the exposure and were selected using appropriate definition of exposure in hospitalized COVID-19 patients [17]. The non-exposed were non-aspirin exposed group with COVID-19 hospitalization. The study also controlled for confounders and 7 of the 10 studies had follow-up long enough for the outcome to occur. All the included studies had scores more than or equal to six which prove that the study was of high quality. Figure 4 depicts the assessment of the risk of bias in the RCT included in our study using the Cochrane risk of bias tool [18]. As mentioned in the Materials and methods section, the summary is colour coded with green representing a low risk of bias, yellowunclear risk of bias (if insufficient detail is reported of what happened in the trial, the judgement will usually be an unclear risk of bias) and red-high risk of bias. Selection bias has two domains-random sequence generation and allocation concealment. These domains in the selection bias describe the methods implemented for randomization to produce comparable groups for the study and concealment of intervention allocation in the comparison groups. Performance bias describes the measures used for blinding the trial participants and researchers regarding the intervention participants received. Detection bias describes measures adopted to blind the outcomes assessment. Attrition bias describes the completeness of the outcome data including the reasons for attrition and exclusions from the analysis. Reporting bias states selective outcome reported in the trial, if any. Other biases illustrate any other important concerns that are not specified in the other domains of the risk of bias tool [18].

Discussion
Our study investigated the role of aspirin on all-cause mortality in COVID-19 hospitalizations, including a subgroup of the role of aspirin in in-hospital mortality of the recently published RECOVERY trial [3] and a large observational study by Chow et al. [11]. Our results show that hospitalized COVID-19 patients taking aspirin had lower all-cause mortality compared with those not taking aspirin. Notably, the RECOVERY trial did not demonstrate improvement in 28-day mortality with aspirin use in hospitalized COVID-19 patients; however, the proportion of patients discharged alive within 28 days was higher in patients who received aspirin (75% versus 74%, rate ratio: 1.06, 95% CI 1.02-1.10; Role of aspirin in COVID-19 hospitalization | 3 P ¼ 0.006) [3]. This finding is similar to our study where aspirin improved in-hospital mortality in patients hospitalized with COVID-19.
COVID-19 contributes to a prothrombotic and hypercoagulable state [8,9]. Increased production of interleukins (IL-6, IL-10) and coagulopathy lead to high fatality rates in hospitalized COVID-19 patients [28,29]. Studies have shown systemic anticoagulation's benefits in reducing mortality in mechanically ventilated patients [30,31]. Aspirin has anti-inflammatory, antiplatelet and antiviral effects which have been shown in both in-vitro and experimental models to reduce replication, propagation and infectivity of many RNA viruses such as MERS-CoV and CoV-229 E [4,[9][10][11][12]. Hence, aspirin was studied as one of the therapeutic options in patients with COVID-19.
While previous meta-analyses have been conducted [15,[32][33][34][35], none have been restricted to hospitalized COVID-19 patients and the outcome of in-hospital mortality. Namely, Osborne et al. [34] conducted a large observational study showing improved mortality in COVID-19 with aspirin; however, the patients were enrolled in the Veterans Affairs health system and were not limited to hospitalized COVID-19 patients.
Our study's results on all-cause mortality in hospitalized COVID-19 patients align with the previous meta-analyses [32][33][34][35]; however, our study is unique in reporting both in-hospital mortality and overall mortality among these patients. Moreover, our study includes updated data with the inclusion of a recent large observational study by Chow et al. [11]. In an earlier study by Chow et al., of the 412 patients hospitalized with COVID-19, low-dose aspirin use was associated with decreased need for mechanical ventilation and intensive care unit admission as well as in-hospital mortality after multivariable adjustment [4]. Similarly, a single-centre study by Liu et al. showed that low-dose aspirin prevents embolic events in patients infected with COVID-19 while decreasing mortality [12]. The studies by Yuan et al. [14], Alamdari et al. [26], Formiga et al. [24] and Vahedian-Azimi et al. [23] were the observational studies in our meta-analysis that did not show improved all-cause mortality. Alamdari et al. [26] conducted a retrospective cross-sectional study with a higher risk of bias and lack of adjustment for potential confounders, which may have contributed to different results. Yuan et al. [14] investigated pre-hospitalization use of low-dose aspirin in COVID-19 patients with coronary artery disease. The group with prehospitalization use of aspirin who continued aspirin in the hospital may have thus been sicker than the non-aspirin cohort leading to higher in-hospital mortality. Moreover, the smaller sample size, differences in comorbidities and non-generalizable populations like Spanish, Italian and Iranian populations could have led to different outcomes [23][24][25][26][27][28].
Our study has several limitations. We had only one RCT; the remaining 10 were observational studies, with 1 cross-sectional study prone to unmeasured confounders. In most of the included studies, the severity of COVID-19 infection is not mentioned; however, all patients required hospitalization, meeting the criteria for moderate-to-severe COVID-19. We also did not report other outcomes like bleeding, which may occur with aspirin. With meta-analyses, there is always a concern for reporting biases such as selective outcome reporting.
Nonetheless, included studies used propensity scores to reduce confounding and selection bias, and we used adjusted HRs for accurate results from the included cohort studies. There was significant heterogeneity among the studies. Furthermore, patients taking aspirin in observational studies are generally more likely to have cardiovascular disease, which may place them at higher risk of mortality, which may have reduced the mortality benefit seen in our study. Finally, the included studies were single-centre studies or registry data from China, Spain, Iran, Italy or the USA, except for the RECOVERY trial [3]. This limitation can affect the generalizability of the study to other ethnic groups.

Conclusion
Our study shows that aspirin decreases in-hospital mortality in patients hospitalized with COVID-19. Further studies are needed to assess which COVID-19 patient populations benefit most, such as patients on aspirin for primary versus secondary prevention of atherosclerotic disease. In addition, significant bleeding also needs to be considered when assessing the risk-benefit of aspirin use.

Funding
None.
Conflict of interest statement. None of the authors have any direct or indirect conflict of interests to report, that might raise the question of bias in the work reported or the conclusions, implications or opinions stated, in the publication of this manuscript.
bleeding among hospitalized patients with coronavirus disease