See Article, p 927
KEY POINTS
Question: Is aspirin use associated with less mechanical ventilation in coronavirus disease-2019 (COVID-19) patients?
Findings: In an observational cohort study of 412 adult patients with COVID-19, aspirin use was associated with a significantly lower rate of mechanical ventilation, intensive care unit (ICU) admission, and in-hospital mortality after controlling for confounding variables.
Meaning: Aspirin may have lung-protective effects and reduce the need for mechanical ventilation, ICU admission, and in-hospital mortality in hospitalized COVID-19 patients.
Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has infected over 54.8 million people and caused 1.3 million deaths worldwide.1 The majority of coronavirus disease 2019 (COVID-19) cases are mild, but critical illness has been reported in 6%–19% of patients.2 , 3 Although pneumonia and acute respiratory distress syndrome (ARDS) are hallmarks of the disease, thrombotic complications have been reported in 25%–42% of patients and are associated with increased mortality.4–6 Evidence of hypercoagulability has been observed on viscoelastic coagulation testing and COVID-19 patients frequently have elevated D-dimer and fibrinogen concentration.7 Deep vein thrombosis (DVT) and arterial thrombosis are relatively common, and on autopsy, megakaryocytes and platelet-rich thrombi have been observed in the heart, lung, and kidneys of COVID-19 patients.7–10 Furthermore, alveolar-capillary microthrombosis has been observed at a rate 9 times higher than in influenza patients and is thought to contribute to the severe lung injury and hypoxemia that occurs in COVID-19 patients.11
A prior study suggested that systemic anticoagulation reduces mortality in mechanically ventilated COVID-19 patients.12 Low dose aspirin has been utilized for prevention of stroke and myocardial infarction in high-risk patients, and the US preventive services task force recommends its use in adults with elevated cardiovascular risk.13 Aspirin reduces cardiovascular events, although a recent randomized control trial and meta-analysis found that it also increases the risk for major bleeding.14 , 15 In ARDS, aspirin has been studied with mixed results, where some studies have demonstrated benefit and others have not.16–21
Given its widespread availability, low-cost, and numerous studies supporting its efficacy and safety in patients with cardiovascular disease, we evaluated aspirin’s impact on clinical outcomes in hospitalized patients with COVID-19.14 , 15 , 22 , 23 Our primary study objective was to evaluate whether aspirin use was associated with the need for mechanical ventilation, and we hypothesized that aspirin use would be associated with a reduced risk of mechanical ventilation. Second, we evaluated whether aspirin use was associated with a reduced risk for intensive care unit (ICU) admission and in-hospital mortality.
METHODS
Patients
COVID-19 patients admitted to the hospital between March 2020 and July 2020 were abstracted from the multicenter collaborative research to understand the sequelae of harm in COVID (CRUSH COVID) registry. This registry contains data on COVID-19 patients from multiple institutions, and 4 hospitals had entered the majority of data as of July 2020: the University of Maryland Medical Center, Wake Forest Baptist Medical Center, Northeast Georgia Health System, and George Washington University Hospital. Data entry for all patients was managed using research electronic data capture (REDCap).24
The study was reviewed and approved by the institutional review board (IRB) at the University of Maryland, Baltimore, which served as the central IRB for all sites. The requirement for written informed consent was waived by the IRB, and the study was conducted in accordance with the ethical principles described in the Declaration of Helsinki. Patients were included if they were ≥18 years old and had laboratory-confirmed SARS-CoV-2 infection by qualitative real-time polymerase chain reaction (PCR). Patients were excluded if they had a do-not-intubate or do-not-resuscitate order at admission or if they remained an inpatient at the time of analysis.
Study Data
Patient demographics, comorbidities, medications, laboratory data, and outcome data were recorded. Major bleeding was defined as bleeding that led to a hemoglobin <7 g/dL and required red blood cell (RBC) transfusion, bleeding that led to transfusion of ≥2 RBC units in 24 hours, intracranial bleeding, gastrointestinal bleeding requiring RBC transfusion, bleeding that required surgical intervention, nasopharyngeal bleeding that required intervention, or ocular bleeding. Overt thrombosis was defined as DVT, pulmonary embolism, peripheral arterial occlusion, ischemic stroke, or ST-elevation myocardial infarction.
Respiratory parameters before intubation were collected, which included peripheral capillary oxygen saturation (Spo 2 ), fraction of inspired oxygen (Fio 2 ), oxygen flow, type of oxygen support, and Pao 2 . Mechanical ventilation parameters after intubation were collected, which included ventilator mode, tidal volume, tidal volume per predicted body weight, set pressure, positive end-expiratory pressure (PEEP), Fio 2 , peak pressure, plateau pressure, Pao 2 , and Pao 2 /Fio 2 ratio.
Aspirin Use Definition
Aspirin use was defined as administration within 24 hours of hospital admission or in the 7 days before hospital admission. This definition was selected based on aspirin’s prolonged duration of action as an irreversible platelet inhibitor and because of aspirin’s rapid onset within 0–4 hours when chewed or swallowed.25 If aspirin was administered after an outcome occurred (ie, after mechanical ventilation), then that patient was categorized in the nonaspirin group for analysis of that outcome.
Study Outcomes
The study’s primary outcome was the need for invasive mechanical ventilation, inclusive of ventilation with an endotracheal or tracheostomy tube, and exclusive of noninvasive ventilation with continuous positive airway pressure (CPAP) or bilevel positive airway pressure (BiPAP). Secondary outcomes were ICU admission and in-hospital mortality. The biologic plausibility for our hypothesis was based on aspirin’s ability to irreversibly inhibit platelet aggregation in the lungs, which could reduce pulmonary microthrombi and subsequent lung injury.
Statistical Analysis
Patients were stratified by whether or not they received aspirin. Nonparametric continuous variables such as body mass index (BMI), admission vital signs, clinical prediction scores, initial laboratory values, adverse events, and mechanical ventilation parameters were summarized as the median and interquartile range (IQR) and were compared between groups using the Mann-Whitney U Test. Categorical variables, such as demographic information, comorbidities, receipt of other investigational therapeutics, type of oxygen support, mechanical ventilation mode, and outcomes, were reported as the number and percentage of patients and were compared between groups using the χ2 test. P values <.05 were considered statistically significant.
Chi-square tests were performed to measure the crude association between aspirin use and the need for mechanical ventilation, ICU admission, and in-hospital mortality. Cox-proportional hazards models were fit to determine the adjusted associations between aspirin use and outcomes after controlling for confounders. Selection of confounders was based on recommendations in the published epidemiology literature.26 , 27 We adjusted for variables that were pre-existing and were associated with aspirin use and the outcome or solely the outcome, including age, sex, BMI, race, hypertension, diabetes mellitus, coronary artery disease, renal disease, liver disease, and home beta-blocker use. Variables that were believed to be in the causal pathway or occurred after aspirin use began were not adjusted for. Because aspirin, as a cyclooxygenase-1 (COX-1) inhibitor, modifies inflammatory and coagulation responses, these variables were considered to be in the causal pathway and were not adjusted for.
The proportional hazards assumption was tested graphically by plotting scaled Schoenfeld residuals and numerically using the Kolmogorov-type Supremum test, and no serious violations of the proportional hazards assumption were observed. To maximize inclusion of confounders in our final model, a liberal P value of .2 was used for likelihood ratio testing to enter and remove variables from the model using forward stepwise regression. Because age, BMI, race, coronary artery disease, hypertension, and diabetes mellitus are established risk factors for poor outcomes in COVID-19 patients, these variables were forced into the final model and were not subject to entry and removal criteria.28 Patients with a negative time to event (ie, mechanical ventilation occurred before hospital admission) were removed from the analysis for that particular outcome. A 95% confidence interval (CI) was calculated for each adjusted hazard ratio (HR).
A subgroup analysis with a Cox-proportional hazard model was performed in the cohort of patients who were not mechanically ventilated at admission. This group included all patients who were admitted on room air, standard nasal cannula oxygen, high-flow nasal cannula oxygen, nonrebreather oxygen mask, CPAP, and BiPap. In addition, another sensitivity analysis with an E value was performed to estimate the magnitude of an unadjusted confounding variable that would be needed to mitigate the association between aspirin use and the need for mechanical ventilation.29
Given the retrospective nature of our study, a priori sample size calculation was not performed; however, post hoc calculations indicated that with the observed mechanical ventilation rate of 48.4% in the nonaspirin group, our sample size allowed for detection of an 11.7% reduction in the rate of mechanical ventilation, given an alpha of 5% and 80% power. In addition, with 9 independent variables in the Cox-proportional hazards model, a minimum of 90 mechanically ventilated patients would be necessary to create a parsimonious model and sufficiently adjust for confounding variables. Statistical analysis was performed using SPSS Statistics 25 (IBM Corporation, Somers, NY) and R 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). The final analysis was confirmed using SAS 9.3 (SAS Corporation, Cary, NC).
RESULTS
Four hundred twelve patients were included in the study. Median age was 55 years (IQR, 41–66 years), and 59.2% of patients were male (Table 1 ). Ninety-eight patients (23.7%) received aspirin, while 314 patients (76.3%) did not. Of those who received aspirin, 75.5% were taking it before admission and 86.7% received it within 24 hours of hospital admission. Median time to aspirin administration in the hospital was 0 days (IQR, 0–1 days), median dose was 81 mg (IQR, 81–81 mg), and median treatment duration was 6 days (IQR, 3–12 days). Patients who received aspirin had significantly higher rates of hypertension, diabetes mellitus, coronary artery disease, and renal disease (P < .001). Furthermore, significantly more patients taking aspirin were on home beta-blockers (P < .001) and had liver disease (P = .04). The proportion of patients in each group receiving other therapeutics, including azithromycin, convalescent plasma, dexamethasone, therapeutic heparin, hydroxychloroquine, remdesivir, and tocilizumab, did not differ.
Table 1. -
Demographic Characteristics at Admission by Aspirin Use
Variable
No aspirin (N = 314)
Aspirin (N = 98)
P
Demographics
Age, y
52 (37–65)
61 (55–72)
<.001
Male
183 (58.3)
61 (62.2)
.49
BMI (kg/m2 )
30.6 (26.5–35.9)
28.4 (25.5–33.6)
.16
Race
.05
Black
172 (54.8)
54 (55.1)
Asian
10 (3.2)
7 (7.1)
White
66 (21)
26 (26.5)
Hispanic/Latino
66 (21)
11 (11.2)
Other
0 (0.0)
0 (0.0)
Comorbidities and medications
Hypertension
165 (52.5)
77 (78.6)
<.001
Diabetes mellitus
91 (29)
54 (55.1)
<.001
Coronary artery disease
18 (5.7)
34 (34.7)
<.001
Renal disease
39 (12.4)
31 (31.6)
<.001
Liver disease
21 (6.7)
13 (13.3)
.04
Home beta-blocker
53 (16.9)
34 (34.7)
<.001
Aspirin within 7 d of admission
0 (0)
74 (75.5)
Dose, mg
0 (0–0)
81 (81–81)
In-hospital aspirin
0 (0)
85 (86.7)
Dose, mg
0 (0–0)
81 (81–81)
Days from admission to initiation
0 (0–0)
0 (0–1)
Duration of therapy, d
0 (0–0)
6 (3–12)
Admission vital signs
Systolic BP, mm Hg
126 (113–142)
130 (115–150)
.14
Diastolic BP, mm Hg
73 (62–85)
74 (62–84)
.70
Heart rate, bpm
92 (81–106)
87 (77–102)
.04
RR
20 (18–24)
20 (18–25)
.89
Spo
2 , %
96 (94–98)
97 (94–99)
.10
Temperature, °C
37.2 (36.8–37.9)
37.1 (36.7–37.7)
.53
Clinical prediction scores
Hospital admission qSOFA
1 (0–1)
1 (0–1)
.28
ICU admission qSOFA
2 (1–2)
2 (1–2)
.98
ICU admission SOFA score
7 (3–9)
6 (3–9)
.56
ICU admission APACHE II score
15 (10–21)
18 (14–25)
.03
Admission oxygen support
Oxygen device
.005
Room air
142 (45.2)
52 (53.1)
Standard nasal cannula
72 (22.9)
32 (32.7)
High flow nasal cannula
24 (7.6)
4 (4.1)
Nonrebreather mask
19 (6.1)
3 (3.1)
CPAP
0 (0.0)
0 (0.0)
BiPaP
0 (0.0)
1 (1.0)
Endotracheal tube
57 (18.2)
6 (6.1)
Initial laboratory values
WBC, K/μL
7.1 (5.3–10.8)
7.5 (5.3–10)
.21
Lymphocytes, K/μL
1.1 (0.8–1.7)
1.1 (0.7–1.7)
.44
Hemoglobin, g/dL
12.7 (11.3–14.1)
12.3 (10.1–13.5)
.01
Platelets, K/μL
209 (159–286)
191 (137–267)
.17
INR
1.1 (1–1.2)
1.1 (1–1.3)
.90
PT, s
14 (13–16)
14 (13–15)
.73
PTT, s
31 (28–36)
31 (28–36)
.07
Ferritin, ng/mL
457 (178–1047)
398 (150–1078)
.76
IL-6, pg/mL
30.3 (7.1–159)
23.1 (5–84)
.48
C-reactive protein, mg/L
15 (5–35)
6 (3–19)
.62
D-dimer, ng/mL
1370 (700–3480)
1340 (890–3180)
.94
Troponin, ng/mL
0.02 (0.01–0.04)
0.02 (0.02–0.13)
.72
Fibrinogen, ng/mL
635 (448–763)
524 (461–630)
.009
Lactate, mmol/L
1.7 (1.3–2.5)
1.7 (1.1–2.9)
.05
Receipt of other therapeutics
Azithromycin
143 (45.5)
40 (40.8)
.41
Convalescent plasma
22 (7)
8 (8.2)
.70
Dexamethasone
18 (6.7)
3 (3.7)
.32
Therapeutic heparin
77 (24.5)
27 (27.6)
.55
Hydroxychloroquine
101 (32.2)
28 (28.6)
.50
Remdesivir
37 (11.8)
12 (12.2)
.90
Tocilizumab
35 (11.1)
10 (10.2)
.79
Outcomes
Major bleeding
24 (7.6)
6 (6.1)
.61
Overt thrombosis
28 (8.9)
8 (8.2)
.82
Hospital LOS, days
8 (3–19)
9 (5–17)
.91
Mechanical ventilation
152 (48.4)
35 (35.7)
.03
ICU admission
160 (51.0)
38 (38.8)
.04
ICU LOS, d
13 (6–23)
17 (7–29)
.25
In-hospital mortality
73 (23.2)
26 (26.5)
.51
Abbreviations: APACHE II, acute physiology and chronic health evaluation II; BiPAP = bilevel positive airway pressure; BMI, body mass index; BP, blood pressure; CPAP, continuous positive airway pressure; ICU, intensive care unit; IL-6, interleukin-6; INR, international normalized ratio; LOS, length of stay; PT, prothrombin time; PTT, partial thromboplastin time; qSOFA, quick sequential organ failure assessment; RR, respiratory rate; SOFA, sequential organ failure assessment; Spo 2 , peripheral capillary oxygen saturation; WBC, white blood cell.
Categorical variables are reported as number (percent). Continuous variables are represented as median (interquartile range). χ2 test for categorical variables, Mann-Whitney U test for continuous variables, pairwise comparison between aspirin and nonaspirin group. Inflammatory marker data not available for all patients (ferritin, 357 of 412 patients; IL-6, 148 of 412; C-reactive protein, 353 of 412; D-dimer, 355 of 412; fibrinogen, 214 of 412).
Admission vital signs and laboratory values did not differ between groups, except for fibrinogen concentration (aspirin median 524 mg/dL versus nonaspirin 635 mg/dL, P = .009), which was significantly lower in patients taking aspirin. Initial fibrinogen concentration was measured at a median of 1 day after hospital admission (IQR, 0–3 days), and there was no difference in timing between groups. Patients receiving aspirin were on significantly less oxygen support at admission compared to patients not receiving aspirin (P = .005). Specifically, a higher proportion of patients receiving aspirin were admitted to the hospital on room air or standard nasal cannula oxygen, while patients not on aspirin were admitted more frequently with high flow nasal cannula oxygen, a nonrebreather oxygen mask, CPAP, BiPAP, or an endotracheal tube and mechanical ventilation (Table 1 ).
The quick sequential organ failure assessment (qSOFA) score at hospital admission, as well as the qSOFA and sequential organ failure assessment (SOFA) scores at ICU admission, did not differ between groups. However, the acute physiology and chronic health evaluation II (APACHE II) score at ICU admission was significantly higher in the aspirin group (Table 1 ). Patients were in severe hypoxemic respiratory failure just before intubation, with a median Pao 2 of 70 mm Hg (IQR, 55–84), median Spo 2 of 92% (IQR, 86%–96%), and a median Fio 2 of 100% (IQR, 80%–100%) on 15 L·min−1 (IQR, 8–40). Immediately before intubation, high flow nasal cannula oxygen was utilized 35.6% of the time, followed by nonrebreather oxygen mask (21.2%), standard nasal cannula oxygen (18.5%), BiPAP (5.5%), and CPAP (1.4%). Preintubation oxygen support was unknown in 15.8% of cases due to the inability to access records from referring institutions.
Once on mechanical ventilation, patients were most frequently ventilated with volume control ventilation (41.2%), followed by pressure regulated volume control (28.9%), pressure control ventilation (16.6%), airway pressure release ventilation (APRV) (8.6%), pressure support ventilation (2.1%), synchronous intermittent mandatory ventilation (SIMV) (1.6%), and other modes (1.1%). Median tidal volume was 440 mL (IQR, 400–460 mL), with a set pressure of 22 cm H2 O (IQR, 14–26 cm H2 O), PEEP of 10 cm H2 O (IQR, 9–14 cm H2 O), and Fio 2 of 80% (IQR, 55%–100%). The median tidal volume per predicted body weight was 6.6 mL/kg (median 6.1–7.3 mL/kg), initial peak pressure was 29 cm H2 O (IQR, 24–33 cm H2 O), and median plateau pressure was 21 cm H2 O (IQR, 16–27 cm H2 O). The initial Pao 2 was a median of 100 mm Hg (IQR, 75–145 mm Hg) for a calculated median Pao 2 /Fio 2 ratio of 146 (IQR, 92–237). There were no statistically significant differences in respiratory parameters between groups.
On unadjusted analysis, patients receiving aspirin had significantly lower rates of mechanical ventilation (35.7% [35/98] aspirin versus 48.4% [152/314] nonaspirin, P = .03) and ICU admission (38.8% [38/98] aspirin versus 51.0% [160/314] nonaspirin, P = .04). There was no crude difference in in-hospital mortality (26.5% [26/98] aspirin versus 23.2% [73/314] nonaspirin, P = .51) between groups. In addition, there was no difference in the rate of major bleeding (6.1% aspirin [6/98] versus 7.6% nonaspirin [24/314], P = .61), or overt thrombosis (8.2% [8/98] aspirin versus 8.9% [28/314] nonaspirin, P = .82) between groups.
After adjusting for confounding variables in the Cox-proportional hazards model, aspirin use was independently associated with a reduced risk for mechanical ventilation, which was the primary study outcome (adjusted HR, 0.56, 95% CI, 0.37-0.85, P = .007). As secondary outcomes, aspirin use was also associated with a reduction in the risk of ICU admission (adjusted HR, 0.57, 95% CI, 0.38-0.85, P = .005) and in-hospital mortality (adjusted HR, 0.53, 95% CI, 0.31-0.90, P = .02) (Table 2 ).
Table 2. -
Cox-Proportional Hazards Model
Variable
B
Adjusted HR
P
Mechanical ventilation
Age (per y)
0.01
1.01 (1.00–1.03)
.02
BMI (per kg/m2 )
0.02
1.02 (1.01–1.04)
.001
Ethnicity
<.001
Black
−0.34
0.71 (0.48–1.06)
.10
Asian
0.34
1.41 (0.65–3.06)
.38
Hispanic/Latino
0.57
1.77 (1.11–2.85)
.02
HTN
0.21
1.23 (0.82–1.83)
.31
DM
0.14
1.14 (0.80–1.65)
.47
CAD
0.19
1.21 (0.72–2.02)
.47
Home beta blocker
−0.33
0.72 (0.45–1.14)
.16
Renal disease
−0.30
0.74 (0.45–1.21)
.23
Aspirin use
−0.58
0.56 (0.37–0.85)
.007
ICU admission
Age (per y)
0.01
1.01 (1.00–1.02)
.07
Sex
−0.35
0.70 (0.52–0.95)
.02
BMI (per kg/m2 )
0.02
1.02 (1.01–1.03)
.002
Ethnicity
Black
−0.35
0.71 (0.48–1.03)
.07
Asian
0.03
1.03 (0.48–2.22)
.94
Hispanic/Latino
0.48
1.62 (1.06–2.49)
.03
HTN
0.16
1.17 (0.82–1.68)
.39
DM
0.30
1.34 (0.97–1.87)
.08
CAD
0.02
1.02 (0.62–1.66)
.95
Renal disease
−0.45
0.64 (0.40–1.01)
.06
Liver disease
−0.50
0.60 (0.32–1.13)
.11
Aspirin use
−0.57
0.57 (0.38–0.85)
.005
In-hospital mortality
Age (per y)
0.04
1.04 (1.02–1.06)
<.001
Sex
−0.41
0.66 (0.42–1.05)
.08
BMI (per kg/m2 )
0.02
1.02 (1.01–1.04)
.004
Ethnicity
.02
Black
0.73
2.07 (1.13–3.81)
.02
Asian
1.44
4.24 (1.62–11.08)
.003
Hispanic/Latino
0.79
2.20 (1.07–4.54)
.03
HTN
−0.18
0.84 (0.46–1.53)
.57
DM
0.23
1.26 (0.80–2.00)
.32
CAD
0.65
1.91 (1.06–3.42)
.03
Home beta blocker
0.73
2.07 (1.22–3.52)
.007
Renal disease
0.44
1.55 (0.90–2.66)
.12
Aspirin use
−0.64
0.53 (0.31–0.90)
.02
Cox-regression analyses performed to examine the effect of aspirin use on mechanical ventilation, ICU admission, and in-hospital mortality after controlling for demographics and comorbidities. For ethnicity, the group is compared to the reference category of White. For sex, the group is compared to the reference category of males. Adjusted HRs are reported with 95% confidence intervals. Patients with a negative time to event (ie, mechanical ventilation occurred before hospital admission) were removed from the analysis for that particular outcome. After cases with negative time were removed, there were 385, 409, and 410 cases in the Cox-proportional hazards models for mechanical ventilation, ICU admission, and in-hospital mortality, respectively.
Abbreviations: BMI, body mass index; CAD, coronary artery disease; DM, diabetes mellitus; HR, hazard ratio; HTN, hypertension; ICU, intensive care unit.
In our subgroup analysis of patients who did not require mechanical ventilation at admission, these favorable associations between aspirin and mechanical ventilation (adjusted HR, 0.63, 95% CI, 0.40-1.00, P = .05), ICU admission (adjusted HR, 0.64, 95% CI, 0.41-0.99, P = .046), and in-hospital mortality (adjusted HR, 0.45, 95% CI, 0.24-0.85, P = .01) persisted. A forest plot was constructed, showing each of the primary and secondary outcomes to graphically depict adjusted differences in hazards between groups (Figure 1 ). The survival function for in-hospital mortality was constructed from the Cox-proportional hazards model, and aspirin use was plotted as a separate line to visualize the cumulative differences in survival between groups (Figure 2 ).
Figure 1.: Forest plot of adjusted HRs with aspirin use. 95% CIs for adjusted HRs are plotted. CI indicates confidence interval; HR, hazard ratio; ICU, intensive care unit.
Figure 2.: Survival function for in-hospital mortality. Patients are stratified by aspirin use. Patients discharged within the study period are right-censored. Aspirin use was associated with a decreased hazard for in-hospital mortality (adjusted HR = 0.53, 95% CI, 0.31-0.90, P = .02). CI indicates confidence interval; HR, hazard ratio.
In our sensitivity analysis, we found that an unexplained confounder would need to be associated with both aspirin use and mechanical ventilation at a risk ratio of 2.35 to mitigate the relationship between these variables and make the HR = 1, while controlling for other covariates in our model (E value 2.35, upper confidence limit 1.48). The upper confidence limit for the risk ratio can be interpreted as the lowest risk ratio that an unobserved confounder would have to have with both aspirin use and mechanical ventilation, after controlling for other covariates in the model, to make the HR = 1.
We performed an additional sensitivity analysis of the primary outcome and stratified patients by the timing of aspirin use. This sensitivity analysis was underpowered but may be helpful for planning future studies. The rates of ICU admission in those receiving aspirin only in the 7 days before hospitalization, only in the first 24 hours of hospitalization, or in both timeframes were 46.2%, 20.8%, and 39.3%, respectively. This resulted in an adjusted HR of 0.78 (95% CI, 0.32-1.95), 0.36 (95% CI, 0.15-0.89), and 0.71 (95% CI, 0.44-1.16) for these 3 timeframes, respectively.
DISCUSSION
In a multicenter cohort study of 412 COVID-19 patients, aspirin use was independently associated with a lower risk of mechanical ventilation, ICU admission, and in-hospital mortality. With an emphasis on avoiding mechanical ventilation in COVID-19 patients, these results are clinically significant. Mechanistically, COVID-19 is associated with hypercoagulability and pulmonary microthrombosis, and aspirin may mitigate these effects.11 , 4 Aspirin is inexpensive, widely available, and has a well-described risk profile. These attributes, in conjunction with our pilot data, support aspirin’s role as a potential adjunctive therapeutic in COVID-19.
The prothrombotic and hypercoagulable state that is induced by COVID-19 is well described, and a prior study has shown that systemic anticoagulation can reduce mortality in mechanically ventilated COVID-19 patients.12 Aspirin may have a similar impact due to its antiplatelet and anti-inflammatory properties as an inhibitor of COX-1, which decreases thromboxane A2 synthesis, platelet aggregation, and thrombus formation.30 Aspirin’s potential benefits in lung injury are thought to be related to reduced platelet-neutrophil aggregates in the lungs, reduced inflammation, and increased lipoxin formation, which restores pulmonary endothelial cell function.21 These protective effects may be augmented in a disease such as COVID-19, where the procoagulant tendency is unusually high, and endothelial cell dysfunction is common.
Aspirin’s anti-inflammatory properties may also contribute to its lung-protective effects in COVID-19. Aspirin has been shown to decrease the production of interleukin-6 (IL-6), C-reactive protein (CRP), and macrophage colony-stimulating factor in patients with cardiovascular disease, and in COVID-19, these actions could reduce the incidence of cytokine storm.31 , 32 In our study, patients on aspirin had significantly lower initial plasma fibrinogen concentration, which might be explained by aspirin’s effect on the acetylation of fibrinogen and acceleration of fibrinolysis.33 Aspirin also has an inhibitory effect on cyclooxygenase-2 (COX-2), which decreases IL-6 and CRP production.31
Several studies have examined aspirin’s possible beneficial effect in ARDS. Erlich et al18 evaluated 161 patients at risk of acute lung injury (ALI), of which 79 received antiplatelet therapy at admission. Of the patients in the antiplatelet group, 75 (94.9%) were on aspirin. The study found that antiplatelet agents were associated with a reduction in the incidence of ALI and ARDS (adjusted risk ratio, 0.34, 95% CI, 0.13-0.88, P = .03).18 Chen et al17 examined 1149 patients at high risk of ARDS and found that prehospital aspirin administration was associated with a reduction in ARDS after adjustment for confounding variables (adjusted odds ratio [OR], 0.66, 95% CI, 0.46-0.94, P = .02). The reduced incidence of ARDS persisted when examining only patients with sepsis (adjusted OR, 0.61, 95% CI, 0.41-0.90, P = .01).17 Another study of high-risk patients found that although aspirin was associated with a reduction in ALI on univariable analysis (OR, 0.65, 95% CI, 0.46-0.90, P = .01), the association was not statistically significant on multivariable analysis (pooled OR, 0.70, 95% CI, 0.48-1.03, P = .07).20 Taken together, a meta-analysis of these 3 studies found that aspirin was associated with an overall reduced incidence of ARDS (pooled OR, 0.59, 95% CI, 0.36-0.98).21
When examining mortality as an outcome, prehospital or in-hospital aspirin use in ARDS patients is associated with a reduction in ICU mortality (adjusted OR, 0.38, 95% CI, 0.15-0.60, P = .04), but not in-hospital mortality (adjusted OR 0.91, 95% CI, 0.46-1.78, P = .78).16 The lung injury prevention study with aspirin (LIPS-A) clinical trial examined 390 patients at high risk for ARDS who were randomized to aspirin or placebo and found that aspirin did not prevent ARDS at 7 days (P = .53) or improve 28-day survival (P = .92).19 However, there are important differences between our study and those previously published. Our study limited enrollment to patients with a common diagnosis of COVID-19, as opposed to other studies, which enrolled all patients at high risk for ARDS. In those that reported baseline risk factors for ARDS, the spectrum of diagnoses was broad: sepsis, noncardiogenic shock, aspiration, pancreatitis, pneumonia, positive shock index, and trauma.18 , 19 This heterogeneity may have diluted and confounded aspirin’s effects.
In addition, not all causes of ARDS are associated with hypercoagulability, as in the case of COVID-19. In a single-center study, SARS-CoV-2 infection was shown to be associated with pulmonary embolism at a rate of 20.6%, which was 3 times higher than in all patients admitted to the ICU during the same time interval (20.6% SARS-CoV-2 versus 6.1% control, absolute risk increase 14.4%, 95% CI, 6.1-22.8). This rate was more than double the rate observed in influenza patients admitted to the ICU (20.6% SARS-CoV-2 versus 7.5% influenza, absolute risk increase 13.1%, 95% CI, 1.9-24.3).34 Furthermore, on autopsy, alveolar microthrombi have been reported to be present at a rate 9 times higher than in influenza patients.11 Because hypercoagulability is common in COVID-19, aspirin’s effect size may be larger in this population and this may explain why our analysis was able to detect statistically significant associations with reduced mechanical ventilation, ICU admission, and in-hospital mortality, whereas some previous studies did not find these benefits.
There is now substantial evidence that COVID-19 is a systemic disease affecting the vasculature and that SARS-CoV-2 causes vascular endothelialitis involving the pulmonary capillary endothelium.10 , 35 Electron microscopy and histological analyses have demonstrated that SARS-CoV-2 infects endothelial cells in multiple organs, causing endothelialitis, impaired microcirculation, and apoptosis, all of which leads to inflammation and microthrombosis.35 Microthrombosis has been well described in autopsies of COVID-19 patients, and excess megakaryocytes have been observed in the heart, lungs, and kidneys of deceased patients.10 , 11 Aspirin, as an irreversible antiplatelet agent, may prevent platelets produced from these megakaryocytes from aggregating and creating microthrombi.10
In our study, we did not observe a lower rate of overt thrombosis in aspirin users, but this may be due to the low number of reported events in each group or due to reporting bias from avoidance of diagnostic imaging. Importantly, the presence of microthrombi does not correlate with overt thrombosis. Microthrombosis is better diagnosed with videomicroscopes, dark-field imaging, and spectral imaging.36 , 37 These noninvasive methods measure the flow and velocity of cells in the microcirculation and allow for the visualization and quantification of flow on a capillary-level. However, these tools are neither widely available nor standard of care. Although a reduction in microthrombosis is a plausible mechanism to explain our findings, the hypothesis that aspirin inhibits microthrombosis in COVID-19 needs to be further explored in mechanistic studies.
Aspirin has been shown to increase bleeding risk in a randomized control trial (HR, 1.38, 95% CI, 1.18-1.62, P < .001) and systematic review (HR, 1.43, 95% CI, 1.30-1.56).14 , 15 Patients in these 2 large studies were treated with aspirin for the primary prevention of cardiovascular disease, and these patients did not have an acute illness, which is an important distinction from our cohort. In our analysis, we did not find a significant increase in major bleeding in patients who received aspirin. This difference may be explained by the fact that patients with COVID-19 are frequently hypercoagulable, thrombocytopenia is uncommon in COVID-19 patients, and the risk of bleeding appears to be low (2%–3%), even with systemic heparinization.12 Nevertheless, aspirin could increase bleeding risk to some degree (particularly gastrointestinal bleeding), and larger studies should be performed to better assess bleeding risk in patients with COVID-19 who are treated with aspirin.
The limitations of our study include its observational design and modest sample size, which limits generalizability and the ability to completely adjust for confounding. At a minimum, 10–20 subjects should be included per independent variable in a multivariable regression model.38 Our Cox-proportional hazard model consisted of 187 mechanically ventilated patients and 9 independent variables. This meets the minimum sample size requirement of 10–20 subjects per included variable but is far from the ideal size of 40 subjects per variable, which is necessary to achieve parsimony.38 However, the confounders that we did include are very specific for their association with exposure and outcome.28 Additionally, patients in the aspirin group may have received different medical care due to their increased comorbidities, leading to treatment biases. Also, we did not record the presence of other medications that are associated with hypercoagulability, such as oral contraceptives and hormone replacement therapy, and differences in the use of these drugs may have confounded our results. Imaging for the detection of thrombotic events was at the discretion of the treating clinician, which may have introduced additional bias and skewed the reporting of thrombotic events. Lastly, inflammatory markers were not measured universally in all patients, which may have skewed our results.
In summary, our analysis suggests that aspirin use may have beneficial effects in patients with COVID-19. Mechanistically, these findings are plausible given aspirin’s irreversible antiplatelet effect and the frequent hypercoagulability observed in COVID-19 patients. The results of our study are intriguing, especially because aspirin has been thoroughly studied in chronic cardiovascular disease, has a well-described safety profile, and is readily available throughout the world. The preliminary, hypothesis-generating nature of our study provides the basis for a larger study, which will be needed to confirm our findings and assess the extent to which the relationships observed in our study are causal. Until a randomized controlled trial of aspirin is performed, it is imperative to exercise cautious optimism and deliberately balance aspirin’s known risks against its potential benefits in patients afflicted by COVID-19.
DISCLOSURES
Name: Jonathan H. Chow, MD.
Contribution: This author helped conceive, design, acquire, analyze, and interpret the study data, draft the manuscript, provide critical revisions and approve the final manuscript, agreed to be accountable for all aspects of the work, and is responsible for archiving the study files.
Conflicts of Interest: J. H. Chow has served on the Speaker’s Bureau for La Jolla Pharmaceutical Company, outside the scope of the submitted work.
Name: Ashish K. Khanna, MD, FCCP, FCCM.
Contribution: This author helped design and acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Shravan Kethireddy, MD.
Contribution: This author helped conceive, design, acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: David Yamane, MD.
Contribution: This author helped design the study, acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Andrea Levine, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Amanda M. Jackson, MD.
Contribution: This author helped analyze and interpret the study data, draft the manuscript, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Michael T. McCurdy, MD.
Contribution: This author helped conceive, design, and acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Ali Tabatabai, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Gagan Kumar, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Paul Park, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Ivy Benjenk, RN, MPH.
Contribution: This author helped design of study, acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Jay Menaker, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Nayab Ahmed, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Evan Glidewell, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Elizabeth Presutto, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Shannon Cain, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Naeha Haridasa, BS.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Wesley Field, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Jacob G. Fowler, BS.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Duy Trinh, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Kathleen N. Johnson, BS.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Aman Kaur, DO.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Amanda Lee, BS.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Kyle Sebastian, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Allison Ulrich, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Salvador Peña, MD, PhD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Ross Carpenter, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Shruti Sudhakar, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Pushpinder Uppal, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Benjamin T. Fedeles, MD, Capt, USAF, MC.
Contribution: This author helped acquire study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Aaron Sachs, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Layth Dahbour, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: William Teeter, MD.
Contribution: This author helped acquire of study data, analyze and interpret the data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Kenichi Tanaka MD.
Contribution: This author helped interpret the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Samuel M. Galvagno, DO, PhD.
Contribution: This author helped acquire and interpret the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Daniel L. Herr, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Thomas M. Scalea, MD.
Contribution: This author helped acquire the study data, provide critical revisions, and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
Name: Michael A. Mazzeffi, MD, MPH.
Contribution: This author helped conceive, design, acquire, analyze, and interpret the study data, draft the manuscript, provide critical revisions and approve the final manuscript, and agreed to be accountable for all aspects of the work.
Conflicts of Interest: None.
This manuscript was handled by: Avery Tung, MD, FCCM.