key: cord-333943-9d93na7s
authors: Jeong, Han Eol; Lee, Hyesung; Shin, Hyun Joon; Choe, Young June; Filion, Kristian B; Shin, Ju-Young
title: Association between NSAIDs use and adverse clinical outcomes among adults hospitalized with COVID-19 in South Korea: A nationwide study
date: 2020-07-27
journal: Clin Infect Dis
DOI: 10.1093/cid/ciaa1056
sha: 
doc_id: 333943
cord_uid: 9d93na7s

BACKGROUND: Non-steroidal anti-inflammatory drugs (NSAIDs) may exacerbate COVID-19 and worsen associated outcomes by upregulating the enzyme that SARS-CoV-2 binds to enter cells. To our knowledge, no study has examined the association between NSAID use and the risk of COVID-19-related outcomes. METHODS: We conducted a cohort study using South Korea’s nationwide healthcare database, which contains data of all subjects who received a test for COVID-19 (n=69,793) as of April 8, 2020. We identified adults hospitalized with COVID-19, where cohort entry was the date of hospitalization. NSAIDs users were those prescribed NSAIDs in the 7 days before and including cohort entry and non-users were those not prescribed NSAIDs during this period. Our primary outcome was a composite of in-hospital death, intensive care unit admission, mechanical ventilation use, and sepsis; our secondary outcomes were cardiovascular complications and acute renal failure. We conducted logistic regression analysis to estimate odds ratio (OR) with 95% confidence intervals (CI) using inverse probability of treatment weighting to minimize confounding. RESULTS: Of 1,824 adults hospitalized with COVID-19 (mean age 49.0 years; female 59%), 354 were NSAIDs users and 1,470 were non-users. Compared with non-use, NSAIDs use was associated with increased risks of the primary composite outcome (OR 1.54 [95% CI 1.13-2.11]) but insignificantly associated with cardiovascular complications (1.54 [0.96-2.48]) or acute renal failure (1.45 [0.49-4.14]). CONCLUSION: While awaiting the results of confirmatory studies, we suggest NSAIDs be used with caution among patients with COVID-19 as the harms associated with their use may outweigh their benefits in this population.

Coronavirus disease 2019 (COVID- 19) , which is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is a global pandemic. [1, 2] Concerns exist that the use of nonsteroidal anti-inflammatory drugs (NSAIDs) may exacerbate COVID-19 by upregulating angiotensin-converting enzyme 2 (ACE2) expressions, [3, 4] the enzyme which SARS-CoV-2 binds to enter cells. In addition, NSAIDs inhibit cyclooxygenase (COX), [5] which could be involved in the pathogenesis of viral infections to result in tissue damage. [6, 7] These concerns were based on unconfirmed anecdotal reports of four young COVID-19 patients who developed serious infectious complications following NSAIDs use. [8] The

Health Minister of France subsequently recommended that paracetamol (acetaminophen) be used as first-line antipyretic agents over NSAIDs.

In contrast, the US Food and Drug Administration, [9] European Medicine Agency, [10] and Australia's Therapeutic Goods Administration [11] stated that there is insufficient evidence to draw conclusions regarding this safety concern and thus, current clinical practice should not be changed until further evidence becomes available. This position is supported by a recent systematic review of randomized trials and observational studies of respiratory viral infections, which concluded that there is currently no evidence to support that NSAIDs are harmful with respect to COVID-19. [12] Despite the widespread use of NSAIDs, to our knowledge, there is currently no published observational study that specifically assessed the association between NSAIDs use and clinical outcomes among COVID-19 patients.

This cohort study therefore aimed to examine the association between NSAIDs use, compared to non-use, and worsened clinical outcomes among adults hospitalized with A c c e p t e d M a n u s c r i p t page | 5 COVID-19 using South Korea's nationwide healthcare database containing all COVID-19 patients.

We used the Health Insurance Review and Assessment Service (HIRA) database of South Korea, provided as part of the #OpenData4Covid19 project, a global research collaboration on COVID-19 jointly conducted by Ministry of Health and Welfare of Korea and HIRA. [13] Briefly, the South Korean government released the world's first de-identified COVID-19 nationwide patient data on March 27, 2020. Owing to South Korea's National Health Insurance system, which is the universal single-payer healthcare provider covering the entire Korean population of 50 million, and its fee-for-service reimbursement system, the database includes information from both inpatient and outpatient settings.

The HIRA COVID-19 database contains data of all subjects who received a test for COVID-19 as of April 8, 2020, linked to their administrative healthcare data from the previous 3 years (January 1, 2017 to April 8, 2020). The HIRA COVID-19 database includes anonymized patient identifiers, sociodemographic characteristics, healthcare utilization history, diagnoses (International Classification of Diseases, 10 th Revision; ICD-10), and drug prescription information (Anatomical Therapeutic Chemical classification codes); use of over-the-counter drugs are not collected in this database (Supplementary Material 1). [14] This study was approved by the Institutional Review Board of Sungkyunkwan University (SKKU 2020-03-012), which waived the requirement of obtaining informed consent.

A c c e p t e d M a n u s c r i p t page | 6

Of 69,793 individuals who received a diagnostic test for COVID-19 between January 1, 2020 to April 8, 2020, 5,707 tested positive for COVID-19 ( Figure 1) 

We defined exposure using inpatient and outpatient prescription records of NSAIDs from the HIRA database, including both oral and intravenous formulations (Supplementary Material 2). We ascertained exposure to NSAIDs according to an intention-to-treat approach, in which exposure was defined in the index period of 7 days before and including cohort entry among hospitalized COVID-19 patients. Patients prescribed NSAIDs during this period were classified as NSAIDs users whereas those not prescribed NSAIDs during this period A c c e p t e d M a n u s c r i p t page | 7 were classified as non-users. To minimize any time-related biases such as immortal time, [17] follow-up was initiated from the date of cohort entry for both NSAIDs users and non-users.

Our primary outcome was a composite endpoint of in-hospital death, intensive care unit (ICU) admission, mechanical ventilation use, and sepsis. Our secondary outcomes were a composite endpoint of cardiovascular complications (myocardial infarction, stroke, heart failure), and acute renal failure. We defined outcomes using in-hospital ICD-10 diagnostic codes and procedures using the national procedure coding system (Supplementary Material 2). Study outcomes were measured between the cohort entry date and the earliest of the date of hospital discharge or end of study period (April 8, 2020).

We assessed sociodemographic and clinical factors considered to be associated with NSAIDs use and risk of the outcomes of interest. For sociodemographic factors, we assessed age, sex, and health insurance type at cohort entry; age was grouped into 10-year bands.

Clinical variables included comorbidities and use of co-medications assessed in the year before cohort entry (Supplementary Material 2). We used the expanded benefit coverage codes in addition to diagnosis codes to define malignancy to minimize false positives.

Baseline characteristics were summarized for NSAIDs users and non-users using counts (proportions) or mean (standard deviation) for categorical or continuous variables, respectively. We calculated the absolute standardized difference (aSD) to determine important imbalances between exposure groups, with aSD ≥0.1 considered important.

A c c e p t e d M a n u s c r i p t page | 8

We estimated the cumulative incidence of the primary and secondary outcomes among NSAIDs users versus non-users. We used three outcome models using logistic regression to estimate odds ratio (OR) with 95% confidence intervals (CIs) of the association of interest. The first model was unadjusted. The second model included all covariates described above. The third model, considered our primary analysis, was weighted by propensity scores (PS) using the inverse probability of treatment weight (IPTW)

approach. [18] The PS, or probability of receiving NSAIDs, was estimated using multivariable logistic regression analysis, with all confounders mentioned above included as independent variables. The c-statistic was used to determine model discrimination, with a value between 0.6 and 0.8 considered adequate to predict treatment status based on covariates included. [19] The IPTW approach involves weighting the inverse probability of receiving NSAIDs (1/PS for NSAIDs, and 1/(1−PS) for non-user groups).

We conducted sex-and age-stratified analyses, with age classified into three groups (<45, 45-65, ≥65 years), for the risk of the primary outcome associated with NSAIDs use. In addition, we stratified by route of administration (oral versus intravenous) and by history of hypertension, hyperlipidemia, or diabetes mellitus. The PS were re-calculated in all subgroup analyses using multivariable logistic regression models.

A c c e p t e d M a n u s c r i p t page | 9

As there is currently no data available on how fast NSAIDs increase ACE2 tissue expressions, we varied the exposure ascertainment window to 14 days and 30 days before and including cohort entry. Patients prescribed NSAIDs during these periods were classified as NSAIDs users whereas those not prescribed NSAIDs were classified as non-users. Follow-up was initiated from cohort entry.

To examine the potential effects of confounding by indication, we compared NSAIDs to paracetamol as these are used for similar indications (Supplementary Material 2).

We classified patients based on their exposure to NSAIDs or paracetamol in the 7 days before and including cohort entry, excluding those not exposed to one of the two drugs of interest and those who received both drugs during this exposure window. Follow-up was initiated from cohort entry for both exposure groups.

As outcome misclassification of sepsis from inaccuracy of coding or reverse causality between NSAIDs use and sepsis is possible, we repeated our main analysis by using a redefined primary outcome that was a composite endpoint of in-hospital death, ICU admission, and mechanical ventilation use.

A c c e p t e d M a n u s c r i p t page | 10

First, to improve comparability between exposure groups, we excluded the most extreme 1% of PS values (IPTW with trimming). Second, we included the estimated PS, in addition to other covariates, into our multivariable logistic regression model. Third, we stratified on the PS in deciles. Last, we applied standardized mortality ratio weights (SMRW)

(1 for NSAIDs, and PS/(1−PS) for non-user groups). [18] All statistical analyses were performed using the SAS Enterprise Guide software (version 6.1).

NSAIDs users (19%) and 1,470 non-users (81%). NSAIDs users were older than non-users There was no difference between the association between NSAID use and the risk of our primary composite endpoint by formulation of NSAIDs, sex, and histories of hypertension and hyperlipidemia ( Figure 2 ). However, we found effect modification in age groups (p-for-interaction <0.0001) and history of diabetes mellitus (p-for-interaction 0.0180). Findings from sensitivity analyses remained largely consistent, where all effect estimates showed positive associations between the primary outcome and NSAIDs users, as compared with non-users, when varying the exposure window, applying other methods involving PS, or redefining the primary outcome. When comparing to paracetamol, our sample size was greatly reduced, and there were no events that occurred in the NSAID group (cumulative incidence for NSAIDs users: 0.0%; paracetamol users 4.1%). Results of sensitivity analyses for the secondary outcomes were also generally consistent ( Figure 3 ).

To the best of our knowledge, this is the first population-based cohort study to have investigated the association between NSAID use and adverse outcomes among patients with Likewise, despite use of NSAIDs known to result in nephrotoxicity [27, 28] , our findings suggest no additional risk of acute renal failure when COVID-19 patients were exposed to NSAIDs.

The underlying pathogenic link between NSAIDs and COVID-19 has yet to be elucidated. However, one animal study found increased ACE2 expressions with NSAIDs (ibuprofen) [29] in various organs such as the lung, heart, and kidneys. [4, 30, 31] Thus, ACE2 upregulation induced by NSAIDs could theoretically heighten the infectivity of SARS-CoV-2 to worsen clinical outcomes, resulting in multiple organ failure in severe cases. Other hypothetical mechanisms have also been suggested. NSAIDs could aggravate infections by upregulating COX-2 in activated B lymphocytes to interfere with antibody productions, [32] or by selectively inhibiting interferon-γ productions that are vital for immunity against foreign pathogens.

[33] However, with inconsistent findings from animal studies and the precise biological mechanisms yet to be understood, it remains unclear as to whether these findings are readily transferable to humans.

We defined exposure using an approach analogous to an intention-to-treat, with exposure assessed in the 7 days before and including the day of cohort entry (hospital admission). We used this approach to avoid time-related biases that could be introduced by assessing in-hospital NSAID use as the date of prescription was not available for ~50% of inhospital prescriptions. The length of hospital stay not only influences the probability of being exposed to NSAIDs while hospitalized but is also associated with worse prognosis. However, with exposure defined using pre-hospital medication use, our exposure assessment was independent of in-hospital outcomes and the duration of hospital stay. Although predicting the direction resulting from this bias is difficult as occurrence of confounders during hospitalization are accounted for in this study, the use of this exposure definition is likely to A c c e p t e d M a n u s c r i p t page | 14 bias our findings towards the null. This is because, by not accounting for NSAID use during hospitalization, the observed increased risk is suggested to be a conservative estimate.

Our study has several strengths. To our knowledge, this is the first population-based study conducted using all hospitalized patients with COVID-19 to assess the association between NSAID use and COVID-19 related outcomes. Moreover, we used a nationwide healthcare database of South Korea that includes information on healthcare utilization of all COVID-19 cases as of April 8, 2020. Therefore, our findings provide real-world evidence that is highly generalizable to everyday clinical practice. With its large source population, our data source was sufficiently large to assess this clinically important issue. In addition, our findings were consistent in sensitivity analyses that extended the index period.

Our study also has some limitations. First, outcome misclassification is possible.

However, misclassification of in-hospital death is likely to be very small, and the validity of procedure codes to define ICU admission or mechanical ventilation use are also expected to be high as these codes are used for reimbursement processes by the health insurance authority. Also, the positive predictive value of diagnosis codes between claims data and electronic medical records was previously reported to be 82%,[34] and we believe its validity to be greater for sepsis, myocardial infarction, stroke, heart failure, and acute renal failure as we restricted to hospitalized patients receiving close monitoring. Second, our findings may have theoretically underestimated the association between NSAIDs users and clinical outcome due to depletion of susceptible patients,[35] as we included prevalent users of NSAIDs. However, our study period included the start of the COVID-19 pandemic in South Korea, making it unlikely that patients who were susceptible to adverse COVID-19 related outcomes were excluded prior to entering our cohort. Third, our results may be affected by confounding by indication given our use of an unexposed reference group. Despite attempting to address this by comparing NSAIDs users to paracetamol users, we were unable A c c e p t e d M a n u s c r i p t page | 15 to provide meaningful results as there were no events among NSAIDs users upon excluding those prescribed both NSAIDs and paracetamol during the exposure window. This therefore suggests that the 23 exposed events from our main analysis were exposed to both drugs during the exposure window, implying that these patients were severe cases who were 

World Health Organization. Coronavirus disease (COVID-19) Situation Dashboard

Rapid asymptomatic transmission of COVID-19 during the incubation period demonstrating strong infectivity in a cluster of youngsters aged 16-23 years outside Wuhan and characteristics of young patients with COVID-19: a prospective contact-tracing study

Are patients with hypertension and diabetes mellitus at increased risk for COVID-19 infection?

SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

Roles of cyclooxygenase (COX)-1 and COX-2 in prostanoid production by human endothelial cells: selective up-regulation of prostacyclin synthesis by COX-2

Hyperinduction of Cyclooxygenase-2-Mediated Proinflammatory Cascade: A Mechanism for the Pathogenesis of Avian Influenza H5N1 Infection

A Tug-Of-War Between Severe Acute Respiratory Syndrome Coronavirus 2 and Host Antiviral Defence: Lessons From Other Pathogenic Viruses

UPDATE -Coronavirus: French health minister and WHO issue warning over taking anti-inflammatories. The Local France

FDA advises patients on use of non-steroidal antiinflammatory drugs (NSAIDs) for COVID-19. US Food and Drug Administration

EMA gives advice on the use of non-steroidal antiinflammatories for COVID-19

Therapeutic Goods Administration. No evidence to support claims ibuprofen worsens COVID-19 symptoms. Therapeutic Goods Administration

Scientific Brief: The use of non-steroidal antiinflammatory drugs (NSAIDs) in patients with COVID-19. World Health Organization

Ministry of Health and Welfare. #opendata4covid19. Available at

Towards Actualizing the Value Potential of Korea Health Insurance Review and Assessment (HIRA) Data as a Resource for Health Research: Strengths, Limitations, Applications, and Strategies for Optimal Use of HIRA Data

World Health Organization. Laboratory testing for coronavirus disease

in suspected human cases: interim guidance. World Health Organization,. 2020. 16. Ministry of Health and Welfare RoK. About COVID-19: Patient Treatment & Management

Immortal Time Bias in Pharmaco-Epidemiology

Alternative approaches for confounding adjustment in observational studies using weighting based on the propensity score: a primer for practitioners

Reducing bias in a propensity score matched-pair sample using greedy matching techniques

Agence nationale de sécurité du médicament et des produits de santé (ANSM)

Anti-inflammatoires non stéroïdiens (AINS) et complications infectieuses graves -Point d'Information. L'Agence nationale de sécurité du médicament et des produits de santé (ANSM)

Risks Related to the Use of Non-Steroidal Anti-Inflammatory Drugs in Community-Acquired Pneumonia in Adult and Pediatric Patients

Non-steroidal anti-inflammatory drugs and covid-19

Acute Respiratory Infection and Use of Nonsteroidal Anti-Inflammatory Drugs on Risk of Acute Myocardial Infarction: A Nationwide Case-Crossover Study

Risk of stroke associated with use of nonsteroidal anti-inflammatory drugs during acute respiratory infection episode

Cardiovascular Safety of Non-Aspirin Non-Steroidal Anti-Inflammatory Drugs: Review and Position Paper by the Working Group for Cardiovascular Pharmacotherapy of the European Society of Cardiology

Vascular and upper gastrointestinal effects of non-steroidal anti-inflammatory drugs: meta-analyses of individual participant data from randomised trials

The Invisible Threat of Nonsteroidal Anti-inflammatory Drugs for Kidneys

Non-steroidal Anti-Inflammatory Drug Induced Acute Kidney Injury in the Community Dwelling General Population and People With Chronic Kidney Disease: Systematic Review and Meta-Analysis

Ibuprofen Attenuates Cardiac Fibrosis in Streptozotocin-Induced Diabetic Rats

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus

A Crucial Role of Angiotensin Converting Enzyme 2 (ACE2) in SARS Coronavirus-Induced Lung Injury

Ibuprofen and Other Widely Used Non

Steroidal Anti-Inflammatory Drugs Inhibit Antibody Production in Human Cells

, and co-medications (angiotensin converting enzyme inhibitors, angiotensinreceptor II blockers, β-blockers, calcium channel blockers, diuretics, nitrates, anticoagulants, inhaled therapy for respiratory for respiratory disease, lipid lowering drugs, opioids, oral glucocorticoids, platelet inhibitors, dipeptidyl peptidase-4 inhibitors, metformin, insulin, sulfonylurea, sodium-glucose cotransporter-2 inhibitors) ‡ IPT weighted multivariable logistic regression model (main model), where the propensity score used was estimated using a multiple logistic regression model that included the following independent variables: age, sex, health insurance type, comorbidities (hypertension, hyperlipidemia, diabetes mellitus, asthma, chronic obstructive pulmonary disease, malignancy, atherosclerosis, chronic renal failure, chronic liver disease, rheumatoid arthritis, osteoarthritis, gastrointestinal conditions, peripheral arterial disease, atrial fibrillation, other arrythmia, ischemic heart disease, pneumonia, psychiatric disorders, depression, thyroid disease), and co-medications (angiotensin converting enzyme inhibitors, angiotensin-receptor II blockers, β-blockers, calcium channel blockers, diuretics, nitrates, anticoagulants, inhaled therapy for respiratory for respiratory disease, lipid lowering drugs, opioids, oral glucocorticoids, platelet inhibitors, dipeptidyl peptidase-4 inhibitors, metformin, insulin, sulfonylurea, sodium-glucose co-transporter-2 inhibitors) (c-statistics: 0.668 for NSAIDs users versus non-users) ⁋ Cardiovascular complications include myocardial infarction, heart failure, and stroke ⁑ Inestimable due to small number of events relative to the large number of confounders