Data-Driven Patterns in Protective Effects of Ibuprofen and Ketorolac on Hospitalized Covid-19 Patients

Benjamin J. Lengerich, Rich Caruana, Yin Aphinayanaphongs

Abstract The impact of nonsteroidal anti-inflammatory drugs (NSAIDs) on patients with Covid-19 has been unclear. A major reason for this uncertainty is the confounding between treatments, patient comorbidities, and illness severity. Here, we perform an observational analysis of over 3000 patients hospitalized for Covid-19 in a New York hospital system to identify the relationship between in-patient treatment with Ibuprofen or Ketorolac and mortality. Our analysis finds evidence consistent with a protective effect for Ibuprofen and Ketorolac, with evidence stronger for a protective effect of Ketorolac than for a protective effect of Ibuprofen.

Introduction

Early in the COVID-19 pandemic, clinical guidance suggested avoiding Non-Steroidal Anti-Inflammatory Drugs (NSAIDs) like Ibuprofen, fearing they might exacerbate symptoms. However, subsequent clinical observations and meta-analyses failed to find evidence of harm, with some even hinting at potential benefits.

The primary obstacle in resolving this debate is confounding by indication. In clinical settings, medication is not assigned randomly; patients receiving NSAIDs may differ significantly from those who do not in terms of baseline health or disease severity. This study utilizes a machine learning framework to "deconfound" these relationships and evaluate the impact of early in-patient NSAID treatment on mortality.

Methodology

The researchers implemented a two-stage machine learning procedure to estimate treatment effects while controlling for baseline risk:

  1. Risk Estimation: A Generalized Additive Model (GAM) was trained on a control group of 3,011 patients who did not receive NSAIDs within the first 24 hours of admission. This model utilized demographics, comorbidities, and lab values available at admission to predict the probability of in-hospital mortality, achieving an AUC of 0.91.
  2. Effect Estimation: This model was then applied to the treatment group to calculate their expected mortality. By comparing the observed mortality to this expected baseline, the researchers derived the Adjusted Risk Difference (ARD) and Adjusted Risk Ratio (ARR).

To prevent data leakage, only lab tests conducted prior to medication administration were used as features for the risk prediction model.

Data and Cohort

The study analyzed deidentified electronic health records (EHR) from 3,108 patients hospitalized for COVID-19 in a New York hospital system between March and August 2020. The total cohort had an average mortality rate of 18.1%. The treatment group consisted of 97 patients who received Ibuprofen or Ketorolac within 24 hours of admission.

Results

The quantitative analysis revealed a significant reduction in mortality risk for patients treated with NSAIDs compared to their predicted outcomes.

Table 1: Observed vs. Expected Mortality and Risk Ratios for NSAID Treatment Groups.

As shown in Table 1, the protective signal was particularly strong for Ketorolac (N=50), which demonstrated an observed mortality of 2.46% against an expected 6.78%, resulting in an ARR of 0.36 (95% CI: 0.24, 0.69). Ibuprofen (N=55) also showed a protective trend with an ARR of 0.65 (95% CI: 0.45, 1.54). When combined, the NSAID group (N=97) showed an ARR of 0.56 (95% CI: 0.40, 0.86), suggesting a substantial reduction in the risk of death. Additionally, the study noted a protective effect for outpatient aspirin use (ARR 0.78).

Discussion and Limitations

The findings suggest that early in-patient treatment with Ibuprofen or Ketorolac is associated with lower mortality rates. The evidence for Ketorolac was statistically significant (p=0.05), warranting further research into its specific mechanisms in respiratory distress.

While the machine learning model successfully adjusted for known confounders like age and comorbidities, the study's observational nature means that unmeasured confounders could still influence the results. Furthermore, the relatively small sample size of the treatment group (N=97) limits statistical power. These results highlight the need for randomized controlled trials (RCTs) to confirm these protective effects and establish formal clinical protocols for NSAID use in COVID-19 patients.

Summary updated

Abstract

Abstract The impact of nonsteroidal anti-inflammatory drugs (NSAIDs) on patients with Covid-19 has been unclear. A major reason for this uncertainty is the confounding between treatments, patient comorbidities, and illness severity. Here, we perform an observational analysis of over 3000 patients hospitalized for Covid-19 in a New York hospital system to identify the relationship between in-patient treatment with Ibuprofen or Ketorolac and mortality. Our analysis finds evidence consitent with a protective effect for Ibuprofen and Ketorolac, with evidence stronger for a protective effect of Ketorolac than for a protective effect of Ibuprofen.
BibTeX
@inproceedings{lengerich2021data,
  author = {Benjamin J. Lengerich and Rich Caruana and Yin Aphinayanaphongs},
  title = {{Data-Driven Patterns in Protective Effects of Ibuprofen and Ketorolac on Hospitalized Covid-19 Patients}},
  booktitle = {AMIA},
  year = {2021},
  url = {https://openreview.net/forum?id=0XlpESrdCL},
}
Figure