Abstract
Acid suppressants are a widely-used class of medications previously linked to an increased risk of aerodigestive infections. However, prior studies of these medications as potentially reversible risk factors for COVID-19 have been conflicting. We performed a case-control study involving clinician-abstracted data from 900 health records across 3 US medical centers. We incorporated sociobehavioral predictors of infectious exposure using geomapping to publicly-available data. We found no evidence for an association between chronic acid suppression and incident COVID-19 (adjusted odds ratio 1.04, 95% CI: 0.92-1.17, P = 0.515). However, we identified several medical and social features as positive (Latinx ethnicity, BMI ≥ 30, dementia, public transportation use, month of the pandemic) and negative (female sex, concurrent solid tumor, alcohol use disorder) predictors of new-onset infection. These results place both medical and social factors on the same scale within the context of the COVID-19 pandemic, and underscore the importance of comprehensive models of disease.
Introduction
Early in the COVID-19 pandemic, a concerning pharmacological hypothesis gained traction: could common acid-suppressing medications, such as Proton Pump Inhibitors (PPIs) and H2-receptor antagonists (H2RAs), increase vulnerability to SARS-CoV-2? The biological rationale suggested that by reducing gastric acidity, these drugs might weaken a natural physiological barrier against viral entry.
However, early observational studies yielded conflicting results. Many of these studies were hampered by "confounding by indication"—the reality that patients prescribed acid suppressants are often older or have more comorbidities, factors that independently increase COVID-19 risk. Furthermore, few studies accounted for Social Determinants of Health (SDOH), the environmental and behavioral factors that dictate actual viral exposure. This study, conducted by researchers across the University of California system (UCSF, UCSD, and UCLA), sought to resolve this debate by integrating granular clinical data with geographic social determinants.
Methodology
To address the limitations of previous research, the study employed a multi-center case-control design using a robust data-gathering framework:
Clinician-Led Data Abstraction
Rather than relying solely on automated database queries, which can be prone to coding errors, physicians manually reviewed 900 health records. This ensured high-fidelity data regarding medication timing, dosage, and the presence of specific comorbidities.
Innovative Geomapping
The study’s most significant methodological contribution was the use of geomapping. By linking patient zip codes to public datasets—including the 2010 US Census, the 2018 American Community Survey, and the US Census Bureau’s Community Resilience Estimates—researchers could estimate sociobehavioral predictors. This allowed the team to factor in local population density, median income, and mobility patterns derived from SafeGraph and Google Community Mobility Reports.
Statistical Framework
The final analysis utilized 861 medical records (428 positive cases and 433 negative controls). The team employed multiple imputation to address missing data and a fixed-effect logistic regression model to identify the most significant predictors of infection.
Key Findings
The Acid Suppression Myth
After adjusting for medical and social confounders, the purported link between acid suppression and COVID-19 infection disappeared. The adjusted odds ratio (aOR) was 1.04 (95% CI: 0.92-1.17, P=0.515), indicating no statistically significant association. This suggests that previous concerns were likely the result of unmeasured confounding variables rather than a true biological vulnerability caused by the medication.
The Impact of Social Determinants
The study demonstrated that social factors were often more predictive of infection than medical ones.
- Public Transportation: Residing in an area with high public transit usage (>4.4% of the population) was a significant predictor of infection (aOR 1.1). This underscores that the ability to socially distance is frequently a function of economic privilege.
- Ethnic Disparities: Latinx ethnicity remained a strong positive predictor of infection (aOR 1.25), even after controlling for income and mobility, highlighting systemic vulnerabilities that traditional medical models often overlook.
Medical Predictors and Protective Behaviors
Several medical conditions were confirmed as risk factors, including a BMI ≥ 30 (aOR 1.11) and dementia (aOR 1.32). Conversely, some factors appeared "protective":
- Female Sex: Associated with a lower risk (aOR 0.89).
- Solid Tumors and Alcohol Use Disorder: Interestingly, patients with concurrent solid tumors (aOR 0.86) or alcohol use disorder (aOR 0.81) showed lower infection rates. The authors hypothesize this is likely due to behavioral modification; these high-risk or vulnerable individuals may have practiced more stringent isolation or adhered more strictly to masking and social distancing guidelines.
Discussion and Limitations
While the study provides a more comprehensive view of infection risk, it is subject to certain limitations. The reliance on patients who were actively tested for COVID-19 introduces potential selection bias, as the results may reflect testing patterns as much as infection patterns. Additionally, the use of zip-code-level SDOH data carries the risk of the "ecological fallacy," where population-level trends may not perfectly describe individual behavior.
Conclusion
This research clarifies that chronic acid suppression is not a significant risk factor for COVID-19. More importantly, it demonstrates the necessity of integrating social and behavioral data into clinical research. By placing medical and social factors on the same analytical scale, the study reveals that the risk of infection is driven as much by a patient's environment and socioeconomic constraints as by their underlying physiology. Future models of infectious disease must adopt this holistic approach to accurately identify and protect vulnerable populations.
Summary updated