key: cord-0891212-natsmzzc
authors: Israelsen, Simone Bastrup; Ernst, Martin Thomsen; Lundh, Andreas; Lundbo, Lene Fogt; Sandholdt, Håkon; Hallas, Jesper; Benfield, Thomas
title: Proton pump inhibitor use is not strongly associated with SARS-CoV-2 related outcomes: A nationwide study and meta-analysis
date: 2021-05-11
journal: Clin Gastroenterol Hepatol
DOI: 10.1016/j.cgh.2021.05.011
sha: 5c6e62d75849d6a3789f31553560ff51e51016e0
doc_id: 891212
cord_uid: natsmzzc

Background and aims Proton pump inhibitor (PPI) use has been associated with increased risk of SARS-CoV-2 infection and severe outcomes. However, meta-analyses show unclear results leading to uncertainty regarding the safety of PPI use during the ongoing COVID-19 pandemic. Methods We conducted a nationwide observational study including all SARS-CoV-2 cases (n=83,224) in Denmark as of December 1, 2020. The association of current PPI use with risk of infection was examined in a case-control design. We investigated the risk of severe outcomes, including mechanical ventilation, ICU admission or death, in current PPI users (n=4,473) compared to never users. Propensity score matching was applied to control for confounding. Finally, we performed an updated meta-analysis on risk of SARS-CoV-2 infection and COVID-19 mortality attributable to PPI use. Results Current PPI use was associated with increased risk of infection; adjusted odds ratio (OR) 1.08 (95% CI 1.03-1.13). Among SARS-CoV-2 cases, PPI use was associated with increased risk of hospital admission; adjusted relative risk (RR) 1.13 (1.03-1.24), but not with other severe outcomes. The updated meta-analysis showed no association between PPI use and risk of infection or mortality; pooled OR 1.00 (95% CI 0.75-1.32) and RR 1.33 (95% CI 0.71-2.48). Conclusion Current PPI use may be associated with an increased risk of SARS-CoV-2 infection and hospital admission, but these results with minimally elevated estimates are most likely subject to residual confounding. No association was found for severe outcomes. The results from the meta-analysis indicated no impact of current PPI use on COVID-19 outcomes.

Acid suppressive drugs -especially proton pump inhibitors (PPI) -are hypothesized to influence the susceptibility to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and affect outcomes in patients diagnosed with Coronavirus Disease 2019 . This concern is based on their suppression of stomach acid and an association with an increased risk of infection and, in particular, with a risk of pneumonia [1] [2] [3] . SARS-CoV-1 has been reported to be inactivated by acidic conditions 4 , and SARS-CoV-2 may directly invade the gastrointestinal epithelium of infected patients 5 .

In July 2020, a large survey found that individuals using PPI had a higher odds of reporting a positive SARS-CoV-2 test 6 . In contrast, Lee et al. reported that current PPI use was associated with an increased risk of severe outcomes of COVID-19, but not with risk of infection 7 . Similarly, Zhou et al. reported an association with severe outcomes, including need for intensive care unit (ICU) admission, intubation or death 8 .

Subsequently, two meta-analyses 9,10 including 14 further observational studies of PPI use in patients with COVID-19 found that PPI use was associated with an increased risk of severe outcomes. However, most of the studies had low statistical precision of estimates, only some controlled fully for confounding and they were all quite heterogenous with study populations from different countries, including patients with few or many comorbidities and with or without requiring hospitalization. Currently, use of PPI and its possible association with risk of infection and disease severity remains uncertain.

In this nationwide study of all individuals tested in Denmark as of December 1 2020, we examined the association between current use of PPI and risk of SARS-CoV-2 infection, and the risk of

The original study protocol and analysis plan is available from the EU PAS Register with identification number EUPAS35835: http://www.encepp.eu/encepp/viewResource.htm?id=37050

Data on all Danish residents tested for SARS-CoV-2 RNA as of December 1, 2020 was retrieved from the Danish Microbiology Database and individually linked to other nationwide health-care registries, as described previously 11 . SARS-CoV-2 infection was verified by a positive real-time polymerase chain reaction (RT-PCR) on an oro-or nasopharyngeal swab or lower respiratory tract specimen. The individuals' medical history included ICD-10 diagnoses registered within 10 years before the date of first positive SARS-CoV-2 test (index date). Comorbidities included were peptic ulcer, chronic obstructive pulmonary disease, asthma, ischemic heart disease, stroke, heart failure, diabetes mellitus, renal failure, and liver cirrhosis. Lifestyle factors included smoking-and alcohol-related diagnoses (Table 1) .

Major psychiatric disorders (schizophrenia, schizoaffective disorders, manic episodes and bipolar disorder) were added along with available frailty markers based on health care utilization (number of admissions within the past three years).

Data on patients' medications included current use (within 90 days before index date) of inhaled corticosteroids and bronchodilators, systemic corticosteroid treatment, immunomodulating treatment, H2-receptor antagonists, non-steroid anti-inflammatory drugs (NSAID), anticholinergic J o u r n a l P r e -p r o o f agents, antibiotics, blood pressure lowering drugs, lipid lowering drugs, glucose lowering drugs, antiplatelets, anticoagulants, treatment to support alcohol abstinence and smoking cessation, and antipsychotic agents (Supplementary Table S1 ). Finally, the total burden of comorbidity was assessed based on the Charlson's Comorbidity Index and classified as 0, 1-2 or ≥3.

The case-control study included all individuals tested for SARS-CoV-2 RNA and examined the risk of infection with current PPI use in cases (test-positive) vs. controls (test-negative). Cases were matched on sex and birth year with up to four controls each. Control subjects were Danish residents alive at the index date of the case and who had been tested negative for SARS-CoV-2 RNA. To account for changing testing criteria and in-hospital capacity during the study period, cases were matched with controls on the week wherein the test was performed.

The cohort study included the test-positive population and investigated the risk of hospital admission and severe outcomes -including ICU admission, mechanical ventilation and deathwithin 30 days of the first positive SARS-CoV-2 RNA test. Patients were followed from date of first positive test until death, migration or end of follow up (30 days).

Current PPI use was defined as having redeemed a prescription of PPI within 90 days before the first positive SARS-CoV-2 RNA test (index date), although only including prescriptions before possible hospitalization. Individuals were classified as former users if they had redeemed a J o u r n a l P r e -p r o o f prescription more than 90 days before the index date. Never use was defined as never having redeemed a prescription since 2005. The specific PPI (pantoprazole, lansoprazole, omeprazole, or esomeprazole) was registered for each user. Dose levels were defined as low or high dose if the prescribed tablet strength was either below or equal to/above 30 mg, respectively. The choice of dose level cut-off was based on usual tablet sizes and a standard once-daily dosing regimen.

In the case-control study, the outcome was a positive SARS-CoV-2 RNA test during the study period.

In the cohort study, the primary outcome was hospital admission within 30 days after a positive In the case-control study, conditional logistic regression was performed to examine a possible association between current PPI use and risk of SARS-CoV-2 infection. Results are presented as odds ratios (OR) with 95% confidence intervals (CI). Confounding by age, sex and calendar time were handled by the risk set sampling and the matched analysis, as described above. Other potential confounders, including comorbidity and current medication use, are listed in Table 1 and included in the multivariable modelling. Comparisons between groups were performed using Fisher's exact test or t-test, as appropriate.

In the cohort study, an individual propensity score of drug exposure was estimated by logistic regression based on age, sex, comorbidities and current medication use, as listed in Table 3 .

Propensity scores were used to match the exposed and unexposed groups to adjust for preexisting differences in risk factors. Covariate balance was quantified by standardized mean differences (SMD) with values below 0.1 considered acceptable. Relative risks (RR) for hospital admission and severe outcomes in the exposed (current PPI use) vs. the unexposed (never PPI use) groups were calculated by log binomial regression and presented as crude (unmatched) and adjusted (propensity-score matched) estimates with 95% CI.

To determine the robustness of the estimates, sensitivity analyses on the chosen PPI exposure window were performed in both studies and included comparisons of current vs. former use and J o u r n a l P r e -p r o o f 13 former vs. never use. Finally, a post-hoc analysis of the dose-effect of PPI was computed comparing low-dose vs. high-dose regimens, as defined above.

In the meta-analyses, we preferred adjusted estimates to unadjusted estimates. In order to include as much information as possible, we extracted estimates for the effect measure most frequently reported in the studies, e.g. OR. For studies using a different effect measure, unadjusted results based on reported events were calculated. Inverse variance random effects models were applied to estimate either OR or RR with 95% CI. We described statistical heterogeneity using I 2 

Cases (n=83,224) and controls (n=332,799) had a median age of 36 and an equal sex distribution.

Less than 15% had a score of one or more in the Charlson's Comorbidity Index (Table 1 ). The predominant comorbidity was ischemic heart disease, followed by asthma, diabetes, stroke and chronic obstructive pulmonary disease. Blood pressure lowering drugs were the most common drugs used in both cases and controls. Other common medications used included lipid lowering drugs, glucose lowering drugs, antibiotics and NSAID.

Current PPI users had a crude OR of 1.04 (95% CI 1.00-1.08) of SARS-CoV-2 infection compared to never PPI users. After including all the covariates in the regression model, the adjusted OR was 1.08 (95% CI 1.03-1.13). The sensitivity analyses (current vs. former use and former vs. never use) yielded similar results with crude and adjusted ORs around 1.0 ( Table S3 ).

In the unmatched SARS-CoV-2 RNA positive population, current users of PPI (n=4,473) were compared to never users of PPI (n=59,413). Current users were older (median 60 vs. 29 years), fewer were male (45% vs. 53%) and they had a greater comorbidity burden with a larger proportion of registered diagnoses across all the included comorbidities ( Table 3 ). Use of other medications was also consistently more frequent in current users compared to never users (Table   3 ).

After propensity score matching, 3,955 individuals persisted in each group and the difference in characteristics was significantly reduced with 89% of the SMDs below 0.05 and none with a SMD above 0.11 (Table 3) . However, important comorbidities and prior health care utilization remained more frequent in current users.

In the propensity score matched analyses, current PPI users had an increased risk of hospital admission compared to never PPI users (19% vs. 16%) corresponding to an adjusted RR of 1.13 (95% CI 1.03-1.24) ( Table 4 ). Current and never PPI users had comparable risks of ICU admission, mechanical ventilation, death and severe outcomes (ICU or death) at around 2%, 1%, 4% and 6%, respectively. The RR for these secondary outcomes in the matched analysis had estimates just below or at 1.0 with 95% CIs on both sides of 1 ( Table 4 ).

The sensitivity analysis comparing current and former users showed risks of 21% and 19% for hospital admission with a similar adjusted RR of 1.08 (95% CI 1.00-1.18). When comparing former users to never users, the risks of hospital admission were comparable between the groups (Supplementary Table S4 ).

In the dose-response analysis, the risk of hospital admission for current PPI users with a high-dose regimen was increased with an adjusted RR at 1.19 (95% CI 1.07-1.32), while current users with a low-dose regimen had an adjusted RR of 1.03 (95% CI 0.90-1.17), when compared to never PPI users. All secondary outcomes were not statistically significant (Supplementary Table S5) .

When comparing the different classes of PPI, neither consistent nor statistically significant differences were observed in risk of hospital admission or severe outcomes (Supplementary Table   S6 ).

In the updated literature search, eight studies 6,7,12-16 , including our current study, investigated the association between current PPI use and risk of SARS-CoV-2 infection. In addition, a recent nationwide study was included alongside a relevant preprint not detected in the initial search 17, 18 .

Study characteristics are shown in Supplementary Table S7 . Analysis of these studies, comprising 730,941 individuals, resulted in a pooled OR of 1.00 (95% CI 0.75-1.32), with considerable between-study heterogeneity (I 2 =98%) (Figure 1 ). When divided into subgroups based on risk of J o u r n a l P r e -p r o o f bias, the analysis of the five studies with high risk of bias showed an OR of 1.13 (95% CI 0.53-2.41) for current PPI use compared to never use ( Figure 1 ). Further, the five studies with low risk of bias showed a decreased risk of infection, though with low statistical precision, OR of 0.86 (95% CI 0.67-1.10) (Figure 1) .

Only four studies 12, 19, 20 reported the association between current PPI use and mortality alone, including 4,150 current PPI users (Supplementary Table S7 ). The overall analysis showed an RR of 1.33 (95% CI 0.71-2.48) in current PPI users compared to never users, though estimates differed between studies with low risk of bias, RR 0.85 (95% CI 0.70-1.03), and high risk of bias, RR 2.37 (1.53-3.67) (Interaction test: p<0.0001) (Figure 2 ). The between-study heterogeneity was substantial (I 2 = 84%).

Overall, current use of PPI was associated with an increased risk of SARS-CoV-2 infection, in the case-control study, and an increased risk of hospital admission, in the cohort study including testpositive individuals. However, both estimates were close to 1.0 and may be caused by residual confounding. Moreover, current use of PPI was not associated with increased risk of severe outcomes that included ICU admission or death.

The lack of a clinically significant association with increased risk of infection in current PPI users in our study is consistent with most previous reports and our updated meta-analysis. In contrast, the meta-analysis by Li et al. 10 showed that current PPI use was associated with increased odds of infection, although the estimate was statistically uncertain.

Our results show a possible association between current PPI use and increased risk of hospital admission in SARS-CoV-2 RNA positive individuals. However, we could not confirm the association with increased risk of severe outcomes of COVID-19 with current PPI use as reported in previous meta-analyses 9, 10, 21, 22 . In addition, a multicenter study from North America and a nationwide UK study, not included in any of the meta-analyses, did not find an association between PPI use and severe outcomes either 18, 23 .

Notably, all the studies reporting the impact of PPI use in SARS-CoV-2 infected individuals differ substantially.

First, the study populations are rather heterogenous including different nationalities, and ranging from young resourceful individuals 6 to elderly with several comorbidities 12, 13 and between hospitalized patients and residents without hospital contact. The SARS-CoV-2 RNA positive population in our study had comorbidities corresponding to previous reports of diagnoses commonly present in infected individuals 24 Although a wide range of relevant comorbidity and medication was used to adjust our analyses there may inherently remain residual confounding by imperfectly measured, unmeasured or unknown factors. Additionally, the propensity score matching failed to fully account for important differences in comorbidities and prior health care use between current and never users. Use of PPI has been associated with socioeconomic deprivation and frailty, but this information was not available through the applied registries. Similarly, information on the indication for use of PPI was unavailable except for a prior diagnosis of peptic ulcer. Interestingly, Luxenburger et al. 20 found that gastroesophageal reflux disease was independently associated with severe courses of COVID-19, thereby raising the question whether the indication for the drug prescription accounts for the association rather than the drug per se. Further, PPI is linked to overprescribing which could be another (unknown) marker of frailty 25 . Low dose PPI is available as over-the-counter medicine in Denmark which could give rise to information bias, affecting the results toward the null.

Finally, for the "test-negative" case-control study of PPI use as risk factor for contracting SARS-CoV-2-infection, there is a potential bias if PPI affects the chance of becoming a control, i.e., being J o u r n a l P r e -p r o o f tested negative. In the early stages of the pandemic, most test-negative individuals had other viral upper respiratory infections, which to our knowledge is not associated with PPI use in general.

Our updated meta-analysis showed a possible increased risk of COVID-19 mortality, but no risk of SARS-CoV-2 infection. However, neither of these results were statistically significant. Indeed, when we restricted our analyses to studies with low risk of bias the point estimates decreased to below 1.0 in both analyses indicating that the conflicting results from the included studies and former meta-analyses arise from between-study differences rather than an actual impact of current PPI use on COVID-19 outcomes.

In conclusion, our data support that PPI use in general is safe with regards to risk of SARS-CoV-2 infection and severe COVID-19 outcomes. The risk of hospital admission was increased for current PPI users, but this minimally elevated RR is seemingly explained by residual confounding.

Following hospital admission there was no association with severity of COVID-19 and use of PPI.

Finally, our updated meta-analysis indicated no impact of current PPI use on COVID-19 outcomes, thereby suggesting that previous conflicting results are more likely due to differences in study design and population.

. J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f

To date, uncertainty prevails regarding the safety of proton pump inhibitor use in relation to SARS-CoV-2 infection as existing evidence has indicated both protective and harmful effects.

In this nationwide observational study, we found a slightly increased risk of infection and hospital admission in 4,473 current proton pump inhibitor users, but no association with other severe outcomes. Our updated meta-analysis showed no association with risk of infection or mortality. 

Use of proton pump inhibitors and the risk of community-acquired pneumonia: a population-based case-control study

Risk of community-acquired pneumonia with outpatient proton-pump inhibitor therapy: a systematic review and meta-analysis

Acid-suppressive medication use and the risk for hospital-acquired pneumonia

Inactivation of the coronavirus that induces severe acute respiratory syndrome, SARS-CoV

Evidence for Gastrointestinal Infection of SARS-CoV-2

Increased Risk of COVID-19 Among Users of Proton Pump Inhibitors

Severe clinical outcomes of COVID-19 associated with proton pump inhibitors: a nationwide cohort study with propensity score matching

Proton pump inhibitor or famotidine use and severe COVID-19 disease: a propensity score-matched territory-wide study

Use of proton pump inhibitors and risk of adverse clinical outcomes from COVID-19: a meta-analysis

Do proton pump inhibitors influence SARS-CoV-2 related outcomes? A meta-analysis

Existing Data Sources in Clinical Epidemiology: The Danish COVID-19 Cohort

COVID-19 in patients with hepatobiliary and pancreatic diseases in East London: a single-centre cohort study

Use of distinct anti-hypertensive drugs and risk for COVID-19 among hypertensive people: A population-based cohort study in Southern Catalonia

Are Proton Pump Inhibitors Contributing to SARS-COV-2 Infection?

Association of Previous Medications with the Risk of COVID-19: A Nationwide Claims-Based Study from South Korea. medRxiv

Exploring Drugs and Vaccines Associated with Altered Risks and Severity of COVID-19: A UK Biobank Cohort Study of All ATC Level-4 Drug Categories. medRxiv

Interest of Proton Pump Inhibitors in Reducing the Occurrence of COVID-19: A Case-Control Study. Preprints

Effect of Acid Suppressants on the Risk of COVID-19: A Propensity Score-Matched Study Using UK Biobank

Pre-hospitalization proton pump inhibitor use and clinical outcomes in COVID-19

Treatment with proton pump inhibitors increases the risk of secondary infections and ARDS in hospitalized patients with COVID-19: coincidence or underestimated risk factor?

Lack of Consistent Associations Between Pharmacologic Gastric Acid Suppression and Adverse Outcomes in Patients With Coronavirus Disease 2019: Meta-Analysis of Observational Studies

A meta-analysis: Proton pump inhibitors current use and the risk of Coronavirus Infectious Disease 2019 development and its related mortality

Digestive Manifestations in Patients Hospitalized with COVID-19

Transmission, Diagnosis, and Treatment of Coronavirus Disease 2019 (COVID-19): A Review

Overprescribing proton pump inhibitors