Proton pump inhibitors and risk of severe COVID-19 in older people

Abstract

Key Points

• Ageing is a risk factor of severe COVID-19.

• PPIs are widely used in older people.

• Our study shows PPIs could be a risk factor of severe COVID-19 in older people.

• There is a dose effect of PPIs on the risk of severe COVID-19.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a respiratory virus, which can cause systemic symptoms and lead to death, particularly in older people who are at a higher risk of developing a severe form of the disease [1].

SARS-CoV-2 can enter the human body by utilising the angiotensin-converting enzyme-2 (ACE2) surface receptor, which is expressed in the respiratory system and gastrointestinal (GI) tracts. Hence, apart from the classical respiratory symptoms in COVID-19, gastritis, enteritis and colitis are present in half of the the patients [2] and the GI system is now thought to be a major site of entry and replication for SARS-CoV-2 [3, 4]. Additionally, COVID-19 is a medical condition that carries a high risk of GI bleeding, particularly upper GI bleeding [5–7] and antiacid drugs may have a protective effect.

Proton pump inhibitors (PPIs) are widely used to treat GI disorders. They are usually considered to have a satisfactory safety profile, with few potential side effects and have therefore become one of the most commonly prescribed medications especially in older people [8, 9].

However, recent studies report potential long-term side effects of prescribed PPIs [10, 11], including malabsorption of vitamin B12 and/or iron, increased risk of bone fracture, chronic kidney failure and increased mortality [12–14]. Prolonged consumption of PPIs has been identified as a risk factor for developing pneumonia, enteric infection including Clostridium difficile colitis and multidrug resistant bacteria carriage [15–18]. Interestingly, the results from a recent double-blinded, randomised, placebo-controlled trial evaluating cardiovascular outcomes in patients with stable atherosclerotic vascular disease have confirmed the increased risk of acute enteric infections, among the suspected harms of PPI therapy [19]. Moreover, in a recent study based on a large French community pharmacy drug dispensation database, Vilcu et al. demonstrated an association between continuous PPIs therapy and the occurrence of acute viral gastroenteritis [20].

Given the involvement of the GI system in the pathogenesis of COVID-19 and the increased risk of GI infections with long-term use of PPIs, others authors have hypothesised that PPIs could be a risk factor for COVID-19 and/or severe COVID-19 [21]. Via an online national health survey conducted in 2020 among community-dwelling Americans, Almario et al. reported evidence of an independent, dose–response relationship between the use of PPI and COVID-19 positivity [22]. In a retrospective study, Luxenburger et al. identified PPIs treatment as a significant risk factor with a 2-fold risk for secondary infection and acute respiratory distress syndrome (ARDS) in hospitalised COVID-19 patients [23]. From a Korean nationwide health insurance database, Lee et al. reported similar results, suggesting that PPIs usage did not increase the susceptibility to SARS-CoV-2 infection but was associated with severe clinical outcomes [24]. All these studies suggest PPIs have a negative effect in COVID-19, however, a recent meta-analysis of randomised control trials and cohort studies had not confirmed these results [25].

Considering the broad use of PPI in older people, the potential role of PPI in COVID-19 could be of dramatic significance. The objective of our study was to evaluate the potential link between PPIs prescription and the severity of COVID-19 infections among older people.

Methods

All patients aged ≥65, admitted to Saint-Etienne University Hospital (France) due to a diagnosis of COVID-19 (confirmed by SARS-CoV-2 RT-qPCR performed on nasopharyngeal swabs) between 1 March 2020 and 15 February 2021 were retrospectively included in the cohort.

Epidemiological, clinical and biological data were extracted from the patients’ hospital medical records. We retrieved information about age, sex, clinical characteristics, biological data at admission and drugs. The Charlson Comorbidity Index (CCI) was calculated as previously described [26].

Specific data regarding PPIs were also collected: PPIs type (pantoprazole, omeprazole, esomeprazole, lansoprazole, rabeprazole), dosage (10, 15, 20, 30, 40 mg) and administration frequency (once a day or twice a day). Then PPIs-users among the cohort were divided into three groups according to the type of PPIs and dosage per day: half dose (pantoprazole 20 mg, omeprazole 10 mg, esomeprazole 20 mg, lansoprazole 15 mg, rabeprazole 10 mg), full dose (pantoprazole 40 mg, omeprazole 20 mg, esomeprazole 40 mg, lansoprazole 30 mg, rabeprazole 20 mg) and double dose (pantoprazole 80 mg, omeprazole 40 mg, esomeprazole 80 mg, lansoprazole 60 mg, rabeprazole 40 mg) of PPIs. PPIs-users among the cohort were also divised into two groups according to the duration of the PPIs usage: long-term (i.e. use of PPIs prior to admission) and recent (i.e. first prescription of PPI during the current hospitalisation) PPIs-users.

The primary outcome was the frequency of severe COVID-19, defined as the composite of poor outcome: death and/or transfer in an intensive care unit (ICU) and/or use of ventilation with an oxygen-requirment ≥9 L/min and/or use of mechanical ventilation.

Statistical analysis

Continuous variables were expressed as mean (standard deviation) when normally distributed or median [interquartile range] when non-normally distributed. Categorical variables were expressed as absolute numbers (percentage).

To assess the association between the use of PPIs and the development of a severe COVID-19, the primary analysis was performed with an Inverse Probability of Treatment Weighing (IPTW) method based on a propensity score. IPTW analysis consists of using a propensity score to balance baseline patient characteristics in the exposed and unexposed groups, creating a pseudopopulation in which confounders are equally distributed across exposed and unexposed groups. We computed each patient’s propensity to have PPIs administered to them by using a logistic regression model including all the potential confounders, under the assumption on no unmeasured confounders.

Patients who received PPIs were assigned a weight of 1/(propensity score), while those who did not receive PPIs were assigned a weight of 1/(1-propensity score). Using those weights, a pseudo-cohort of patients who were similar with respect to every measured characteristic, except PPIs intake, was created.

To assess the balancing performances of the propensity score, absolute standardised difference between the treatment groups was calculated, and displayed on a love plot. A standardised difference for a given variable of 10% or less was considered a well-balanced result. In addition, to validate and assess the capability of the propensity score to correctly predict the treatment group, we calculated the Area Under Curve (AUC) of the Receiver Operator Characteristics (ROC) curve of the model and its 95% confidence interval (CI). An AUC > 0.6 would be considered as satisfying, an AUC > 0.7 as optimal and an AUC close to 1 as unusable (the covariates define perfectly the treatment group).

Three sensitivity analyses were performed to assess the robustness of the results. First, to detect an eventual residual confounding bias, we performed a negative control outcome analysis, using gastric ulcer as an independant variable instead of PPIs intake in the same IPTW model as previously. In this analysis, gastric ulcer should not be associated with a severe COVID-19 infection, validating our negative control.

The second sensitivity analysis consisted of a model only adjusted on the propensity score, with the use of PPIs as an independant variable and the development of a severe form of COVID-19 as the dependant variable.

Third, a multivariate model was built a-priori, before accessing the data. Potential confounders were included in the model based on previous knowledge from the literature and expert’s opinion, as well as clinically relevant variables, while selecting the most important variables to respect the principle of parsimony. The final multivariate model included the following variables as adjustment variables: age, sex, cardiovascular diseases, cerebrovascular diseases, chronic kidney disease, chronic hepatic disease and diabetes mellitus, which are known to be risk factors for severity or mortality in COVID-19. We also adjusted for the cycle threshold (CT) value, which is the number of cycles required for detectable fluorescence above a set threshold, indicating the presence and quantity of viral RNA in the sample (with a lower CT suggesting a higher viral load).

Secondary analyses were performed, including sub group analysis with patients divided into four groups of different PPIs dosages (PPIs-non users, half dose, full dose and double dose PPIs-users) to identify if the dosage increased the risk. We also divided PPIs usage into two groups, to compare long-term PPIs-users and recent PPIs-users to PPIs-non users.

Results are reported as odds ratio (OR) along with their 95% CI. Patient with missing values in at least one of the covariates in a model had to be excluded from said model.

All statistical tests were two-sided, with P < 0.05 considered as statistically significant. Analysis were performed using the R software for statistical computing and graphics, version 4.2.2 (with the packages ‘cobalt’ version 4.3.1, and ‘pROC’ version 1.17.01).

This study was approved by the ethics committee of the Saint-Etienne University Hospital (Ethical reference: IRBN1292021/CHUSTE).

Results

From 1 March 2020 to 15 February 2021, a total of 1,352 patients aged ≥65 years infected with COVID-19 were hospitalised in our hospital. Among them, 518 patients were excluded (Figure 1).

Figure 1

Open in new tabDownload slide

The final cohort included 834 hospitalised inpatients, with a median age of 83, IQR [76–89] and 52.8% were male (Table 1). A total of 410 patients had a PPIs prescription, 358 (87.3%) were long-term PPIs-users and 52 (12.7%) were recent PPIs-users. In our cohort, pantoprazole was the most prescribed PPIs (69.3%) then esomeprazole (13.9%), omeprazole (8.5%), lansoprazole (6.3%) and rabeprazole (2.4%). Both groups were comparable regarding demographic, co-morbidities and clinical data. Lymphocytes were significantly lower and D-dimer significantly higher in PPI-users.

Among PPIs-users, 163 (39.8%) patients developed severe COVID-19 versus 113 (26.7%) in PPIs-non users (OR = 1.59 [1.18–2.14]; P < 0.05), with 110 (26.8%) and 85 (20.0%) deaths, respectively (Figure 2). Details of the composite endpoint are shown in Appendix S1. No absolute standardised differences >10% between the two pseudo-populations were observed (Appendix S2–S3), and the AUC of the ROC curve = 0.610 [0.571–0.649] (Appendix S4).

Figure 2

Forest plot of the risk factors for severe COVID-19.

Open in new tabDownload slide

For the sensitivity analysis (Figure 2), concerning the negative control outcome analysis with the same IPTW model, the occurring of a severe form of COVID-19 was not associated with the presence of a gastric ulcer (OR = 1.27 [0.51–2.3], P = 0.596). With the model only adjusted on the propensity score and the multivariate model, the results were similar to those of the IPTW analysis, with OR = 1.6 [1.18–2.17], P = 0.003 and OR = 1.93 [1.42–2.63], P < 0.001, respectively.

In the secondary analysis, the double dose PPI-users had a higher risk of developing severe COVID-19 (OR = 3.36 [1.17–9.66]; P < 0.05) than the full dose PPI-users (OR = 2.15 [1.22–3.76]; P < 0.05) and the half dose PPI-users (OR = 1.64 [1.13–2.37]; P < 0.05). Likewise, recent PPIs-users (OR = 3.13 [1.69–5.79], P < 0.001) and long PPIs-users (OR = 1.79 [1.30–2.46], P < 0.001) had a higher risk of developing severe COVID-19 compared with PPIs-non users (Figure 2 and Appendix S5).

Discussion

In this study based on a large cohort of older people hospitalised for COVID-19, PPIs use was associated with an increased risk of severe COVID-19. To our knowledge, this is the first study specifically investigating this association in patients older than 65.

Since the onset of the pandemic, numerous studies have been conducted to analyse the potential association between PPIs exposure and the risk of COVID-19 occurrence or development of severe COVID-19 [27–31]. Such an association could be explained in several ways. First, PPIs may facilitate secondary infection [23]. Indeed, co-infections are frequent in COVID-19 and can be responsible for severe outcome [32]. Second, PPIs have been identified in some studies as an independant risk factor for severe COVID-19 with increased risk of ICU admission and mechanical ventilation use [24, 33, 34]. In a study investigating the impact of medication on severe COVID-19, Mc Keigue et al. reported that severe COVID-19 was associated with polypharmacy and more precisely with antipsychotic agent and PPI use [35].

In our cohort, the increased risk of severe COVID-19 was PPIs dose-dependent, indicating a dose-effect relationship. Similarly, Almario et al. showed an increased risk with a dose-dependent effect [22]. In a nationwide database study, Israelsen et al. reported similar findings, showing a slight increase in COVID-19 infection among PPIs users compared with individuals who were PPIs-non users. In the dose–response analysis, individuals with current low-dose PPIs use had a lower risk of SARS-CoV-2 infection [36].

Interestingly, in our cohort, recent PPIs-users have an increased risk of severe COVID-19 than long term PPIs-users. An explanation could be that admitted inhospital older patients with severe COVID-19 receiving steroids were systematically treated with PPIs.

However, these results have not been confirmed. Other retrospective cohort studies did not confirm the relationship between PPIs and COVID-19 [37, 38], PPIs and mortality [39, 40] or adverse clinical outcomes in COVID-19 [38, 41, 42], even in older people in which a protective effect has even been discussed [43]. Other national database did not confirm a positive association between PPIs and COVID-19 [24, 41] leading to ongoing debates and discussions [44–46]. One hypothesis to explain these discrepancies could be that PPIs represent a marker of frailty rather than an independant risk factor for serious disease.

The association between PPIs consumption and the risk of viral infection is supported by previous studies, and various mechanisms have been suspected [47, 48]. Gastric acidity serves as a nonspecific defence mechanism for the stomach to eradicate ingested pathogens by preventing infectious agents from reaching the intestine. PPIs covalently bind to the sulfhydryl groups within the cysteine residues of the H+/K+ adenosine triphosphatase (H⁺/K⁺ ATPase) on the gastric parietal cells’ membrane and reduce acid gastric secretion, consequently impairing virus inactivation [49]. Moreover, reduction of gastric acid secretion leads to hypochlorhydria, which was suspected to be responsible for reducing microbial diversity and eradication of ingested pathogens [48, 50, 51]. Likewise, PPIs could be responsible for microbiota dysbiosis in COVID-19 patients, which may increase the potential for secondary infections [52], but this hypothesis needs to be confirmed. Moreover, hypochlorhydria leads to reduce micronutrients absorption such as vitamin B12, iron and magnesium [53–55]. Micronutrients deficiency impairs immune system and makes individuals more susceptible to infectious disease [56, 57]. Additionally, PPIs can inhibit the functions of polymorphonuclear neutrophil, T lymphocytes and natural killer cells [58–60]. This could partially explain why PPIs could be associated with severe COVID-19. PPIs are also known for increasing community-acquired pneumonia risk by potentially impairing the immune system [61]. SARS-CoV-2 can infect host cells and enter the human body when viral spike protein binds to the cell surface receptor ACE2 which is mainly expressed in the respiratory system and the GI tract, especially in the intestine [62]. Faecal-oral transmission has secondarly emerged as a plausible route for SARS-CoV-2 infecton, as COVID-19 patients have detectable infectious viruses in their stools suggesting the digestive tract might be a viral replication site [63]. A higher expression of ACE2 allows for higher viral entry into cells, resulting in higher viral loads, and previous study reported that PPIs could increase ACE2 expression and facilitate the internalisation of SARS-CoV-2 into human cells [64, 65]. All these mechanisms could lead to high viral loads, which are associated with high mortality.

As we discussed, gastric acid reduction could be an explanation of an increased susceptibility to infections. H2-receptor antagonists (H2RAs), like famotidin, are others antiacid drugs. Previous studies reported various results with better clinical outcomes [66] but not others [67]. Moreover, some research teams hypothetise a beneficial effect of famotidine in COVID-19 symptoms resolution [68–70] which has not been confirmed by a recent meta-analysis [71]. Finally, in a recent meta-analysis, Kim et al. reported evidence that PPIs use, but not H2RAs use, was associated with increased risk of adverse outcomes in COVID-19 patients [72]. In subgroups analyses, this result was still significant for the population aged ≥60 years, which is consistent with our study.

Our study has some strengths. First, the primary outcome was a composite endpoint that included death, admission to the ICU and oxygen requirement exceeding 9 L/min or mechanical ventilation use. During the COVID-19 pandemic, because of the ICU capacity crisis, older patients were often recused to ICU or invasive ventilation. To avoid misclassfying patients, we have considered the use of high oxygen levels as an indicator of COVID-19 severity. Second, we have used an IPTW analysis using a propensity score to balance baseline patient characteristics in the exposed and unexposed groups, creating a pseudopopulation in which confounders are equally distributed across exposed and unexposed groups [73, 74]. Moreover, we conducted several sensitivity analysis, a negative control that considered an eventual residual confounding bias, a multivariate analysis and a model only adjusted on the propensity score, which yielded the same results as the primary IPTW analysis. Higher risk of severe COVID-19 in PPIs-users remained statistically significant in all these analyses.

Our study also has limitations. First, we have only included hospitalised patients. Hence, PPIs effect cannot be extrapoled to COVID-19 outpatients. However, these patients are of special interest due to a higher risk of complications. Our cohort is monocentric and we excluded a large proportion of patients, approximately a third, due to missing data. Data were collected retrospectively from hospital medical records with potential reporting bias and errors. Moreover, there is a chance of residual unmeasured confounding. Exact duration of PPIs use was often difficult to assess, hence we have created two groups of PPIs-users, depending on their consumption of PPIs prior to hospital admission during the considered hospitalisation. Indication of PPIs has not been searched for and we can hypothetise that some indications might be an independent risk factors of severe COVID-19, such as obesity which is a risk factor of gastroesophageal reflux disease. We tried to limit this biais with multivariate analysis.

Finally, our study was carried out during the first and second waves of COVID-19 in France with a large majority of historical SARS-CoV-2 strains and before vaccination was implemented. Despite these limitations, our data clearly point to a negative effect of PPIs on the progression of COVID-19 in older people.

Further studies are needed to answer the remaining question about PPIs and infectious disease. It is of major importance to identify infections the most at risk for severe forms in older patients taking PPIs such as influenza infection. Moreover, it is unknown if stopping PPIs early at the point of diagnostic of COVID-19, or other infectious disease, might improve outcome.

Conclusion

Our study reports evidence that the use of PPIs was associated with an increased risk of severe COVID-19 in older people. These findings hold significant clinical implications. In older patients who are taking PPI, the onset of COVID-19 necessitates increased clinical monitoring. These novel COVID-19-related data fuel the importance to rationalise PPIs prescription in this population. Additional studies are warranted to address the ongoing debate on adverse effects of PPIs in older people.

Declaration of Conflicts of Interest:

None.

Declaration of Sources of Funding:

None.

References