1 Introduction

Patients with cirrhosis are vulnerable to debilitating and potentially deadly infections [1, 2]. The use of proton pump inhibitors (PPIs) may increase the risk of spontaneous bacterial peritonitis, Clostridioides difficile infections, and infections in general; however, the results are conflicting [3-8]. Adding to this risk is the comparatively higher serum concentration of PPIs because of compromised hepatic elimination [9]. For these reasons, caution is recommended when prescribing PPIs to patients with alcohol-related cirrhosis (ALD cirrhosis) [9-11]. Even so, PPIs are widely used among patients with ALD cirrhosis for acid-related disorders and for prophylaxis when using non-steroidal anti-inflammatory drugs (NSAIDs). However, indications for PPI initiation are often unclear, causing concerns about overprescription [12-14].

While worries are increasing, it remains unknown if the prevalence of PPI use is increasing too. Several studies have reported the cross-sectional prevalence of PPI use in smaller cohorts of cirrhotic patients, but none have examined the prevalence of PPI use in nationwide cohorts of patients with ALD cirrhosis [5, 9, 12]. The primary aim of this descriptive study was to calculate the prevalence and the cumulative incidence of PPI use among patients with ALD cirrhosis in the last two decades, where the studies of PPI use and cirrhosis complications have shown conflicting results (Table S1). Our secondary aim was to identify predictors of PPI use to better understand which patients initiate PPI treatment.

2 Patients and Methods

2.5 Statistical Analysis

2.5.1 Prevalence of PPI Use

Analyses of prevalence included all patients diagnosed with ALD cirrhosis, between 1 January 1997 and 31 December 2022 (Figure 1, ‘Main cohort’).

2.5.1.1 Prevalence

For each year, on 1 January, we computed the prevalence of PPI use as the number of patients who were using PPIs divided by the total number of patients on that date. Furthermore, we calculated the prevalence of PPI use based on time since cirrhosis diagnosis, on each day since diagnosis. We did not present prevalence estimates for 1997, 1998 and 1999 because the cohort was small in those early years and consisted exclusively of newly diagnosed patients (Figure S1).

2.5.1.2 Quantitative Analysis of Annual DDD of PPI

For each PPI user, we calculated the average annual number of pills as the total number of pills from filled prescriptions during the follow-up divided by the patient's follow-up time from diagnosis. Similarly, we calculated the average annual dose of PPIs as the total dose divided by follow-up time and then converted to DDD using the WHO's definition [23].

2.5.1.3 Sensitivity Analysis of the Grace Period

To examine if the grace period caused us to overestimate the true prevalence of PPI use, we additionally computed the prevalence with no grace period and with a double-length grace period. We further computed the proportion of patients who were or had been PPI users within the last 6 months and the prevalence of PPI users supposing patients took two pills per day.

2.5.2 Initiation of PPI Use

We examined the initiation of PPI use after diagnosis of ALD cirrhosis including predictors of initiation. These analyses included only patients diagnosed with ALD cirrhosis after 1 January 2000, to ensure all included patients had at least 5 years without any filled PPI prescriptions prior to ALD cirrhosis diagnosis, and follow-up ended on 31 December 2022 (Figure 1, ‘PPI naïve patients’).

2.5.2.1 Cumulative Incidence

The cumulative incidence function was used to estimate the probability of becoming a first-time PPI initiator with respect to time since ALD cirrhosis diagnosis while considering death as a competing event [24].

2.5.2.2 Predictors

Cox proportional hazards regression was used to estimate the association between a candidate predictor and PPI initiation. The predictors included age, sex, region, history of decompensation, GERD, peptic ulcer and NSAID use. All variables were included in both a univariable and a multivariable analysis.

2.5.2.3 Proportion of PPI Users Without a Detectable PPI Indication

We calculated the proportion of patients who initiated PPI treatment without any diagnosis of GERD, peptic ulcer disease or active NSAID prescriptions at the time of their first PPI prescription.

3 Results

We identified 41 263 patients with ALD cirrhosis between 1997 and 2022 (Figure 1). At the time of ALD cirrhosis diagnosis, 10 823 patients (26%) were using PPIs, and PPI users were slightly older than nonusers (median age 60 years vs. 57 years). More PPI users had a history of decompensation (44.1% vs. 27.3%), GERD (19.9% vs. 5.2%), peptic ulcer (22.4% vs. 9.1%), used NSAIDs (45.4% vs. 27.8%) or were from outside the Capital Region (66.1% vs. 61.3%). PPI users and nonusers had the same low prevalence of hepatitis B or hepatitis C infection: 3.2% (Table 1).

3.1 Prevalence of PPI Use

3.1.1 Prevalence

The prevalence of PPI use rose from 13.4% (95% CI: 12.2%–14.5%) in 2000 to 39.7% (95% CI: 38.1%–40.0%) in 2016 and plateaued at this level until 2022 (Figure 2). The prevalence of PPI use peaked at 38.2% (95% CI: 37.7%–38.7%) within the first 3 months after a diagnosis of ALD cirrhosis, followed by a drop to 32.7% (95% CI: 32.2%–33.2%) at 1 year after diagnosis, after which the prevalence remained stable (Figure 3).

3.1.2 Quantitative Analysis of Annual DDD of PPIs

Among patients who used PPIs after diagnosis, 78.8% filled more than 30 DDD of PPIs on average each year, and 23.5% filled more than 365 DDD each year (Figure 4). The median number of pills from filled prescriptions was 163 pills per patient per year (interquartile range [IQR]: 38–329) among PPI users. The median number of DDD was 170 per patient per year (IQR: 41-348).

3.1.3 Sensitivity Analysis of Grace Period

The prevalence of PPI use reached a plateau in 2016 independently of how we defined the grace period. With a grace period equal to the pack size (i.e., considered user for two times the pack size) a plateau was reached at 40%. When discarding the grace period, the plateau was reached at 34%, and when doubling the grace period, the plateau was reached at 43%. With a grace period of 6 months, a plateau was reached at 44% (i.e., 44% were or had been PPI users within the last 6 months). If patients were only considered PPI users for the number of days corresponding to half the pack size, a plateau was reached at 21% (i.e., 21% were PPI users if PPI dose was assumed to be two pills per day).

4 Discussion

We identified all patients with ALD cirrhosis in Denmark and described the prevalence, incidence and predictors of PPI use: The prevalence rose to around 40% and plateaued at this level from 2016 to the end of 2022. Among PPI users, 78.8% filled more than 30 DDDs of PPIs per year, and 23.5% filled more than 365 DDDs per year. Many patients were put on PPIs after cirrhosis diagnosis, the prevalence peaked at 38.2% within 3 months after diagnosis, and the cumulative incidence was 29.7% after the first year. Patients were more likely to initiate PPI treatment if they had an indication such as GERD, peptic ulcer disease or were using NSAIDs, but 55.3% of the patients who initiated PPI treatment did not have any of those indications. We found that patients living in the Region of Southern Denmark were less likely to initiate PPI treatment compared to patients living in the Capital Region of Denmark. We could not determine whether this difference resulted from regional variation in unmeasured patient characteristics or from regional variation in physicians' willingness to prescribe PPIs.

This population-based study included all patients with a hospital diagnosis of ALD cirrhosis in Denmark, and all Danish citizens can be followed to death or emigration; hence, selection bias in cohort studies like ours is essentially eliminated. Our results are generalisable to countries with comparable healthcare systems, prescription practices and reimbursement of prescribed medications.

We identified patients with ALD cirrhosis in the NPR. The positive predictive value of a cirrhosis diagnosis code in the NPR is high, estimated at 85.4% in a study from 1997 (95% CI: 79.8–89.6) [25]. In 2011, a study of the NPR found that a diagnosis code of moderate/severe liver disease had a positive predictive value of 100% (95% CI: 92.9–100) compared to the patients' discharge notes [26]. The completeness is considered high as well, as health care is fully tax-financed in Denmark, and all patients with suspected severe liver disease are referred from the general practitioner to a hospital for diagnostic workup and treatment. Therefore, any possible misclassification is minor and unlikely to have biased our results.

The DNPR was used to identify all PPI prescriptions. Valid, high-quality data were available since 1994, and it is considered complete since 1995 [20]. Although the actual use of medicine is unknown, the patient filled the prescription and had the medicine in their possession on the date defined as the date of PPI initiation. Unfortunately, the indications listed for the PPI prescriptions are incomplete, as well as the intended duration and dosage. We therefore presumed that patients took one pill each day, but the prescription could also have been intended as ‘use as necessary’. To further investigate the magnitude of this uncertainty, we conducted a sensitivity analysis with five different interpretations of the length of PPI use. Even when we made considerable changes in our definition of PPI use, the estimates prevalence of PPI use remained around 34%–44%. Supposing all patients took two pills daily, pack size representing half the length of use, the prevalence would be lower, around 21%. However, this would be unlikely in the majority of patients, since they already receive high doses of PPIs and a double dose would be improbable (the DDD of pantoprazole is 40 mg).

We found the prevalence of PPI use among patients with ALD cirrhosis in Denmark to reach a plateau at 40% from 2016 to 2022. This is in line with previous studies, which found that PPIs are widely prescribed to patients with severe liver disease and cirrhosis, ranging from 40% to 78% depending on the population and the definition of PPI use [5, 12, 13]. By comparison, in the adult Danish population, only 7% were using PPIs in 2014 [27]. PPIs are predominantly prescribed in primary care settings (94%), where physicians may be less aware of the risks associated with PPI use among patients with cirrhosis [28]. The need to de-prescribe could have been overlooked or could have been unsuccessful because of rebound acid hypersecretion [29]. A study showed that 71% of long-term PPI users with cirrhosis who discontinue PPI treatment for ≥ 90 days return to PPI use [30].

Established indications for PPI initiation include GERD, peptic ulcer disease and use of NSAIDs, and patients were more likely to initiate PPI treatment if they had one of these diagnoses or used NSAIDs. GERD is a common and unspecific symptom; consequently, we expect some patients without the ICD-10 code to have GERD symptoms and get relevant treatment. Still, more than half of the patients who initiated PPI treatment had no detectable indication for PPI use in our data. This is in line with several previous studies: A Swedish study from 2008, of 128 patients with liver cirrhosis of any cause, examined the indications behind the many PPI prescriptions and found that 63% did not have an adequate indication for PPI use [12]. Similarly, a recent Romanian study reviewed the hospital records of PPI users with cirrhosis of any aetiology, and 47% of patients had no relevant indication for PPIs [31]. In both studies, indications were similar to ours, including GERD, peptic ulcer and NSAID use.

Studies on PPI use and the association with cirrhosis complications and bone fractures have shown conflicting results in the last two decades [4, 5, 7, 8, 32-35]. These studies have mainly been observational, efficient in showing a potential association between PPIs and cirrhosis complications, but with the risk of uncontrolled confounding (Table S1). Causation should ideally be deduced from well-designed RCTs. Another important limitation of previous studies is the fact that PPI use may vary over time, so defining PPI users only at inclusion may result in misclassification and underestimation of effects (Table S1). Our nationwide study considered changes in PPI use during the follow-up and showed that PPIs continue to be prescribed frequently. Future studies evaluating PPI treatment among patients with ALD cirrhosis should account for the morbidity associated with PPI use. Predictors of PPI initiation included a history of decompensation, as well as GERD and peptic ulcer. For now, clinicians should be cautious when prescribing PPIs to patients with ALD cirrhosis and always consider de-prescribing [29]. Our results provide essential background information for future RCTs on the risks and benefits of prescribing and deprescribing PPI in patients with ALD cirrhosis.

5 Conclusion

Our nationwide cohort study of the prevalence of PPI use from filled prescriptions among patients with ALD cirrhosis showed that 40% were using PPIs in 2022. The prevalence has remained stable, although the controversy about PPIs and their association with cirrhosis complications persists. Reassuringly, the survival for Danish patients with ALD cirrhosis has improved despite the increasing prevalence of PPI use: The relative risk of 1-year all-cause mortality in 2014–2018 versus 1994–1998 was 0.83 (95% CI: 0.78–0.87) [36]. Regardless of the continuing debate on PPI use and the risk of cirrhosis complications, many patients initiate PPI treatment after they are diagnosed with cirrhosis.

Author Contributions

All authors made a significant contribution to the work reported. Marine Søllinag Ramsing: writing – original draft (lead), formal analysis (lead). Morten Daniel Jensen: conceptualization (supporting), formal analysis (supporting), interpretation (supporting), writing – review and editing (equal). Peter Jepsen: conceptualization (lead), acquisition of data (lead), interpretation (lead), writing – review and editing (equal). All authors approved the final version of the article, including the authorship list.

Ethics Statement

The authors have nothing to report.

Consent

This nationwide registry-based cohort study was conducted in accordance with Danish law. No individual consent was needed since the data from the Danish healthcare registries was pseudonymized.

Conflicts of Interest

The authors declare no conflicts of interest.