#NephJC Chat
Tuesday June, 25th, 2024 at 9 pm Eastern (AEST = June 12th, 11am)
Wednesday June 26th, 2024, at 9 pm Indian Standard by Time and 3:30 pm GMT (AEST = June 13th, 2am)
J Am Soc Nephrol. 2024 Apr 11. doi: 10.1681/ASN.0000000000000356. Online ahead of print.
The Effects of Pantoprazole on Kidney Outcomes: Post Hoc Observational Analysis from the COMPASS Trial
Lonnie Pyne, Andrew Smyth, Amber O Molnar, Paul Moayyedi, Eva Muehlhofer, Salim Yusuf, John Eikelboom, Jacqueline Bosch, Michael Walsh
PMID: 38602780
Introduction
Proton pump inhibitors (PPIs), for the treatment of GI disorders, have been around since the 1990s and are now over-the-counter in many countries. In fact, PPI use could be described as ‘fairly common and ubiquitous', but also as one of the most inappropriate prescribed medications globally. (Muheim L, et al, 2021, Schietzel S, et al, 2024) PPIs are indicated in treating various conditions, including GERD, Barrett’s esophagitis, peptic ulcers, Zollinger-Ellison Syndrome (ZES), and are part of the triple therapy for Helicobacter pylori infections. However, as nephrologists, nothing makes our GERD worse than taking unnecessary medications that may have negative consequences on kidney function. Long-term PPI use (at least in observational studies) has been linked to serious complications including calcium and magnesium malabsorption, Clostridium difficile diarrheal disease, dementia, increased cardiovascular events, and community-acquired pneumonia. In addition, their use has also been linked to adverse kidney outcomes in several large observational population-based studies. PPIs causing acute interstitial nephritis (AIN) has been demonstrated in many case reports and series to be an established cause, but subsequent epidemiological studies also suggest it’s association with AKI in general.
Beyond AKI, could it play a role in CKD as well? As seen below, several observational studies also have described association with incident CKD and faster GFR loss (in the past we have discussed Lazarus et al JAMA Int Med 2016| NephJC Summary). The problem, as we alluded to above, is that PPI use is common, and kidney disease is common, but observational studies just don’t allow you to prove a causal relationship. The same risk factors for CKD (e.g. increasing age, diabetes, obesity) are also associated with the need of PPI use (eg GERD). It is difficult to untangle this relationship (see Tomlinson et al, NDT 2017 for more discussion). Until now, we had no large randomized controlled data linking PPI use with kidney disease. Perhaps this latest study will help direct us down the right path.
The original Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) trial was an international, blinded, placebo-controlled, randomized trial comparing the effects of rivaroxaban and aspirin, alone or in combination, on cardiovascular outcomes (Eikelboom JW, et al., NEJM 2017). Additionally, safety outcome data of PPI use in this cohort showed no increased adverse events (pneumonia, enteric infections, fractures, CKD, diabetes, COPD, dementia, cardiovascular disease, cancer, hospitalizations, or all-cause mortality) over 3 years (Moayyedi P, et al., Gastroenterology 2019). However CKD was not defined on the basis of GFR measurements but rather tas an ‘adverse event’. The current authors (Pyne L, et al., J Am Soc Nephro 2024), were able to do a post hoc analysis of with cleaner GFR data from the COMPASS cohort and provide a more detailed relationship of PPI use and its association with changes in eGFR and the incidences of AKI and CKD.
The Study
Methods
The effects of pantoprazole on kidney outcomes was a prespecified secondary analysis of the COMPASS trial. The COMPASS trial was an international, multicenter, blinded, placebo-controlled, randomized trial using a partial factorial design. A partial factorial design uses a subset of patients to estimate effects, making them very efficient. The downside is that this method may confound some effects, so not all can be accurately estimated.
The primary results had already been published before this post hoc analysis (see above). The mean duration between randomization and open-label extension measured eGFR values was 3.3 years. There were 27,000 patients enrolled in the original cardiovascular outcomes COMPASS trial. All patients had an initial eGFR done at enrollment. Nearly 10,000 patients from the original trial were excluded, many of whom were on PPIs prior to the run-in period, or needed continuous PPI use for GI pathology. That left 17,598 patients randomized to either pantoprazole or placebo. Nearly 50% of patients had eGFR measurements at enrollment into the extension study, with 9,218 patients included in the final analysis. The flow diagram below shows the populations for the primary and secondary outcomes analysis. Note the N is different for the various outcomes as shown in the figure.
Figure 1 Flow diagram illustrating the analysis population for each kidney outcome. (A) Mean rate of eGFR change. (B) CKD composite outcome. (C) Other kidney outcomes (AKI, acute nephritis, and nephrotic syndrome), from Pyne, et al. JASN 2024.
Inclusion and Exclusion Criteria
Participants were included if they had chronic coronary artery disease and/or peripheral arterial artery disease and were at least 65 years of age or had evidence of atherosclerosis of at least two vascular beds or two additional risk factors featured as current smokers, diabetes mellitus, kidney dysfunction with eGFR <60 ml/min per 1.73 m2, heart failure, and non-lacunar ischemic stroke within 1 month ago. The study excluded individuals with a high risk of bleeding, stroke within 1 month or history of hemorrhagic or lacunar stroke, and patients with severe heart failure who had an ejection fraction less than 30% or NYHA class 3 or 4 symptoms. Again, this was a cardiovascular study, so although eGFR data was collected the inclusion/exclusion criteria were not concerned with CKD stage, albuminuria or risk for AKI.
Randomization
There was a run-in period of 30 days for participants in the rivaroxaban part of the study, which was not used for the pantoprazole study. However, participants had to have at least 80% adherence to treatment during the run-in to be eligible for PPI randomization. The subgroup of patients not taking a PPI at baseline were randomized 1:1 by computer to pantoprazole 40 mg daily or pantoprazole-matched placebo, stratified by center. Participants, site personnel, sponsor personnel, and data analysts were blinded. The study pantoprazole/placebo was not administered during the 30-day run-in period.
Primary outcome
It was represented by the rate of eGFR change (in ml/min/1.73m²), calculated on the base of two eGFR values, one at baseline and another at the start of open-label extension. The prespecified statistical analysis plan considered a significant difference in the primary outcome if the eGFR decline was higher than 40%.
Secondary Outcomes
The investigators defined CKD based on a hierarchical composite outcome, which included:
1. eGFR <60 ml/min/1.73m² at the start of the open-label extension study.
2. CKD as exclusion criterion from open-label extension (eGFR <15 ml/min/1.73m²).
3. Newly diagnosed CKD or kidney failure-related death occurring during follow-up of the trial.
AKI, acute interstitial nephritis, and nephrotic syndrome events were tallied based upon records obtained during the trial follow-up, derived from case report form billing codes.
Statistical Analysis
Baseline characteristics were summarized with continuous variables as means and standard deviations, and categorical variables as counts and percentages. Primary analyses followed the intention-to-treat principle, with additional on-treatment analyses excluding participants who discontinued the treatment.
The primary outcome, the rate of change in eGFR, was analyzed using a mixed-effects linear regression model. If significant, the secondary outcome of a 40% decline in eGFR was examined using mixed-effect logistic regression. Other secondary outcomes were analyzed using mixed-effects logistic regression.
To address potential biases, a sensitivity analysis was performed. For participants excluded from the open-label extension due to kidney function, eGFR of 15 ml/min/1.73m2 was imputed, and the mean duration between randomization and open-label extension was used.
Partially adjusted models accounted for stratification factors (center and rivaroxaban/ASA randomization). At the same time, fully adjusted models also included age, baseline eGFR, diabetes, hypertension, age, sex, and race, with centers as random intercepts and other variables as fixed effects. Unadjusted, partially adjusted, and fully adjusted models reported point estimates with 95% confidence intervals.
Subgroup effects on eGFR changes were explored based on trial eligibility, sex, age, baseline eGFR, diabetes, hypertension, and geographic region, without adjustment for multiplicity. In the fully adjusted model, linear regression was used to determine the rate of eGFR change, adjusting for age, sex, and baseline eGFR.
Funding
The COMPASS trial was sponsored by Bayer AG and overseen by a steering committee of Population Health Institute investigators, national study leaders, and Bayer representatives. Although Bayer funded the trial, this post hoc analysis was not part of the original statistical analysis plan.
Results
Data from more than 8,000 patients was used for the primary outcome and more than 17,000 participants’ data was analyzed for the secondary outcomes (see Figure 1). Mean age was 67 years old, 20% were females, approximately one-third of the population was diabetic, and 90% had coronary artery disease. The mean baseline eGFR was 75 ml/min/1.73m2.
Table 1. Baseline characteristics from Pyne, et al. JASN 2024
Primary outcome
In the pantoprazole group, in the fully adjusted analyses, the rate of decline in eGFR was measurably faster when compared with their placebo counterparts. In terms of absolute numbers, the placebo group had an eGFR decline of 1.14 ml/min/1.73 m2 per year, and the pantoprazole group had a decline of 1.64 ml/min/1.73 m2 per year. This was an absolute difference of only 0.27 ml/min/1.73 m2 per year, or about 20%. This was an overall low-risk CKD population and 80% of the cohort had CKD 1 or 2, with an eGFR >60 ml/min/1.73 m2.
Figure 2. Distribution (percent of participants) of the rate of change of eGFR (ml/min per 1.73 m2 per year, CKD-EPI) by treatment group (placebo versus pantoprazole) from Pyne, et al. JASN 2024.
The sensitivity analysis, which imputed the eGFR of 15 ml/min/1.73 m2 for those excluded from the open-label extension due to kidney function, yielded results consistent with primary analyses.
Supplemental Table 4. Partially adjusted and fully adjusted intention-to-treat analysis results examining the effect of pantoprazole vs placebo on eGFR decline rate, from Pyne, et al. JASN 2024
Having in mind that subgroup analyses are hypothesis-generating at most, we should not over interpret this forest plot. Overall the results appear quite consistent in most subgroups.
Figure 3. Mean rate of eGFR change (ml/min per 1.73 m2 per year) pantoprazole as compared with placebo by subgroup and overall population in fully adjusted analyses from Pyne, et al. JASN 2024.
Secondary Outcomes from the COMPASS Trial
Chronic Kidney Disease
There were 1,352 incident CKD composite events, including 1,261 (93%) with open-label extension eGFR <60 ml/min/1.73 m2, 10 (1%) excluded for impaired kidney function (eGFR <15 ml/min), one death due to kidney failure (0.1%), and 80 (6%) with CKD recorded during follow-up. In the placebo group, 646 participants (10%) had a CKD composite outcome as opposed to 706 (10%) in the pantoprazole group, corresponding to an odds ratio of 1.11 (95% CI 0.98–1.25), which did not attain statistical significance.
Given the significant difference in mean eGFR change rate, the effect of pantoprazole on a 40% eGFR decline was examined. There were 275 such events, with 139 (3%) in the placebo group and 136 in the pantoprazole group (or 1.03, 95% CI 0.81 to 1.32).
Table 2 from Pyne, et al. JASN 2024
Acute Kidney Injury (AKI) and other kidney outcomes
There were 173 AKI events, with 92 (1%) in the placebo arm and 81 (0.9%) pantoprazole group (OR 0.89, 95%CI 0.89 to 1.21). Nephrotic syndrome occurred in four participants in the pantoprazole group (0.05%) and one in the placebo group (0.01%). Acute nephritis was reported in one participant from the pantoprazole group (0.01%) and none in the placebo group.
The incidence of AKI was only a secondary outcome in the COMPASS trial. Cases of AKI were identified by Medical Dictionary for Regulatory Activities coding on case report forms, and events were not identified by specified criteria, for example, using the Kidney Disease Improving Global Outcomes definition. There is obviously a lot of space to interpret what “AKI” means on billing and coding data. Furthermore, acute interstitial nephritis and nephrotic syndrome cases were also evaluated but had a very low rate making any interpretation of causality difficult.
Low event rates, in a low-risk population, and reliance on MedDRA coding likely resulted in significant underreporting and imprecision of the AKI outcome assessment. Without more robust AKI data, it is difficult to draw firm conclusions about the effect of pantoprazole on AKI risk. Further research with more objective methods of AKI assessment would be required to better elucidate any potential association.
Discussion
This is the largest and possibly only RCT data which provides unconfounded data on the relationship between PPI use and subsequent CKD, GFR change and AKI. There was a statistically significant 20% faster GFR loss with PPI use, albeit in this low risk and slowly progressive cohort, most of them with GFR > 60. The ~ 10% higher incident CKD outcome did not translate into a statistical significance (95% CI 0.98 to 1.25) despite ~ 1300 events. The number of AKI events (from coding, not meticulous ascertainment) was too low to be meaningful.
Strengths
1. Randomized Controlled Trial Design: The COMPASS trial was a large, well-designed, randomized controlled trial; it is considered a gold standard for evaluating the effects of an intervention on outcomes. Randomization helps eliminate confounding.
2. Holistic Renal Outcomes: The trial measured a wide range of kidney-related outcomes, including eGFR decline rate, incident chronic kidney disease, acute kidney injury, acute nephritis, and nephrotic syndrome. This makes for a more comprehensive assessment of how pantoprazole could affect kidney health.
3. Long-term Follow-up: The study had an average 3.3-year follow-up from randomization to open-label extension, making it possible to identify more long-term effects of pantoprazole on kidney function.
4. Robust Endpoint Definitions: The rate of eGFR decline, which is one of the primary outcome measures of the trial, was computed based on the standardized eGFR measures at baseline and during an open-label extension; therefore, through this measurement, a change in kidney function was gauged reliably.
Limitations
1. Post-hoc Analysis: Such analyses of kidney outcomes were not pre-specified in the original protocol but were post-hoc. This increases the risk of false-positive findings.
2. Incomplete Kidney Data: Only 51% of the participants randomized either to pantoprazole or a placebo had complete data on eGFR for primary outcome analysis. It is unclear if data were missing at random or whether people with renal concerns would be more likely to have GFR measured and whether that introduces a information bias for outcome measurement.
3. Suboptimal CKD outcome definitions: the definition of incident CKD was based on a composite outcome that did not make consistent use of laboratory data, hence there may have been an underreporting of CKD events. This was an inevitable outcome of the post hoc renal-focused analysis from a primary CV trial.
4. Possible underreporting of AKI cases: AKI was based on MedDRA coding, which can be different from the standardized definitions for AKI and thus may represent a gross underestimation of the burdens of such events.
5. No mechanistic Insights: It is not explored in the trial that defines the possible underlying mechanisms by which pantoprazole may act on kidney function, such as calcium/magnesium absorption, gut microbiome, and chronic interstitial disease.
The COMPASS trial does provide valuable insights into the potential effects of pantoprazole on kidney outcomes but has limitations from its post-hoc design and incomplete data collection. Further studies, including prospective studies assessing the metabolic effects of PPIs on kidney function, would help consolidate the generalizability and extend these findings.
Extra Information on the ELAIA-2 Trial
An article published last year (Wilson, F.P., et al. Nat Commun 2023) evaluated changes in kidney outcomes in hospitalized patients on specific high-risk medications. This was a pragmatic, open-label, parallel-group, randomized controlled trial of patients hospitalized with AKI. The study was designed to investigate whether an automated clinical decision support system affected discontinuation rates of potentially nephrotoxic medications (NSAIDs, RAASi, PPIs), and improved outcomes. The PPI subgroup was the only one to show a significant overall effect of the alert, and underwent further analysis. It was observed that the alerts decreased the relative odds of death, dialysis, and progression of AKI by 18%, and that 10.7% of that total effect was mediated through PPI cessation (95% CI, 2.9%- 44.7%). The observed benefits among those receiving PPIs may have been due to the fact that PPI use flagged a distinct population of patients that benefited from AKI alerts in general. For example, those receiving PPIs were more acutely ill, with a higher proportion of ICU patients, and overall worse outcomes. Although underpowered for this subgroup analysis, it is reasonable to consider further studies to determine the mechanism of PPI-associated AKI and the benefits of withholding these medications in the setting of changes in kidney function.
Figure 2. Medication of interest discontinuation rate, from Wilson, F.P., et al. Nat Commun 2023
Implications for Clinical Practice
In this post hoc analysis of the COMPASS trial data we have our first well designed, large, blinded, randomized study looking at pantoprazole and kidney function. We are left, however, wondering how does this new information help inform clinical decision-making? This trial was piggybacked on patients enrolled in a cardiovascular risk trial, and thus the study population did not have significant CKD at baseline. In addition, proteinuria was not quantified (likely very low prevalence in this population), the number of eGFR data points was limited, and the incidence of AKI was very rare. These are not the typical CKD clinic patients that many of us will need to advise about deprescribing PPI use.
Many observational studies have warned about a potential association between PPI use and AKI/CKD. There are many potential reasons why PPIs may worsen kidney disease. Theories have been brought forth that PPIs may cause advanced arterial calcification, oxidative stress and tubular cell death, altering or diminishing the gut microbiome to undetectable levels, and chronic interstitial nephritis. Although this new study does not give us any further insight into the mechanisms of disease, it does finally give us strong quantifiable evidence of PPIs effect on eGFR decline.
Is a decline of 0.27 ml/min/1.73 m2 per year clinically significant? The authors opined on the recent JASN podcast that they were not sure that it was, particularly in low-risk CKD patients. On the other hand, the authors wondered if this 1.2-fold (20%) difference persisted in a high-risk CKD group. For example, if a patient with rapidly declining kidney function is losing 5 ml/min/1.73 m2 per year of eGFR, would PPI-exposed patients lose 6 ml/min/1.73 per year? Most nephrologists would agree that this could be significant and accelerate the need for kidney replacement therapy by months or even years, depending on the length of drug exposure. The authors ultimately concluded this is an area for further research.
In addition, the authors emphasize that kidney function should be more closely monitored when high-risk CKD patients are treated with long-term PPIs. A risk/benefit analysis should be done (just as it should with any medication) to determine if PPI use is necessary and appropriate. It is hard to argue against the use of PPIs in high-risk disorders like Barrett’s esophagitis (pre-cancerous) and ZES (gastrinomas), even in patients with advanced CKD. However, in more mild diseases like GERD, a full discussion with patients about the risks of PPI use may be undertaken. This research emphasizes that it is important to have shared decision-making by patients with their healthcare providers.
In the end, insights from the COMPASS trial into the potential quantifiable effects of pantoprazole on kidney outcomes showed a trend toward a faster decline in eGFR and a higher incidence of CKD composite outcomes. Although clinically modest, these findings are relevant and emphasize the need for careful consideration of the risks associated with long-term PPI use, particularly in high-risk patients. Frequent kidney monitoring and shared decision-making with patients are important considerations in PPI prescribing.
De-prescribing PPIs-A Prickly Problem
PPI deprescribing requires careful consideration and individualization. Although PPIs are very effective in the symptomatic treatment of GERD and other diseases, PPIs have now definitively been shown to have, albeit a small but measurable impact on kidney function. The question remains whether we can stratify this risk, particularly among high-risk kidney disease patients.
PPIs are among the most commonly prescribed medications in the US, with a prevalence of almost 20% among community-dwelling older adults. In hemodialysis patients taking PPIs, more than a quarter of the time the indication is unclear or unknown. Additionally, polypharmacy is already a significant problem for many patients with CKD and ESKD. Inappropriately dosed medications, drug–drug and drug–disease interactions, morbidity and mortality are just a few of the complications physicians and patients must navigate. Polypharmacy is also associated with nonadherence, which leads to recurrent hospitalizations and poorer outcomes in CKD patients. Perhaps most importantly, the current study is really just one more reason to discontinue PPIs in CKD.
Given the potential risks associated with long-term PPI use, it is recommended to: (1) reviewing the indication for PPI use, (2) weighing the risks versus benefits of continued use in those with a valid indication, and (3) considering deprescribing if clinically indicated. Once deprescribing eligibility is established, a deprescribing strategy that tapers PPI use is recommended, since abrupt PPI discontinuation could potentially result in rebound symptoms of acid hypersecretion. The majority of the tapering strategies support a reduction of the PPI maintenance dose by 50% in 1- to 2-week intervals. Non-pharmacologic symptom management methods, including dietary and lifestyle modifications, are also highly encouraged (i.e., eating smaller meals, elevating the head of the bed, avoiding lying down after eating, avoiding alcoholic/caffeinated beverages and fried/fatty foods). Given the risks of deprescribing (not just GERD symptoms, but GI bleeding, see Czikk et al, Can JKHD 2022) an argument could be made that we need a deprescribing RCT before widespread PPI stoppage becomes fashionable.
Conclusion
The COMPASS trial post hoc analysis is the first RCT to objectively measure the decline in eGFR seen with pantoprazole use in a predominantly low-risk CKD population. This information may help inform clinical shared decision-making. Further studies in high-risk CKD patients are also needed to clarify the risk/benefits of PPIs in CKD.