Use of Proton Pump Inhibitors Increases Risk of Incident Kidney Stones

Abstract

Background and Aims:

Proton pump inhibitors (PPIs) are widely prescribed and have effects on gut ion absorption and urinary ion concentrations. PPIs might therefore protect against or contribute to development of kidney stones. We investigated the association between PPI use and kidney stones.

Methods:

We performed a retrospective study using data from the Women’s Veteran’s Cohort Study, which comprised men and women, from October 1, 1999 through September 30, 2017. We collected data from 465,891 patients on PPI usage over time, demographics, laboratory results, comorbidities, and medication usage. Time-varying Cox proportional hazards and propensity matching analyses determined risk of PPI use and incident development of kidney stones. Use of histamine-2 receptor antagonists (H2RAs) was measured and levothyroxine use was a negative control exposure.

Results:

PPI use was associated with kidney stones in the unadjusted analysis, with PPI use as a time-varying variable (hazard ratio [HR], 1.74; 95% CI, 1.67–1.82), and persisted in the adjusted analysis (HR, 1.46; CI, 1.38–1.55). The association was maintained in a propensity score-matched subset of PPI users and nonusers (adjusted HR, 1.25; CI 1.19–1.33). Increased dosage of PPI was associated with increased risk of kidney stones (HR, 1.11; CI, 1.09–1.14 for each increase in 30 defined daily doses over a 3-month period). H2RAs were also associated with increased risk (adjusted HR, 1.47; CI 1.31–1.64). We found no association, in adjusted analysis, of levothyroxine use with kidney stones (adjusted HR, 1.06; CI 0.94–1.21).

Conclusions:

In a large cohort study of veterans, we found PPI use to be associated with a dose-dependent increase in risk of kidney stones. H2RA use also has an association with risk of kidney stones, so acid suppression might be an involved mechanism. The effect is small and should not change prescribing for most patients.

Introduction

Proton pump inhibitors (PPIs), a class of medications that reduce gastric acid production, are used to treat several conditions, including gastroesophageal reflux disease (GERD) and peptic ulcers^{1, 2}. PPIs are among the most widely used medications in the United States and abroad, available by prescription and over-the-counter^3-5. While generally well-tolerated, an increasing body of literature has found associations of PPI use with a variety of conditions including chronic kidney disease and enteric infections^{6, 7}.

Decreased gastric acid secretion with PPIs is associated with gut absorptive effects including impaired absorption of calcium, iron, and magnesium⁸. Additionally, PPIs are associated with decreased urinary excretion of calcium and magnesium^9-11. Given these effects, there is interest in understanding the effects of PPIs on nephrolithiasis (‘kidney stones’). Decreased calcium gut absorption and decreased urinary calcium excretion should protect against nephrolithiasis, specifically against calcium oxalate stones, the most common stone type^{12, 13}. However, decreased urinary excretion of magnesium, an inhibitor of stone formation, could increase risk^{14, 15}. Given conflicting mechanisms for a physiologic effect of PPIs on stone formation, we performed a retrospective cohort analysis to clarify the relationship of PPI use and nephrolithiasis.

Methods

Setting

All data were from the Women’s Veterans Cohort Study (WVCS) cohort, a national mixed male/female cohort of veterans within the Veterans Health Administration (VHA) who served in recent conflicts in Iraq and Afghanistan¹⁶. We obtained patient encounter, medication prescribing, and outcomes data. Dispensing information for medications received outside the VHA, including over-the-counter medications, were limited to start and end dates of these medications as entered by providers.

Population

In this retrospective cohort study, we considered patients with at least two outpatient encounters between October 1, 1999 and September 30, 2017 for analysis. We defined ‘pre-observation’ as the time between the first primary care encounter to one year after and gathered baseline patient characteristics including demographics, comorbidities, and renal function.

The number of outpatient and inpatient encounters during pre-observation were measured as a marker of health care utilization. Baseline creatinine was defined as the mean of outpatient creatinine values during pre-observation with estimated glomerular filtration rate (eGFR) calculated using CKD-EPI¹⁷.

Exclusions included <1 year of follow-up, diagnosis of nephrolithiasis during or prior to pre-observation, PPI usage during or prior pre-observation, incomplete date of birth or sex, missing outpatient creatinine, and baseline eGFR<30 mL/min.

Observation end-point was the earliest of the first episode of nephrolithiasis or the last primary care visit.

Outcome Measure and Covariates

We used the ninth and tenth revisions of the International Classification of Diseases (ICD-9/ICD-10) to identify nephrolithiasis and comorbidities. Comorbidities included diabetes, gout, gastroparesis, GERD, Barrett’s esophagus, peptic ulcer disease, gastritis, functional dyspepsia, and gastrointestinal (GI) bleeding. Significant GI surgery was identified with Current Procedural Terminology (CPT) codes. We utilized previously validated lists of diagnosis codes for nephrolithiasis, diabetes, gout, and GI bleeding^18-21. Relevant codes may be found in Supplemental Table 1. Covariates included the number of inpatient and outpatient encounters during pre-observation and observation and outpatient creatinine values during observation.

We collected data for age, sex, and self-reported race. Medication utilization was measured for thiazide diuretics, loop diuretics, histamine-2 receptor antagonists (H2RAs), and gout-related medications. Both scheduled and as-needed prescriptions for H2RAs were included in H2RA use measurement. Supplemental Table 2 contains a list of measured medications. Prescription data was collected with medication fill dates. The earliest date of PPI usage was defined as the PPI start time. Defined-daily-dosage (DDD) for PPIs was collected to measure an aggregate total of PPI exposure during a patient’s observation period. One DDD was equal to omeprazole 20mg, pantoprazole 40mg, lansoprazole 30mg, rabeprazole 20mg, esomeprazole 30mg, and dexlansoprazole 30mg as per the WHO Drug Statistics Methodology²². Once-daily and twice-daily dosages could not be distinguished. Thiazide and loop diuretic, H2RA, and gout-related medication use were measured with usage start time equal to the first fill date. Patients were deemed to be continuously on a medication after it had been filled at least once, approximating the intention-to-treat principle in a clinical trial.

Survival Analysis Models

Given the large number of observational studies associating PPI usage with outcomes, two models were used to assess robustness of results.

Model 1: Time-varying Cox Analysis

We developed a time-varying Cox proportional hazards model to study the PPI-nephrolithiasis association. This method performs an adjusted analysis with respect to time-varying covariates, e.g., changes in clinical diagnoses or medication usage²³. In our model, time-varying covariates included initiation of any of the medications listed above, diagnosis of the conditions listed above, creatinine, number of outpatient visits and hospitalizations in the preceding year. Most recent outpatient creatinine at timepoints when covariates changed estimated a patient’s renal function.

Baseline covariates were included in this model as non-time-varying covariates. Unadjusted analysis measured ‘events-per-patient-time’ for individuals on and not on a PPI. Patients who start a PPI during observation period contribute non-PPI-time prior to their first filled PPI prescription and contribute PPI-time afterwards.

Adjusted analysis was performed by sequentially adding covariates to the time-varying model in the order of baseline demographics, medications, comorbidities, laboratory studies, and inpatient and outpatient utilization. Separate adjusted analysis including cumulative DDD over the preceding 3 and 12 months of observation evaluated dose-dependence. Hazard ratios (HRs) with 95% confidence intervals are reported.

Model 2: Propensity-Score Analysis

A randomized controlled study was emulated via matching patients at the start of PPI usage with a similar unexposed control. We performed a 1:1 matched analysis using a propensity score generated for each patient measuring the odds that the individual would be prescribed a PPI. Using the logarithm of the odds (logit) of being prescribed a PPI, a patient was selected randomly from the PPI-exposed group and matched to the individual from the non-exposed group with the closest logit value. Matches could differ by at most 0.2 standard deviations of the logit scores; otherwise the PPI-exposed individual was excluded from matching as no close match could be found. To maintain the ‘intention-to-treat’ principle, patients in the non-PPI group could initiate PPI therapy after the start date. These patients remained in the non-PPI group for analyses. Analyses were performed using R (R Foundation; Vienna, Austria)²⁴.

Patients in the matched cohort were right-censored at the earlier of the last outpatient visit or nephrolithiasis episode. Unadjusted analysis in the propensity-matched cohort used the Kaplan-Meier estimator with log-rank testing. Adjusted analysis accounting for demographics, medications, comorbidities, laboratory studies and healthcare utilization was performed via a Cox proportional hazards model over the matched cohort.

Statistical Analyses

Baseline differences between PPI users and non-users were compared via chi-square statistic for discrete variables and Kruskal-Wallis test by ranks for continuous variables.

Differences between PPI and non-PPI groups for propensity-matched samples were compared with standardized mean difference with 0.2 for small differences²⁵. P-values for propensity matched samples were calculated using paired statistics with McNemar’s testing for categorical data and Wilcoxon signrank testing for continuous variables.

Sensitivity Analysis

Several sensitivity analyses were performed to ensure validity. The first analysis included PPIs prescribed outside the VHA; these prescribing data rely on manual entry and self-report, and therefore have limited completeness and accuracy than VHA prescription data. We set a patient’s date of first PPI as the earliest of the first VHA and non-VHA PPI.

Second, to assess potential protopathic bias we analyzed an alternate PPI usage definition where patients were considered PPI-exposed after 90 days of daily dosage (90 DDD) were filled as well as evaluating the effects of introducing a 90, 180 and 365-day lag from study entry date to accrual of follow-up time.

A secondary analysis with respect to H2RA evaluated whether results were secondary to PPI usage or factors common to acid suppression medications. A cohort was developed in identical fashion to the PPI cohort except patients who previously took H2RAs during pre-observation or earlier were excluded. In the time-varying Cox model, patients who ever received PPIs during observation were excluded. In the propensity-matched model, matches were between H2RA users not on a PPI to control individuals on neither medication. To assess the potential for an additive/synergistic effect of concurrent PPI and H2RA use, we also compared those PPI patients on concurrent H2RAs (greater than one H2RA prescription from 30 days before to 30 days after PPI prescription) in the primary PPI cohort to the control cohort.

Lastly, a negative control exposure tested the internal validity of our study. This method involves assessing a treatment which is expected to have no impact on the outcome²⁶. We assessed the hazard ratio of levothyroxine, a hypothyroidism medication, for the development of incident nephrolithiasis. Literature review shows no established relationship between levothyroxine and nephrolithiasis.

The STROBE checklist was used to ensure reporting guidelines were followed. Approval for this study was obtained from the Human Studies Subcommittee from the Human Research Protection Program of the VA Connecticut Healthcare System.

Results

The WVCS cohort contained a total of 1,065,962 initial patients for analysis. 465,891 patients remained eligible for analysis after applying exclusion criteria (Figure 1). From this cohort, 89,329 (19.2%) were exposed to PPIs during observation. Baseline demographics for patients ever and never prescribed a PPI are displayed in Table 1.

Figure 1.

Cohort construction flow diagram with exclusion criteria and final cohort size.

Table 1.

Baseline characteristics for patients who did and did not receive PPI during observation period^*

	PPI-Unexposed	PPI-Exposed
Total	376,562	89,329
Male (%)	87.2	85.9
Race/Ethnicity (%)
White	58.2	54.3
Black	15.3	15.7
Hispanic	10.9	12.3
Other	5.2	4.7
Unknown	10.4	13.1
Age (median [IQR])	30.25 [26.20, 40.12]	31.58 [26.52, 41.85]
Baseline Creatinine (median [IQR])	1.00 [0.90, 1.10]	1.00 [0.90, 1.10]
H2RA (%)	2.8	8
Thiazide Diuretic (%)	5	7.3
Loop Diuretic (%)	0.1	0.2
Gout Medication (%)	0.4	0.6
GERD (%)	3.3	9.9
Peptic Ulcer Disease (%)	0.1	0.3
Barrett's Disease (%)	0	0.1
GI Bleed (%)	1.7	2.5
Gastritis (%)	0.3	0.7
Functional Dyspepsia (%)	0.4	1
GI Surgery History (%)	0	0
Diabetes (%)	2.1	2.8
Gout (mean (SD))	0.01 (0.08)	0.01 (0.09)
Outpatient Encounters During Pre-Observation (median [IQR])	5.00 [3.00, 9.00]	7.00 [4.00, 11.00]
Inpatient Encounters During Pre-Observation (median [IQR])	0.00 [0.00, 0.00]	0.00 [0.00, 0.00]

PPI=proton pump inhibitor; H2RA=histamine-2 receptor antagonist; GERD=gastroesophageal reflux disease; GI=gastrointestinal; SD = standard deviation; IQR = interquartile range

^*p-value for comparison of all characteristics is <0.001 with exception of GI surgery history which carries a p-value < 0.05

All baseline characteristics were statistically different between populations. Individuals ever prescribed PPIs were older at the start of observation, had more outpatient visits and hospitalizations, more H2RA use, and more GERD diagnoses. Patients who were ever-prescribed a PPI had higher prevalence of all measured comorbidities.

Our cohort had 11,224 cases of nephrolithiasis over 2,544,638 patient-years of observation: 4,219 cases in patients with PPI exposure and 7,005 cases in those without exposure. PPIs carried an unadjusted incidence rate ratio of 1.74 (95% CI 1.67-1.82). Based on these analyses, 365 (95% CI 333-399) patients would need to be exposed to a PPI for one additional episode of nephrolithiasis in one year. The proximity of diagnosis of nephrolithiasis to prescription of a PPI was measured; 0.7% of cases were within one week of prescription, 3.3% within one month and 26% within one year.

Sequential adjusted analysis was performed using the time-varying model with hazards listed in Supplemental Table 3. Full multivariable adjustment revealed an adjusted HR=1.46 (95% CI 1.38-1.55) for PPIs and nephrolithiasis. Analysis restricted to those in the PPI cohort on concurrent H2RA (representing 3% of PPI users) revealed a similar adjusted hazard: 1.41 (95% CI 1.28-1.56), providing no evidence of an additive or synergistic effect of PPIs with H2RAs.

Additional adjusted analysis examined the effects of cumulative DDD on nephrolithiasis. Cumulative DDD measured in the preceding 3 months and 12 months for each patient time interval were used as covariates in the time-varying model. The 3-month cumulative DDD covariate had an adjusted HR=1.11 (95% CI 1.09-1.14) per 30-DDD change in cumulative dose. The 12-month cumulative DDD covariate had an adjusted HR=1.04 (95% CI 1.02-1.06) per 90-DDD change in cumulative dose.

For the propensity-matched model, the logistic regression model for propensity score had an area under the operator receiving curve of 0.80; 96.5% of the PPI group were matched with 86,264 patients in both groups. Matched cohort characteristics are displayed in Table 2.

Table 2.

Characteristics of the patients who did and did not receive PPI in the propensity-matched cohorts.

	PPI-Unexposed	PPI-Exposed	Standardized Mean Difference*
Total	86,264†	86,264
Male (%)	86	85.8	0.006
Race/Ethnicity (%)			0.019
White	54.7	54.1
Black	15.8	15.8
Hispanic	11.9	12.3
Other	4.9	4.7
Unknown	12.8	13.1
Age (median [IQR])	32.01 [26.81, 42.49]	31.58 [26.52, 41.84]	0.044
Most Recent Creatinine (median [IQR])	1.00 [0.90, 1.10]	1.00 [0.88, 1.10]	<0.001
H2RA (%)	19.8	17.4	0.061
Thiazide Diuretic (%)	1.9	1.5	0.034
Loop Diuretic (%)	14.7	13.8	0.025
Gout Medication (%)	1.2	0.9	0.03
GERD (%)	40.9	40.9	<0.001
Peptic Ulcer Disease (%)	1	1.3	0.029
Barrett's Disease (%)	0.4	0.4	0.005
GI Bleed (%)	9.1	8.4	0.024
Gastritis (%)	5	5.7	0.031
Functional Dyspepsia (%)	3.6	3.7	0.006
GI Surgery History (%)	0.3	0.2	0.02
Diabetes (%)	7.8	6.1	0.068
Gout (%)	2.7	1.9	0.052
Outpatient Encounters in Preceding Year (median [IQR])	5.00 [3.00, 10.00]	5.00 [2.00, 10.00]	0.019
Inpatient Encounters in Preceding Year (median [IQR])	0.00 [0.00, 0.00]	0.00 [0.00, 0.00]	0.035

PPI=proton pump inhibitor; H2RA=histamine-2 receptor antagonist; GERD=gastroesophageal reflux disease; GI=gastrointestinal

^*Standardized mean difference < 0.2 suggests good matching with respect to the specific variable.

^†3.4% of PPI-users were unmatched because no control individual meeting matching criteria was identified; these patients were excluded from analysis

In the propensity-matched cohort, there were 4,735 episodes of nephrolithiasis over 765,154 patient-years of observation with median observation time of 4.01 years (IQR 1.92 years-6.52 years). The PPI group developed nephrolithiasis at a rate of 69.6 cases per 10,000 patient-years versus 55.3 cases per 10,000 patient years in the non-PPI group. Unadjusted rate ratio in the propensity-matched cohort was 1.25 (95% CI 1.19-1.33) and the Cox-model adjusted hazard was 1.28 (95% CI 1.21-1.36). Figure 2 displays the unadjusted survival per the Kaplan-Meier estimator.

Figure 2.

Kaplan-Meier curve for PPI and non-PPI propensity-matched cohort with respect to percentage free from nephrolithiasis. Log-rank test p<1.0 x 10⁻¹⁶.

Sensitivity analysis including PPIs prescribed outside the VHA showed an unadjusted HR=1.71 (95% CI 1.63-1.79) and adjusted HR=1.42 (95% CI 1.34-1.51). The sensitivity analysis in which individuals accrued at least 90 DDD to be defined as a PPI user showed unadjusted HR=1.62 (95% CI 1.54–1.70) and adjusted HR=1.24 (95% CI 1.17–1.33). Finally, analyses in which individuals needed to accrue 90, 180 and 365 days prior to the start of PPI observation time showed unadjusted HRs of 1.66 (95% CI 1.58-1.75), 1.63 (95% CI 1.54-1.72) and 1.57 (95% CI 1.48-1.66) respectively and adjusted HRs of 1.39 (95% CI 1.30-1.48, 1.36 (95% CI 1.27-1.45) and 1.30 (95% CI 1.21-1.40) respectively.

Secondary analysis evaluating the H2RAs-nephrolithiasis relationship showed unadjusted HR=1.70 (95% CI 1.53-1.89) and adjusted HR=1.47 (95% CI 1.31-1.64). This H2RA cohort was generally older, more female, more Black/Hispanic, had a lower diagnosis of GERD/GI bleeding during pre-observation, and had more outpatient visits than the PPI cohort. A propensity-matched cohort contained 21,907 patients with H2RA but not PPI exposure matched 1:1 to patients without exposure to either H2RAs or PPIs. Unadjusted and adjusted analyses revealed HR 1.32 (95% CI 1.17-1.50).

Negative control exposure analysis evaluating levothyroxine showed unadjusted HR=1.28 (1.13-1.46) and adjusted HR=1.06 (0.94-1.21). A propensity-matched cohort was developed in a similar fashion as with PPI and H2RA cohorts with 10,630 levothyroxine users matched 1:1 to non-users. Unadjusted HR was 97 (0.82-1.16) with adjusted HR=1.02 (0.86-1.22). Table 3 summarizes the unadjusted/adjusted hazards of the above models.

Table 3.

Hazard ratios for development of nephrolithiasis with medication exposure versus no medication exposure in the time-varying Cox and propensity-matched models.

Model	Unadjusted (95% CI)	Adjusted (95% CI)
Model 1: Time-Varying Cox
PPI	1.74 (1.67 - 1.82)	1.46 (1.38 – 1.55)
PPI Sensitivity Analysis: Non-VHA Medication Inclusion	1.71 ( 1.63 – 1.79)	1.42 ( 1.34 – 1.51)
PPI Sensitivity Analysis: At least 90DDD Definition of PPI use	1.62 ( 1.54 – 1.70)	1.24 ( 1.17 – 1.33)
H2RA	1.70 (1.53 – 1.89)	1.47 (1.31 – 1.64)
Negative Control Exposure: Levothyroxine	1.28 (1.13 - 1.46)	1.06 (0.94 - 1.21)
Model 2: Propensity-Matched
PPI	1.25 (1.19 – 1.33).	1.28 (1.21 – 1.36).
H2RA	1.32 ( 1.17 – 1.50)	1.32 ( 1.17 – 1.50)
Negative Control Exposure: Levothyroxine	0.97 (0.82 – 1.16)	1.02 (0.86 – 1.22)

PPI=proton pump inhibitor; DDD=daily defined dose; H2RA=histamine-2 receptor antagonist Covariates in adjusted analyses included: sex, race/ethnicity, age, most recent creatinine, use of the medications (H2RAs, thiazide diuretics, loop diuretics, gout medications), medical history (gastroesophageal reflux disease, peptic ulcer disease, Barrett’s disease, gastrointestinal bleed, gastritis, functional dyspepsia, gastrointestinal surgical history, diabetes, gout) and total number of inpatient/outpatient encounters in the previous year.

Discussion

This study evaluated the relationship between PPIs and development of incident nephrolithiasis. Given numerous observational studies linking PPIs to various comorbidities, we used a rigorous approach using two survival analysis models. Both models found modest association of PPI use with nephrolithiasis which was preserved after adjustment for pertinent covariates including demographics, laboratory studies, diagnoses, medications, and healthcare utilization. Furthermore, the effect was dose-dependent with modest increases in risk with increased PPI prescribing. Dose-dependence was less pronounced when assessed over a longer time window of 12 versus 3 months.

This relationship was preserved across sensitivity analyses, including over-the-counter use. A small percentage of patients were found to be diagnosed with nephrolithiasis within one week and within one month of PPI prescribing. However, the relationship was maintained with modified start times for the observation date, ranging from 90 to 365 days after initial PPI prescription. These findings suggest against a protopathic bias, i.e. the association is not primarily due to prescription of PPI for symptoms of nephrolithiasis preceding the diagnosis of the condition. Secondary analysis evaluating H2RA usage revealed an increased hazard for nephrolithiasis. These data suggest the possibility that the PPI-nephrolithiasis association is due to generalized acid suppression therapy. Though the hazard ratios for PPIs and H2RAs are similar, the populations used in the two analyses are very distinct and therefore it is difficult to directly compare hazards between the populations. Levothyroxine, our negative control exposure, carried no association in adjusted or propensity-matched analysis, suggesting that important confounders were captured in this study.

A MEDLINE search through October 2019 found no studies evaluating the PPI-nephrolithiasis association. A study using the FDA Adverse Event Reporting System reported increased odds of nephrolithiasis in PPI users compared to H2RA users, but these are self-reported data that don’t allow incidence estimates or adjusted analyses and standard disproportionality methods were not used²⁹. The only other report of this association is a conference abstract several years ago which reported an increased nephrolithiasis with acid suppression medications²⁷.

A mechanism to explain this increased risk is uncertain. PPIs decrease intestinal magnesium absorption and reduce urinary magnesium, a well-known nephrolithiasis inhibitor historically prescribed for patients with frequent stones^{10, 28, 29}. H2RA effects on urinary magnesium are less studied with scant evidence available suggesting against an effect³⁰. The pathophysiology of stone formation is believed to be due to an imbalance in stone-forming and stone-inhibiting molecules within the urine³¹. In addition to magnesium, PPI use potentially might decrease urinary concentration of other inhibitory molecules and/or increase concentrations of stone-facilitating molecules.

Our study has several strengths. We used a large cohort from a national dataset. Propensity matching was done to ensure the two study cohorts were similar in characteristics other than use of PPIs. Analyses were adjusted for a variety of potential confounders including demographics, laboratory values, and medical comorbidities. We also adjusted for healthcare utilization. It has been suggested that PPI prescribing is related to interactions with the healthcare system; this is generally an unmeasured confounder and may contribute to the multitude of health conditions that are now associated with PPI use^32-34.

Additionally, we performed a sensitivity analysis accounting for PPI prescriptions obtained outside the VHA. Given that PPIs are commonly obtained over-the-counter, accounting for non-prescribed use is important. Finally, we performed analyses with a negative control exposure. If a medication not known to have an association with an outcome is found to be positively associated with that outcome, this may indicate bias. We show no association reducing the likelihood that residual confounding accounts for our findings with PPIs.

Our study should be viewed in light of several limitations. First, we acknowledge that our study is retrospective and unmeasured confounders may contribute to the association we observed between PPIs and nephrolithiasis; adjustment with numerous measured potential confounding factors reduced the hazard ratio compared to the unadjusted analysis, and inclusion of unmeasured confounders might further reduce the association. Such skepticism is important in light of a recent large randomized control trial in individuals randomized to receive PPI therapy or placebo which showed increase in enteric infections but otherwise similar rates of commonly-associated outcomes such as pneumonia, fracture, and CKD³⁵. This study did not evaluate nephrolithiasis and with the incidence we found in our analyses, such a study would need to enroll around 40,000 patients to have sufficient power (alpha=0.05 and beta=0.2) to identify a difference; this study of around 18,000 participants is insufficient to evaluate this outcome. Additionally, this study evaluated patients with stable angina who otherwise did not meet criteria for PPI prescription. Our population was all-comers with incident prescription of a PPI, which improves the generalizability of our findings.

Other limitations include that daily dosage data was only obtained for PPI prescriptions within the VHA. Additionally, the study population was younger at the start of observation and mostly male with over half being white. One should be careful when applying these results to significantly different populations.

Conclusions

We analyzed a large retrospective dataset with two robust survival analyses to evaluate the relationship between PPI use and nephrolithiasis. A modest increase in incident nephrolithiasis was seen in patients on PPI in both models with a dose-dependent response. The mechanism may be related to acid suppression given that H2RAs also showed an increase. The increased hazard is small; these findings should not alter prescribing of PPIs in the general population. However, these findings may be of interest for patients with history of nephrolithiasis. Further studies are needed to determine the direct effects of PPI use and nephrolithiasis.

What You Need to Know.

Background:

Proton pump inhibitors (PPIs) are widely prescribed and have effects on gut ion absorption and urinary ion concentrations, so they might protect against or contribute to development of kidney stones.

Findings:

In a large cohort study of veterans, PPI use was associated with a dose-dependent increase in risk of kidney stones. H2RA use also has an association with risk of kidney stones, so acid suppression might contribute to kidney stone development.

Implications for patient care:

The association between PPI use and kidney stone development is small and should not change prescribing patterns for most patients

Abbreviations:

CI

confidence interval

CPT

Current Procedural Terminology

DDD

defined daily dose

GERD

gastroesophageal reflux disease

GI

gastrointestinal

H2RA

histamine-2 receptor antagonist

HR

hazard ratio

ICD

International Classification of Diseases

IQR

interquartile range

PPI

proton pump inhibitor

SD

standard deviation

VHA

Veterans Health Administration

WVCS

Women's Veterans Cohort Study

eGFR

estimated glomerular filtration rate