Proton Pump Inhibitors and Hyporesponsiveness to Erythropoiesis-Stimulating Agents in Hemodialysis Patients: Results from the Japan Dialysis Outcomes and Practice Patterns Study

Regardless of erythropoiesis-stimulating agent (ESA) use, renal anemia remains frequent among hemodialysis patients. Depending on the definition, about 5–15% of end-stage renal disease patients are ESA hyporesponsive [1‒3]. Higher ESA dose is a risk factor for cardiovascular disease and mortality in dialysis patients, and causes increased medical costs as well, and ESA response to anemia also affects several outcomes [4, 5]. ESA hyporesponsiveness refers to the constant need for high doses of ESA to maintain target hemoglobin (Hb) levels [3]. There are several factors of ESA hyporesponsiveness, such as blood loss, iron deficiency, chronic inflammation, malnutrition, vitamin D deficiency, renin-angiotensin inhibitor (ACE-i), angiotensin-II receptor blocker (ARB), and severe secondary hyperthyroidism. On the other hand, the causes of ESA hyporesponsiveness remain as unresolved problems, and searches for new causes and approaches are required [6‒8]. Proton pump inhibitors (PPIs) are widely prescribed throughout the world, including for patients with chronic kidney disease (CKD). Due to an increase in the number of patients taking PPIs, the risk of its use has been reported recently [9]. Gastric acid promotes absorption of non-heme iron by converting it from a non-absorbable ferric form to an absorbable ferrous one [10]. As chronic PPI use reduces gastric acid and keeps gastric pH high, it seems possible that iron deficiency and anemia result from taking PPIs chronically. Iron deficiency is a critical complication in dialysis patients because it is a risk factor for cardiovascular events and mortality regardless of anemia status. Iron deficiency occurs frequently among dialysis patients, with a prevalence of 10–30% [11, 12]. In Japan, its prevalence is as high as about 40% [13]. Several studies have reported iron deficiency anemia caused by PPI use [14‒17]. In addition, earlier studies reported that PPIs also inhibit absorption of vitamin B12 and cause macrocytic anemia [18, 19]. Another study reported that PPIs have direct effects on erythrocytes and cause apoptosis of red blood cells and hemolysis [20]. Although the relationship of PPIs and anemia has recently been clarified, it is not clear whether PPIs cause ESA-resistant anemia in hemodialysis patients. The purpose of this study is to clarify how PPI use is associated with the erythropoietin resistance index (ERI), Hb level, ESA dosage, and iron status in hemodialysis patients.

The Dialysis Outcomes and Practice Patterns Study (DOPPS) is a world-wide cohort study in a random representative sample of selected dialysis which facilities within each participating country in more than 20 countries [21, 22]. The present study was a cross-sectional study using repeated 4-month observations, up to eight observations/patient, from Japan DOPPS (J-DOPPS) Phase 6 (2015–2018).

The inclusion criteria were patients aged 20 years or older and at least 3 months’ duration since the start of the hemodialysis. Patients were excluded if they had polycystic kidney disease because anemia management may differ considerably in these patients [23]. Other exclusion criteria were serious infections and hemorrhagic diseases at baseline. Patients taking histamine-2 blockers were excluded in order to evaluate the effects of PPIs only.

The exposure of interest was the use of PPIs, defined as any prescription of PPIs including esomeprazole, lansoprazole, omeprazole, rabeprazole, and vonoprazan in each 4-month observation period.

We used the erythropoietin resistance index (ERI) as the primary outcome, defined as ERI (IU/week/kg/[g/dL]) = ESA dose (IU/week)/Hb (g/dL)/BW (kg) by references to previous reports [24‒26], where ESA dose is the average weekly dose of an ESA, Hb is the average pre-dialysis blood Hb concentration, and body weight is the average pre-dialysis body weight, within each observation period. The ESA doses were converted into an erythropoietin alfa dose equivalent (erythropoietin alfa: darbepoietin alfa: epoetin beta pegol = 1: 200 [27]: 225 [28]).

We further examined five secondary outcomes: (1) Hb (g/dL), (2) ESA dose (IU/week), (3) transferrin saturation (TSAT) (%), (4), serum ferritin (ng/mL), and (5) the prevalence of erythropoietin-resistant anemia (being present if Hb <10 g/dL and ESA dose ≥6,000 IU/week).

Baseline characteristics were expressed either as frequencies, means (standard deviation) or medians (interquartile range), for the total population and subpopulations taking or not taking PPIs. The outcome values were summarized in a box-whisker plot or a bar plot.

For the primary analysis, a linear regression model was used with generalized estimating equations accounting for the repeated measurements. The outcome variable was the ERI, and the exposure was the use of a PPI. The following variables were included as potential confounders: age, sex, hemodialysis vintage, comorbidities (cardiovascular disease, diabetes mellitus, and gastrointestinal bleeding), medications (antiplatelet agents, anticoagulants, non-steroidal anti-inflammatory drugs, and renin-angiotensin system inhibitors including angiotensin-converting enzyme inhibitors [ACE-Is] and angiotensin receptor blockers [ARBs]), and laboratory values (C-reactive protein (CRP), platelet counts, aspartate aminotransferase, serum albumin, creatinine, and parathyroid hormone) (model 1). Results were shown as point estimates and 95% confidence intervals (95% CIs).

Other outcomes, including Hb, ESA dose, TSAT, serum ferritin, and the prevalence of erythropoietin-resistant anemia were analyzed using the same models and covariates [29]. Missing data were imputed using multiple imputation by chained equations with 100 repetitions using predictive-mean matching [30, 31]. All available cases were included rather than performing a formal sample size calculation. All statistical analyses used Stata SE17 (StataCorp, College Station, TX, USA) and were two-sided with level of significance set at 0.05 without correction for multiple testing.

For the ERI, we did subgroup analyses stratified by each of eight predetermined variables: age category (<75, ≥75 years), sex, BMI (<25, ≥25), serum albumin (<3.8, ≥3.8 g/dL), normalized protein catabolism rate (<1.0, ≥1.0), single-pool Kt/V (<1.0, ≥1.0), history of gastrointestinal bleeding within 12 months before registration, and use of an antithrombotic drug, including antiplatelets and anticoagulants. Effect modifications were tested by assessment of the interaction term.

We hypothesized that TSAT, serum ferritin, and iron drugs affect iron metabolism and may act as an intermediate. We therefore did not include them in our primary model (model 1). To assess intermediate effects of these factors, we performed two sensitivity analyses: as model 2, we included serum ferritin and TSAT as covariates in addition to model 1, and as model 3, the use of oral and/or intravenous iron drugs in addition to model 2, and assessed whether the direction or magnitude of the effect size was altered or not.

The study included 1,644 patients after excluding 406 (20%) (online suppl. Fig. 1; for all online suppl. material, see https://doi.org/10.1159/000534701). The total number of 4-month observations was 12,048. At the baseline, 867 patients (52.7%) received PPI prescriptions (PPI[+]); the remaining 777 patients did not (PPI[−]). Characteristics of patients at the baseline are shown in Table 1. Data were complete for sex, dialysis vintage, comorbidities, cause of end-stage renal disease, and medications. For up to about 25% of the cohort, data for age, BMI, and laboratory values (CRP, platelet counts, aspartate aminotransferase, serum albumin, creatinine, and parathyroid hormone) were missing.

Distributions of the outcomes are shown in Figure 1. Means (standard deviations) in PPI(+) and PPI(−) patients were 7.1 (7.0) and 8.7 (9.9) for ERI, 11.0 (1.0) and 10.9 (1.1) for Hb, 4,140 (3,661) and 4,843 (4,809) for ESA dose, 122 (147) and 121.3 (151) for ferritin, and 26.5 (12.2) and 24.7 (12.1) for TSAT, respectively. Proportions of erythropoietin-resistant anemia in the two groups were 13.8% and 20.1%, respectively. When looking at the raw values, ESA dose and erythropoietin-resistant anemia were higher, and Hb, serum ferritin, and TSAT were lower in PPI(+) than in PPI(−) patients.

Crude and adjusted differences in ERI are shown at the top of Table 2. PPI(+) patients had significantly greater ERI than PPI(−) patients (adjusted difference of 0.95 IU/week/kg/[g/dL] [95% CI: 0.40–1.50]). The results of subgroup analyses are shown in Figure 2. There was a significant interaction for gastrointestinal bleeding (p = 0.03), but not for other subgroups (p for interaction = 0.72 for sex, 0.96 for age, 0.45 for body mass index, 0.81 for normalized PCR, 0.72 for single-pool Kt/V, 0.03 for gastrointestinal bleeding, and 0.96 for antithrombotics.).

Crude and adjusted differences in secondary outcomes are shown after the second row of Table 2. Among other measures representing anemia, there were significant differences in ESA dose and the prevalence of erythropoietin-resistant anemia even after adjusting for potential confounders (336 IU/week [95% CI: 70–602] and a risk difference of 0.04 (0.02–0.06), respectively), but not in Hb. Among mediators of the association of PPIs and anemia, there was a significant difference in TSAT (−0.82% [95% CI: −1.56 to −0.07]), but not in serum ferritin.

Adjusted differences derived from model 1, model 2, and model 3 are shown in Figure 3. The inclusion of TSAT and serum ferritin (model 2) (0.87, 95% CI: 0.32–1.43) among the covariates and the addition of iron drugs (model 3) (0.85, 95% CI: 0.30–1.41) gave estimates similar to model 1.

This study examined whether PPIs play a harmful role in response to ESAs using J-DOPPS data collected from multiple centers in a real-world setting of Japanese dialysis patients. We found a significant association between PPI use and higher ERI. PPI use was also associated with higher ESA doses and lower TSAT levels, but not with serum Hb and ferritin levels. There were no interactions of clinical parameters examined, except for gastrointestinal bleeding, on the association of PPI use and ERI. Further adjustment using iron-related TSAT, ferritin, and use of iron drugs did not alter the association of PPI use and ERI.

The background of patients with hemodialysis makes them more likely to be prescribed PPIs. In this study, 53% of the patients were prescribed PPIs. Previous studies have also reported that more than 50% of dialysis patients are taking PPIs [32, 33]. Gastrointestinal disorders, such as gastrointestinal bleeding or GERD, are more frequently seen in dialysis patients [34]. In addition, as PPIs undergo extensive hepatic metabolism, there is no need for dose adjustment according to renal function, as there is, for example, with H2 receptor antagonists [35]. For that reason, although earlier studies have reported that overuse of PPIs is seen in hemodialysis patients, PPI prescriptions are rarely reviewed in clinical practice [36, 37]. Epidemiological studies have revealed that PPIs produce side effects, including pneumonia, diarrhea, Clostridium difficile colitis, bone fractures, hypomagnesemia, and interstitial nephritis [38‒42]. Recent meta-analysis also reported that PPIs use in hemodialysis patients is independently associated with adverse reactions such as fractures and all-cause mortality. In this study, the relationship between PPI and ESA dose and ERI was clarified for the first time in hemodialysis patients.

Several former studies have been reported on PPI, anemia, and iron metabolism. Termanini et al.’s [43] cohort study of patients with Zollinger-Ellison syndrome concluded there is no association between long-term use of PPIs and iron malabsorption. On the other hand, Lam et al. [16] evaluated PPI use and iron deficiency in a community-based case-control study and showed longer use of PPIs was a risk factor for iron deficiency. Tran-Duy et al. [17] also found that chronic PPI use increases the risk of iron deficiency in a population-based case-control study. The study of Sarzynski et al. [14] reported that patients taking PPIs have lower Hb levels and lower mean corpuscular volume. Previous studies of PPIs and anemia have reported controversial results [14‒17, 43]. However, earlier studies that analyzed PPIs and anemia excluded CKD patients and did not evaluate ESA doses nor was information obtained in earlier studies on Hb, TSAT, or ferritin. Our study, which did include laboratory test values, did, however, show associations between PPIs and ERI and iron status.

Anemia is usually evaluated using Hb or hematocrit. It is possible, indeed likely, that no differences were found in Hb levels with or without PPI use in this study because clinicians routinely make adjustments to ESA doses to maintain stable Hb levels. In this study, as Hb levels were very similar in patients whether they took PPIs or not, most of the difference of ERI appears to result from differences in ESA dosage. Although various causes of ESA resistance have been reported, such as malnutrition [6], vitamin D deficiency [7], renin-angiotensin inhibitor (ACE-i) use, ARB use [8], and severe secondary hyperthyroidism, many of the causes have not yet been elucidated. Patients with hemodialysis using high-dose ESAs, may pose a clinical problems. This study showed that the use of PPIs may cause an increase in the ESA dose and cause ERA-resistance anemia.

Several mechanisms may be involved in an association between PPIs and anemia. First, PPIs could suppress iron absorption and cause iron deficiency anemia. Gastric acid promotes absorption of non-heme iron to an absorbable ferrous form, and this condition is dependent on pH in the stomach. Earlier, animal studies reported omeprazole inhibits iron absorption in rats that were fed an iron-deficient diet [44]. In addition, several clinical studies have reported an association between PPIs and iron deficiency anemia [14‒16]. If dialysis patients are regularly receiving intravenous iron, the impact of decreased iron absorption due to changes in gastric pH by PPIs may be mitigated, lowering the ability to detect associations between PPI use, iron deficiency, and PPI-induced ESA resistance. Although the use of intravenous iron preparations is less common in Japanese hemodialysis patient [13], it is important in future studies to consider the extent to which intravenous iron is administered. Second, PPIs could inhibit vitamin B12 absorption and cause macrocytic anemia. Vitamin B12 in food is bound to protein and is released by gastric acid and pancreatic proteases. Vitamin B12 released from protein binds to the glycoprotein intrinsic factor secreted [18] by gastric parietal cells and is absorbed in the ileum. PPIs inhibit gastric acid secretion and decrease the absorption of vitamin B12 [19]. Vitamin B12 is essential for blood cell production and causes macrocytic anemia in vitamin B12 deficiency [45, 46]. Third, although the detailed mechanism of effects on the bone marrow is unknown, several reports have shown that cytopenia is induced by PPIs [47, 48].

Because PPI use was significantly associated with TSAT, suggesting an association between PPIs and iron metabolism, it is possible that the iron factor may be partially associated in the mechanism of PPIs’ influence on anemia. On the other hand, there were no significant differences in ferritin levels between PPI use and PPI non-use. Although there was no difference between PPI(−) and PPI(+) groups in factors that lead to high ferritin levels, dialysis patients have many factors other than iron metabolism that affect ferritin; it is possible that there was no difference in this study. Because the results of this study showed that PPIs and the ERI were significantly associated even after the analysis included the iron factor, it is suggested that other mechanisms such as vitamin B12 or direct myelosuppression due to PPIs should also be considered. In the subgroup analysis of the ERI in this study, there were significant interactions affecting the ERI in cases with or without GI bleeding. Although the number of cases with a past history of GI bleeding is small, the effect of PPIs on the ERI may differ in patients depending on whether there was GI bleeding or not. A causal statement cannot be made in the current study because of the possibility that patients with active gastrointestinal bleeding who usually use PPIs were using high doses of ESA. However, since no interaction was observed in other subgroup analyses, the relationship between PPI and ERI is considered to be uniform regardless of patient background. In Japan, PPIs are commonly used when prescribing antiplatelet drugs and anticoagulants. Therefore, it is thought that there is a relationship between PPIs and these drugs. Prescription of anticoagulants or antithrombotic drugs may be a marker of potential bleeding. Adjusting for these two regimens in the analysis is thought to reduce the effect of potential bleeding, an unmeasured confounding.

The present study has one notable strength: patients in Japan, unlike other countries, cannot obtain PPIs without a physician’s prescription [49]; we therefore had information about PPI therapy without missing data. In clinical practice in Japan, not only Hb but also TSAT and ferritin levels of dialysis patients are measured periodically (once every 1–3 months); this has the advantage that data on iron metabolism is more detailed than what is found in previous studies using medical records or conducted in a local resident cohort. This study had several limitations. First, because this was an observational study, the association found between PPI exposure and anemia status could not establish causality. Most importantly, it was threatened by confounding by indication: the reasons for PPI prescription could be a gastric symptom, which was not captured in the data and may have caused occult hemorrhage, and anemia eventually. On the other hand, a previous study in which the upper gastrointestinal tract was examined using video capsule endoscopy for screening purposes in dialysis patients [16] reported that asymptomatic gastrointestinal bleeding was not observed. Therefore, the impact of such occult hemorrhage on the association observed in this study might not be critical. Second, the cross-sectional design may cause the reverse causality. However, this is less likely in this study because physicians do not change PPI prescriptions because of anemia in routine practice. Third, in this study, we used ERI as an indicator of ESA usage, but it is not a commonly used indicator in daily clinical practice. On the other hand, the results regarding ESA dosage were also lower in the PPI use group. In the future, we believe it is necessary to conduct more detailed studies that analyze PPI and ERI, as well as PPI and ESA doses. Fourth, chronic inflammation is an important factor involved in renal anemia and also plays a role in ESA dosage. One of the limitations of CRP is that there are many missing values, so we used multiple imputations as a countermeasure. Fifth, although the diagnosis of gastrointestinal bleeding was adjusted in the analysis model as a past history, it cannot be denied that latent and asymptomatic chronic bleeding was a confounding factor. The effects may be minimal as occult bleeding has been reported to be low in hemodialysis patients [50]. However, it cannot be said with certainty that there will be no impact.

In conclusion, the present study revealed associations between PPIs and the ERI, ESA dose, and TSAT in hemodialysis patients. The possible associations of anemia and PPI use should be considered in high ERI and high-dose ESA hemodialysis patients.

ESA-resistant anemia is a major problem in hemodialysis patients, and few studies have analyzed the association between PPIs and anemia or ESA in hemodialysis patients. Associations were found between PPIs and ERI, ESA dose, and TSAT in hemodialysis patients. In patients with high ERIs and high-dose ESAs, the possibility of anemia associated with PPIs should also be considered.

Christopher Holmes at the Institute for Health Outcomes & Process Evaluation Research (iHope International) provided comments, suggestions, and editorial advice on the manuscript.

All participants provided written informed consent and the study protocol of J-DOPPS was approved by the Ethics Committee of Tokyo Women’s Medical University (approval number: 2388-R4).

None declared.

This manuscript was directly supported by the Kidney Foundation, Japan. Global support for the ongoing DOPPS Program is provided without restriction on publications by a variety of funders (details in https://www.dopps.org/AboutUs/Support.aspx). All grants were made to Arbor Research Collaborative for Health and not to coauthors directly.

A.N., Y.M., H.K., N.K., Y.O., T.Y., and M.F. contributed to the study design and coordinated the study. Y.M. conducted the statistical analysis. A.N., Y.M., and Y.O. wrote the first draft of the manuscript. All authors read, reviewed, edited, and finally approved the final manuscript.

The data underlying this article were provided by Arbor Research Collaborative for Health under license and are not publicly available. However, data requests can be sent to Arbor Research via their website (http://www.arborresearch.org/AboutUs/ContactUs.aspx). Further inquiries can be directed to the corresponding author.