Background
Proton pump inhibitors (PPIs.) are widely used for the treatment of gastrointestinal disorders such as peptic ulcer disease and dyspepsia. Although PPIs have been recognized to show high efficacy , there are growing safety concerns about PPI causing a potential risk of acute kidney injury (AKI), acute interstitial nephritis (AIN) and chronic kidney disease [
1‐
12]. The effect of PPI use on the development of AKI has been detected in several large observational studies [
1‐
6,
12]. However, those studies defined AKI based on the medical records with ICD-9 or ICD-10 [
1‐
6,
12]. Because AKI is commonly under-recorded in hospital data [
13], the absolute risk of AKI in the target population may be underestimated in those studies. Other studies have used laboratory data to detect AKI in PPI users [
7‐
11]. However, these studies have the following limitations. First, they identified H
2-receptor antagonist users as active comparator [
7‐
10]. With this design, the effect of PPIs could be significantly affected by confounding by indication. Second, it was not investigated whether AKI occurred during the exposure period to PPIs [
11]. In addition, no studies have classified the stage of AKI. Therefore, further studies are required to validate whether PPI use is associated with the risk of any stage of AKI or more severe AKI.
PPIs play a pivotal role in the standard of care for
Helicobacter pylori eradication and are often concomitantly used with antibiotics, such as macrolides and penicillins. Although macrolide antibiotics are considered to have relatively low nephrotoxicity [
14], they are well known to increase the risk of AKI due to drug-drug interactions. For example, the concomitant use of macrolide antibiotics with calcium channel blockers [
15] or HMG-CoA reductase inhibitors (statins) [
16] cause a decrease in the clearance of these drugs, and a resultant increase in circulating drug, thereby increasing the risk of drug-induced AKI. For PPIs, previous studies have shown that clarithromycin, a macrolide antibiotic, decreases the clearance of omeprazole [
17], lansoprazole [
18], and esomeprazole [
19]. However, it is unclear whether the risk of AKI is affected by the interactions between PPIs and macrolide antibiotics. Therefore, further studies are required to investigate the effect of macrolide antibiotic use on the risk of AKI in PPI users.
Based on the above backgrounds, we investigated the association between PPI use and the development of AKI using a self-controlled case series (SCCS) design [
20‐
22]. This design allowed us to assess the effects of PPI use on renal function while minimizing chance of potential and unmeasured confounding factors. We also evaluated how the relative risk of AKI in PPI users was associated with concomitant use of macrolide antibiotics.
Discussion
With PPIs being one of the most commonly used classes of drugs, the potential for adverse reactions to PPI use should be clarified. Recently, an association between PPI use and increased risk of AKI has been reported [
1,
2,
4‐
12]. However, these studies have the following shortcomings in their study design: First, the severity of the AKI was not evaluated [
1,
2,
4‐
6,
12]. Second, the effect of confounding by indication was not eliminated [
6‐
10]. Third, temporal proximity between PPI exposure and AKI onset was not considered [
11]. Therefore, the main objective of this study was to confirm the association between PPI use and the risk of any stage of AKI or more severe AKI (stage 2 or higher). To our knowledge, this is the first study to use a SCCS study and serum creatinine data to investigated the association between PPI use and the risk of AKI. We also evaluated the effect of the concomitant use of macrolide antibiotics on the risk of AKI in PPI users.
This study confirmed that PPI use is associated with an increased risk of AKI. The direction and significance of the results are consistent with those of previous studies [
1,
2,
4‐
8,
10‐
12]. However, the severity of AKI has not been evaluated in previous studies. Some studies measured AKI using ICD codes [
1,
2,
4,
5,
12], but this method does not provide information on the severity of AKI [
13]. Other studies used serum creatinine data to measure the incidence of AKI [
7‐
11] but only determined whether it was stage 1 or higher. On the other hand, we found that more severe AKI (KDIGO stage 2 or 3) was associated with PPI use, which expands on the findings of previous studies. Taken together, these results indicated that PPI use is independent risk factors for the development of AKI. Given this notion, further attention should be paid to the development of AKI during the course of treatment with PPIs.
Macrolide antibiotics have the potential to alter the pharmacokinetics and risks of adverse reactions to commonly used drugs because they are enzyme inhibitors [
15,
16]. In previous studies, macrolide antibiotics increased the serum concentration of PPIs [
17‐
19], implying that its concomitant use with macrolide antibiotics may alter the risk of AKI in PPI users. Therefore, we investigated the effects of PPIs and macrolide antibiotics on the risk of AKI. The results of this study suggest that the association between macrolide antibiotic use and AKI risk among PPI users is not significant. Although the pharmacokinetics of PPIs was not evaluated in this study, our results suggest that the exposure levels of PPIs are not related to their nephrotoxicity. This is supported by a previous study showing that the genotype or phenotype of CYP2C19, a PPI-metabolizing enzyme, does not affect the risk of AIN in PPI users [
31]. AIN is the most frequently reported pathology in patients with PPI-related AKI [
1‐
3,
32,
33]. These results suggest that indirect mechanisms rather than direct nephrotoxicity may be involved in the development of PPI-related nephropathy. In previous studies, it was suspected that several mechanisms of drug hypersensitivity reactions play a role in the pathogenesis of PPI-related AKI and that these mechanisms overlap [
34]. On the other hand, there is growing evidence that the reduction of gastric acidity induced by PPI use leads to changes in the gut microbiome and that PPI-induced dysbiosis is associated with the progression of PPI-related adverse effects [
35]. For example, omeprazole has been shown to increase the expression of inflammatory cytokines in liver tissue and increases its vulnerability to hepatic injury, which occurs via changes in the gut microbiota [
36]. It should also be investigated whether changes in the gut microbiota and subsequent changes in inflammatory status are related to PPI-associated nephropathy.
This study has several strengths. First, we used the SCCS study which compares risk within individuals at different periods and thus, the results are less influenced by potential confounding between comparison groups than cohort studies or case control studies. In the present study, we only included patients who developed AKI and used PPIs and macrolide antibiotics at least once, which minimized confounding by indication. Second, the IRR was adjusted by the time method using a regression model. Third, we used serum creatinine data to measure the development of AKI with reference to the National Health Service (NHS) England AKI algorithm criteria [
23‐
25]. We also assessed the stage of AKI. A recent study validated the AKI algorithm by comparing a clinical diagnosis by experienced nephrologists with the NHS England criteria [
37]. The study indicated that 90.5% of patients with a clinical diagnosis of AKI was detectable using the AKI algorithm of the NHS. Therefore, we assumed that AKI could be detected with a high sensitivity in this study. Fourth, since low renal function is itself a risk factor for AKI, we adjusted the incidence ratio by baseline renal function.
This study has several limitations. First, we only included patients who underwent blood examinations at least once a year in the study hospital and did not filter by the frequency of the examination. The frequency of blood examinations in patients who did not develop AKI was 11.0 ± 14.9 times a year and that of those who did develop AKI (named the ‘cases’) was 27.2 ± 25.8 times a year. It is possible that few patients underwent blood examination at other hospitals. Moreover, the data on the urine volume is missing. Therefore, it is possible that there was a delay in identifying the timing of AKI onset or misclassification of non-cases. Second, we did not consider liver function or systemic inflammation to be time-varying factors. The reason for this setting was to avoid missing data, which could decrease the number of cases and the power of detection. However, the criteria for selection of the study population may have led to decreased sensitivity of AKI detection and modification of the results by unmeasured confounders. For example, rapid changes in renal functions due to primary disease or comorbidities could influence the development of AKI. Third, we did not include comorbidities and administration of nephrotoxic agents other than antibiotics as confounding factors. Fourth, severe AKI could lead to death or dialysis dependence [
29,
30] and observation may have been terminated due to the occurrence of severe AKI. However, the results of main analysis were consistent in the sensitivity analysis that excluded cases in which observation was terminated due to death (Tables S
3 and S
4). Fifth, the timing of drug administration was based on prescription records. Therefore, the actual timing of drug administration may be different. In addition, adherence to medication is unknown. Future studies should include prospective measurements of serum creatinine levels and use databases to accurately determine the duration of drug exposure.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.