Proton pump inhibitors associated acute kidney injury and chronic kidney disease: data mining of US FDA adverse event reporting system

Proton pump inhibitors (PPIs) were widely used. Observational studies suggested increasing risk of kidney injury in patients with PPIs treatment. We gathered six PPI regimens and adverse reports of acute kidney injury (AKI) and chronic kidney disease (CKD) based on US FDA Adverse Event Reporting System (FAERS) database from 2004 to 2019. We employed reporting odds ratio (ROR) to detect signals. Finally, we identified 3187 PPIs-associated AKI cases and 3457 PPIs-associated CKD cases. We detected significant signals between PPIs and AKI as well as CKD. The signal strength was stronger for CKD (ROR = 8.80, 95% CI 8.49–9.13) than AKI (ROR = 3.95, 95% CI 3.81–4.10), while dexlansoprazole performed stronger association for CKD (ROR = 34.94, 95% CI 30.89–39.53) and AKI (ROR = 8.18, 95% CI 7.04–9.51) than the other five PPIs. The median time from PPIs use to event occurrence was 23 days for AKI and 177 days for CKD. PPIs-associated AKI resulted larger proportion of death, life-threatening, hospitalization and disability events than PPIs-associated CKD. By mining the FAERS big data, we provided more information between PPIs use and the AKI and CKD events. PPIs rational use should be repeatedly stressed.

Signal detection. We first conducted signal detection based on all PPIs, detected significant AKI and CKD signals. The strength of CKD signal (ROR = 8.80, 95% CI 8.49-9.13) was stronger than AKI (ROR = 3.95, 95% CI 3.81-4.10). Then we conducted signal detection in individual PPI, detected significant AKI and CKD signals in all the six PPIs (Table 2).

Time interval calculation.
The time interval analysis included a total of 848 cases, including 562 cases for AKI and 286 cases for CKD.
The median time from PPIs use to AKI occurrence was 23 (interquartile range (IQR) 4 to 179) days (Fig. 4a). The longest AKI occurrence median time was 476 (IQR 51 to 917) days for dexlansoprazole, while the shortest time was 6 (IQR 2 to 20) days for esomeprazole.
The median time from PPIs use to CKD occurrence was 177 (IQR 15 to 1528) days (Fig. 4b). The longest CKD occurrence median time was 1495 (IQR 114 to 2883) days for lansoprazole, while the shortest time was 9 (IQR 3 to 22) days for esomeprazole. Prognosis analysis. We analyzed outcomes of PPIs-associated AKI and CKD events. Compared with PPIsassociated CKD, PPIs-associated AKI resulted larger proportion of death, life-threatening, hospitalization and disability events. The individual PPI-associated AKI and CKD outcomes were shown in Table 3.    of the WHO DDD, followed by esomeprazole (CKD group), 1.33-fold of the WHO DDD, however the dose ranges were within that recommended by drug label. The median daily doses of other PPIs were equal to or lower than the WHO DDD (Table 4).

Discussion
The present study investigated six PPIs-associated AKI and CKD, as well as indications and outcomes of reported cases, time interval to events occurrence and dosage of PPIs using. The results indicated significant association between AKI or CKD events and PPIs, including dexlansoprazole, lansoprazole, pantoprazole, omeprazole, esomeprazole and rabeprazole. The signal strength was stronger in CKD than AKI and varied across specific PPI regimens. PPIs-associated renal injury had caught professionals' attention, as well as the public and the media. Recently, many studies focused on PPIs associated AKI 8,16,17 and CKD 9,18-21 . To reduce the influence of the public and the media, we had removed cases reported by consumers and lawyers. However, more than half the cases were reported in 2019, the risk of stimulated reporting could not be ruled out.
Overuse of PPIs was a worldwide problem. Our study found more than 20% PPI cases did not report clear indications. A PPI prescription survey of 45 hospitals published in 2019 indicated that between 32 and 56% of the PPI prescriptions did not have appropriate indications in China 2 . And an multicenter observational study in Table 3. Outcomes of cases reported as AKI and CDK using PPIs from FAERS database. PPIs proton pump inhibitors, AKI acute kidney injury, CKD chronic kidney disease.  www.nature.com/scientificreports/ 2020 showed a range from 22 to 63% of PPIs treatment in Italian nursing homes were inappropriate 3 . To decrease drug related kidney injury, improper PPI use should be avoided firstly.

Renal injury Drugs Cases with renal injury/N
Omeprazole is the PPI with the longest history of clinical use and first report of PPIs-associated kidney injury 22 . However, the most cases reported of PPIs-associated AKI and CKD was lansoprazole, and the strongest signal of PPIs-associated AKI and CKD was dexlansoprazole from FAERS. Welch analyzed drug-associated AKI compared with all other drugs using FAERS data from 2004 to 2015, found the ROR was 2.35 (95% CI 2.19-2.53) for omeprazole 23 , being consistent with the result of our study. Chen detected PPIs-associated AKI based on cases reported by all reporters in FAERS from 2004 to 2019, the ROR value of each PPI was bigger than our study 17 , indicated over-estimation could be reduced to a certain extent by removing non-health professionals' reports.
Our study found the highest yearly proportion of PPIs-associated AKI and CKD was within the first years after PPIs treatment and the median time was 23 days for AKI and 177 days for CKD. Chen found a longer median time (446 days) for PPIs-associated AKI 17 . Studies revealed the risk of AKI or CKD increased with cumulative or prolonged PPIs use 20,24 . Unnecessary long-term PPIs use should be avoided.
Studies revealed PPIs was associated with increased risk of death [25][26][27] . Research reported the all cause death ratio of PPIs was 13.40%, 4861 deaths out of 36,282 PPIs users, and PPIs showed higher odds (OR 1.76) for mortality after logistic analysis 25 . Another study reported the estimated death ratio due to PPIs-induced CKD was 0.419% 26 . In our study, outcome analysis revealed all cause death proportion (PPIs-associated renal injury number as denominator) was 10.85% for PPIs-associated CKD and 12.27% for PPIs-associated AKI. Because the FAERS data did not include PPI users without adverse event occurrence, the incidence rate could not be calculated. However, such a high proportion of death and hospitalization could still attract our attention.
The findings of the present study and the existing evidence should arouse vigilance of PPI use. PPIs should be prescribed or sold following proven indication, low dose and short course 28 . PPIs users with no or unknown indication should stop PPI treatment, and those with indications should avoid long term PPI use if unnecessary 26 .
Patients with primary kidney disease should pay more attention using PPIs. Monitoring of kidney function for long-term PPI users was necessary 26 .
Our study revealed significant association between AKI, CKD and PPIs treatment based on the FAERS real world big data, however, certain limitations existed. Like other spontaneous reporting system, FAERS is voluntary and opened to health professional as well as general public, so under-reporting, over-reporting, varied or missing information was inevitable 29 . Some calculation, especially PPIs duration and PPIs dose, just included cases with data reported. It was also difficult to perform risk factors analysis between PPIs and kidney injury for the limited information. Although non-health professionals' reports excluded, the risk of stimulated reporting could not be eliminated.

Conclusion
The present study identified six PPIs-associated AKI and CKD signals based on the FAERS data. Significant signal strength was stronger for CKD than AKI, while dexlansoprazole performed stronger association for AKI and CKD than the other five PPIs. A high proportion of renal injury occurred after the first year of PPI treatment and the median time was longer for CKD than AKI. Besides, more than 20% PPI treatment did not have clear indications. Our findings suggest PPIs-associated AKI and CKD should be taken into account, and the importance of PPIs rational use should be reemphasized.

Methods
Data source. We downloaded 64 quarters of adverse event report datasets from the FAERS Quarterly Data Extract Files website 30 , covering the period from January 2004 to December 2019.
Each quarterly data was consisted of seven data tables, including the DEMO table for patient demographic  and administrative information, the DRUG table for the drug information, the REAC table for adverse events  information, the OUTC table for patient outcomes information, the RPSR table for report sources information,  the THER table for drug therapy start and end dates information and the INDI table for the indications for drug use. We imported all data into the local FAERS database, managed by Microsoft SQL server 2017 software.
We de-duplicated cases from the original datasets following the FDA recommendations. We removed the same records from the DEMO table and kept one, then removed the earliest FDA_DT when the CASEIDs were the same and removed the lower PRIMARYID when the CASEID and FDA_DT were the same. To make the evaluation more accurately, we only included cases reported by health professionals, including physician, pharmacist and other health professional in the OCCP_COD column of the DEMO table.
PPIs identification. In the DRUG table, drugs could be documented in brand names, synonymous names or their abbreviations besides generic names in the DRUGNAME or the PROD_AI columns. We used the MedEx software (MedEx UIMA 1.3.7, Vanderbilt university, US) to transform different forms of drug name into the "generic name", and add an identification code named RXNORM_RxCUI 31,32 .
We tried to identify the seven single component PPI regimens with the WHO Anatomical Therapeutic Chemical (ATC) code of A02BC from the local FAERS database. The seven PPIs included omeprazole (ATC code: A02BC01), pantoprazole (ATC code: A02BC02), lansoprazole (ATC code: A02BC03), rabeprazole (ATC code: A02BC04), esomeprazole (ATC code: A02BC05), dexlansoprazole (ATC code: A02BC06) and dexrabeprazole (ATC code: A02BC07). We restricted the drug role as primary suspected (PS) drug in the DRUG One case could be reported more than one PTs of the same SMQ, duplicate records was removed. For example, one case reported two records of "acute kidney injury" and "acute phosphate nephropathy", we counted the two records as one AKI case. The PTs details could be found in Supplementary Table S3.
Data mining. We counted the characteristics of AKI and CKD cases with PPIs use, including cases attributed to different PPIs, age and sex of patient, occupation and country of reporter, year of event reported and indications for PPIs application. We employed reporting odds ratio (ROR) to detect the signals of AKI and CKD relevant to PPIs. In current study, ROR was the degree of disproportionate reporting of AKI or CKD event for PPIs compared to the same event for all other drugs in the FAERS database. The calculation method of ROR, 95% confidence interval (95% CI) for ROR were shown in Table 5. A significant signal is detected when the report cases ≥ 3 and the lower limit of 95% CI exceed one 33 . Basically, the higher the ROR value, the stronger the strength of the signal 13 .
We further estimated the time interval from PPIs use to AKI and CKD events occurrence. We unified the time format as yyyy-mm-dd. The time interval was calculated using event date (EVENT_DT) in the DEMO table minus drug start date (START_DT) in the THER table. To make the calculation more accurately, we excluded cases not in the period of 2004 to 2019, cases without year, month or day data in either EVENT_DT or START_DT field, and cases with earlier event date than drug start date.
We also analyzed patient outcomes for the events, including death, life-threatening, hospitalization and disability for the PPIs associated AKI and CKD cases. If a case reported more than one outcome, we kept the worst one, for instance, one case reported both death and life-threatening, we removed the life-threatening one.
Finally, we analyzed PPIs daily dose and compared to the defined daily dose (DDD) recommended by the world health organization (WHO) or the label dose.
The statistical analyses were conducted by SPSS version 20.0 (IBM corporation, Armonk, New York, USA) and GraphPad prism version 8.0.2 (GraphPad Software, San Diego, California, USA). Table 5. Two-by-two contingency table for reporting odds ratio analyses. PPIs proton pump inhibitors, AKI acute kidney injury, CKD chronic kidney disease, ROR reporting odds ratio, 95% CI 95% confidence interval.

Drugs AKI/CKD cases All other adverse event cases
PPIs a b All other drugs c d ROR = a/b c/d , 95% CI for ROR = exp (ln (ROR) ± 1.96 1 a + 1 b + 1 c + 1 d )