A pharmacovigilance study of association between proton pump inhibitor and dementia event based on FDA adverse event reporting system data

Proton pump inhibitor (PPI) was widely used around the world. Studies suggested conflicting results between PPI treatment and dementia event. This study examined the association between six PPI agents and dementia event by mining the US FDA Adverse Event Reporting System (FAERS) database from 2004 to 2020. We employed proportional reporting ratio (PRR) and information element (IC) methods to detect the signals of dementia relevant to PPI. We also analyzed characteristics of PPI and positive control reports, compared dementia event between long- and short-duration of PPI treatment. Finally, we identified 2396 dementia cases with PPI treatment. We did not detect significant signal between PPI and dementia event: PRR = 0.98, 95%CI 0.94 to 1.02, IC = −0.03, 95%CI − 0.17 to 0.10, even in gastroesophageal reflux disease cases: PRR = 0.65, 95%CI 0.59 to 0.72, IC = −0.62, 95%CI − 0.97 to − 0.27. No significant differences of dementia event were detected between long- and short- duration groups, the OR (95%CI) of the 3 years, 5 years and 10 years comparison were 0.70 (0.48 to 1.02), 0.72 (0.45 to 1.15) and 1.65 (0.75 to 3.63), respectively. Based on the current FAERS data mining, we discovered no association between PPI use and dementia event, even in long-term PPI therapy case.

We then conducted PPI signal detection based on cases with indication of GERD, 355 dementia cases were gathered out of 9053 PPI users with GERD indication reported by health professionals in FAERS. However, no significant signal between PPI treatment and dementia event was detected, PRR = 0.65, 95%CI 0.59-0.72, IC = −0.62, 95%CI − 0.97 to − 0.27 (Table 2).
We further performed sensitivity analysis of signal detection for the association between PPI use and dementia event in three independent methods (Supplementary Table S1). The first way, we excluded cases with antidementia drugs as co-therapy from the previously included cases (Supplementary Table S1-1). The second way, we excluded cases with age blow 65 years from the previously included cases (Supplementary Table S1  www.nature.com/scientificreports/
Time event comparison. We estimated the time interval from PPI use to adverse event onset, comparing dementia event between long-and short-time interval groups. 754 dementia cases were identified out of 11,033 PPI users with time interval data reported by health professionals in FAERS. We divided different long-and short-time interval with 3 years, 5 year and 10 years. However, no significant difference was found between each long-and short-term groups ( PPI dosage. The dosage analysis found no difference of daily dosage for each PPI agent between dementia group and non-dementia group. Moreover, the median (IQR) daily dose of each PPI agent was within the range recommended by drug label (Table 4).

Discussion
The current study investigated the association between six PPI agents and dementia event, compared different time interval of PPI treatment and dementia event. The results indicated no association between dementia event and PPI agents, including dexlansoprazole, esomeprazole, lansoprazole, omeprazole, pantoprazole and rabeprazole. To the best of our knowledge, this was the first pharmacovigilance study concerned the association between PPI use and dementia event based on FAERS database. With the widespread use of PPI agents, PPI-associated adverse event had caught health professionals' attention, as well as the public and the media. In FAERS database, more than half of the PPI adverse event cases were reported by non-health professionals. To reduce the influence of non-health professionals, we only included cases reported by health professionals. However, the proportion of case reported by physician and pharmacist was less than a half for PPI, smaller than AC and BD. Moreover, the number of dementia case treated by PPI was increasing by years, especially from the year of 2018. The suddenly increased case number might be caused by the suddenly increased annual FAERS reports in 2018 and concerns of PPI safety in recent years. The risk of stimulated reporting, both by clinical evidence and media influence, could not be ruled out.
Based on the β-amyloid enhancement 8 , vitamin B12 deficiency phenomena 9 and the widespread PPI use, the association between dementia event and PPI use had become a hot topic. Professor Akter et al. first revealed five different PPI agents had varying degrees of influence on different cognitive domains associated with dementia based on the Cambridge Neuropsychological Test Automated Battery (CANTAB) software test 10 . Professor Haenisch et al. conducted the first epidemiological investigation, indicated PPI might have an impact on dementia Table 3. Comparison of dementia event between long-term and short-term proton pump inhibitors treatment cases. AE adverse event, OR odds ratio, 95%CI 95% confidence interval.

Groups Time interval Dementia cases (n/%) All AE cases (N)
OR  www.nature.com/scientificreports/ based on the German Study on Aging, Cognition and Dementia in Primary Care Patients (AgeCoDe) 11 . Then, professor Gomm et al. conducted the first prospective cohort study, revealed regular PPI treatment had a significantly increased risk of dementia using data derived from the largest German statutory health insurer, Allgemeine Ortskrankenkassen (AOK) 12 , which had been hotly commented. However, conflicting results had been gradually published. Professor Lochhead et al. conducted a nationwide prospective cohort study and divided PPI users into four groups based on duration of PPI treatment, revealed a modest association between duration of PPI use and cognitive function, however, Lochhead stated that the result could not support PPI use increased dementia event 31 . Professor Taipale et al. finished a nationwide nested case-control study which set a lag window of different duration, found PPI use was not associated with risk of Alzheimer's disease with a 3-year lag window 32 . Professor Gray et al. reported a prospective cohort study and found no association between PPI exposure and dementia event after a mean follow-up of 7.5 years 20 . Professor Cooksey et al. conducted a large population-based study based on electronic health-data from the Secure Anonymised Information Linkage (SAIL) Databank from 1999 to 2015, could not confirm an association between PPI use and dementia event 17 . The current study based the FAERS big data, indicated no significant signal between PPI use and dementia event. Even compared in different time duration, no significant difference of dementia event was found between long-and short-term groups.
Our study revealed no association between PPI use and dementia event based on the FAERS real world big data, however, certain limitations existed. FAERS is a spontaneous reporting system, voluntary and opened to health professional as well as the public, so under-reporting, over-reporting or missing data was inevitable 33 . The time event comparison only included limited cases with time data reported. Although non-health professionals' reports excluded, the risk of stimulated reporting could not be eliminated.
In summary, the current study revealed no association between six PPI agents and dementia event based on the FAERS data mining. Our findings suggested that dementia event might not be considered as a factor in discontinuing PPI treatment.

Materials and methods
Data source. We downloaded FAERS data from January 2004 to December 2020 in the FAERS Quarterly Data Extract Files website 34 . FAERS data was processed anonymously, no ethical review was required.
The FAERS datasets consisted of seven data tables as follow: "DEMO" table for patient demographic and administrative information, "DRUG" table for the drug information, "REAC" table for adverse events information, "OUTC" table for patient outcomes information, "RPSR" table for report sources information, "THER" table for drug therapy start and end dates information and "INDI" table for the indications for drug use. We managed FAERS data in local by Microsoft SQL server 2017 software.
We first removed duplicated cases from the original data as the FDA recommended. We removed the same records from "DEMO" table and left one, then deleted the earliest FDA_DT when the CASEIDs were the same and removed the lower PRIMARYID when the CASEID and FDA_DT were the same. We further removed cases listed in the FAERS deleted files. In the current study, we only included cases reported by health professionals, including physicians, pharmacists and other health professionals, for both target drug data and background data. Target drug identification. In "DRUG" table, drugs could be documented in various forms, such as generic names, brand names, synonymous names or their abbreviations. We used the MedEx software (MedEx UIMA 1.3.8, Vanderbilt university, US) to standardize different names of the same drug into the "generic name" 35,36 .
We also identified anticholinerigc drugs (AC) and benzodiazepine drugs (BD) as positive controls. We included 56 AC drugs following the Coupland et al. study 37 and 30 BD drugs following three BD data mining studies [38][39][40] , detailed in Supplementary Table S2.
Dementia event identification. According to Medical Dictionary for Regularly Activities (MedDRA) and Standardised MedDRA Queries (SMQs) version 23.1. We identified dementia cases in "REAC" table using SMQ (code: 20,000,073) broad searching, including 105 Preferred Terms (PTs). For cases reported more than one PT of the same SMQ, we removed duplicate records and kept one. The PTs details could be found in Supplementary Table S3.
Data mining. We gathered the characteristics of dementia case with PPI, AC and BD, including age and sex, reporter and report country, annual case reported, indications and daily dosage.
We employed both proportional reporting ratio (PRR, a frequency method) and information component (IC, a Bayesian method) to detect signals of dementia event relevant to PPI, as well as AC and BD positive controls. The calculation method of PRR, IC and the 95% confidence interval (95% CI) were shown in Supplementary  Table S4. A significant signal was defined as both PRR and IC signal detected. The PRR signal criteria was case number ≥ 3, PRR ≥ 2 and χ 2 ≥ 4 41 . The IC signal criteria was IC > 0 and the lower limit of 95% CI > 0 28 .
We further calculated signal between PPI use and dementia event in GERD cases who might receive longterm PPI treatment.