Proton Pump Inhibitors and Risk of Mild Cognitive Impairment and Dementia

Felicia C. Goldstein, PhD; Kyle Steenland, PhD; Liping Zhao, MSPH; Whitney Wharton, PhD; Allan I. Levey, MD, PhD; Ihab Hajjar, MD, MS


J Am Geriatr Soc. 2017;65(9):1969-1974. 

In This Article



Information available in the National Alzheimer's Coordinating Center (NACC) database[8] from 33 ADCs from September 2005 through September 2015 was used. Written consent was obtained using forms that the institutional review boards at each site approved. Participants included persons with a baseline diagnosis of normal cognition or MCI according to their ADC clinicians. Clinical diagnosis at each center relies on ADC coding guidelines using not only performance on a core battery of neuropsychological tests,[9] but also the Clinical Dementia Rating[10] score, which provides an index of cognitive and functional status using a structured interview with the participant, a separate interview with the study informant, and a behavioral and neurological examination. A diagnosis of normal cognition requires that neuropsychological test scores be within expectation for age and that persons be able to perform instrumental activities of daily living independently. A diagnosis of MCI requires impairment in a single cognitive domain or multiple domains but also evidence of independence in performing activities of daily living.[11]

Criteria for inclusion of cognitively normal controls and participants with MCI in the current study were information about PPI medication use at every visit, aged 50 and older at baseline, and a consistent conversion diagnosis over time (e.g., conversion from normal cognition to MCI but not back to normal cognition at the next visit).

PPI Medications

Participants were classified as PPI users or nonusers at each annual visit based on information on self- and informant-reported PPIs. PPIs included omeprazole (Prilosec) and omeprazole-sodium bicarbonate (Zegerid), esomeprazole (Nexium), lansoprazole (Prevacid), pantoprazole (Protonix), rabeprazole (AcipHex), and dexlansoprazole (Dexilant, Kapidex).

Statistical Analyses

Analyses included participants taking PPIs at every visit, those taking PPIs intermittently, and those not taking PPIs at any visit. Multivariable Cox regression analyses using PROC PHREG were performed to evaluate the association between PPI use (always or intermittently vs never) and decline in cognitive status to MCI or dementia in the combined population of those who had normal cognition or MCI at baseline. Follow-up time was the time variable in the regression, beginning with the baseline visit. This was defined as time to decline for those who declined and follow-up time for those who did not decline. A similar regression was run in which the outcome was decline in cognitive status specifically to an underlying AD etiology (conversion to MCI with suspected AD etiology or probable AD), in which MCI due to other suspected etiologies and non-Alzheimer's dementias were treated as censored. Finally, separate regressions were run for persons with normal cognition at baseline versus those with MCI at baseline to determine whether one diagnostic group was more vulnerable to the adverse effects of PPIs on risk of decline. Proportional hazard assumptions were tested using an interaction term between PPI use and follow-up time. No violation of the proportional hazard function was found for PPI. Two-tailed statistical significance was set at P < .05. All models controlled for potential confounders, including demographic variables (age at baseline, race, sex, education), vascular comorbidities (self-reported hypertension, diabetes mellitus, heart disease, stroke or transient ischemic attack), mood (depression), and anticholinergic medications and H2RAs. Information on vascular comorbidities and mood were taken from the NACC Uniform Data Set Subject Health History form, which enquires about whether a person has had a diagnosis in the past 2 years of these conditions. Information about medications was taken from the NACC Uniform Data Set Subject Medications form, which enquires about use of prescription and commonly prescribed over-the-counter medications within the 2 weeks before the current visit. The form lists the 100 drugs that NACC participants most commonly report. Additional drugs and their associated DrugIDs are also available at https// In the current study, medications were classified into drug classes including H2RAs (cimetidine, ranitidine, famotidine, nizatidine)[7] and those with moderate to severe anticholinergic effects.[12]