Association Between Proton Pump Inhibitor Use and Cognitive Function in Women

Abstract

Background & Aims

Studies have reported associations between proton pump inhibitor (PPI) use and dementia. However, data are lacking on long-term PPI use and cognitive function. We therefore examined associations between PPI use and performance in tests of cognitive function. Because of shared clinical indications, we examined associations for H2 receptor antagonists (H2RAs) as a secondary aim.

Methods

We used prospectively collected data on medication use and other potential risk factors from 13,864 participants in the Nurses’ Health Study II who had completed a self-administered computerized neuropsychological test battery. Multi-variable linear regression models were used to examine associations between medication use and composite scores of psychomotor speed and attention, learning and working memory, and overall cognition.

Results

We observed a modest association between duration of PPI use and scores for psychomotor speed and attention (mean score difference for PPI use of 9–14 yrs vs never users, −0.06; 95% CI, −0.11 to 0.00; Ptrend = .03). After controlling for H2RA use, the magnitude of this score difference was attenuated. Among individuals who did not regularly use PPIs, duration of H2RA use was associated with poorer cognitive scores, with the strongest association apparent for learning and working memory (mean score difference for H2RA users of 9–14 years vs never users, −0.20; 95% CI, −0.32 to −0.08; Ptrend < .001).

Conclusions

In an analysis of data from the Nurses’ Health Study II, we did not observe a convincing association between PPI use and cognitive function. Our data do not support the suggestion that PPI use increases dementia risk. Since our primary hypothesis related to PPI use, our findings for H2RAs should be interpreted with caution.

INTRODUCTION

Proton pump inhibitors (PPIs) are potent suppressors of gastric acid secretion that have been marketed for over 25 years for the treatment of acid-related upper gastrointestinal disorders.¹ Since their introduction, the prevalence of PPI use has increased substantially,^2,3 and global total cost expenditure on PPIs ranks second only to statins.⁴ PPIs are well-tolerated, having few acute side effects,⁵ and are now available without prescription in many countries.⁶ However, observational studies of PPI use have reported associations with several adverse outcomes including community acquired pneumonia,⁷ hip fracture,^8,9 hyomagnesemia,¹⁰ chronic kidney disease,¹¹ and Clostridium difficile-associated diarrhea.¹² A recent pharmacoepidemiologic analysis conducted using a German claims database found that patients prescribed PPIs had a 44% increased risk of incident dementia diagnosis compared to those not receiving PPIs.¹³ The existence of a causal mechanism linking PPI use to dementia is suggested by observations from cellular and animal models of Alzheimer’s disease (AD), where PPI exposure appears to influence amyloid-β metabolism.¹⁴ However, other preclinical data on PPIs and AD are conflicting.¹⁵

Cognitive function has been shown to predict the risk of dementia in later life.¹⁶ We therefore tested the hypothesis that PPI use would be associated with poorer cognitive function utilizing a subgroup of participants in the Nurses’ Health Study II (NHS II) with available data on mediation use and cognitive function. As a secondary aim, we also examined associations between use of H2 receptor antagonists (H2RAs) and cognitive outcomes given the shared clinical indications with PPIs.

SUBJECTS AND METHODS

The NHS II is a nationwide prospective cohort that began in 1989 with the enrollment of 116,430 female nurses, aged 25–42 years.¹⁷ Follow-up has been achieved by the return of biennial questionnaires, which update information on a variety of lifestyle and health-related factors. Follow-up exceeds 90% of total potential person years. The institutional review boards of the Brigham and Women’s Hospital and Harvard Medical School approved this study. Consent was implied by return of questionnaires.

Assessment of PPI exposure

Starting in 2001, the medication section of each questionnaire queried whether or not participants had regularly used a PPI or H2RA during the preceding two years. Those who answered yes were considered regular users for that 2-year cycle for the purposes of the current analysis. Those who completed the medication section, but did not affirm regular PPI or H2RA use, were considered non-users for that 2 year period. Participants were not asked about the dose, type or brand of PPI or H2RA taken.

Assessment of cognitive function

Between 2014 and 2016, a subset of 39,960 participants in the NHS II, who had previously responded to supplementary questionnaires related to violence, trauma and mental health, and who had a known e-mail address, were invited to complete the Cogstate cognitive battery.¹⁸ A total of 14,151 women completed the neuropsychological battery, which is self-administered online using a home computer. The characteristics of women who accepted the Cogstate invitation were similar to those who did not respond.¹⁹ The battery comprises four tasks presented in the following order: Detection (DET), which measures psychomotor function and information processing speed; Identification (IDN), which measures visual attention and vigilance; One Card Learning (OCL), which measures visual learning and short-term memory; and One Back (ONB), which measures attention and working memory. All tasks involve images of playing cards, due to their familiarity across cultures and ages, and require participants to indicate responses by pressing either of two assigned keys on their computer keyboards.^18,20 The entire battery takes 15–20 minutes to complete. Consistent with previous studies,²⁰ mean response times for correct trials of DET, IDN and ONB were normalized by log₁₀ transformation, while arcsine of the square root was applied to the proportion of correct OCL responses. We generated three composite scores by averaging standardized scores (z-scores) from individual Cogstate tasks.¹⁹ A composite score of psychomotor speed and attention was derived from DET and IDN, a composite measure of learning and working memory from OCL and ONB, and a composite score for overall cognition from the z-scores for all four tasks. It has been suggested that the use of these composite scores may increase power by enhancing precision and sensitivity.^21,22 The validity of Cogstate has been previously established.^18,23 Furthermore, deficits in cognitive domains assessed by Cogstate are recognized to be predictors of dementia.²⁴ To provide a point of reference for interpretation of Cogstate scores, mean scores for overall cognition across the whole study population were 0.24, 0.12, −0.03, and −0.18 for age groups <55 years, >55–60 years, >60–65 years, and >65 years respectively.

Study population

Consistent with previous research,²⁵ we excluded the small number of participants who failed integrity checks on all four Cogstate tasks or had insufficient data to calculate at least one composite cognitive score (n=125; 0.9% of Cogstate completers). Data on PPI and H2RA use were available from 7 biennial study questionnaires covering the period 1999 through 2013. Among the 14,026 women with valid Cogstate data, we excluded participants who failed to complete three or more of the questionnaire medication sections during follow-up (n=162), leaving 13,864 women for inclusion in the analysis. Thirty-two participants (0.2% of the final study population) had sufficient data to calculate only one composite score; thus, for each cognitive outcome, the number of participants in the analysis differs very slightly.

Statistical analysis

Our primary outcome measures were composite scores of psychomotor speed and attention, learning and working memory, and overall cognition. We employed multivariable-adjusted linear models to evaluate the association between duration of acid suppressant medication use and each cognitive score, after adjusting for known cognitive risk factors described in the models below. To test for linear trend, we used duration of PPI or H2RA use as a continuous variable, representing cumulative 2-year cycles in which regular use was reported (range 0–14 years). We also generated a categorical variable for each medication (never regular use vs. regular use of 1 to 4 years, 5 to 8 years, or 9 to 14 years). Never users were individuals who had never reported regular PPI or H2RA use on any study questionnaire. Relative to the referent category of never users, we computed estimates and 95% confidence intervals (CI) for the mean differences in cognitive scores for each category of use duration.

Our base model (model 1) was adjusted for age at cognitive assessment (continuous), and surrogate indicators of educational attainment and socioeconomic status: husband’s highest educational level, queried in 1999 (high-school, undergraduate, post-graduate, unmarried/unknown), and highest educational level of either parent, queried in 2005 (high school, undergraduate, post-graduate, unknown). Educational attainment has not been directly queried in participants because all are registered nurses and likely to have similar educational backgrounds. A second multivariable model (model 2) was additionally adjusted for body mass index (BMI) (continuous, kg/m²), regular use of antidepressants (yes, no), smoking status (never, past, current), regular use of multivitamins (yes, no), history of high blood pressure (yes, no), high cholesterol (yes, no), stroke (yes, no), myocardial infarction (yes, no), type 2 diabetes (yes, no), and regular use of aspirin or non-steroidal anti-inflammatory drugs (both yes, no). Information for these covariates was obtained from the questionnaire returned most proximate to the date of cognitive testing (i.e. 2013). For physical activity (continuous, MET-hours/week), alcohol consumption (none, 1–14 g/day, ≥15 g/day) and diet quality (continuous, Alternative Healthy Eating Index 2010,²⁶ excluding alcohol), we used cumulative average values to minimize variance and represent long-term patterns, consistent with prior studies.^26–28

Because duration of medical conditions, such as diabetes or cardiovascular disease, may be associated with both cognitive function and duration of acid-suppressant use, we constructed a third multivariable model containing the same covariates as model 2, but with adjustment for time since the first report of each medical condition (as continuous variables with zero representing never reported), pack-years of smoking, and cumulative cycles of antidepressant, aspirin, and NSAID use (model 3).

Based on a sample size of 13,864 and an alpha of 0.05, we estimated that we would have 80% power to detect a difference in overall cognitive score of 0.007 for each year of PPI use in a test of linear trend. All analyses were performed using SAS (version 9.3; SAS, Cary, NC). All P values were 2-sided and values <.05 were considered statistically significant.

RESULTS

The mean age at the time of cognitive testing was 61 years (range 50–70 years). Compared to individuals who had never reported regular PPI use, PPI users tended to be slightly older, were more likely to report a history of chronic medical conditions, such as cardiovascular disease and diabetes, and more likely to regularly use antidepressants and aspirin (Table 1). PPI users also tended to be less physically active, had higher BMIs, poorer surrogate measures of educational attainment, and slightly lower dietary quality scores.

Table 1.

Age-standardized characteristics of NHS II cognitive sub-study participants according to duration of PPI use in 2013.

	PPI never-users	PPI use 1–4 years	PPI use 5–8 years	PPI use 9–14 years
n (% of study population)	9,252 (67)	2,334 (17)	1,181 (8.5)	1,097 (7.9)
Mean age at cognitive assessment, years (SD)a	60.9 (4.6)	61.5 (4.6)	61.7 (4.4)	62.2 (4.4)
Husband’s education level, %
High school or less	14	16	17	18
College degree	43	42	41	39
Graduate school	32	28	27	28
Unmarried/unknown	12	14	15	16
Highest parental education level, %
High school or less	46	50	53	54
College degree	23	23	22	23
Graduate school	26	23	22	20
Unknown	4.2	3.9	3.2	2.7
Mean BMI, kg/m2 (SD)	26.6 (6.1)	27.8 (6.2)	28.8 (6.5)	29.9 (7.0)
Smoking status, %
Never	67	63	61	65
Former	30	33	35	33
Current	2.9	4.2	4.0	2.2
Alcohol consumption, %
No alcohol	22	23	22	27
1–14 g/day	64	64	67	62
≥15 g/day	14	13	12	12
Hypertension, %	32	42	50	57
High cholesterol, %	54	63	71	75
Myocardial infarction, %	0.8	1.0	1.9	2.7
Stroke, %	1.0	0.8	1.5	1.5
Type 2 diabetes, %	4.9	8.5	11	16
Mean physical activity, MET-h/wk (SD)b	24.5 (22.3)	22.7 (21.6)	20.3 (18.1)	18.6 (19.9)
Regular aspirin use, %c	7.8	11	9.8	13
Regular NSAID use, %c	37	40	42	37
Regular multivitamin use, %c	57	57	57	55
Regular anti-depressant use, %c	15	26	33	40
Healthy eating index scored	54.2 (10.0)	53.6 (10.5)	52.5 (10.4)	52.3 (9.7)

Values are means (standard deviation) or percentages standardized to the age distribution of the study population

Values of polytomous variables may not sum to 100% due to rounding

^aValue is not age-adjusted.

^bMET-h/wk, metabolic equivalent hours per week.

^cParticipants who responded ‘yes’ when asked whether they had regularly used aspirin, NSAIDs, anti-depressants, or multivitamins over the preceding two years.

^dAlternative Healthy Eating Index 2010, higher scores are associated with lower risks of chronic diseases.

BMI, body mass index; NSAID, non-steroidal anti-inflammatory drug; PPI, proton pump inhibitor.

Compared to never-users, increasing duration of regular PPI use was associated with trends toward poorer scores for all three cognitive domains (Table 2). However, a modest statistically significant association was apparent only for psychomotor speed and attention after multivariable adjustment in model 3, which included duration of chronic medical conditions. Mean score differences were 0.00 (95% CI, −0.04 to 0.04) for PPI users of 1–4 years, −0.03 (−0.08 to 0.03) for 5–8 years, and −0.06 (−0.11 to 0.00) for 9–14 years compared to never users (P_trend=.03). For comparison, in multivariable models, a one year increase in age was associated with mean score decreases of 0.03 for psychomotor speed and attention, 0.02 for learning and working memory, and 0.03 for overall cognition.

Table 2.

Mean differences in cognitive scores according to duration of PPI use

Composite score	Estimates (95% CI) for mean difference in cognitive scores according to duration of PPI use
	No PPI use	1–4 years	5–8 years	9–14 years	Ptrenda
Psychomotor speed, attention (n)	(9,235)	(2,328)	(1,181)	(1094)
Model 1b	Reference	−0.02 (−0.06, 0.02)	−0.05 (−0.11, 0.00)	−0.10 (−0.15, −0.04)	<.001
Model 2c	Reference	−0.01 (−0.05, 0.03)	−0.04 (−0.09, 0.02)	−0.06 (−0.12, −0.01)	.01
Model 3d	Reference	0.00 (−0.04, 0.04)	−0.03 (−0.08, 0.03)	−0.06 (−0.11, 0.00)	.03
Learning, working memory (n)	(9,248)	(2,334)	(1,181)	(1095)
Model 1b	Reference	−0.02 (−0.05, 0.01)	−0.03 (−0.07, 0.02)	−0.08 (−0.12, −0.03)	<.001
Model 2c	Reference	−0.01 (−0.04, 0.02)	0.00 (−0.04, 0.05)	−0.03 (−0.08, 0.01)	.31
Model 3d	Reference	0.00 (−0.04, 0.03)	0.01 (−0.03, 0.05)	−0.03 (−0.07, 0.02)	.50
Overall cognition (n)	(9,231)	(2,328)	(1,181)	(1092)
Model 1b	Reference	−0.02 (−0.05, 0.01)	−0.04 (−0.08, 0.00)	−0.08 (−0.13, −0.04)	<.001
Model 2c	Reference	−0.01 (−0.04, 0.02)	−0.02 (−0.06, 0.02)	−0.05 (−0.09, 0.00)	.02
Model 3d	Reference	0.00 (−0.03, 0.03)	−0.01 (−0.05, 0.03)	−0.04 (−0.08, 0.00)	.07

^aP for trend calculated using the total number of questionnaire cycles where regular PPI use was reported, as a continuous variable in the model.

^bModel 1: Adjusted for age and educational attainment of parents and husband.

^cModel 2: Adjusted for variables included in Model 1 plus smoking status, body mass index, alcohol intake, physical activity, antidepressant use, aspirin use, NSAID use, Alternative Healthy Eating Index score, multivitamin use, and history of hypertension, stroke, type 2 diabetes, myocardial infarction, or high cholesterol.

^dModel 3: Adjusted for variables included in Model 2, but using duration of chronic medical conditions, if present, cumulative duration of aspirin, NSAID or multivitamin use, and pack-years of smoking.

CI, confidence interval; PPI, proton pump inhibitor.

Considering the shared clinical indications for PPIs and H2RAs, we were interested in determining whether the associations observed between PPI use and cognitive function would be observed among H2RA users. To minimize the influence of regular PPI use, we restricted the analysis to individuals who had reported PPI use in no more than one follow-up cycle (n=10,778). The differences in characteristics between regular H2RA users and non-users were broadly similar to those for PPI use (Supplementary Table 1). Increasing duration of H2RA use was associated with poorer scores in all three cognitive domains, which remained statistically significant after multivariable adjustment in models for learning and working memory and overall cognition (both P_trend ≤.002; Table 3). The magnitudes of mean score differences were larger than those observed in the analysis of PPI use, particularly for learning and working memory (mean score difference between H2RA users of 9–14 years and never users was −0.20 (95% CI, −0.32 to −0.08) (Table 3).

Table 3.

Mean differences in cognitive scores according to duration of H2 receptor antagonist use among participants who did not regularly use PPIs^a

Composite score	Estimates (95% CI) for mean difference in cognitive scores according to duration of H2 receptor antagonist use
	No H2RA use	1–4 years	5–8 years	9–14 years	Ptrendb
Psychomotor speed, attention (n)	(9,261)	(1,100)	(265)	(132)
Model 1c	Reference	−0.03 (−0.08, 0.03)	−0.09 (−0.19, 0.02)	−0.11 (−0.26, 0.04)	.02
Model 2d	Reference	−0.02 (−0.08, 0.03)	−0.07 (−0.18, 0.03)	−0.09 (−0.24, 0.06)	.06
Model 3e	Reference	−0.02 (−0.07, 0.04)	−0.07 (−0.18, 0.03)	−0.10 (−0.25, 0.05)	.06
Learning, working memory (n)	(9,276)	(1,101)	(265)	(132)
Model 1c	Reference	−0.05 (−0.09, 0.00)	−0.07 (−0.16, 0.02)	−0.22 (−0.34, −0.10)	<.001
Model 2d	Reference	−0.03 (−0.08, 0.01)	−0.04 (−0.13, 0.04)	−0.19 (−0.31, −0.07)	<.001
Model 3e	Reference	−0.03 (−0.08, 0.01)	−0.04 (−0.13, 0.04)	−0.20 (−0.32, −0.08)	<.001
Overall cognition (n)	(9,257)	(1,100)	(265)	(132)
Model 1c	Reference	−0.04 (−0.08, 0.00)	−0.08 (−0.16, 0.00)	−0.17 (−0.28, −0.05)	<.001
Model 2d	Reference	−0.03 (−0.07, 0.01)	−0.06 (−0.14, 0.02)	−0.14 (−0.25, −0.03)	.002
Model 3e	Reference	−0.03 (−0.07, 0.02)	−0.06 (−0.14, 0.02)	−0.15 (−0.26, −0.04)	.002

^aParticipants who never reported regular PPI use, or reported use during one 2-year questionnaire cycle only.

^bP for trend calculated using the total number of questionnaire cycles where regular PPI use was reported, as a continuous variable in the model.

^cModel 1: Adjusted for age and educational attainment of parents and husband.

^dModel 2: Adjusted for variables included in Model 1 plus smoking status, body mass index, alcohol intake, physical activity, antidepressant use, aspirin use, NSAID use, Alternative Healthy Eating Index score, multivitamin use, and history of hypertension, stroke, type 2 diabetes, myocardial infarction, or high cholesterol.

^eModel 3: Adjusted for variables included in Model 2, but using duration of chronic medical conditions, if present, cumulative duration of aspirin, NSAID or multivitamin use, and pack-years of smoking.

CI, confidence interval; H2RA, H2 receptor antagonist; PPI, proton pump inhibitor.

Because of the rise in the prevalence of PPI use over time (33% of participants reported regular PPI use at some point during follow-up), restricting the H2RA analysis to participants who never used PPIs substantially reduced the size of the population for analysis and limited the number of participants in the longest duration of use category (n=89). Nonetheless, similar trends toward poorer cognitive scores with increasing duration of H2RA use were observed across all three cognitive domains when PPI users were excluded (Supplementary Table 2).

In light of the finding of associations between H2RA use and cognitive function, we examined the association of PPI use among participants who never reported regular H2RA use (n=10,795; Table 4). The characteristics of PPI users compared to non-users in this subgroup were similar to those for PPI users overall (Supplementary Table 3). Although mean scores for the longest duration category of PPI use were modestly lower than for never users, there were no statistically significant trends across duration categories in multivariable models (all model 3 P_trend ≥.34).

Table 4.

Mean differences in cognitive scores according to duration of PPI use among participants who never used H2 receptor antagonists.

Composite score	Estimates (95% CI) for mean difference in cognitive scores according to duration of PPI use
	No PPI use	1–4 years	5–8 years	9–14 years	Ptrenda
Psychomotor speed, attention (n)	(8,286)	(1,412)	(554)	(520)
Model 1b	Reference	−0.01 (−0.06, 0.04)	−0.02 (−0.09, 0.06)	−0.06 (−0.14, 0.01)	.07
Model 2c	Reference	0.00 (−0.05, 0.05)	0.00 (−0.08, 0.08)	−0.03 (−0.11, 0.05)	.44
Model 3d	Reference	0.00 (−0.05, 0.05)	0.00 (−0.07, 0.08)	−0.03 (−0.11, 0.05)	.47
Learning, working memory (n)	(8,298)	(1,417)	(554)	(521)
Model 1b	Reference	−0.01 (−0.05, 0.03)	0.00 (−0.06, 0.06)	−0.10 (−0.17, −0.04)	.007
Model 2c	Reference	0.00 (−0.04, 0.04)	0.03 (−0.03, 0.09)	−0.06 (−0.12, 0.01)	.37
Model 3d	Reference	0.01 (−0.03, 0.05)	0.03 (−0.03, 0.09)	−0.06 (−0.12, 0.01)	.39
Overall cognition (n)	(8,282)	(1,412)	(554)	(519)
Model 1b	Reference	−0.01 (−0.04, 0.03)	−0.01 (−0.07, 0.05)	−0.08 (−0.14, −0.03)	.007
Model 2c	Reference	0.00 (−0.03, 0.04)	0.01 (−0.04, 0.07)	−0.04 (−0.10, 0.02)	.31
Model 3d	Reference	0.01 (−0.03, 0.04)	0.01 (−0.04, 0.07)	−0.04 (−0.10, 0.02)	.34

^aP for trend calculated using the total number of questionnaire cycles where regular PPI use was reported, as a continuous variable in the model.

^bModel 1: Adjusted for age and educational attainment of parents and husband.

^cModel 2: Adjusted for variables included in Model 1 plus smoking status, body mass index, alcohol intake, physical activity, antidepressant use, aspirin use, NSAID use, Alternative Healthy Eating Index score, multivitamin use, and history of hypertension, stroke, type 2 diabetes, myocardial infarction, or high cholesterol.

^dModel 3: Adjusted for variables included in Model 2, but using duration of chronic medical conditions, if present, cumulative duration of aspirin, NSAID or multivitamin use, and pack-years of smoking.

CI, confidence interval; H2RA, H2 receptor antagonist; PPI, proton pump inhibitor.

To avoid biases that may be introduced by excluding either PPI or H2RA users, we also examined multivariable models that included both duration of PPI and H2RA use (Supplementary Tables 4 & 5). The results were generally compatible with those of our main analyses, although the estimates for mean differences in cognitive scores across categories of H2RA use were attenuated somewhat. Nonetheless, only duration of H2RA use remained statistically significantly associated with poorer scores for learning and working memory and overall cognition in both multivariable models (all P_trend ≤.03).

We did not find that regular PPI or H2RA use nearest the time of cognitive testing (i.e. use in 2013 alone) was associated with statistically significant differences in any cognitive domain (all P ≥.84 for PPI use and P ≥ .11 for H2RA use).

DISCUSSION

In an analysis nested within a population-based cohort of middle-aged and older women, we found no convincing evidence of an association between duration of PPI use and cognitive function. While a modest association (equivalent to approximately two years of age-associated decline) was observed for psychomotor speed and attention among PPI users of 9–14 years in our main analysis, the magnitude of this association was attenuated after controlling for H2RA use. Furthermore, although we did not hypothesize that an association would be observed for one particular composite score, it is noteworthy that it is the learning and working memory score that has been most strongly associated with AD-type cognitive impairment.^29,30

Our results relating PPI use and cognitive differ from the findings of a recent pharmacoepidemiologic analysis conducted using a large German health insurance database.¹³ Among 73,679 elderly individuals, regular use of PPI was associated with an increased risk of incident dementia (HR = 1.44; 95% CI, 1.36–1.52) over 7 years of follow-up. Since this analysis used claims data, it was less able to account for potential confounding by differences in education and health-related characteristics between PPI users and non-users, which we observed in our analysis and have been described in other populations.^9,31 Additionally, as an outcome, incident dementia diagnosis may be prone to misclassification and ascertainment bias. The clinical signs and symptoms of dementia are insidious, leading to a high prevalence of missed and delayed diagnoses.³² Moreover, 25–50% of PPI prescriptions in ambulatory and in-patient settings lack appropriate indications^33–35 and non-judicious PPI prescribing is especially frequent among the elderly and those with cognitive impairment.³⁶ Therefore, elderly individuals who have frequent contact with health providers are at increased risk of both PPI prescription and dementia diagnosis. This bias may not be completely mitigated by adjustment for comorbidities or polypharmacy.

Our results are also inconsistent with an analysis conducted using data from the German Study on Aging Cognition and Dementia in Primary Care Patients (AgeCoDe), which found that PPI use among 3,076 individuals aged 75 years or older was associated with increased risk of any incident dementia (HR = 1.38; 95% CI, 1.04–1.83) and AD (HR = 1.44; 95% CI, 1.01–2.06).³⁷ The cohort design employed in the AgeCoDe study allowed for repeated evaluation of cognitive function and collection of detailed information on dementia risk factors, including educational level. PPI use was treated as a time-varying exposure and dose or duration-dependent associations with dementia diagnosis were not evaluated. A proximal association between PPI use and dementia diagnosis may arise where more frequent contact with healthcare providers occurs prior to formal dementia diagnosis, increasing the attendant risk of risk of PPI prescription.

Although several mechanisms have been proposed to explain associations between PPI use and dementia, none has been robustly validated by independent analyses. For example, PPIs, which readily cross the blood brain barrier,³⁸ appear to increase amyloid-β production in vitro and promote its accumulation in the brains of transgenic AD mice.¹⁴ However, lansoprazole has actually been shown to exert a protective influence on memory dysfunction in an alternative mouse model of dementia.¹⁵ Observational data suggest that malabsorption of micronutrients, such as vitamin B12, may be a mechanism linking PPIs or H2RAs with dementia.³⁹ In contrast, analysis of prospectively-collected safety data from two, long-term, randomized controlled trials did not demonstrate a difference in B12 levels in patients assigned to PPIs compared with anti-reflux surgery.⁴⁰

Contrary to expectation, we observed poorer cognitive function associated with increasing duration of regular H2RA use, with the strongest association evident for learning and working memory. However, our findings for H2RA use should be interpreted with some caution considering that this was not our primary hypothesis. Although a number of early cross-sectional analyses reported null or protective associations between H2RA use and dementia,^41–43 our findings are consistent with the results of several smaller cohort studies. In a longitudinal analysis involving 1,558 community-dwelling African Americans, compared to non-use, continuous use of H2RAs was associated with an increased risk of incident cognitive impairment (OR = 2.49; 95% CI, 1.17–5.04), as ascertained by cognitive and neurophsychologic testing.⁴⁴ Similarly, in the Duke Established Populations for Epidemiologic Studies of the Elderly cohort, H2RA use among 2,082 elderly individuals was associated with statistically non-significant increases in the risk of cognitive impairment and cognitive decline over repeated assessments of cognitive performance.⁴⁵ Finally, in the Adult Changes and Thought prospective cohort, compared to the infrequent or no-use group, individuals with the highest cumulative H2RA exposure experienced an increased risk for AD (HR = 1.41; 95% CI, 1.00–1.97), although no dose-response relationship was observed.⁴⁶ Neither the German claims database analysis nor the AgeCoDe study examined H2RA use.^13,37

All H2RAs are known to be associated with central nervous system (CNS) side effects and are a recognized cause of delirium in the elderly.⁴⁷ H2RAs, such as ranitidine and cimetidine, have been shown to exhibit anti-cholinergic properties, which could theoretically also pose a risk for adverse cognitive effects with long-term use.⁴⁸ Compared to PPIs, H2RAs are much less potent acid suppressors and exhibit tachyphylaxis with regular use; ⁴⁹ thus, one would expect the risk of malabsorption and micronutrient deficiency to be lower with H2RAs than PPIs.

Our study possesses several strengths. First, our outcomes were based on a validated tool for cognitive function assessment, administered to a large population-based sample of middle-aged and older women. Second, we prospectively collected data on PPI and H2RA use prior to cognitive testing, minimizing the potential for recall bias. Third, our participants were all health professionals; thus, the accuracy of self-reported PPI and H2RA use is likely to be high and is more likely to capture actual use from all sources (including over-the-counter) over an extended time period than prescription databases. Fourth, we were able to comprehensively adjust for potential confounding using prospectively collected data on health and lifestyle factors that may be associated with cognitive function, including educational attainment, antidepressant use and chronic illnesses, such as diabetes.

We recognize that our study has some limitations. We cannot exclude the possibility that PPI use was associated with a more modest reduction in cognitive function, and we lacked power to detect it. However, even if PPI use were associated with a reduction in overall cognitive function score of 0.01 per year, which we had >99% power to detect, we estimate that this would only explain around 0.1% of the variation in cognitive function score. This is difficult to reconcile with the risk estimates for dementia seen over relatively short follow-up in prior analyses.^13,37 Although our analysis was observational, and uncontrolled or residual confounding could have influenced our findings, definitive data from a randomized trial of cognitive outcomes in relation to PPI or H2RA use are unlikely to be forthcoming given the large number of participants and long follow-up that would be necessary. Our database lacked information on the specific drug type, and frequency and dose taken by regular users. Additionally, we were not able to adjust for the indication for acid-suppression medication use in our analysis. Nonetheless, we are not aware of acid related disorders being associated with cognitive function and one might expect confounding by indication to be greater for PPIs, the more efficacious of the two anti-secretory medications, for which we essentially observed null findings. Antiplatelet drug use could have been a confounder on account of associations with anti-secretory medication use and cerebrovascular disease. However, our multivariable models carefully adjusted for aspirin use, stroke, and cardiovascular risk factors. Furthermore, our estimates were not altered by additionally adjusted for ever use of thienopyridines (e.g. clopidogrel, prasugrel), data for which were available from 2009, when use of these agents became more prevalent. Since our data on cognitive function were derived from a single round of testing, we were unable to examine trajectories of cognitive decline. Repeated cognitive testing in this cohort in the future may allow for such analyses. Although the previously reported association between PPI use and dementia was evident for both sexes,¹³ the fact that all participants in our study were female nurses may influence the generalizability of our findings.

In conclusion, after adjusting for multiple potential confounding factors, including H2RA use, we did not observe a convincing association between use of PPIs and cognitive function in middle-aged and older women. Our data should provide some reassurance to individuals who require these highly effective medications for long-term treatment. In the absence of data on cognitive outcomes from studies where PPI use has been randomly assigned, replication of our findings in other observational studies is desirable.

Abbreviations

AD

Alzheimer’s disease

AHEI

alternative healthy eating index

BMI

body mass index

CI

confidence interval

H2RA

H2 receptor antagonist

MET

metabolic equivalent of task

NSAID

non-steroidal anti-inflammatory drug

PPI

proton pump inhibitor

SD

standard deviation