Background
Proton pump inhibitors (PPIs) are widely used in the treatment of gastric acid-related disorders such as peptic ulcer disease, gastroesophageal reflux disease (GERD), non-steroidal anti-inflammatory drug (NSAID)-associated ulcers, and eradication of
Helicobacter pylori [
1]. In many countries, including the UK, some PPIs are available for over-the-counter purchase, which increases public accessibility. Furthermore, PPIs are often prescribed in and out of the hospital for incorrect indications or long-term use that does not meet guidelines [
2,
3]. With the increasing use of PPIs [
4], more attention has been paid to the research on its side effects [
5]. A series of studies have reported associations between PPI use and cardiovascular disease [
6,
7], fracture [
8], kidney disease [
9], infections [
10], and diabetes [
11], but the association with dementia is controversial [
12].
Dementia is characterized by inexorably progressive impairment in cognitive and independent living functions. Alzheimer’s disease (AD), vascular dementia (VaD), Lewy body, and frontotemporal dementia are the most common pathologies. It is estimated that there were 35.6 million dementia patients worldwide in 2010, and the number may reach 65.7 million in 2030 [
13]. Meanwhile, the worldwide costs of dementia were estimated at $818 billion in 2015 [
14]. To prevent dementia, reducing risk factor exposure is vital in the circumstance of limited treatment. Several cohort studies reported the association between PPI use and all-cause dementia or AD among the elderly, and the hazard ratios (HRs) of PPI users were 1.38 to 1.44 [
15,
16]. However, other studies showed conflicting conclusions, and most of these studies did not observe any associations [
17‐
22]. Therefore, the association between PPI use and dementia is still uncertain.
PPIs are aimed to reduce the gastric acid secretion of the parietal cell by inhibiting (H(+), K(+))-ATPase [
23]. Similar enzymes are also found in microglia lysosomes [
24], and the lysosomal acidic environment is essential for amyloid-β (Aβ) clearance, the disorder of which may lead to neurodegeneration and dementia [
25]. A study reveals that PPIs may increase Aβ deposition in the mouse brain by affecting the β- and γ-secretases [
26]. However, precision measurement of Aβ metabolism in a large population would be difficult. A measurable proxy for Aβ is required to infer whether the PPIs promote dementia via affecting Aβ metabolism.
Apolipoprotein E (Apo-E) is a primary cholesterol carrier involved in lipid transport, and
APOE ε4 alleles are the main genetic risk factors for AD and dementia due to their reduced capacity for Aβ transport [
27].
APOE ε4 may also promote AD by reducing the ability of astrocytes to remove toxic fatty acids from the extracellular milieu [
28]. Possible mechanisms for the potential association between PPIs and dementia and whether PPIs can interact with
APOE require evidence from population-based studies.
To further explicitly whether regular PPI use is associated with incident all-cause dementia and pathological specific dementia (AD and VaD), we conducted a large prospective cohort study in the UK Biobank. Furthermore, we also tried to explore the differences in the associations among different APOE ε4 genotypes, a potential regulatory gene of Aβ metabolism, to suggest biological mechanisms.
Discussion
In this population-based prospective cohort study of half a million participants, we found that regular PPI use was associated with an increased risk of incident all-cause dementia, AD, and VaD. Meanwhile, we found an interaction between PPI use and APOE ε4 genotype for all-cause dementia, and the association was more significant among APOE ε4 heterozygotes.
Our results were consistent with the previous studies that reported the association between PPI use and increased risk of dementia [
15,
16,
29]. The study, which followed 70,000 participants over 75 years of age for 7 years, showed that PPI users had 1.44 times the risk of incident dementia as non-users [
16]. Another study of more than 15,000 participants over 40 years of age without prevalent dementia followed for 8.44 years found that the adjusted HR for PPI users was 1.22 [
29]. However, a separate series of observational studies reported that the associations were absent [
19‐
22]. For example, a prospective study including over 70,000 participants showed that PPI use was not associated with dementia [
21]. A systematic review and meta-analysis pooling 11 observational studies did not observe the association between short-term PPI use and dementia [
30]. After that, the current study of more than 500,000 participants suggests that significant associations with incident dementia were still emerging with regular PPI use after adjusting for a wide range of lifestyle, comorbidity, and clinical indications. To our knowledge, this is the most extensive prospective study of PPI-dementia associations in the general population while providing some validation of the possible biological mechanisms of the association. Therefore, this study offers high-quality population-based evidence to assess the side effects accompanying regular PPI use.
Aβ aggregation to form plaques triggers neuronal dysfunction and death in the brain, which is the critical pathological feature of AD [
31]. Studies of mouse models showed that PPIs might cross the blood-brain barrier [
32,
33] and exacerbate Aβ production [
26] to promote the development of dementia. Another mechanism was that the PPIs increase the accumulation of fibrillar Aβ by inhibiting the acidification of the degradation process in microglia [
34,
35]. Aβ clearance from the brain requires the involvement of membrane cholesterol, and glial-derived
APOE is a critical cholesterol transporter in the brain [
31].
APOE ε4 is a determining risk factor of AD by promoting Aβ aggregation, associated with a 4-fold increased risk for a single allele [
36]. We unprecedentedly reported the interaction between PPI use and
APOE ε4 genotype in dementia risk. Compared to the
APOE ε4 noncarrier, the risk of dementia among ε4 heterozygotes may be further amplified with regular PPI use. We speculated that PPIs might affect Aβ metabolism and synergize with the
APOE ε4 to promote Aβ accumulation and increase dementia risk. PPIs may reduce lysosomal acidification by inhibiting V-ATPase activity, which is critical for Aβ clearance [
25,
37].
Notably, the association between PPI use and dementia was not presented among
APOE ε4 homozygotes. The
APOE ε4 homozygote is a validated risk factor with significant effects, and its HRs of all-cause dementia and AD were 6.93 (95% CI, 6.05–7.92;
P < 0.001) and 12.91 (95% CI, 10.59–15.75;
P < 0.001) in this study. We hypothesized that the ε4 homozygotes are more likely with high loading of Aβ level, which may mask the relatively modest effects of PPI use by the mechanism like the epistasis effect [
38]. When we investigated the combined effect of PPI use and
APOE ε4, the results showed a significantly increased risk of dementia in ε4 homozygotes, regardless of whether they used PPI or not (Additional file
1: Fig. S2). In addition, PPIs may promote dementia by inducing vitamin B
12 deficiency [
39] or inhibiting choline acetyltransferase [
40], but this has not been verified in this study.
Sex factors play an unavoidable role in the development of dementia. This study showed that the PPI-dementia association was more pronounced in females. Previous studies reported that females are more likely to develop dementia due to carrying
APOE ε4 [
41], which may be explained by the increased sensitivity of females to Aβ [
42]. Thus, based on the hypothesis that PPIs promote dementia by increasing Aβ accumulation, we speculated that PPIs would synergize with the high Aβ sensitivity to increase the risk of dementia among females. In addition, the results of the subgroup analysis also suggested that the association between PPI use and dementia was more substantial in the non-smokers and participants without stroke. Smoking and stroke are often concomitant with cerebral oxidative stress and vascular inflammation, which are potential mechanisms for increased risk of AD [
43]. Meanwhile, functional studies on primary human tissues and animal models showed PPIs had antioxidant and anti-inflammatory properties [
44]. Therefore, we speculate that PPI use may neutralize the risk effect of smoking and stroke.
Our results showed that the association between different types of PPIs and dementia might differ, with lansoprazole being associated with dementia with greater strength than omeprazole at a relatively close statistical power. Consistent with earlier studies, results based on the AD cell model showed that the increase in Aβ levels after lansoprazole stimulation was more pronounced than omeprazole [
29]. Lansoprazole also profoundly limits the retention of spatial information and the capacity to manipulate remembered memory to develop a strategy and execute a complex task [
45]. In addition, there were more adverse effects of headaches after lansoprazole use [
46]. Therefore, we believe that attention should be paid to the potential differences in PPIs in the nervous system.
Our study has several significant strengths, including the prospective population-based study design, the large sample size, and detailed information on related covariates, which provided adequate confounding adjustment and robust statistical power. In addition, individual genotype data set the stage for investigating drug-gene interactions. Thus, we demonstrated that PPI use and dementia associations might vary across APOE ε4 genotypes for the first time.
Some limitations should also be considered. First, PPI use was self-reported at baseline, and accurate dosage, duration, and validation by other sources were lacking. These may lead to recall bias and obscure within-group heterogeneity. This issue obstructed us from performing further analyses on these important factors. The primary exposure was based on data from a single baseline assessment only, and it cannot be excluded that a few participants only used the PPIs for a short period around the survey. Second, PPI use was not randomly assigned. Although we corrected for as many confounding factors and clinical indications as available, there may still be unmeasured confounding. Third, dementia consists of a complex set of symptomatic, and there may be diagnostic inaccuracies through ICD coding in electronic health records, while information on severity may be lost [
47]. Due to the high under-diagnosis in the natural population, defining dementia based on hospital admissions and death registers may lead to missed diagnoses, and recorded dementia in these systems is often in an advanced stage. Besides, participants with comorbidities and prescription of PPI may have more contact with the health system and thus have a greater chance of being diagnosed with dementia. Fourth, considering the interpretability of the biological mechanisms, only one genetic risk factor,
APOE ε4, was included in this study. In contrast, dementia and AD have complex genetic susceptibility factors, and the Aβ metabolism has complex regulatory mechanisms, and these may be the effect modifiers on the role of PPIs. Fifth, the UK Biobank study population may have intrinsic characteristics and limit the generalization of the results to other populations or nations.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.