The use of proton pump inhibitors (PPIs) has increased strongly, especially in the elderly. Recently, there is evidence that PPI consumption might affect cognition. An association of PPI use with vitamin B12 deficiency was observed in a population-based sample. Poor vitamin B12 status has been described to negatively affect cognition and promote neurological damage. PPIs were also detected to enhance amyloid beta peptide (Aβ) levels in mouse brain by affecting the enzymes β- and γ-secretase which leads to increased Aβ levels in mice.

Using longitudinal data from a large statutory health insurance we observed an association between long-term PPI prescription and occurrence of dementia in the elderly aged 75 years or older. While the analysis includes a large unselected patient population, there are also limitations with the use of claims data. As we mention in the discussion, residual confounding in claims data is a concern that cannot be ruled out completely. We adjusted our analysis by inclusion of several potential confounders such as age, sex, comorbidities and polypharmacy. Other risk factors for dementia, e.g. ApoE4 allele status or education, could not be integrated into the analysis as the claims data lack genetic information or detailed sociodemographic parameters. As we mention in the conclusion, our study can only provide a statistical association. The study does not prove that PPIs cause dementia. To evaluate cause and effect relationships between long-term PPI use and possible effects on cognition in the elderly randomized, prospective clinical trials are needed.

References

Badiola N, Alcalde V, Pujol A, et al. The proton-pump inhibitor lansoprazole enhances amyloid beta production. PLoS One. 2013;8(3):e58837.

Lam JR, Schneider JL, Zhao W, Corley DA. Proton pump inhibitor and histamine 2 receptor antagonist use and vitamin B12 deficiency. JAMA. 2013;310(22):2435-2442.

Gomm W, von Holt K, Thomé F, et al. (2016) Association of Proton Pump Inhibitors With Risk of Dementia: A Pharmacoepidemiological Claims Data Analysis. JAMA Neurol 73: 410-416.

Big data analysis holds great promise for the rapid discovery of small effects, associations and phenomena that could take years of painstaking work using more traditional epidemiological methods.  When the subject under investigation is one of the most common and devastating but least well understood conditions of old age, the potential benefits can be emphatically expressed in health economic terms.  In the present case, Gomm et al. attempt to use a nationwide database of diagnostic and prescription records to look for evidence that regular use of the widely prescribed proton pump inhibitors may be a (modifiable) risk factor for the development of dementia in later life.  There is much to commend in the authors’ adoption of this approach to studying this drug class at this time. There are also, however, reasons to be cautious about the details of methods and results, as that is where the devil is fond of hiding.

To begin with the positives: the link in question did not pop up from an undirected data mining exercise but emerged instead from an earlier study (the German Study on Aging, Cognition and Dementia in Primary Care Patients (AgeCoDe). It is reassuring to note that the statistical effects associated with PPI use in each case not only pointed in the same direction, but were of similar magnitude (hazard ratios of 1.38 [95% CI, 1.04-1.82] in AgeCoDe, and 1.44 [95% CI, 1.36-1.52] in this study). The effect was independent of other differences (age, sex, cerebrovascular disease, depression and polypharmacy) in both studies. Biological plausibility is provided by a body of in vitro work implicating PPI as an enhancer of a key enzyme – Beta-site Amyloid precursor protein Cleaving Enzyme 1 (BACE1) – involved in amyloid production, which is a critical initiating step in the development of the neuropathology of Alzheimer’s disease (AD). Inhibition of BACE1 has become one of the most ‘druggable’ targets in the search for the first genuine disease modifying treatment for AD, and several large trials of such agents are currently in progress. When the data come to be analysed, investigators will now be primed to look for an interaction between treatment effect and concomitant use of PPI agents.

Yet examine the case selection criteria and definitions more closely and the study takes on an altogether more procrustean character.  To qualify for inclusion in the dementia group there had to have been at least 2 references to a dementia diagnosis in an individual’s record. At some level this is understandable, as the early stages of dementia are notoriously difficult to detect and may be confused with transient cognitive complaints or states of reduced performance (such as depression). But not reporting a previously made diagnosis is not the same as rescinding it, and does reporting the same condition twice add that much more certainty about its presence? Figure 1 (a flowchart of patients included for analysis) is also worth a closer look: an enviable total of nearly 220,000 case records was available to the study, but two thirds were excluded from the analysis. Although some exclusions are unassailable (e.g. due to incomplete data), the decision to consider those with ‘no regular use of PPI’ as neither users nor non-users of the drug is puzzling, as the proposed amyloidogenic mechanism need not depend on regular exposure to the drug, and a dose-response effect is one of the most convincing patterns to demonstrate when cause and effect are being considered.

A more open and comprehensive exploration of the data, incorporating different criteria for inclusion as a case of incident dementia and different levels of exposure to the putative risk factor would have made the study more convincing as a replication of a previous association.