Introduction
Alzheimer’s disease (AD) is the leading cause of dementia and the sixth leading cause of death in the USA. Retrospective cohort studies of the associations between use of selected medications and occurrence of probable AD have been conducted using large claims databases to examine associations between medications that may impact positively with lower risk of dementia and possible AD. These classes of medications include angiotensin-converting enzyme inhibitor (ACEI) [
1,
2], simvastatin [
3,
4], beta blockers [
5], and metformin [
6,
7]. Currently, there is no effective treatment to prevent or slow the progression of AD.
Possible association of traumatic brain injury (TBI) with dementia has been extensively investigated. TBI is a broad term defining neurological injury resulting from an external impact involving the head. TBI is a leading cause of long-term disability in the USA and is common in military personnel and veterans, especially in service members with blast exposure from improvised explosive devices. Since 2000, over 350,000 US service members have been clinically diagnosed with TBI; a total of 23% of 4000 soldiers from a brigade combat team deployed in Iraq were reported to have blast-related TBIs [
8]. TBI history is associated with significant comorbidities, including posttraumatic stress disorder [
9,
10], mood disorders [
11], sleep disturbances [
12], and increased risk for neurodegenerative diseases such as AD [
13] and Parkinson’s disease (PD) [
14].
History of TBI has been associated with an increased risk of dementia [
15]. A retrospective study of 188,000 veterans found that history of TBI was associated with a 60% increase in the risk of developing dementia [
15]. Another retrospective study found a 1.68 times greater risk of dementia in those with TBI after adjusting for sociodemographic characteristics and comorbidities [
16]. Several early reports described associations between TBI and AD [
17‐
19]. Lee et al. found a significant association between mild traumatic brain injury (mTBI) and AD based on International Statistical Classification of Diseases (ICD) 9 code in medical records [
20]. Based on the data from the National Alzheimer’s Coordinating Center (NACC) that collects AD patients’ data from 39 past and present AD clinical centers (ADC) since 1999 [
21,
22], a retrospective study supports a significant association of the early-onset AD with self-reported head injury [
23]. A systematic review from 18 studies on the association of TBI severity and AD between 1998 and 2014 revealed that 55% of patients with TBI develop cognitive deficits that meet the clinical diagnostic criteria of AD [
24]. At the molecular level, elevated tau-containing neurofibrillary tangle formation and cerebral atrophy are apparent in AD mouse models after TBI exposure compared to mice without TBI [
25]. Moreover, progressive spreading of tau proteinopathy has been triggered in wild-type mice by impact or blast TBI [
26‐
29]. Quantitation of tau and phosphorylated tau (p-tau) in mice show that tau and p-tau levels are significantly changed within 1 day after blast, then return to the pre-blast stage when mouse brains were examined at 15 weeks post-blast [
30], consistent with previous findings in mini-pigs [
31] and mice [
32]. Other studies conducted in mice have shown long-term persistence and even progression of blast- and impact-induced p-tauopathy [
26‐
28,
33,
34]. In humans, effort has been made to search for an association (or a lack of) between cognitive decline and overlapping neurodegenerative pathologies [
35,
36], and possible biomarkers that reflect those changes [
37]. Repetitive mild TBI was found to be associated with tau pathology and allied neurodegenerative changes in the brain [
38]. AD-like pathology, amyloid plaques, and neurofibrillary tangles have also been shown to be more prevalent in patients following a single TBI relative to uninjured, age-matched controls [
39]. Currently, it is not clear whether TBI-induced molecular changes (such as tau protein phosphorylation) cause irreversible damages leading to AD onset, and equally important, whether available medications can prevent or suppress AD initiation and progression.
On the other hand, analysis of the data from the Religious Orders Study (ROS), Memory and Aging Project (MAP), and Adult Changes in Thought study (ACT) reveals that TBI with loss of consciousness (LOC) is associated with risk for Lewy body accumulation, progression of parkinsonism, and PD, but not dementia, AD, pathologic neuritic plaques, or neurofibrillary tangles [
40]. The Adult Changes in Thought data (ACT) is a community-based database to reveal the correlation between clinical characteristics and biochemical/structural features of dementia [
41]. Both the Nun study and ACT data were explored to seek correlations between the pathological factors and cognitive function [
42]. Similarly, in ROS and the Rush MAP studies, pathology alteration and clinical observations were explored and correlated [
43,
44]. A recent study on the dataset from NACC reveals failure of self-reported TBI to predict neuropathologic changes in postmortem brain tissue, indicating that self-reported TBI may not be an independent risk factor for clinical or pathological AD [
45]. The self-reported TBI did not predict AD neuropathological changes and was not associated with dementia severity or cognitive function in subjects [
45].
In this study, we use an existing VA national database which contains rich longitudinal patient medical information to address two questions, namely, whether TBI contributes to the development of dementia with possible AD, and for those with a history of TBI, whether treatment with ACEI, beta blockers, metformin, statins, or a combination of these medications prolongs the interval between the occurrence of TBI and the onset of possible AD. Previous retrospective cohort studies using claims databases support a protective role for statin treatment with respect to possible AD onset [
3]. Therefore, we used statin treatment as the control condition in comparing the association of single and combination of medications on onset of possible AD after TBI. The choices of our medications are based on previously published studies on these medication classes related to AD therapeutics and with the consideration of the sample sizes of our databases. For example, possible benefit of ACEI [
2], beta blockers [
5], and metformin [
7] for AD patients has been explored. Statins with higher blood–brain barrier (BBB) penetration capacity (e.g., lipophilic statins) may have more influence on AD progression [
4]. The use of ACEI, beta blockers, metformin, and statins was assessed using Cox proportional hazard models [
46] with survival time from initial TBI to occurrence of dementia with possible AD, adjusted for demographics and comorbidities. In previous studies, we found reduced hazard rates or relative risks for dementia occurrence in veterans using ACEI [
1] or simvastatin [
3]. In this study, we explored the associations of selected concomitant medication use with the occurrence of dementia with possible AD after TBI. We found that combination treatments with ACEI and statins reduce the risk of AD and prolongs the time between the occurrence of TBI and onset of dementia with possible AD.
Discussion
Our study results showed that a history of TBI was associated with increased hazard risk of developing dementia with possible claims-based AD (defined by clinical diagnosis) compared to patients without TBI history. This finding supports emerging data implicating TBI as one of the risk factors for AD [
53,
54].
We performed statistical analyses of medical records from a large national VA database of over 1.6 million veterans to investigate the relationship of prior TBI on onset of dementia with possible AD and the therapeutic potential of monotherapy and combination therapy with ACEI, beta blockers, metformin, and statins to postpone onset of dementia with possible AD.
We selected four commonly prescribed medication classes—ACEI, statins, metformin, and beta blockers—based on prior reports suggesting potential prophylactic effectiveness for AD [
1,
3,
5,
6]. To estimate the association of two-medication combination therapy on first claims-based diagnosis of dementia with possible AD after initial occurrence of TBI, we used multivariable Cox regression models with propensity score weighting for demographic factors and other covariates. Subjects treated with two-medication combination therapy were compared to reference groups without medication treatment. We performed similar analyses for non-TBI patients. Beta blocker users had significantly lower possible AD risk compared to statin users and to the no medication group in our Cox regression model. Interestingly, the impact of ACEI + statin combination was less clear, and the ACEI + statin combination does not show a significant association with AD risk in the multivariate Cox regression model after controlling for all other variables. Our results suggest that this combined medication regimen is specific for the TBI + dementia/AD population compared to the general population of dementia with possible AD without a history of TBI.
Our analyses indicate that patients prescribed a combination of two of these medication classes were at a lower risk of the initial occurrence of possible claims-based AD after the first claims-based diagnosis of TBI. However, we note one important exception, TBI patients treated with metformin and statins had a claims-based diagnosis of dementia with possible AD sooner than TBI patients in the other medication treatment groups (Fig.
5) and also sooner than TBI patients without medication treatment. This interesting finding may point to an unidentified synergistic pathway that potentiates dementia with possible AD and, as such, may provide clues to mechanisms that link TBI and AD. For individual medication, there was no significant association among metformin users (
p = 0.12, Table
3B). This result is consistent with a recent study based on pooled analyses from multiple studies [
55]. The result for statin was also not significant compared with the reference group (
p = 0.07, Table
3B). Therefore, there is not sufficient evidence derived from our analysis to support their respective impact on possible claims-based AD after the first claims-based diagnosis of TBI.
Patients taking a combination of ACEI and statins showed prolonged time to development of claims-based diagnosis of dementia with possible AD following TBI. The validity of this finding is supported by other modeling approaches that yielded similar conclusions. Other combinations also had a significant association, namely, the combinations of beta blockers + metformin and ACEI + metformin, albeit exhibited wider variation than ACEI + statins.
There were several important limitations to our study. First, the study identified ACEI + metformin as a combination of possible interest, having the protective, lowest risk ratio in the Cox regression model (HR [95% CI] 0.18 [0.04–0.75];
p = 0.02, see Table
5). However, owing to the small sample size of this medication group, this low HR was not estimated with significance as we were limited by the small sample sizes from selected subgroups. Furthermore, we do not have sufficient statistical power for subgroups to adjust for other important covariates including hypertension and diabetes as well as other drugs used for those conditions, since medications often are administered for off-label use and are prescribed for more than one indication in routine practice. A second limitation is that clinical severity measures for TBI are not available in this claims database. In our study, we used two or more clinical visits with diagnosis of TBI as criteria to screen TBI patients for inclusion. This criterion was used as a proxy measure of TBI severity. Currently, it is not possible to create subpopulations among TBI subjects corresponding to hemorrhage, concussion, or other indicators of TBI severity, which is not available in this dataset. Larger datasets with this rich clinical information would be needed in the future. A third limitation is the presence of unmeasured variables in our analysis, including indicators of direct medication adherence and measures related to selection bias due to provider and patient preferences for prescribing specific medications. The significant effect of medication switching (including adding or dropping of one medication) needs to be explored in future investigations. The fourth limitation is that our study was based on a Department of Veterans Affairs database with a population that is predominantly male with a smaller fraction of females, with higher proportion of those veterans with physical and mental health conditions than in a general community-wide population. This would limit the generalizability of the findings. The fifth limitation is the sparse information on biomarkers related to amyloid, tau, and neurodegeneration from the existing database we used. NIA-AA criteria on clinical diagnosis of AD published in 2011 [
56,
57] and the guidelines under the research framework published in 2018 refer AD to an aggregate of neuropathologic changes and define it by biomarkers and by postmortem examination in vivo [
58]. In this study, we obtained clinical records between 1998 and 2018 with minimum data on amyloid, tau, and neurodegeneration (ATN biomarkers). Information of preclinical AD and genetic information (such as
apoE) is not available for our analysis, and we could not create a simulation study to predict its influence. Finally, the occurrence of TBI attributable to medications promoting falls was not considered; our analysis only focused on risk of dementia with possible AD after the TBI occurrence.
At the molecular level, our findings of candidate medications (ACEI and statins) to be used in delaying onset of dementia with possible AD are supported by mechanistic studies. Previous studies have shown that ACE activity is elevated in AD and correlates with the Braak stage, and ACE density is 70% higher in the temporal cortex of AD patients relative to control subjects [
59]. ACE activity is also correlated with Aβ load in AD patients [
60], and cultured neurons exposed to Aβ showed increases in levels and activity of ACE [
61]. Population studies have shown that antihypertensive medication usage is associated with reduced occurrence of AD [
62]. Statins have been tested as potential therapeutics, as high cholesterol levels are implicated as a risk factor for AD [
63]. Cholesterol depletion in neuronal cells resulted in complete inhibition of Aβ formation, and this inhibition was fully reversible upon re-addition of cholesterol to the neurons [
64]. Transgenic mice subject to cholesterol-lowering drugs showed reduced brain Aβ peptides and amyloid load by greater than twofold relative to non-treated mice [
65]. Additionally, treatment with simvastatin has been shown to reduce Aβ peptide levels in both cellular models and Guinea pigs [
66].
In addition to brain pathology manifested in a number of neurocognitive disorders, which are diagnosed using ICD-9 and 10 codes, it is well known that cognitive and brain reserve plays an important role in the maintenance of cognitive function and prevention of neurodegeneration [
67]. Difference in cognitive reserve may enable individuals to be more resilient to neural dysfunction, while difference in brain structure may differentiate individuals who may have higher tolerance to brain pathological alteration [
67].
Our results have implications in examining FDA-approved drugs for the purposes of testing novel hypotheses and introducing new therapeutics for possible future indications for AD. The challenge of addressing treatment of AD could be targeted in part by the wide-scale use of “big data” sources that have integrated electronic health record information with rich clinical data and provide a valuable data source to link medication use with clinical manifestations. This ultimately gives a viable alternative for future investigations of AD treatment [
68]. Similar approaches are already being used in improving detection, treatment, and prevention of complex diseases such as cancer [
69] and type 2 diabetes [
70]. Further in vivo studies are needed to determine the effectiveness of our drug combinations in Alzheimer’s treatment.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.