Antihypertensive Treatment is associated with MRI-Derived Markers of Neurodegeneration and Impaired Cognition: A Propensity-Weighted Cohort Study

Data source

This observational analytic cohort study used prospectively-collected data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu, downloaded April 6, 2015) [26]. The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial MRI, positron emission tomography, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early AD. For up-to-date information, see http://www.adni-info.org.

Study participants gave written informed consent at the time of enrollment and completed questionnaires approved by each participating site’s institutional review board.

Study cohort and exposure

For this study, we identified a cohort of elderly hypertensive individuals (N = 886) enrolled in ADNI (all phases), who had a current or prior history of treatment for hypertension, with current or prior exposure to the following anti-hypertensive treatments groups of interest at the time of enrollment: 1) ACEIs only, 2) ARBs only, 3) ACEIs and ARBs, or 4) other hypertensive medications, including beta blockers, calcium channel blockers, or diuretics. Exposure status was retrospectively ascertained via manual analysis of prescription history and associated usage dates provided in the recent medication table in the ADNI database. This cohort was then restricted to those with data for study outcomes (N = 497) for the analysis of primary and secondary outcomes as described below.

Outcomes

The primary outcomes were cross-sectional MRI-derived volumetric biomarkers of neurodegeneration and SVD from the ADNI baseline visit. Neurodegenerative markers were indexed by measures of global atrophy, including brain parenchymal fraction (BPF), defined as the ratio of total brain volume to the total intracranial volume, and total hippocampal volume. Hippocampal volumes were generated in-house, using T1 MRI images obtained from ADNI and applying a previously described and validated methodology for automated hippocampal parcellation [27]. Markers of SVD were indexed using WMH volume, obtained from the ADNI database as previously derived volumes analyzed according to published protocols of the UCSD ADNI study site [28].

Secondary outcomes included cross-sectional standardized scores on a battery of neuropsychological assessments from the ADNI-1 baseline visit. Measures included tests of attention (WMS-R Digit Span total score), episodic memory (Delayed recall Logical Memory Story A [LM-II], Delayed recall and Recognition trials of the Rey Auditory Verbal Learning Test [RAVLT]), language (Category Fluency, Boston Naming Test [BNT]), executive functioning (Trails A & B total time), and psychomotor processing speed (WAIS-R Digit-symbol substitution total score). Raw test scores for all measures were obtained from the ADNI-1 neuropsychological battery and standardized to Z-scores or scaled scores based on published age-adjusted norms [29]. For all measures, higher raw scores represent better performance, with the exception of Trails A and B, where higher scores represent worse performance. Higher standardized scores represent better performance.

Demographic and clinical variables

Demographic and clinical variables included: age; gender; years of education; diagnostic classification, including normal control (NC), MCI, or AD; clinical history of hyperlipidemia, diabetes, atrial fibrillation, stroke, or transient ischemic attack (TIA); smoking status; baseline scores on the MMSE and the Global Depression Scale; and apolipoprotein E (ApoE) ɛ4 genotype.

Statistical analyses

Descriptive statistics were generated to characterize the study cohort with respect to all study variables. We used one-way analysis of variance and Chi-square or Fisher exact testing to compare treatment groups on continuous and categorical variables.

To address the potential for selection biases due to non-randomized treatment allocation and confounding by indication within this observational cohort, we employed a propensity score methodology to adjust for potentially confounding selection factors between our treatment groups. We identified the following factors that impact anti-hypertensive treatment indications from the 2014 update of clinical guidelines from the Canadian Hypertension Education Program: race (African American versus other) and history of coronary artery disease, myocardial infarction, or chronic kidney disease [30]. Geographic location, indexed using ADNI site identifiers, was also included to account for potential regional differences in prescribing practices associated with different classes of anti-hypertensives. To estimate the effect of these factors on treatment allocation, we used a multinomial regression model for multiple treatment classes and applied the parameter estimates for each treatment group as propensity weights in the outcome analysis [31].

To estimate the associations between treatment class and MRI volumetric outcomes and performance on neuropsychological assessments, we conducted a series of propensity-weighted multivariate linear regression models. Natural logarithm transformations were applied to non-normally distributed variables prior to analysis and quantile-quantile (q-q) plots were used to assess the validity of all transformed distributions. All multivariate models were adjusted for treatment propensity weights and all demographic and clinical variables shown to be significant in the bivariate analyses.

Associations with MRI volumetrics

The present study provides novel evidence for a neuroprotective effect of ARB anti-hypertensive treatment on multiple MRI-derived volumetric markers of neurodegeneration, including global atrophy and hippocampal volume in elderly hypertensive adults. Specifically, those treated with ARBs demonstrated less global atrophy and increased hippocampal volumes, two MRI markers of AD pathology, [12] compared to those treated with ACEIs or other anti-hypertensives. These findings are consistent with recent ADNI analyses evaluating cerebrospinal (CSF) markers of neurodegeneration in dementia-free adults, which showed that ARB use was associated with increased CSF Aβ and decreased CSF phosphorylated tau compared to other or no antihypertensive medications [32]. However, a noted limitation of this previous work was the potential for confounding for indication in assessing treatment differences. The findings of the present study specifically address this source of bias and offer new data to support the protective benefits of ARBs for brain biomarkers in addition to CSF biomarkers of neurodegeneration.

The present study showed no associations between treatment class and SVD pathology in the overall cohort. Two clinical trial sub-studies have previously evaluated the effects of anti-hypertensive treatment on WMH burden. The MRI sub-study of PROGRESS demonstrated that treatment with the ACEI perindopril stopped or delayed progression of WMH in patients with cerebrovascular disease [33], while the imaging sub-study of the PRoFESS trial reported no preventive effect for WMH progression association with the ARB telmisartan [24]. Unlike PROGRESS, median WMH burden was reduced for those treated with ARBs in the overall cohort of the present study (although not significantly so). Potential differences between our findings and previous work may, in part, relate to the use of semi-quantitative rating scales for WMH assessment in these trials or the use of a cohort with mixed disease progression in the present study.

To assess the potential influence of disease progression, we stratified these analyses by diagnostic classification into early (NC and MCI) versus late (AD) disease stage and demonstrated that, for AD patients only, reductions in median WMH burden were significantly associated with ARBs use. While this relationship emerged for AD patients only, associations with neurodegenerative markers remained significant only among normal elderly adults or patients with MCI. The finding that treatment with ARBs impacted global and hippocampal atrophy only prior to progression to dementia is consistent with evidence indicating that the trajectories of AD biomarkers, including hippocampal atrophy, plateau over time even though clinical progression may continue [34]. The fact that significant differences in WMH volume were observed only for AD patients may suggest that anti-hypertensive treatment has differential effects on vascular versus neurodegenerative processes, reflecting the potential contributions of multiple pathogenic mechanisms to disease progression in these patients.

Associations with cognition

Emerging evidence from observational studies and clinical trials suggests a potential benefit of treatment with ARBs for cognition. In a recent large prospective hospital-based cohort study (N > 800,000), ARBs were associated with a significant reduction in the incidence and clinical progression of AD compared to ACEIs and other anti-hypertensives [18]. Similarly, a retrospective postmortem analysis of 710 hypertensive patients showed a protective association between treatment with ARBs and pathological hallmarks of AD relative to those treated with ACEIs and other anti-hypertensives [35]. Recently, two small clinical trials showed improvement on a cognitive battery for treatment with the ARB losartan, especially in elderly hypertensive patients [19, 20] and some larger trials have also shown modest support for the benefit of ARBs. ONTARGET compared telmisartan with ramipril and showed a significant 9% reduction in a secondary composite neurological outcome of stroke and dementia associated with the ARB telmisartan; [21] however, it is unclear to what extent this absolute reduction was attributable to a reduction in incident dementia. In a subsequent pooled analysis of data from ONTARGET and TRANSCEND, the benefit of telmisartan for global cognition (measured using the MMSE) was only observed when a stringent 18-point threshold for impairment was used [25]. The MOSES trial also evaluated global cognition (MMSE) as a secondary endpoint and demonstrated that eprosartan was equivalent to nitrendipine, which reduced dementia risk by 50% in the Syst-Eur trial, [5, 22] suggesting a potential risk-modifying effect for ARBs.

Consistent with those findings, hypertensive adults treated with an ARB in the present study had better performance on several measures of cognitive function compared to those treated with ACEIs alone or other anti-hypertensives. Unlike previous studies evaluating cognition using only the MMSE, which is insensitive to mild impairments, these analyses also offer novel insights into specific cognitive domains sensitive to this effect, with significant differences among treatment groups observed for several measures of memory and executive function, including tests of verbal episodic memory, language, set shifting, and psychomotor speed. Deficits in memory and executive functions are commonly reported in patients with MCI and AD, [36, 37] with domain-specificity varying according to clinical stage and extent of history of vascular risk. Notably, those treated with ARBs in the present study showed higher delayed recall scores on the RAVLT and LM-II, two commonly used measures of verbal episodic memory. As deficits in delayed recall are one of the best predictors of progression from MCI to AD, [38] preserved function in this domain supports the potential clinical relevance of this effect.

Proposed mechanisms of ARB neuroprotection

Proposed mechanisms of neuroprotection for ARBs relate to potentially critical differences between the effects of ACEIs and ARBs on the metabolism of the Aβ peptide in the brain [39]. Aβ is toxic to both the vascular endothelium and neurons, [40] and Aβ abnormalities are observed early in the cascade of AD pathological processes [41]. Aβ accumulates in perivascular spaces to form fibrillar deposits as parenchymal amyloid plaques, a hallmark of AD pathology, and in addition to neurotoxicity, the deposition of Aβ in the cerebrovasculature increases inflammation [42] and enhances vasoconstriction [43] leading to higher levels of cortical Aβ and lower levels of plasma Aβ. Although centrally-acting ACEIs, such as perindopril, stimulate and support the cholinergic system and reduce the incidence of dementia from recurrent stroke, these benefits may be offset by its amyloidogenic actions. In vitro, ACE inhibits amyloid aggregation and fibrillation, thus the inhibition of ACE leads to increased levels of cortical Aβ. By contrast, most ARBs are brain-penetrating and block the angiontensin 1 (AT1) receptor pathway, attenuating Aβ oligomerization and increase insulin degrading enzymes, which catabolize Aβ and have been shown to decrease amyloid plaque counts in a transgenic mouse model almost four-fold [44]. In addition, telmisartan, a third generation ARB activates the peroxisome proliferator-activated receptor-gamma (PPAR-gamma), which provides additional benefits of blood-brain barrier protection, and anti-inflammatory and anti-atherosclerotic effects beyond angiotensin AT1 blockade and in high-doses reduces cellular Aβ and phosphorylated tau [45].

These differences in amyloidgenic effects for ACEIs versus ARBs may have important implications for amyloid clearance in the brain. The deposition of collagen in the deep cerebral venular system, particularly in oligemic dorsal periventricular regions underperfused by the cerebral arterioles penetrating into the deep white matter, known as venous collagenosis, [46, 47] is influenced by genetic and vascular risk factors [48]. Recent data from a spontaneously hypertensive rat model confirmed an association between arterial hypertension and venous collagenosis [49] and, in humans, imaging findings from AD patients suggest that WMH may actually represent vasogenic edema related to leakage from venules stenosed or occluded with collagen [50]. The resulting periventricular venous insufficiency interferes with interstitial cerebral fluid circulation, important for clearance of amyloid, and promotes amyloid deposition in the cortex as plaques and around the vessels as amyloid angiopathy. Future investigations are required to confirm this pathway as a potential mechanism of impaired Aβ clearance and evaluate the potential neuromodulatory role of ARBs in this process.

Mitigating bias and limitations

We employed several strategies to mitigate potential selection biases to strengthen these analyses. To address the potential for confounding by indication and bias due to non-random treatment allocation, we used a propensity score analysis to adjust for several selection factors that were based on current clinical guidelines for hypertension and were determined to be significantly associated with treatment. In addition to weighting the multivariate outcome analysis for this propensity score, we also adjusted all models for potential demographic, clinical, and genetic confounders in the estimation of treatment effects. However, this study also has limitations. There is a potential for recall bias and/or misclassification in this study due to the retrospective ascertainment of exposure status. Due to the high degree of variability in systolic and diastolic blood pressure measurements available in the ADNI data, these factors were not included in the present analysis. Information on duration of hypertension and the duration and dose of treatment were not available in the ADNI database. As a result, we were unable to adjust for length of exposure or evaluate potential dose-dependent interactions for these treatment associations. In addition, the lack of valid information on medication start and end dates limited our ability to explore potential predictive associations between treatment exposure and the study outcomes and all reported associations should not be considered causal in nature.

Conclusions

The present study has important potential implications for the treatment of elderly adults with hypertension, particularly those at risk of cognitive decline and progression to AD. Our findings generate several hypotheses about the potential neuroprotective effects of treatment with ARBs and support the need for clinical trial data to explicitly test these effects. As ACEIs are generally more commonly used in clinical practice, choosing a drug class that has equivalent benefits for blood pressure control and peripheral target organs and a similar side effect profile, but may have neuromodulating properties would be desirable in clinical practice when treating patients with hypertension, a major risk factor for SVD and AD. This study also provides clinical evidence to inform the understanding of the pathophysiological mechanisms underlying amyloid deposition and clearance and its potential impact on the structure and cognitive function in the aging brain.