Epilepsy, Vascular Risk Factors, and Cognitive Decline in Older Adults

Abstract

Background and Objectives

Recent studies have shown that global cognitive ability tends to decline faster over time in older adults (≥65 years) with epilepsy compared with older adults without epilepsy. Scarce data exist about the role of vascular risk factors (VRFs) on cognitive course in epilepsy. We assessed whether the associations of individual VRFs with cognitive trajectory differed depending on the presence of prevalent epilepsy.

Methods

The Cardiovascular Health Study is a population-based longitudinal cohort study of 5,888 US adults aged ≥65 years. Cognitive function was assessed annually with modified Mini-Mental State Examination (3MS; global cognitive ability) and Digit Symbol Substitution Test (DSST; information processing speed). We used linear mixed models to estimate the individual and joint associations of epilepsy and VRFs with cognitive decline by modeling epilepsy × VRF interactions one by one, each adjusted for all other VRFs considered, including demographics, health behaviors, clinical characteristics, and comorbid diagnoses. From these models, we estimated excess mean cognitive decline due to interaction of epilepsy with each VRF.

Results

We observed excess mean decline in global cognitive ability (3MS) due to interactions of epilepsy with hypertension (6.6 points greater mean 8-year decline than expected if no interaction; 95% CI 1.3–12.0) and with abstaining from alcohol (5.8 points greater than expected; 95% CI 0.3–11.3). We also observed excess mean decline in information processing speed (DSST) due to interactions of epilepsy with prior stroke (18.1 points greater mean 9-year decline than expected; 95% CI 7.6–28.5), with abstaining from alcohol (6.1 points greater than expected; 95% CI 2.5–9.8), and with higher triglyceride levels (2.4 points greater than expected per SD; 95% CI 0.4–4.3).

Discussion

Associations of some VRFs with cognitive decline in older adults are stronger in the presence of epilepsy, suggesting a need for greater attention to vascular protection for preserving brain health in older adults with epilepsy.

In large population-based longitudinal studies, older adults with epilepsy experienced more rapid cognitive decline on average than expected for their age.^1,2 Longitudinal studies of aging that combine epilepsy ascertainment with thorough risk factor characterization and repeated cognitive measurements are well suited to elucidating modifiable factors that hasten cognitive decline in epilepsy. Identification of such factors could help to guide development of interventions for preserving cognitive health in older adults with epilepsy.

In the general aging literature, vascular risk factors (VRFs) such as hypertension,³ diabetes,⁴ and smoking⁵ are associated with development of cognitive impairment and dementia. Approximately 35% of dementia is attributable to a combination of potentially modifiable risk factors.⁶ However, a striking lack of data are available to address the relationship of VRFs with cognitive course in older adults with epilepsy. Older adults who have epilepsy in combination with another risk factor may experience greater-than-expected cognitive decline, not only greater than those who have either epilepsy or the other risk factor alone but even greater than would be expected from the sum of excess declines associated with epilepsy alone and the other risk factor alone. We call this greater-than-expected cognitive decline an excess decline due to interaction of epilepsy and the other risk factor.⁷ As an example of this excess decline due to interaction of epilepsy and another factor, in the Cardiovascular Health Study (CHS), older adults who had prevalent epilepsy in combination with APOE ε4 allele experienced substantially greater-than-expected mean decline in global cognitive ability.²

Here, we extend the scope of prior work² to identify modifiable risk factors that have stronger associations with cognitive decline when combined with epilepsy, with an emphasis on VRFs. Are associations of individual VRFs, such as hypertension, with cognitive decline comparable between older adults with and without epilepsy, or are they stronger? The previous finding of excess cognitive decline due to interaction of epilepsy and APOE ε4 allele² underlies our present hypothesis: older adults with epilepsy who also have a given VRF experience cognitive decline that is (1) greater than in older adults with epilepsy but without the VRF, (2) greater than in older adults without epilepsy but with the VRF, and (3) even greater than what would be expected from summing excess declines associated with epilepsy alone and the VRF alone.

Methods

Design, Setting, and Participants

The CHS is a population-based multicenter observational cohort study of coronary heart disease (CHD) and stroke in adults aged 65 years and older in whom data were collected prospectively.^8,9 Details are in eMethods (links.lww.com/WNL/C287).

Outcome Variables

Primary outcomes were cognitive functioning assessed by modified Mini-Mental State Examination (3MS) and Digit Symbol Substitution Test (DSST). 3MS, a test of global cognitive performance,¹⁰ was administered annually up to 9 times from 1990/1991 to 1998/1999 (initial measure followed by 8 annual follow-up measures). Its score ranges from 0 (worst) to 100 (best), with a score <78 as a cutoff for screening for dementia.¹¹ A 5-point decline in the 3MS score over the course of 5–10 years is considered clinically meaningful in older adults.¹² DSST, a 90-second test of information processing speed, visuomotor coordination, and attention,¹³ was administered annually up to 10 times from 1989/1990 to 1998/1999 (initial measure followed by 9 annual follow-up measures). Its score ranges from 0 (worst) to 90 (best). Clinically meaningful decline in DSST scores among older adults is not clear. Participants included in our analysis contributed a mean (SD) of 6.7 (2.6) 3MS scores and a mean (SD) of 7.2 (2.9) DSST scores.

Exposure Variables

Epilepsy

We developed an algorithm, as detailed previously,¹⁴ to identify CHS participants who had epilepsy, relying on previously validated International Classification of Diseases, ninth revision (ICD-9) codes for epilepsy submitted during hospitalizations that occurred during CHS follow-up, outpatient ICD-9 codes for epilepsy submitted during CHS follow-up, antiepileptic medication use, and self-report of seizures during study follow-up. Details are in eMethods (links.lww.com/WNL/C287). For this analysis, we compared the prevalent epilepsy group with participants who had no history of epilepsy at enrollment, censoring participants at occurrence of incident epilepsy. Because of the small number of incident epilepsy cases occurring through 1998/1999 when annual cognitive testing ended (n = 21), we did not include incident epilepsy in this analysis.

VRFs of Interest

Demographics

We examined 4 demographic exposures including age (years), sex (male vs female), race (Black vs White or other), and level of education (any post-high school education vs none and years of education among those who did not complete high school). We did not consider Hispanic ethnicity in our analysis, as only 62 of 5,888 CHS participants (1%) reported Hispanic ethnicity, and only 2 of those Hispanic participants had prevalent epilepsy, too few for meaningful statistical adjustment, subgroup analysis, or interaction models.

Health Behaviors

Health behaviors included smoking status at study enrollment (former smoking vs not; current smoking vs not) and alcohol use at enrollment (abstaining vs light drinking [4.9 drinks/wk]; drinks per week among those with any use).

Clinical Characteristics

Clinical characteristics, all measured at enrollment, included body mass index (kg/m²), waist circumference (cm), systolic blood pressure (mm Hg), triglycerides (mg/dL), high-density lipoprotein (HDL) cholesterol (mg/dL), depressive symptom score measured with the Center for Epidemiologic Studies Depression Scale, a scale ranging from 0 to 30, with higher scores indicating more depressive symptoms, and self-rated health (fair or poor vs good to excellent).

Comorbid Cardiovascular or Metabolic Diagnoses

Comorbid diagnoses included a history or presence of the following conditions at enrollment: hypertension, diabetes mellitus, CHD, atrial fibrillation, heart failure, stroke, and chronic kidney disease. Details are in eMethods (links.lww.com/WNL/C287).

Statistical Analysis

For participant characteristics, we calculated frequencies and percentages for categorical variables and mean and SDs for continuous variables. For each VRF in turn, we examined whether its association with cognitive decline was stronger when combined with prevalent epilepsy (Table 1). We analyzed 3MS and DSST separately, each as the dependent variable in a set of models. Details are in eMethods (links.lww.com/WNL/C287).

Table 1.

Structure of Calculations for Estimates of Mean Decline in the Cognitive Score by Epilepsy and Vascular Risk Factor Status

Standard Protocol, Approvals Registrations, and Patient Consents

Informed consent was obtained from all participants in the study.

Data Availability

Deidentified data will be shared through the CHS Neurology Working Group by request from any qualified investigator.

Results

Among 5,672 CHS participants included in our 3MS analysis, 208 had prevalent epilepsy at enrollment. The mean age at enrollment was 73 years, and 58% of participants were women (Table 2). Over the ensuing 8-year follow-up period with annual 3MS assessments, 10% of longitudinal 3MS scores sought while participants were living were not obtained due to loss to follow-up. Loss to follow-up differed by epilepsy and VRF status. For example, percentages of 3MS scores not obtained due to loss to follow-up were 9% among participants without epilepsy or hypertension, 9% among those with epilepsy but no hypertension, 11% among those with hypertension but no epilepsy, and 14% among those with both epilepsy and hypertension.

Table 2.

Characteristics at Study Enrollment of Participants Included in 3MS Analysis

Trajectory of Global Cognition (3MS)

Interaction of Epilepsy and Hypertension

The combination of having epilepsy and hypertension together was associated with greater-than-expected mean 8-year decline in the 3MS score (Table 3 and Figure 1A). As outlined in Table 3, mean 8-year decline in the 3MS score was 5.3 points in older adults with neither epilepsy nor hypertension (reference group), 8.0 points in those with epilepsy but no hypertension, 7.4 point in those with hypertension but no epilepsy, and 16.7 points in those with both epilepsy and hypertension. The group with only epilepsy declined on average 2.7 points more than the reference group, and the group with only hypertension also declined on average 2.1 points more than the reference group. Therefore, we would expect the group with both epilepsy and hypertension to decline on average 4.8 points more than the reference group over 8 years—the sum of excess mean declines observed in epilepsy only and hypertension only groups—for an expected 8-year decline of 10.1 points if no interaction (d₁₀ + d₀₁ − d₀₀ = 8.0 + 7.4 − 5.3 = 10.1). However, the group with both epilepsy and hypertension actually did not decline on average 10.1 more points, but rather 16.7 more points than the reference group, which is 6.6 points greater decline over 8 years than would be expected from the sum of excess mean declines associated with epilepsy alone and hypertension alone. Thus, excess 8-year decline in the mean 3MS score due to interaction of epilepsy and hypertension was 6.6 points (95% CI 1.3–12.0). Given that the mean 8-year decline among people with no epilepsy and no hypertension was 5.3 points (95% CI 4.5–6.1), the excess mean 8-year decline of 6.6 additional points due to interaction (95% CI 1.3–12.0) represents approximately a doubling in the rate of cognitive aging due to the interaction, relative to what was expected with chronological aging.

Table 3.

Estimated Mean 8-Year Decline in the 3MS Score by Prevalent Epilepsy and Selected Vascular Risk Factor Status

Figure 1. Trajectories of the Mean 3MS Score by Prevalent Epilepsy and Selected Vascular Risk Factor Status.

(A) Hypertension. (B) Alcohol abstention. (C) Coronary heart disease. (D) Current smoking. Estimates adjusted for age, sex, race, education, smoking, alcohol use, body mass index, waist circumference, systolic blood pressure, triglycerides, HDL cholesterol, depressive symptoms, self-rated health, diabetes, CHD, atrial fibrillation, heart failure, stroke, and chronic kidney disease. 3MS = modified Mini-Mental State Examination; CHD = coronary heart disease; HDL = high-density lipoprotein.

Trajectories of the mean 3MS score by epilepsy and hypertension status are visualized in Figure 1A across 8 years of follow-up, showing a steeper rate of cognitive decline over time in the group having both epilepsy and hypertension (purple) than in the groups having hypertension only (blue), epilepsy only (red), or neither condition (black). In relation to 3MS score <78 as a suggested screening threshold for dementia,¹¹ we observed that among participants with no epilepsy and no hypertension, 7.6% scored <78 at the first assessment, increasing to 15.7% at the last assessment. Among those with epilepsy but not hypertension, 13.6% scored <78 at the first assessment, increasing to 28.4% at the last assessment. Among those with hypertension but not epilepsy, 10.8% scored <78 at the first assessment, increasing to 21.3% at the last assessment. Finally, among those with both epilepsy and hypertension, 14.2% scored <78 at the first assessment, increasing to 29.2% at the last assessment.

Interaction of Epilepsy and Alcohol Abstention

As shown in Table 3 and Figure 1B, the combination of epilepsy and alcohol abstention was associated with mean 8-year decline in the 3MS score of 16.9 points, which was 5.8 points greater (95% CI 0.3–11.3) than the 11.1 points of mean 8-year decline that would be expected if there were no interaction. In Figure 1B, the groups with epilepsy alone (red) or alcohol abstaining alone (blue) each showed faster mean decline than the reference group of light drinkers without epilepsy (black), and estimated decline among those who had epilepsy and abstained from alcohol (purple) was substantially faster than in the other groups.

Possible Interactions of Epilepsy and CHD and of Epilepsy and Current Smoking

As shown in Table 3 and Figure 1C, the combination of epilepsy and CHD was associated with mean 8-year decline in the 3MS score of 18.3 points, which was 5.5 points greater (95% CI −1.5 to 12.4) than would be expected if there were no interaction. Similarly, as shown in Table 3 and Figure 1D, the combination of epilepsy and current smoking was associated with mean 8-year decline in the 3MS score of 19.8 points, which was 6.2 points greater (95% CI −2.0 to 14.3) than would be expected if there were no interaction. These interactions were not statistically significant, as shown by the 95% CI overlapping the null value of 0. Unlike for current smoking, we did not observe an excess cognitive decline due to interaction of epilepsy and former smoking (not shown).

Summary of All VRFs Examined

In Figure 2, estimates in purple falling to the right of the vertical dashed line indicate excess mean 8-year decline in the 3MS score due to interaction of epilepsy with each VRF, with the strongest findings being for hypertension and alcohol abstaining as described above. Estimates falling to the left indicate less-than-expected decline, and estimates falling along the vertical line indicate observed decline equal to what would be expected if no interaction. We did not observe interactions of epilepsy with demographic factors, except for somewhat less-than-expected mean 8-year decline among men with epilepsy, which would imply greater-than-expected mean 8-year decline among women with epilepsy. Contrary to what we expected, we also observed less-than-expected mean 8-year decline among people with epilepsy who were former smokers, who reported fair or poor general health, or who had diabetes. Finally, for some comorbid diagnoses, including atrial fibrillation, heart failure, and stroke, estimates of excess decline due to interaction with epilepsy were imprecise, having wide 95% CIs consistent with either substantially greater- or less-than-expected mean 8-year decline, or with no interaction.

Figure 2. Excess Mean 8-Year Decline in the 3MS and 9-Year Decline in DSST Scores Due to Interactions of All Vascular Risk Factors Examined With Epilepsy.

Horizontal bars accompanying each estimate are 95% CIs. Contrasts across risk factor levels are age per SD increment; men vs women; Black race vs White race; no post-high school education vs any; years of education through 12th grade per SD decrement; former smoking vs never; current smoking vs never; alcohol abstaining vs 4.9 drinks/week; drinks/week per SD increment among drinkers only; BMI per SD increment; waist circumference per SD increment; SBP per SD increment; triglycerides per SD increment; HDL cholesterol per SD decrement; depressive symptoms per SD increment; fair/poor health vs good/excellent; and each comorbidity diagnosis vs not. Each estimate adjusted for all other risk factors listed in the figure. 3MS = modified Mini-Mental State Examination; AF = atrial fibrillation; BMI = body mass index; CHD = coronary heart disease; CKD = chronic kidney disease; DSST = Digit Symbol Substitution Test; HDL = high-density lipoprotein; HF = heart failure; HS = higher secondary; SBP = systolic blood pressure.

Trajectory of Information Processing Speed (DSST)

Interaction of Epilepsy and Stroke

As shown in Table 4 and Figure 3A, the combination of epilepsy and stroke was associated with mean 9-year decline in the DSST score of 28.1 points, which was 18.1 points greater (95% CI 7.6–28.5) than the 10.0 points of mean 9-year decline that would be expected if there were no interaction.

Table 4.

Estimated Mean 9-Year Decline in the DSST Score by Prevalent Epilepsy and Selected Vascular Risk Factor Status

Figure 3. Trajectories of the Mean DSST Score by Prevalent Epilepsy and Selected Vascular Risk Factor Status.

(A) History of stroke. (B) Alcohol abstention. (C) Triglyceride levels. (D) HDL cholesterol levels. Estimates adjusted for age, sex, race, education, smoking, alcohol use, body mass index, waist circumference, systolic blood pressure, triglycerides, HDL cholesterol, depressive symptoms, self-rated health, diabetes, CHD, atrial fibrillation, heart failure, stroke, and chronic kidney disease. CHD = coronary heart disease; DSST = Digit Symbol Substitution Test; HDL = high-density lipoprotein.

Interaction of Epilepsy and Alcohol Abstention

As shown in Table 4 and Figure 3B, the combination of epilepsy and alcohol abstaining was associated with mean 9-year decline in the DSST score of 13.0 points, which was 6.1 points greater (95% CI 2.5–9.8) than the 6.9 points of mean 9-year decline that would be expected if there were no interaction. This finding for DSST was similar to that for epilepsy and alcohol abstention for 3MS.

Interactions of Epilepsy and Lipid Measures: Triglyceride Level and HDL Level

As shown in Table 4 and Figure 3C, in the absence of epilepsy, the triglyceride level (adjusted for HDL in addition to other factors) was not associated with DSST decline, which was about 6 points over 9 years regardless of the triglyceride level. However, in the presence of epilepsy, mean 9-year decline in DSST was greater when the triglyceride level was 1 SD above the mean (11.8 points) than at the mean triglyceride level (9.3 points). This was 2.4 points excess mean 9-year decline in the DSST score due to interaction of epilepsy with a higher triglyceride level (95% CI 0.4–4.3; apparent discrepancy 11.8 − 9.3 = 2.4 not 2.5 is due to rounding results to 1 decimal place for display). As shown in Table 4 and Figure 3D, we observed a possible interaction of epilepsy with low HDL cholesterol (adjusted for triglyceride in addition to other factors), with an excess mean 9-year decline in the DSST score of 1.4 points due to interaction of epilepsy with a lower HDL level (95% CI −0.4 to 3.2); this was not statistically significant.

Summary of All VRFs Examined

In Figure 2, estimates in orange show excess mean 9-year decline in the DSST score due to interaction of epilepsy with each VRF. The strongest findings were for stroke, alcohol abstention, and higher triglycerides. We observed no interactions of epilepsy with demographic variables except for the association of epilepsy with greater-than-expected decline among men with epilepsy. Unlike for the 3MS score, we did not observe an interaction of epilepsy with hypertension for mean 9-year decline in the DSST score. We observed no interactions of epilepsy with diabetes, CHD, AF, HF, or CKD; some of those estimates had wide 95% CIs that would not allow us to rule out the possibility of either greater-than-expected or less-than-expected mean 9-year decline due to interaction.

Discussion

In this longitudinal study of older adults with and without epilepsy, we examined whether associations of VRFs with cognitive decline were stronger when combined with prevalent epilepsy, namely with epilepsy onset before CHS enrollment. We found that older adults with epilepsy who also had hypertension or who abstained from alcohol at study enrollment experienced global cognitive decline (3MS) at a faster rate than what would be expected from summing excess declines associated with epilepsy alone and the VRF alone. We also found that older adults with epilepsy who also had prior stroke, abstained from alcohol, or had higher triglycerides at study enrollment experienced decline in information processing speed (DSST) at a faster rate than expected. We also observed some possible interactions, though not statistically significant, of other VRFs and epilepsy, including CHD or current smoking for global cognition and lower HDL levels for information processing speed. Each epilepsy × VRF interaction that was associated with greater cognitive decline was independent from other VRFs. Although our findings support mechanistic interaction in which cognitive decline was faster when both exposures were present together, instead of just one or the other,¹⁵ our analysis was not designed to confirm or rule out biological interaction in which 2 exposures physically interact in the body.¹⁶ Further research with different methods, such as animal models, may shed further light on biological mechanisms that underlie the interactions we observed.

Hypertension, especially when present in midlife, increases the risk for cognitive impairment and dementia,¹⁷ likely through a mechanism of subclinical strokes.¹⁸ The mechanism by which the effect of hypertension on cognition is more potent in older adults with epilepsy will need to be investigated. One possible explanation is that older adults with epilepsy had a higher burden of covert vascular brain injury (VBI) that was not accounted for even when we adjusted for other VRFs. Although emerging data show a higher prevalence of covert VBI in older adults with late-life epilepsy (onset after 60 or 65 years) when compared with those without epilepsy,^19,20 little is known about the burden of covert VBI in older adults whose epilepsy began earlier in life.

We found that older adults with epilepsy who abstained from alcohol experienced greater-than-expected decline in both 3MS and DSST. This finding extends prior knowledge about the relationship between alcohol consumption and cognitive decline. Systematic reviews indicate that light to moderate drinking is associated with a lower risk for dementia compared with abstaining from alcohol.²¹ In a cohort study of 1,341 participants with a mean follow-up of 21 years, never drinkers in midlife had a poorer performance (than infrequent drinkers) on various assessments of learning, memory, executive function, and psychomotor speed, assessed during late life.²² In a prior analysis of CHS participants, those who consumed 1–6 drinks per week had a lower risk for incident dementia than nondrinkers.²³ The reason that light to moderate alcohol use is associated with good cognitive outcomes may be because of healthy effects on cerebral blood vessels and a lower risk of clinical and subclinical strokes.^23-25 Light drinking may also be beneficial through increasing the level of brain-derived neurotrophic factor, which regulates neuroprotection and synaptic plasticity.²⁶ Unlike heavy drinking, which may exacerbate seizures, light to moderate drinking has no known ill effects in epilepsy.²⁷ Although this finding may be helpful to patients with epilepsy who drink in moderate amounts, it is not adequate to justify alcohol consumption among those with epilepsy who abstain. We were not able to assess whether certain epilepsy-related characteristics, such as high seizure frequency or antiepileptic medication polytherapy, affected alcohol consumption behavior and confounded the relationship between epilepsy and alcohol. For example, some patients with more severe epilepsy may have chosen to abstain from alcohol. Thus, epilepsy severity could be a potential confounder in the association of alcohol abstention and epilepsy with cognitive decline.

Midlife dyslipidemia increases the risk of Alzheimer disease and vascular dementia,²⁸ whereas dyslipidemia in late life seems to be associated with reduced risk of dementia.²⁹ This discrepancy may be related to cholesterol level decreasing in the early stages of dementia.^28,30 We found that decline in DSST was greater than expected in older adults who had epilepsy and elevated triglyceride levels. Similarly, we observed greater-than-expected decline in DSST in older adults with epilepsy and low HDL levels. Increased levels of triglycerides in midlife have been associated with the risk of dementia,³¹ with a study showing elevated triglycerides levels in midlife being predictive of brain Aβ and tau pathology 20 years later.³² Higher levels of HDL have been inversely associated with dementia compared with low HDL levels³³ and may be a stronger predictor of cognitive decline³⁴ and Alzheimer disease than other lipid measures.³³

Although individuals with prior stroke can experience cognitive deficits in several domains including memory and language, impairment in information processing speed is the most pronounced deficit seen among those with poststroke cognitive decline.³⁵ Periventricular white matter lesions and lacunar infarcts have been associated with decline in information processing speed.³⁶ Greater-than-expected decline in information processing speed in patients with epilepsy who also had prior stroke raises the possibility that older adults with epilepsy have a greater degree of covert VBI.

The association of epilepsy with global cognitive decline was stronger in current smokers. Although the difference in the rate of cognitive decline for current vs never smokers was not statistically significant, the 95% CI did not overlap the null value by much, raising the possibility that a true difference may exist in the population, warranting additional study. In the general aging population, smoking has been independently associated with accelerated cognitive decline and higher risk of dementia. For example, a meta-analysis showed that older adult smokers have a higher risk for any dementia and greater annual decline in the Mini-Mental State Examination (MMSE) score (a measure similar to 3MS).⁵ Smoking causes vascular injury, including endothelial injury, arterial stiffness, and increased platelet aggregation,^37,38 and reduces cortical gray matter density.³⁹ Approximately 1 in 4 adults with epilepsy currently smoke cigarettes compared with 1 in 6 adults who do not have epilepsy.⁴⁰ The mechanism by which older adults with epilepsy who smoke have accelerated cognitive decline is unclear. However, health care providers should inform adults with epilepsy who smoke about excess risk for cognitive decline and provide smoking cessation resources.

Older adults with epilepsy who also had CHD experienced global cognitive decline at a faster rate than expected. As with epilepsy and current smoking, the finding for epilepsy and CHD was not statistically significant but may warrant further study. A meta-analysis of prospective cohort studies suggests that individuals with CHD have 45% increased risk of cognitive impairment or dementia than those without CHD.⁴¹ The link between CHD and cognitive impairment or dementia may not be directly causal but possibly reflects widespread atherosclerosis. Studies also show that individuals with asymptomatic CHD have increased white matter hyperintensities.⁴² Whether older adults with epilepsy who also have CHD have a higher burden of covert VBI than older adults without epilepsy who also have CHD may explain excess risk seen in older adults with epilepsy.

We found different VRFs associated with accelerated cognitive decline when tested with 3MS vs DSST. The differences may be because 3MS and DSST have different sensitivity for detecting cognitive changes. The 3MS captures global cognitive function by testing a broad range of cognitive domains including attention, expressive language, and short-term memory. A 3MS score <78 has 88% sensitivity and 90% specificity for detecting dementia.¹¹ A 5-point decline in the 3MS score is considered clinically meaningful.¹² By that metric, we observed clinically meaningful excess declines in 3MS scores due to interactions of epilepsy with hypertension, alcohol abstaining, and possibly other VRFs. In the aging population, DSST may be more sensitive to changes in cognition than 3MS, given that 3MS, like its parent assessment the MMSE, has a ceiling effect in people with high levels of cognitive ability.⁴³ As a measure of executive cognitive function, including processing speed, visuomotor coordination, and attention,¹³ the DSST is sensitive to cognitive impairment that precedes dementia,⁴⁴ and lower scores are associated with higher white matter rating scores⁴⁵ and other measures of covert VBI on MRI.⁴⁶ Changes in DSST could be useful to track covert VBI. However, clinically meaningful decline in DSST scores among older adults is not clear.

Our study has limitations. First, the epilepsy sample in this study was likely too small to adequately investigate interactions of epilepsy with certain VRFs, although CHS is a large population-based study. Second, because CHS is a study of cardiovascular disease in older adults, with epilepsy cases ascertained via claims data, our study did not have data on seizure types, seizure control, or duration of epilepsy, which could affect cognitive function. Third, we analyzed VRFs ascertained only at study enrollment; therefore, our findings pertain to differences in cognitive decline across groups defined by different VRF status at a given point in time, but not in relation to changes in VRF status over time within individuals. Fourth, we did not examine whether our analyses of dyslipidemia or BMI were affected by antiseizure medication use. Older antiseizure medications like carbamazepine and phenytoin, which are inducers of cytochrome P450 enzymes, involved in cholesterol synthesis, have been shown to alter lipid profile, increasing total cholesterol and triglycerides.⁴⁷ Fifth, we used data collected in the 1990s for this analysis because those were the years during which CHS participants underwent annual cognitive testing. The relationships of epilepsy and VRFs with cognitive health are likely similar now to what they were then. Advances in treatment for epilepsy and VRFs in recent decades may have important implications for cognitive health in epilepsy, which is an avenue for further research using more recently collected data. Sixth, nonrandom loss to follow-up occurred in this study, whereby participants with cognitive decline and with a greater burden of adverse health conditions, such as epilepsy combined with hypertension, dropped out earlier than relatively healthy participants. However, with such nonrandom loss, we would expect that the true rate of cognitive decline among people having both epilepsy and a given VRF would be even faster than what we observed. Seventh, we did not consider prior alcohol consumption history to distinguish between former vs never drinkers among alcohol abstainers. Abstainers at study enrollment likely include former drinkers, possibly some with heavy drinking, a known risk factor for dementia, who might have quit alcohol before study enrollment.

Our findings raise questions about specific interventions needed to mitigate excess risk of cognitive decline in older adults who have epilepsy and VRFs. Better cognitive outcomes during aging are associated with ideal cardiovascular health,⁴⁸ as with the American Heart Association's Life's Simple 7 health factors including nonsmoking, healthy diet, adequate physical activity, healthy body weight, blood pressure, total cholesterol, and fasting glucose levels. Individuals with epilepsy have a higher burden of VRFs such as hypertension and hyperlipidemia compared with individuals without epilepsy.⁴⁹ If epilepsy and VRFs combined lead to faster cognitive decline, should older adults with epilepsy be treated more aggressively to control modifiable VRFs? The presence of multiple VRFs at midlife substantially increases the risk of dementia in a dose-dependent manner. For example, the hazard ratio of dementia increases from 1.27 in individuals with 1 midlife VRF compared with none to 2.37 for individuals with 4 midlife VRFs vs none.⁵⁰ With epilepsy being a chronic condition with typical age at onset at age 20 years or younger, when should patients with epilepsy be screened for modifiable VRFs? Studies addressing such questions are urgently needed, not only for epilepsy that onset early in life but also for incident late-life epilepsy cases, which were shown in prior work to experience faster mean decline in the 3MS score than their peers without epilepsy.²

Our findings offer some insight into the role of VRFs in the relationship between epilepsy and accelerated cognitive decline in older adults. Our findings will need to be confirmed in additional studies. More research in vascular protection in epilepsy is needed, so that we may come to understand biological mechanisms that would promote optimal brain health in older adults with epilepsy throughout late life.

Glossary

3MS

Modified Mini-Mental State Examination

CHD

coronary heart disease

CHS

Cardiovascular Health Study

DSST

Digit Symbol Substitution Test

HDL

high-density lipoprotein

ICD-9

International Classification of Diseases, ninth revision

MMSE

Mini-Mental State Examination

VBI

vascular brain injury

VRF

vascular risk factor

Appendix. Authors

Study Funding

This research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, and 75N92021D00006 and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided by R01AG023629 from the National Institute on Aging (NIA).

Disclosure

H. Choi receives funding from the NIH as a coinvestigator. M.S.V. Elkind receives ancillary funding from Roche and study drug in kind from the BMS-Pfizer Alliance for Eliquis, both for a federally funded trial of stroke prevention, and honoraria from UptoDate for chapters related to stroke. W.T. Longstreth receives funding from the NIH as a coinvestigator. The other authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.