Abstract
Alcohol consumption is a potentially modifiable risk factor for dementia, but the literature is not completely consistent. This inconsistency may be partly due to an interaction with the apolipoprotein E (APOE) genotype, an established risk factor for Alzheimer’s dementia. The aim of this study was to examine whether alcohol consumption is associated with incident dementia or decline in specific cognitive domains over 4 years, and if this effect is modified by APOE ɛ4 status. Non-demented community dwelling older adults (70-90 years) from an ongoing longitudinal study were assessed for cognitive impairment in attention/processing speed, language, executive function, visuospatial ability, and memory. Incident dementia was diagnosed according to DSM-IV criteria. Compared to those who did not drink in the previous 12 months, neither low consumption (HR 0.64 95% CI 0.3-1.4) or risky consumption (HR 0.58 95% CI 0.2-1.5) was associated with incident dementia. Carriers of the APOE ɛ4 allele were more likely to develop dementia, but there was no significant interaction with alcohol consumption.
INTRODUCTION
Excessive alcohol consumption can have negative health effects, including a detrimental impact on cognition [1]. However, there are multiple studies demonstrating that moderate alcohol consumption reduces the risk of some diseases and a similar protective effect has been suggested for cognitive decline in late life [1, 2].
The role of alcohol in cognitive decline is of interest as it represents a potentially modifiable risk factor for dementia. Two meta-analyses suggest low alcohol consumption may lower the risk for dementia in late life [3, 4]. This reflects the often reported finding that moderate alcohol consumption is protective against cognitive decline compared to nil or heavy consumption (e.g., [5, 6]). Despite this, multiple reviews on the effect of alcohol on cognition note that alcohol consumption is often estimated through retrospective self-report which can be unreliable; there is little consistency in the literature in how levels of alcohol consumption are defined or which other factors are controlled for in analyses [3, 6–9]. In addition to dementia risk, reviews have also considered the effect of alcohol on cognition as measured by test scores. One review showed moderate drinking reduced the risk of cognitive impairment (defined as low test scores) but that there was no significant benefit in the rate of cognitive decline over time [8]. By contrast, another review concluded that although there was evidence for light to moderate alcohol being protective against dementia, there was no clear support for a similar effect on cognitive decline [9].
Apolipoprotein E (APOE) genotype status might account for some of the variation in the literature. The apolipoprotein ɛ4 allele is a known genetic risk factor for Alzheimer’s disease [10–12], but there is conflicting evidence about its link to Alzheimer’s disease progression [13] and how it interacts with modifiable risk factors for cognitive decline [14]. The presence of the ɛ4 allele did not significantly moderate the effect of alcohol consumption on cognitive decline in some studies [15–19]. Other studies have shown different effects: when stratified by presence of the ɛ4 allele, the protective effect of alcohol was limited to those without APOE ɛ4, but no significant associations between alcohol and dementia were found among those with APOE ɛ4 [20, 21]. When comparing directly between those with and without APOE ɛ4, those with the ɛ4 allele had a greater risk of dementia with increased alcohol consumption [22–24]. These existing studies provide limited evidence to support that the APOE ɛ4 allele influences dementia risk with alcohol consumption. Additionally, studies with the longest follow-up periods are also in conflict. While there are two studies showing that the APOE ɛ4 allele increased the risk of dementia with alcohol intake, they are both drawn from the same longitudinal study [22, 23]. Only one long-term study has shown that an increased dementia risk was limited to those with no ɛ4 alleles for abstainers and heavy drinkers [25].
The aims of this study were to examine whether alcohol consumption increases the risk for incident dementia in an older community-living population, if this is modified by APOE ɛ4 status, and if alcohol consumption is associated with cognitive decline in specific domains (memory, language, attention/processing speed, executive function, and visuospatial ability).
METHODS
Participants
Participants were recruited as part of an ongoing longitudinal study, the Sydney Memory and Ageing Study (MAS). Full details of the Sydney MAS have been reported previously [26]. Briefly, individuals aged 70-90 years on the Electoral Roll in Eastern Sydney, New South Wales, Australia were invited to participate. To be eligible, participants needed to be sufficiently fluent in English to complete the psychometric assessment. Participants were excluded if they had a Mini-Mental State Examination (MMSE) score <24 (age and education adjusted) [27, 28], psychotic symptoms, a diagnosis of dementia, schizophrenia, bipolar disorder, multiple sclerosis, motor neuron disease, developmental disability, or progressive malignancy. Written, informed consent was obtained from all participants and the study was approved by the ethics committees of the University of New South Wales and the South Eastern Sydney and Illawarra Area Health service.
Neuropsychological testing
In the Sydney MAS, a comprehensive neuropsychological test battery was administered by trained research psychologists every 2 years (baseline, 2 years, and 4 years). The tests were categorized into five cognitive domains (memory, language, visuospatial, attention/processing speed, and executive functioning) based on the principal cognitive function represented according to convention and consistent with our previous studies (see details [26, 29]). The domains and tests used were: attention/processing speed (Digit Symbol-Coding [30], Trail Making Test (TMT) A [31]); memory (Logical Memory Story A delayed recall [32], Rey Auditory Verbal Learning Test (RAVLT) [31], RAVLT total learning, sum of trials 1–5, RAVLT short-term delayed recall, trial 6, RAVLT long-term delayed recall, trial 7, Benton Visual Retention Test recognition [33]); language (Boston Naming Test –30 items [34], Semantic Fluency (Animals) [31]; Visuo-spatial (Block Design [35]); Executive Function Controlled Oral Word Association Test (FAS) [31], Trail Making Test (TMT) B [31]).
Raw test scores were transformed into quasi-z scores, calculated using the mean and standard deviation of participants with normal cognition at baseline. Domain scores were calculated by averaging the quasi-z scores of the component tests. At baseline, cognitive impairment was defined as a quasi-z score of at least 1.0 SD below the mean. At follow-up, decline was defined as a greater than 1.0 SD decline from the baseline score. The most recent follow-up score was used to calculate decline.
Dementia classification
Neuropsychological tests measuring the cognitive domains of memory, language, visuospatial, attention/processing speed, and executive functioning were administered by trained research psychologists (see details [26, 29]). Full details of dementia diagnoses have been described previously [26]. Briefly, dementia classifications were made by specialist clinicians, derived from a panel of geriatric psychiatrists, neuropsychiatrists, clinical neuropsychologists, and clinical psychologists based on cognitive data collected every 2 years. All participants included in the Sydney MAS were non-demented, therefore all dementia cases reported are incident cases. Dementia was diagnosed based on DSM-IV criteria [36].
Alcohol consumption
Alcohol consumption was estimated from self-reported data collected in an interview at baseline. In the interview, participants were shown examples of what constituted an Australian standard drink (10 grams alcohol) for common types of alcoholic drinks (e.g., beer, wine, spirits). Participants were asked to report if they had consumed alcohol in the past 12 months, how frequently they drank (monthly, weekly, daily) and the amount of drinks per drinking session (number of standard drinks). This information was used to estimate the number of standard drinks per day a participant consumed over the past 12 months. If they reported not consuming alcohol in the past year, they were asked if they had ever consumed alcohol. To estimate whether this pattern of drinking was consistent with past behavior, participants were also asked if there was a period in their life when they drank more heavily than in the last 12 months. If they responded yes, they were asked to report the frequency (monthly, weekly, daily) and amount (number of standard drinks) during that period when they drank more heavily.
Using Australian National Health and Medical Research Council guidelines [37] on long term risks associated with drinking, the following drinking categories were used: abstainers (no alcohol in past 12 months), low risk (men ≤4/day, women ≤2 /day), and risky (men > 4/day, women > 2/day). A standard from the National Institute on Alcohol Abuse and Alcoholism (NIAAA; part of the National Institutes of Health, USA) was also used (US standard drink = 14 g alcohol). This categorized drinking levels in terms of amount consumed on an average day as well as amount consumed across a week. Alcohol consumption was categorized using the following cut-offs for men (≤4/day and ≤14/wk) and women (≤3/day and ≤7/wk). These criteria were adjusted for the difference with Australian standard drink size. Low risk consumption adhered to both these recommended amounts, increased risk was consumption above the recommended amount for either daily or weekly amounts; highest risk was consumption greater than both the daily and weekly recommended amounts.
Cardiovascular disease (CVD) risk score
A CVD risk score was calculated for each participant using a similar method to the Framingham Heart Study [38]. Variables included were: current smoking status, diabetic status, systolic blood pressure, total cholesterol level, high-density lipoprotein (HDL) level, and current anti-hypertensive medications. Where blood analyses were unavailable, body mass index was used instead of cholesterol or HDL data.
Depression
Symptoms of depression were measured using the Geriatric Depression Scale (15 item [39]). In the Sydney MAS, we used the GDS-15 with item 9 as described in Brink [40] (here item 12) instead of the more common one. A cut-off of ≥6 was used to indicate depression.
APOE genotyping
Genomic DNA was extracted using standard methods. APOE ɛ2/3/4 genotyping was undertaken using Taqman assays as described in Song and colleagues [41]. Carriers of the ɛ4 allele were defined as those individuals with least one ɛ4 allele.
Statistical analysis
Sample characteristics were determined using des-criptive statistics. Group differences were assessed using two-tailed independent t-tests for continuous variables and chi-square tests for categorical variables. Cox regression was used to determine risk of incident dementia during follow-up. For the cox regression analysis, the event was dementia and time to event was calculated to be halfway between cognitive assessment dates. For those who were withdrawn, deceased or did not have a follow-up assessment, a withdrawal date was estimated based on the sample mean for months in study. For cognitive domain data, logistic regression models were used to assess the impact of alcohol consumption at baseline and follow-up. All models adjusted for age, sex, and years of education.
RESULTS
Study sample
The Sydney MAS study consisted of 1,037 participants at baseline. For this study, participants were excluded if they did not learn English before they were 10 years old (n = 164), had fewer than three (out of five) valid scores for neuropsychological domains (n = 3), did not have baseline alcohol data (n = 1), or if their APOE ɛ4 status was unknown (n = 48). Characteristics of included and excluded participants are shown in Table 1. Excluded participants were older, had lower MMSE scores (unadjusted), had higher cardiovascular risk score, consumed fewer alcoholic drinks, but did not differ by years of education, sex, APOE ɛ4 status, or depressive symptoms. Participants who were not Caucasian were included as this was only a small number in our sample (n = 5). The APOE allele frequencies were 8.0%, 79.4%, and 12.6% for the ɛ2, ɛ3, and ɛ4 alleles. At the latest follow-up, 68 participants were deceased and 103 had withdrawn from the study. Those who were deceased or had withdrawn at the final follow-up but had at least one follow-up assessment (n = 33 and n = 49, respectively) were included in follow-up analysis. Those included with valid neuropsychological data (3/5 tests n = 640) were also included at follow-up. There were 722 participants included in the follow-up sample. Those excluded were older, had lower MMSE scores, were more likely to be men, and had higher depressive symptoms. Mean time in study was 38.3 months (SD = 14.0, range 10.5-57.1). Baseline characteristics for incident dementia compared to no dementia are shown in Table 2.
Alcohol consumption
Overall, as shown in Table 3, a minority of participants reported not drinking in the past year (n = 100, 12%). Of these, almost half (n = 46) were lifetime non-drinkers. About 40% of participants reported a change in drinking pattern and drinking more heavily in the past (n = 329). There was a sex difference in alcohol consumption; women (Median = 0.4, IQR 0.04-2.1) consumed significantly fewer drinks per day than men (Median = 2.6, IQR 0.4-3.0) (U = 52814.0, p < 0.001). Alcohol consumption did not differ by APOE ɛ4 status (abstainers 17%, low risk 24%, risky 25%; χ2(2) = 2.57, p = 0.28).
Cognitive impairment by domain and incident dementia
Participants in the baseline sample (N = 821) were classified as having cognitive impairment in a domain (attention/processing speed, executive function, language, visuospatial ability, memory) at baseline if their score was 1.0 SD or greater below the mean for the MAS sample (Table 4). Participants were classified as declining in a cognitive domain over follow-up if their follow-up score was 1 SD or more below their baseline score (Table 4). Over the 4-year follow-up period, 48 participants developed dementia (5.8%). APOE ɛ4 carriers were more likely to develop dementia (10.0%) than those who were not (4.6%) (χ2(1) = 7.7, p = 0.005).
Alcohol consumption and baseline cognitive domain impairment
Baseline logistic regression models of the association between recent alcohol consumption and cognitive impairment are shown in Table 5. There was no difference between odds of cognitive impairment at baseline based on amount of recent alcohol consumption (NHMRC standard). Covariates (APOE ɛ4, depression, cardiovascular disease risk score (CVD)) were included in separate models for each cognitive domain (Table 4). APOE ɛ4 was associated with executive function impairment (OR 1.61, 95% CI: 1.1, 2.4, p = 0.02) and memory (1.89, 95% CI: 1.3, 2.7, p = 0.001), while depression was associated with visual spatial ability (OR 2.56, 95% CI: 1.4,4.7, p = 0.003). However, the inclusion of these covariates in models did not impact on the association between alcohol consumption and cognition.
Alcohol consumption and longitudinal cognitive outcomes
Hazard ratios were calculated for incident dementia using cox regression (Table 6 and Fig. 1). Including age, years of education, and sex as covariates there was no association between drinking category and incident dementia. Neither the addition of CVD risk score, depression or the interaction term for APOE ɛ4 and alcohol consumption impacted on risk for incident dementia.
The association between alcohol consumption and decline in cognitive domains was tested using logistic regression models (Table 5). There was no association between drinking at baseline and cognitive decline. The inclusion of CVD risk and depression did not impact these findings.
Alternative drinking guidelines and past drinking
The above analysis was repeated using NIAA guidelines (adjusted for difference between Australian and US standard drink size). Using these guidelines, there was no cross sectional association between alcohol consumption and cognitive impairment, nor was there any association with cognitive decline or incident dementia (Supplementary Tables 2 and 3).
Participants who reported a change in drinking habits and who used to drink more heavily in the past also reported how much and how frequently they consumed alcohol earlier in their life (n = 329). The above analysis was repeated with just this sub-group and there was no association between those who used to drink risky amounts compared to low risk amounts either cross-sectionally or at follow-up (all p values >0.05; data not shown).
DISCUSSION
We did not find an association between recent alcohol consumption and incident dementia over 4 years. Additionally cognitive performance in specific domains was not associated with alcohol consumption either cross-sectionally or over time. The expected association between APOE ɛ4 status and incident dementia did not impact the effect of alcohol on incident dementia or cognitive decline. Inclusion of baseline cognition, cardiovascular risk, and depression symptoms did not influence these results. Our results showing no association between late-life alcohol consumption and incident dementia are consistent with some studies using a similar follow-up duration [42, 43] but not others [44, 45]. The absence of a relationship between alcohol consumption and dementia could not be explained by APOE ɛ4 carrier status.
In addition to dementia risk, our study examined the effect of alcohol consumption on cognitive decline in attention/processing speed, executive function, language, visuospatial ability, and memory. We found no evidence for alcohol being associated with decline in cognitive domains. This is consistent with the finding that there is no difference for men between abstainers and low/moderate drinkers, but inconsistent with the finding that men who consumed excessive alcohol declined faster in memory and executive function than those with light to moderate alcohol consumption [46].
Previous studies refer to categories of alcohol consumption (light, moderate, etc.) but there is little consistency in how these were operationalized [3, 7]. This variability makes it difficult to know what an ideal amount of drinking might be. To address this, we assessed alcohol consumption using two alcohol guidelines set by national health bodies (Australia and USA), neither of which provided evidence for a benefit of low alcohol consumption over abstaining from alcohol. Further, there was no evidence that those who consumed higher amounts of alcohol were at greater risk of dementia, although few in our sample had a high alcohol intake.
Given the relatively small number of people with dementia in our sample, the lack of an association in our study is not enough to offer guidance about the association between amount of alcohol consumption and dementia. Although there is existing evidence to support the claim that low amounts of alcohol can have cognitive benefits [3], there remains no firm evidence about the actual amount of most benefit in terms of standard drinks or differences by individual characteristics such as sex and APOE ɛ4 carrier status. There are studies that have included measures of mid-life alcohol consumption with a follow-up over approximately 20 years [23, 25]. These have shown protective effects of low alcohol consumption for cognitive impairment and incident dementia amongst carriers of APOE ɛ4 [23, 25]. Our results indicate that for older people (70 years and over) alcohol consumption in late life is not a useful modifiable risk factor. However, owing to the small number of people with dementia in our sample, there is a possibility that with more power in the analysis a significant effect could be observed.
The impact of alcohol consumption on incident dementia was independent of APOE ɛ4 status, which is consistent with some previous studies [15–18]. Not all studies investigating alcohol and cognitive decline report APOE ɛ4 status and those that do are not consistent. Some studies report that moderate alcohol consumption is protective against cognitive decline/dementia only for those with the APOE ɛ4 allele but no protective effect on those without the allele [22–24, 47]; other studies observed a protective effect for alcohol consumption but only in those without the APOE ɛ4 allele and no increased risk for those with the allele [20, 25].
There are possible limitations to this study. This study used retrospective estimates that are unlikely to be as reliable as those from prospective studies [23, 25]. That being said, prospective studies have not reported consistent findings with regards to alcohol consumption, dementia, and APOE ɛ4 allele status. The reported 12-month drinking period may not be a long enough time to observe effects on cognitive decline or dementia for our sample. Where alcohol has been measured in midlife, heavier drinking has been shown to be associated with faster cognitive decline compared to low/moderate drinkers [46]. Although we included an estimate of how much alcohol people consumed when they drank the most, analysis using these estimates also showed no association between alcohol consumption and dementia.
Another possible limitation is that if presence of the APOE ɛ4 allele is necessary for alcohol to moderate the risk of cognitive decline, our small sample (n = 137 with APOE ɛ4 allele) could limit our ability to observe this effect. Further, Harwood and colleagues [48] showed that dementia onset was much earlier for those with lifestyle risk factors (smoking, drinking) and APOE ɛ4. Since our sample only included those who were older than 70 years and did not have dementia at baseline, those most likely to have an increased risk due to these lifestyle factors may have already developed dementia and not been included in our study (no dementia at baseline). Also, as our study has focused on late life dementia, our findings may not apply to those with younger onset dementia. A further effect of the sample size is that we were unable to explore possible differences for men and women. This was accounted for to an extent as both alcohol guidelines applied recommended different amounts for men and women. However, no associations were directly tested by sex.
Studies which have shown protective effects for low alcohol consumption have used non-drinkers as the comparison group, and the small number of abstainers in our sample may have limited our ability to find significant associations (for reviews, [3, 4]). Our sample included few people who were lifetime abstainers, making categorical comparisons with this group difficult. An alternative to using lifetime abstainers only is to include a comparison with people who had not consumed alcohol in the past 12 months (lifetime abstainers and former drinkers) and/or those who consume alcohol in small amounts [23, 47]. However, the inclusion of people who are former drinkers has been criticized as those who stop drinking for health reasons may already be at increased risk of cognitive decline due to other risk factors [49]. The relatively small group of lifetime abstainers in our sample could reflect a cultural difference in drinking patterns compared to studies conducted in other countries. This is unlikely to be the case here as it is inconsistent with data on alcohol consumption in Australia. In a recent survey, the percentage of people over 70 years of age who had not consumed alcohol in the past 12 months was estimated to be 30.4% [50]. This includes 17% who were lifetime abstainers. By contrast, only 6% of our sample were lifetime abstainers. The difference between our sample and this broader Australian survey suggests our sample underestimates the number of non-drinkers among older people and may be a limiting factor for our study. Additionally, the generalizability of our study findings may be hindered by the exclusion of those from a non-English speaking background. This exclusion was applied owing to the use of neuropsychological tests that have not been sufficiently validated amongst those from a non-English speaking background.
We conclude that we could find neither an association between dementia risk and (i) alcohol intake nor (ii) any impact of including APOE ɛ4 allele status, in a cohort of older adults.
