Skip to main content

Table 4 Detailed characteristics of studies examining the effects of different levels of alcohol consumption

From: Long-term effects of alcohol consumption on cognitive function: a systematic review and dose-response analysis of evidence published between 2007 and 2018

Study details

Sample

Alcohol exposure categories

Details of the included article

Study dates

Arntzen 2010 [44]

Norway

Cohort name: the Tromsø Study

Serious risk of bias

Based on 5,033 men and women (56% female) with a mean age of 58 years at point of first alcohol measure (T0) and ~65 years at final cognitive assessment.

Substudy of the Tromsø Study cohort which was established in 1974 to examine cardiovascular risk among people aged 25-85 years.

Teetotaller: not defined

Category (referent): < 1 glass per fortnight for women or men (midpoint = 0.5 g/day)

Category: 1–2 glasses per fortnight for women or men (midpoint = 1.4 g/day)

Category: 3–4 glasses per fortnight for women or men (midpoint = 3.4 g/day)

Category: > 5 glasses per fortnight for women or men (midpoint = 5.0 g/day)

Grams per drink: Not reported. Assumed 12–15 g (RARHA 2015).

Observational cohort examining associations between different levels of alcohol consumption and cognitive function.

Inclusion criteria: Eligible participants were aged 25–85 years at start of T0 (December 1994; 100% of those aged 55–74 and 5–10% of other birth cohorts were invited).

Exclusion criteria: self-reported stroke; incomplete alcohol data; incomplete covariate data; no data for any of the 4 cognitive tests.

Alcohol ascertainment: Current: self-report questionnaire asking about frequency (“how many times a month do you normally drink alcohol” and quantity “how many glasses of (beer/wine/spirits) do you normally drink” in a fortnight. Recall: not reported. Lifetime: “Are you a teetotaller”.

Cognitive function: Learning and memory (immediate and delayed recall of 12 nouns), complex attention (Digit Symbol-Coding test from Wechsler adult intelligence scale (WAIS); Tapping test scores for dominant and non-dominant hand). Mean difference in raw scores. Higher score = better cognition.

Study period: 1994-2001

Alcohol exposure: single assessment at baseline (T0: 1994-95)

Outcome measures: single assessments at ~7 year from T0. (T1: 2001)

Length of outcome follow-up: ~ 7 years

Downer 2015 [45]

United States

Cohort name: Framingham Heart Study Offspring Cohort

Serious risk of bias

Based on 664 men and women (56% female) mean age of 41.8 years at point of first alcohol measure (T0) and 74.8 at final measure of cognition (T2).

Substudy of the Framingham Heart Study Offspring Cohort among those actively participating in the cohort when cognitive testing was first introduced (1999)

Abstainer (referent): 0 drinks per week

Light: 1–6 drinks per week (mean = 5.6 g/day)

Moderate: 7–14 drinks per week (mean = 20 g/day)

Heavy: 15–34 drinks per week (mean = 41.6 g/day)

Grams per drink: Not reported. Assumed 14 g based on US standard.

Observational cohort examining associations between different levels of alcohol consumption in midlife and cognitive function at late life.

Inclusion criteria: Eligible participants were 60 years or older at first measure of cognition (T1).

Exclusion criteria: stroke, Alzheimer’s disease, other dementia; did not receive cognitive testing or an MRI within 6 months, no APOE genotype data; history of consuming ≥ 5 drinks almost daily (based on screening question administered at T2).

Alcohol ascertainment: Current: self-report questionnaire asking about quantity (“how many bottles/glasses/drinks of beer/wine/cocktails”) consumed per week. Recall: 12 months. Lifetime: screening question to exclude those who had drunk ≥ 5 drinks almost daily at any time of life.

Cognitive function: Global cognitive function based on average of standardised individual scores from 11 tests measuring: language (Boston naming test), complex attention (TMT-A and B), perceptual motor (Hooper Visual Organisational test), learning and memory (tests of immediate and delayed recall assessing: visual memory, verbal memory, and learning), abstract reasoning. Test results converted to Z-scores ((individual score - sample mean)/SD). Higher scores = better cognition. Other outcomes reported: learning and memory, executive function, brain volume.

Study period: 1971-2008

Alcohol exposure: single assessment at baseline = ‘midlife’ (T0: 1971). (‘Late life’ measure exclude from the SR because analyses are cross-sectional)

Outcome measures: two assessments, at ~ 6 year interval (range 1.5-8 years) (T1, T2: 1999-2002; ~2005-2008)

Length of outcome follow-up: ~ 34 years from T0.

Hassing 2018 [46]

Sweden

Cohort name: none—data from Swedish Twin Registry

Critical risk of bias

Based on 305 men and women (56% female) age ~ 56 to 66 years at point of first alcohol measure (T0) and mean age of 83 years at first measure of cognition (T1).

Analysis of data from the Swedish Twin Registry (established late 1950s) and the OCTO-Twin study on cognitive ageing (started 1991-93)

Abstainer: excluded from analyses

Occasional: < 1 drink per week

Low: ~4 drinks per week (midpoint g/day, not estimable)

Moderate: ~8 drinks per week (midpoint g/day, not estimable)

Heavy: > 15 drinks per week (no heavy drinkers in the sample)

Categories were reported for descriptive purposes only. Alcohol consumption was analysed as a continuous variable (grams per week).

Grams per drink: 12 g

Observational cohort examining associations between different levels of alcohol consumption in midlife and cognitive function at late life.

Inclusion criteria: Eligible participants were twins on The Swedish Twin Registry, aged ≥80 years at first measure of cognition (T1; birth years 1901-1911).

Exclusion criteria: non-drinkers (at T0: no information on how abstention was measured), dementia diagnosis at T1 (first measure of cognition), missing cognition data (T1), missing alcohol data (T0).

Alcohol ascertainment: Current: self-report questionnaire asking about frequency (whether drank alcohol or not; how often) and quantity (how much consumed on a typical occasion, by type). Recall: not reported. Lifetime: not reported.

Cognitive function: Global cognitive function (MMSE). Raw scores converted to T-scores (mean=50; SD=10). Smaller change in mean score over time = less cognitive decline. Other outcomes reported: learning and memory (subscale of Wechsler adult intelligence scale (WAIS): prose recall; Thurstone’s picture recognition test; Information task), perceptual motor ability (Block design test).

Study period: 1967-2001

Alcohol exposure: single assessment ‘midlife’ (T0: 1967)

Outcome measures: 5 assessments, at ~ 2 year intervals (T1-T5: 1991-93, 1993-95, 1995-97, 1997-99, 1999-2001)

Length of outcome follow-up: ~ 32 years from T0 to T5 (final cognition measure )

Heffernan 2016 [47]

Australia

Cohort name: Sydney Memory and Ageing Study

Serious risk of bias

Based on 821 men and women (55% female) aged 70-90 years at point of first alcohol measure (T0) and ~74-94 years at final cognitive assessment.

Abstainers (referent): no alcohol (last 12 months)

Low risk: > 0 to ≤ 2 drinks per day for women; > 0 to ≤ 4 drinks per day for men (weighted midpoint based on proportion of women in low risk group = 15 g/day)

Risky: > 2 drinks per day for women; > 4 drinks per day for men (weighted midpoint based on proportion of women in risky group = 43 g/day)

Grams per drink: 10 g based on Australian standard.

Data also re-analysed using NIAAA categories (results not presented in SR).

Observational cohort examining associations between different levels of alcohol consumption and cognitive decline in specific domains.

Inclusion criteria: Eligible participants were aged 70-90 years at T0, community dwelling.

Exclusion criteria: MMSE <24; health conditions (psychotic symptoms, dementia, schizophrenia, bipolar disorder, multiple sclerosis, motor neuron, developmental disability, progressive malignancy); learnt English after age 10; 2 or fewer valid scores for measured domains; no alcohol data; unknown APOE.

Alcohol ascertainment: Current: self-report in interview asking about frequency of drinking (monthly, weekly, daily) and “amount of drinks per drinking session”. Recall: last 12 months. Lifetime: ever “drank more heavily than in the last 12 months”; if no alcohol in last 12 months “had they ever consumed“.

Cognitive function: Learning and memory (immediate and delayed recall: Logical Memory Story A; Rey Auditory Verbal Learning; Benton Visual Retention), executive function (Controlled Oral Word Association; Trail making test B), complex attention (Digit Symbol-Coding; Trail making test A), language (Boston Naming, Semantic fluency - animals), perceptual motor ability (Block design test). Scores transformed to quasi-z scores (using baseline mean and SD of participants with cognition ≥1 SD from mean) and averaged across tests. Higher z score = better cognition (change from baseline > -1.0 SD = decline).

Study period: 2005-2011

Alcohol exposure: single assessment at baseline (T0: 2005-07)

Outcome measures: baseline (T0) and 2 follow-up assessments at ~2 year intervals. (T0-T2: 2005-2007; 2007-2009; 2009-2011)

Length of outcome follow-up: ~ 4 years (mean 38 months)

Hogenkamp 2014 [48]

Sweden

Cohort name: Uppsala Longitudinal Study of Adult Men (ULSAM)

Serious risk of bias

Based on 652 men aged 70 years at point of first alcohol measure (T0).

Substudy of the ULSAM cohort which was established to identify metabolic risk factors for CVD.

Non-drinker: 0 drinks per day

Category: > 0 to ≤ 1 drinks per day (mean = 5.4 g/day)

Category: > 1 to ≤ 2 drinks per day (mean = 16.7 g/day)

Category: > 2 drinks per day (mean = 28.9 g/day)

Alcohol analysed as continuous variable, examining linear trends, so no referent.

Grams per drink: 12 g

Observational cohort examining associations between different levels of alcohol consumption and cognitive function in older men.

Inclusion criteria: Eligible participants were healthy males aged 70 years (T0),.

Exclusion criteria: Cognitively unhealthy (MMSE <25), missing data on alcohol intake.

Alcohol ascertainment: Current: self-report of usual intake of types of alcohol per week. Recall: not reported. Lifetime: not measured.

Cognitive function: Specific cognitive domains (2 outcomes). Executive function (Trail making test part B [TMT-B]) and complex attention (Trail making test part A [TMT-A]). Higher raw scores = worse cognition (these are timed tests).

Study period: 1990-2001

Alcohol exposure: single assessment (T0: 1990-1994)

Outcome measures: baseline and follow-up assessment ~7 years later. (T0-T1: 1990-1994; 1997-2001)

Length of outcome follow-up: ~7 years from baseline (T0)

Horvat 2015 [49]

Eastern Europe (Russia, Poland, Czech Republic)

Cohort name: HAPIEE (Health, Alcohol, and Psychosocial Factors in Eastern Europe) prospective cohort study

Serious risk of bias

Based on 28,947 men and women (54.7% female) aged 45–69 years at point of first alcohol measure (T0).

Non-drinker: 0 g/day

Light (referent): < 5 g/day for women (midpoint = 2.5 g/day); < 10 g/day for men (midpoint = 5 g/day)

Moderate: ≥ 5 to < 20 g/day for women (midpoint = 12.5 g/day); ≥ 10 to < 40 g/day for men (midpoint = 25 g/day)

Heavy: ≥ 20 g/day for women (midpoint = 27.5 g/day); ≥ 40 g/day for men (midpoint = 55 g/day)

Observational cohort examining associations between different levels and patterns (frequency, binge, problem drinking) of alcohol consumption and cognitive function in older adults.

Inclusion criteria: Eligible participants were aged 45–69 years (T0), randomly selected from population registers and electoral lists.

Exclusion criteria: none reported.

Alcohol ascertainment: Current: self-report graduated frequency questionnaire (GFQ) asking about frequency of consumption and number of drinks (by alcohol type; not specified whether asked in relation to a typical occasion/week/other). Recall: last 12 months. Lifetime: not measured.

Cognitive function: Specific cognitive domains (4 outcomes). Learning and memory (immediate recall of words in 3 x 1 minutes trials; delayed recall of words after other tests administered), language (verbal fluency, number animals named in 1 minute), complex attention (letter cancelled test for attention, mental speed, concentration). Test results were converted to Z-scores (mean =0; SD = 1) using whole sample means and SDs. Higher scores = better cognition.

Study period: 2002-2008

Alcohol exposure: single assessment (T0: 2002-2005; second assessment made at follow-up, but not used in prospective analysis)

Outcome measures: baseline and follow-up assessments at ~ 4 year intervals. (T0-T1: 2002-2008)

Length of outcome follow-up: 4 years from baseline (T0)

Kesse-Guyot 2012 [50]

France

Cohort name: SU.VI.MAX 2 cohort

Serious risk of bias

Based on 3,088 men and women (46% female) aged 45-60 years (mean 52) at point of first alcohol measure (T0).

Observational follow-up of SU.VI.MAX randomised trial of dietary supplements for prevention of cancer, heart disease and mortality.

Non-drinker (referent): 0 g/day for women or men

Category: ≥ 0.1 to ≤ 4.9 g/day for women or men (midpoint = 2.5 g/day)

Category: ≥ 5.0 to ≤ 14.9 g/day for women or men (midpoint = 9.95 g/day)

Category (referent): ≥ 15.0 to ≤ 29.9 g/day for women or men (midpoint = 22.45 g/day)

Category: ≥ 30.0 to ≤ 59.9 g/day for women or men (midpoint = 44.95 g/day)

Category: ≥ 60.0 g/day for women (midpoint = 74.95 g/day); ≥ 60.0 to ≤ 89.9 g/day for men (midpoint = 74.95 g/day)

Category: ≥ 90.0 g/day for men (midpoint = 119.9 g/day)

Observational cohort examining associations between different levels of alcohol consumption in midlife and cognitive function 13 years later.

Inclusion criteria: Eligible participants were healthy adults aged 45-60 years (T0), and agreed to participate in the observational follow-up SU.VI.MAX.

Exclusion criteria: incomplete cognitive tests, < 3/12 dietary records, missing values for any covariables.

Alcohol ascertainment: Current: 24 hour dietary record (bimonthly over 2 years, randomly assigned across 2 weekend days and 4 week days) asking about the number alcoholic drinks (by type) and portion size (validated photographs of 7 portion sizes, including 2 extreme). Recall: 24 hours. Lifetime: not measured.

Cognitive function: Global cognitive function based on mean of standardised individual scores from 4 tools measuring: learning and memory (RI-48 test - a delayed cued recall test), language (verbal fluency, number animals named and number words beginning with P in 2 minutes), executive function (forward and backward digit span; Delis- Kaplan trail-making test). Test results converted to T scores (rescaled to SD = 10; 1 point difference in score = 1/10 difference in SD). Higher scores = better cognition. Also reported results for specific domains.

Study period: 1994-2009

Alcohol exposure: single assessment at baseline (T0: 1994-1996)

Outcome measures: single assessment (T1: 2007-2009)

Length of outcome follow-up: ~ 13 years from T0.

Kitamura 2017 [51]

Japan

Cohort name: Murakami Cohort Study

Serious risk of bias

Based on 1,814 men and women (60% female) aged 44-79 years at point of first alcohol measure (T0).

Substudy of the Murakami Cohort Study which was established to examine risk factors for age-related disease.

Non-drinker or rare drinker (referent):

< 1 g of alcohol per week

Category: 1–149 g of alcohol per week (midpoint = 11 g/day)

Category: 150–299 g of alcohol per week (midpoint = 32 g/day)

Category: 300–449 g of alcohol per week (midpoint = 54 g/day)

Category: ≥ 450 g of alcohol per week (midpoint = 75 g/day)

Observational cohort examining association between different levels of alcohol consumption (and other lifestyle factors) and cognitive impairment

Inclusion criteria: Eligible participants were those aged 44-79 at T0, and participating in the Murakami Cohort. No information on eligibility criteria for cohort.

Exclusion criteria: None reported.

Alcohol ascertainment: Current: self-report questionnaire asking about frequency of consumption and amount (by alcohol type). Lifetime: no information. Recall period: not reported.

Cognitive function: Global cognitive function (MMSE). Results reported as binary outcome in which cognitive impairment was defined as score <24.

Study period: 2011-2016

Alcohol exposure: single assessment at baseline (T0: 2011-2013)

Outcome measures: single assessment. (T1: 2014-2016)

Length of outcome follow-up: not reported. Assumed to be ~ 3 years from baseline (T0)

Lang 2007 [52]

United States, United Kingdom

Cohort name: English Longitudinal Study of Ageing (ELSA); U.S Health and Retirement Study (HRS)

Serious risk of bias

Based on 13,333 men and women (57% female) aged 65 years or above at point of first alcohol measure (T0).

Non-drinker: 0 drinks per day

Category (referent): > 0–1 drinks per day for men or women (midpoint = 7 g/day)

Category: > 1–2 drinks per day for men or women (midpoint = 21 g/day)

Category: > 2 drinks per day for men or women (midpoint = 35 g/day)

Grams per drink: not reported (assumed 14 g based on USA standard and [62])

Pooled data from two observational cohorts examining the association between different levels of alcohol consumption and cognitive function, and between alcohol consumption and physical disability, and among older people.

Inclusion criteria: Eligible participants were aged 65 years or above (T0).

Exclusion criteria: none reported.

Alcohol ascertainment: Current: self-report questionnaire asking “how many days per week” they drank alcohol and number of drinks consumed “on average” on drinking days (HRS: last year; ELSA: last 3 months). Lifetime: no information (HRS); non-drinkers who had quit were asked if they had done so for health reasons (ELSA).

Cognitive function: Global cognitive function based on the sum of scores on three tests, word recall (mean of immediate and delayed word recall scores, score out of 10), numeracy (score out of 4), and specifying the date (day, date, month, year; score out of 4). A score in the bottom quintile was assessed as “poor cognitive function”.

Study period: 1998-2002

Alcohol exposure: single assessment at baseline (T0: 1998)

Outcome measures: single assessment. (T1: 2002)

Length of outcome follow-up: ~ 4 years (median: 50 months for HSE, 45 months for ELSA)

McGuire 2007 [53]

United States

Cohort name: Second Longitudinal Study of Aging (LSOA II)

Critical risk of bias

Based on 2,572 men and women (66% female) aged 70 years or above at point of first alcohol measure (T0; mean age 76).

Substudy of the LSOA II cohort which was established to examine health and, and the causes and consequences of health events among older persons (9447 men and women).

Non-drinker (referent): zero drinks per day (past year)

One drink per day or less: ≤ 1 drink/day men or women (≤ 12 g/day, midpoint = 6 g/day)

More than one drink per day: > 1 drink/day for men or women (> 12 g/day, midpoint = unknown)

Categories based on NIAAA guideline s[63].

Grams per drink: not reported (assumed 12 g based on NIAAA guidelines)

Observational cohort examining association between different levels of alcohol consumption and cognitive impairment among people 70 years and over.

Inclusion criteria: Eligible participants were aged 70 years (T0), and community-dwelling.

Exclusion criteria: Cognitively impaired (1.5 SD units less than the cohort mean at T1) on measures of cognitive function (below).

Alcohol ascertainment: Current: self-report questionnaire asking “how many days they drank alcoholic beverages, on average, in the last year” and number of drinks consumed on drinking days. Lifetime: no information.

Cognitive function: Global cognitive function based on the sum of scores on two tests, one of mental status (0-10 points: e.g. questions ‘who is the president’, ‘what is used to cut paper’; ‘what is desert plant’; ‘what is the day, date, month, year’; counting backward from 20 and 86) and one of immediate memory (0-10 points: 10 item list of concrete nouns). Function was dichotomised as low (score of 9.5-13) or high (score of 14-20).

Study period: 1994-2000

Alcohol exposure: two assessments, ~2 years apart (T0, T1: 1994, 1997-1998)

Outcome measures: 2 assessments, baseline and ~ 2 years later. (T1, T2: 1997-1998, 2000)

Length of outcome follow-up: ~2 years from baseline (T1)

Piumatti 2018 [54]

United Kingdom

Cohort name: UK Biobank prospective cohort

Serious risk of bias

Based on 13,342 men and women (54.7% female) aged 40-73 years at point of first alcohol measure (T0).

Substudy of the UK Biobank cohort involving those who had undergone a repeat assessment.

Alcohol consumption was treated as a continuous variable in analyses (mean grams of alcohol per day), so categories were not defined.

The analysis was limited to ‘weekly drinkers’: those who consumed alcohol at least once per week.

Observational cohort examining associations between different levels of alcohol consumption and change in cognitive function in middle and older populations.

Inclusion criteria: Eligible participants were aged 40-73 years (T0), from a population sample from those registered for the UK National Health Service and living within 40 km of a Biobank research centre.

Exclusion criteria: Consumed alcohol less frequently than once a week, self-disclosed history of neurological disorder (e.g. stroke, head trauma), only one valid score (from 7 tests) at baseline or follow-up.

Alcohol ascertainment: Current: self-report questionnaire asking about frequency of consumption and number of drinks consumed on average per week (by alcohol type). Recall: not reported. Lifetime: not measured.

Cognitive function: Specific cognitive domains (2 outcomes). Complex attention (processing speed based on a ‘stop-go’ reaction time task). Results reported for reaction time (mean of completed test trials) and intra-individual variation (IIV; standard deviation of each participant’s reaction time over 7 trials). Lower scores = better cognition.

Study period: 2006-2015

Alcohol exposure: single assessment (T0: 2006-2010; second assessment made at follow-up, but not used in prospective analysis)

Outcome measures: baseline and follow-up assessments at ~ 5 year intervals. (T0-T1: 2006-2010; 2011-2015)

Length of outcome follow-up: ~ 5 years from baseline (T0; mean 4.31)

Richard 2017 [55]

United States

Cohort name: The Rancho Bernardo Study

Serious risk of bias

Based on 1334 men and women (54% female) aged 55-84 years at point of first alcohol measure (T0).

Substudy of the Rancho Bernardo Study cohort which was established to examine heart disease risk factors.

Non-drinker (referent): ‘no past alcohol use’ or 'did not drink in last year'

Moderate: ≤ 1 drink/day for men 65 and older and women; ≤ 2 drinks/day for men (midpoint = 6 g/day for women; midpoint = 12 g/day for men)

Heavy: > 1–3 drinks/day for men age 65 and older and women; > 2–4 for men under 65 (midpoint = 24 g/day for women; midpoint = 36 g/day for men)

Excessive: > 3 drinks/day for men age 65 and older and women; > 4 drinks/day for men under 65 (midpoint = 48 g/day for women; midpoint = 60 g/day for men)

Grams per drink: 12 g; NIAAA guidelines

Observational cohort examining association between different levels and patterns (by frequency) of alcohol consumption and cognitively healthy longevity (survival to age 85).

Inclusion criteria: Eligible participants were those with potential to reach 85 years during follow-up period (55-84 years at T0).

Exclusion criteria: Those who did not have ‘intact cognitive function’ at any assessment prior to 85th birthday (or had not had an assessment 2 years prior to birthday). Missing data on education status. Missing data on education status.

Alcohol ascertainment: Current: self-report questionnaire asking about frequency of consumption and number of drinks (by alcohol type) in a typical week. Lifetime: asked about any ‘past alcohol use’.

Cognitive function: Global cognitive function (MMSE). Raw scores converted to Z-scores (adjusted for sex, age, education) using normative data. Cognitive impairment: Z-scores below −1.5. Outcomes reported: Cognitively Healthy Longevity (CHL: survival to age 85 without cognitive impairment), Cognitively Impaired Longevity (CIL: survival to age 85 with cognitive impairment).

Study period: 1984-2009

Alcohol exposure: single assessment at baseline (T0: 1984-1987)

Outcome measures: up to 6 assessments at ~ 4 year intervals. (T1-T6: 1988-2009)

Length of outcome follow-up: median of 13.9 years from baseline (T0)

Sabia 2011 [56]

France

Cohort name: GAZEL cohort study

Serious risk of bias

Based on 4073 men aged ~45-55 years at point of first alcohol measure (T0) and 55-65 years at point of cognition measure (T10).

Substudy of GAZEL cohort study which was established to examine disease and health-related factors among workers in France’s national electricity and gas company.

No-consumption: 0 drinks per week

Occasional: 1–3 drinks per week (midpoint = 3 g/day)

Light (referent): 4–14 drinks per week (midpoint = 14 g/day)

Moderate: 15–21 drinks per week (midpoint = 28 g/day)

Heavy: > 21 drinks per week (midpoint = 38 g/day)

Grams per drink: reported as 10–12 g (11 g assumed in SR analyses)

Observational cohort examining association between average level of alcohol consumption (measured over 10 years) and cognitive function at age ≥ 55 years. Also examines the effect of the trajectory of consumption (decreasing, stable, or increasing over 10 years) on cognition.

Inclusion criteria: Eligible participants were men aged ≥ 55 years at the time cognition was measured (T10) and working for the electricity and gas company in which the GAZEL cohort was based.

Exclusion criteria: Women (due to small number in the GAZEL cohort: ~10%); had no measure of alcohol consumption from T0-T4, T5-T9, or both; did not have full covariate data; did not participate in cognitive tests (n = 4525, 48.2%).

Alcohol ascertainment: Current: self-report questionnaire asking about frequency of consumption and number of drinks per day (by alcohol type) in last 7 days. Calculated mean consumption per week over 10 years using annual measures of consumption (T0-T9). Lifetime: no information.

Cognitive function: Specific cognition domain - complex attention measured by the Digit Symbol Substitution Test (DSST; subtest of the Weschler Adult Intelligence Scale). Mean scores reported for number of correct responses on 93 items (score range 0-93; higher score=better cognition).

Study period: 1992-2004

Alcohol exposure: 10 assessments at ~ 1 year intervals (T0-T9: 1992-2001, or 1993-2002, or 1994-2003; period determined by year of cognitive testing)

Outcome measures: single assessment (T10: 2002, or 2003, or 2004)

Length of outcome follow-up: 12 months from baseline (T9)

Sabia 2014 [57]

England

Cohort name: Whitehall II cohort study

Serious risk of bias

Based on 7153 men and women (29 % female) aged 35-55 years at point of first alcohol measure (T0) and 55-80 years at final cognition measure.

Analysis from the Whitehall II cohort which was established to examine social determinants of health among British civil servants.

Alcohol abstainers in the last 10 years: 0 grams in last 12 months (T1, T2, and T3)

Alcohol cessation in the last 10 years (quitters): 0 g in last 12 months (T2), > 0 grams at T0 or T1

Occasional drinkers: > 0 g in last 12 months, none in the last week (T0, T1 and T2)

0 to 70th percentile (referent): 0.1–9.9 g/day for women (median = 3.4 g/day); 0.1–19.9 g/day for men (median = 8.4 g/day)

70th to 90th percentile: 10–18.9 g/day for women (median = 13.3 g/day); 20–35.9 g/day for men (median = 26.3 g/day)

> 90th percentile: 19–66 g/day for women (median = 23.8 g/day); 36–112 g/day for men (median = 46.9 g/day)

Observational cohort examining association between average level of alcohol consumption in midlife (mean age 44 years) and subsequent cognitive decline.

Inclusion criteria: Eligible participants were British public servants age 35–55 years at cohort inception (T0).

Exclusion criteria: Missing alcohol or covariate data. Did not participate in the any of the baseline or follow-up assessments of cognition.

Alcohol ascertainment: Current: self-report questionnaire asking about frequency of consumption (last 12 months) and number of drinks (by alcohol type) in last 7 days. Calculated mean consumption over 10 years from data collected at T0, T1 and T2. Lifetime: no information.

Cognitive function: Global cognitive function (average of scores on 4 tests, each standardised using the mean and SD of scores at T2). Tests were of executive function (Alice Heim 4-I timed test of inductive reasoning; recall of “S” words; recall of animal names), learning and memory (recall of 20 words). Higher GCF score = less cognitive decline (over 10 years). Other outcomes reported: executive function; learning and memory

Study period: 1985-2009

Alcohol exposure: multiple assessments; baseline and then 2 assessments at ~5 year intervals (T0-T2: 1985-88, 1991-93, 1997-99)

Outcome measures: 3 assessments at ~5 year intervals (T2-T4: 1997-99, 2002-04, 2007-09)

Length of outcome follow-up: ~ 10 years from baseline (T2)

Samieri 2013a [58]

United States

Women’s Health Study

Serious risk of bias

Based on 6174 women aged ≥ 60 at point of first alcohol measure (T0).

Observational substudy of Women’s Health Study randomised trial of aspirin and vitamin E for prevention of CVD and cancer.

Non-drinker (referent): 0–1 drinks per day (median = 0 g/day)

Category: ≥ 1 to ≤ 14.9 g/day (median = 2.9 g/day; range 1.2–6.0)

Category: ≥ 15 g/day (median = 25.4 g/day; range 16.8–37.8)

Observational cohort examining association between a Mediterranean diet and specific components (including different levels of alcohol consumption) and cognitive function over time.

Inclusion criteria: Eligible participants were those aged ≥65 years at cognitive assessment (T1; ~60 at T0),

Exclusion criteria: complete dietary data (‘complete’ was not defined).

Alcohol ascertainment: Current: self-report food frequency questionnaire asking about frequency of consumption of foods and beverages, including alcohol, and portion size (no information reported). Recall period: last 12 months. Lifetime: no information.

Cognitive function: Global cognitive function average of z-scores from 5 tests: Telephone Interview for Cognitive Status (overall, including delay recall of 10-word list), East Boston Memory Tests (immediate and delayed recall), category fluency test. Other outcomes reported: Learning and memory (average of z-scores on the 4 tests). Higher mean scores = better cognition (inferred, not reported). No information on SD for average of Z scores, so scores are difficult to interpret.

Study period: 1992-2004

Alcohol exposure: single assessment (T0: 1992-1995)

Outcome measures: 3 assessments, first at T1 (average of 5.6 years from T0) and then at ~2 year intervals (T1-T3: 1998-2004)

Length of outcome follow-up: ~10 years from T0.

Solfrizzi 2007 [59]

Italy

Cohort name: Italian Longitudinal Study on Aging (ILSA)

Serious risk of bias

Based on 1445 men and women (44% female) aged 65 to 84 years at point of first alcohol measure (T0).

Substudy of ILSA which aims to examine common chronic conditions in the older population, and identify risk and protective factors

None: zero in last 5 years (current abstainer; referent) (former= zero in last 5 years, but some over lifetime)

Category: < 1 drink per day (midpoint = 7.5 g/day)

Category: 1–2 drinks per day (midpoint = 22.5 g/day)

Category: > 2 drinks per day (midpoint = 37.5 g/day)

Grams per drink: 15 g of alcohol

Observational cohort examining association between different levels of alcohol consumption and incidence of mild cognitive impairment (also progression to dementia).

Inclusion criteria: Eligible participants were 65-84 years at baseline (T0), independent or institutionalised.

Exclusion criteria: Confirmed diagnosis of dementia at T0 (structured clinical assessment for all participants with score on MMSE <24), refusal to perform MMSE or other neuropsychological test, unknown level of education.

Alcohol ascertainment: Current: self-report food frequency questionnaire asking about frequency of consumption (number of times per day/month/year) and number of drinks per day (by alcohol type; 3 portion sizes). Recall: last 12 months. Lifetime: asked ‘when they had begun to drink’ and ‘how much beer or wine per day ever since’ (to identify former drinkers, and changed patterns).

Cognitive function: Incidence of mild cognitive impairment diagnosed by trained neurologist using diagnostic criteria based on Petersen [64] (did not require subjective memory impairment; allowed for neurocognitive disabilities and comorbidities). Other outcomes: progression from MCI to dementia.

Study period: 1992-1996

Alcohol exposure: single assessment at baseline (T0: 1992)

Outcome measures: 2 assessments, baseline and then ~3.5 years later (T0, T1: 1995-1996)

Length of outcome follow-up: 3.5 years from baseline alcohol measurement (T0).

Stott 2008 [60]

United Kingdom, Netherlands

Prospective Study of Pravastatin in the Elderly at Risk (Prosper)

Serious risk of bias

Based on 5804 men and women (52 % female) aged 70-82 years at point of first alcohol measure (T0) and ~73–85 years at final cognition measure.

Observational study using data collected from the PROSPER randomised trial of Pravastatin.

Non-drinker (referent): not defined. Assumed 0 to < 1 unit/week for men and women (midpoint = 0.6 g/day; )

Low intake: ≥ 1 to ≤ 3 units/week for women; ≥ 1 to ≤ 7 units/week for men (midpoint = 2.3 g/day for women; midpoint = 4.6 g/day for men)

Moderate intake: > 3 units/week for women ; > 7 units/week for men (midpoint = 4.6 g/day for women; midpoint = 11.4 g/day for men)

Grams per drink: not reported (assumed 8 g based on UK standard, but study includes participants from Netherlands where 10 g is a standard drink)

Observational cohort examining association between different levels of alcohol consumption and cognitive function over time.

Inclusion criteria: Eligible participants were those aged 70-82 years (T0), with good cognitive function (see exclusion) and evidence of vascular disease or major vascular risk factors (hypertension, smoking, diabetes).

Exclusion criteria: MMSE 24 or below at T0. Alcohol or drug abuse.

Alcohol ascertainment: Current: very little information reported about the measurement of alcohol here or in the trial protocol or report except “alcohol intake was … quantified in terms of usual alcohol intake in units per week for the previous month”. Lifetime: no information; assume not collected.

Cognitive function: Global cognitive function (MMSE; higher scores means better cognition). Mean scores are reported for MMSE and other measures (below). Other outcomes reported: complex attention (Stroop Color–Word test; Letter Digit Coding test); learning and memory (immediate and delayed recall on Picture-Word Recall test).

Study period: Dec 1997- Mar 2002

Alcohol exposure: single assessment at baseline (T0: Dec 1997 to ~May 1999)

Outcome measures: 5 assessments, first at baseline then at ~1 year intervals (T0-T4: years not reported)

Length of outcome follow-up: mean 3.2 years from baseline (T0).

Wardzala 2018 [61]

United States

Cohort name: Oregon Brain Aging Study (OBAS); Intelligent Systems for Assessing Aging Changes (ISAAC) study

Serious risk of bias

Based on 486 men and women (75% female) aged ~80 years or above at point of first alcohol measure (T0).

Substudy involving participants from OBAS and ISAAC cohorts that met eligibility criteria for the current study.

Rare/never-drinker (referent): zero drinks per week (for any period ≥ 3 month over lifetime)

Moderate: for women: < 3 drinks/day and < 7 drinks per week (mean = 8 g/day); for men: < 4 drinks/day and < 14 drinks/week (mean = 9 g/day)

Heavy: for women: ≥ 3 drinks/day and ≥ 7 drinks per week (mean = 27 g/day); for men: ≥ 4 drinks/day and ≥ 14 drinks/week (mean = 24 g/day)

Categories based on NIAAA guidelines.

Grams per drink: not reported (assumed 12 g based on NIAAA guideline s[63])

Observational cohort examining association between different levels of alcohol consumption and cognitive function among people ~80 years and over.

Inclusion criteria: Eligible participants were aged ≥80 years at T0 (≥70 years for non-Caucasian, who comprised <10-20% of participants), living independently in the community with better than average health for age.

Exclusion criteria: Cognitively impaired (Clinical Dementia Rating (CDR) of > 0.5 and a Mini-Mental State Examination (MMSE) score of ≤24). No alcohol data; missing outcome data.

Alcohol ascertainment: Current: self-report questionnaire in interview asking about “frequency of drinking (days per week) and drinks per drinking day". Lifetime: asked if "ever consumed > 1 drink per week for > 3 months". Asked about drinking (same quantity/ frequency questions) at age ’40-current’, ’19-39’ and ‘0-18’ years.

Cognitive function: Global cognitive function MMSE score. Specific cognitive domains: learning and memory (word list: delayed recall), executive function (Trail making test B), complex attention (Digit Symbol-Coding), language (Semantic fluency - animals). Results reported as change in mean score over time (smaller change = better outcome).

Study period: 2004 to ~2011 (OBAS); 2007 to ~2017 (ISAAC)

Alcohol exposure: single assessments for most participants at baseline (T0: ~2004 OBAS; 2007-09 ISAAC)

Outcome measures: not reported, ~6-7 annual assessments based on time in study (mean 6-8 years). (No information on time points. Assume-T0-T7: 2004 to ~2011 (OBAS); 2007 to ~2017 (ISAAC)

Length of outcome follow-up: no information. ~5–7 years from alcohol measurement (T0)

  1. *Content is replicated for studies that examined levels and patterns, except details of alcohol categories/ascertainment. Denotes a study that also contributed data on patterns of alcohol consumption. ††National Institute on Alcohol Abuse and Alcoholism (NIAAA), United States [63]