Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The relationship between alcohol use and long-term cognitive decline in middle and late life: a longitudinal analysis using UK Biobank

The relationship between alcohol use and long-term cognitive decline in middle and late life: a... Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 Journal of Public Health | Vol. 40, No. 2, pp. 304–311 | doi:10.1093/pubmed/fdx186 | Advance Access Publication January 9, 2018 The relationship between alcohol use and long-term cognitive decline in middle and late life: a longitudinal analysis using UK Biobank 1,2 3 4 5 Giovanni Piumatti , Simon C. Moore , Damon M. Berridge , Chinmoy Sarkar , John Gallacher Department of Psychiatry, University of Oxford, Oxford, UK Unit of Development and Research in Medical Education (UDREM), Faculty of Medicine, University of Geneva, Geneva, Switzerland Violence & Society Research Group, School of Dentistry, Cardiff University, Cardiff, UK Farr Institute—CIPHER, Swansea University Medical School, Swansea UK Healthy High Density Cities Lab, HKUrbanLab, The University of Hong Kong, Hong Kong, China Address correspondence to Simon C. Moore, E-mail: MooreSC2@cardiff.ac.uk ABSTRACT Background Using UK Biobank data, this study sought to explain the causal relationship between alcohol intake and cognitive decline in middle and older aged populations. Methods Data from 13 342 men and women, aged between 40 and 73 years were used in regression analysis that tested the functional relationship and impact of alcohol on cognitive performance. Performance was measured using mean reaction time (RT) and intra-individual variation (IIV) in RT, collected in response to a perceptual matching task. Covariates included body mass index, physical activity, tobacco use, socioeconomic status, education and baseline cognitive function. Results A restricted cubic spline regression with three knots showed how the linear (β = −0.048, 95% CI: −0.105 to −0.030) and non-linear effects (β = 0.035, 95% CI: 0.007–0.059) of alcohol use on mean RT and IIV in RT (β = −0.055, 95% CI: −0.125 to −0.034; β = 0.034, 2 1 2 95% CI: 0.002–0.064) were significant adjusting for covariates. Cognitive function declined as alcohol use increased beyond 10 g/day. Decline was more apparent as age increased. Conclusions The relationship between alcohol use and cognitive function is non-linear. Consuming more than one UK standard unit of alcohol per day is detrimental to cognitive performance and is more pronounced in older populations. Keywords alcohol, alcohol consumption, public health The suggested curvilinear association between alcohol Introduction 16–18 and cognition, however, is controversial. Recent reviews The neurodegenerative effects of excessive alcohol con- 19–21 and meta-analyses indicate that there is little consensus 1–4 sumption are well documented. Alzheimer’s disease and on the level of alcohol consumption at which the harmful dementia have replaced ischaemic heart disease as the lead- effects of alcohol on cognition emerge. Furthermore, a ing cause of death in England and Wales, and death rates Mendelian Randomization study of alcohol and cognitive 6–8 for neurological disease are increasing worldwide. A lim- ited number of studies suggest a J- or U-shaped relationship between the volume of alcohol consumed and the long-term Giovanni Piumatti, Postdoctoral Researcher 9–11 cognitive decline, suggesting light to moderate alcohol Simon C. Moore, Professor of Public Health Research Damon M. Berridge, Professor of Applied Statistics consumption is a positive predictor of health status in older Chinmoy Sarkar, Assistant Professor adults, protects cognition and may reduce the risk of 13–15 John Gallacher, Professor of Cognitive Health dementia in later life. © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), 304 which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 305 performance found evidence of benefit from reduced alco- bottle of wine’; ‘there are 25 standard measures in a normal hol intake at all levels of self-reported consumption. sized bottle’). Respondents who declared that they drank alco- The current study examined the shape of the association hol ‘one to three times a month’ or on ‘special occasions only’ between alcohol consumption and change in cognitive per- (henceforth monthly drinkers) were also asked to record how formance. Data were drawn from UK Biobank, a large many drinks they consumed on average each month. However, cohort of middle and older aged adults. Respondents who these questions were not included at baseline and therefore consumed alcohol at least once a week or more frequently more than half of the sample of monthly drinkers at baseline were eligible for inclusion to reduce selection and reporting were not assessed. Accordingly, only participants who declared biases. A reaction time (RT) task was used as a robust test that they drank at least once a week (henceforth, weekly drin- of central processing speed. Cognitive performance was kers) were included in primary analyses. measured using mean RT and IIV in RT. Cognitive performance Cognitive performance was assessed using a ‘stop-go’ RT task Methods in which participants were shown two cards simultaneously on Sample a computer screen. Each card had a symbol on it and partici- Between 2006 and 2010, a heterogeneous population sample pants were asked to respond as quickly as possible, using a of 502 649 adults aged 40–73 years participated in the UK button-box, if both symbols matched. RT, from the presenta- Biobank prospective cohort study at 22 research centres tion of the cards to their press of the button, was recorded in located across the UK. Participants were registered with milliseconds (ms). Each participant was presented with 12 the UK National Health Service (NHS) and lived within a pairs of cards, the first five of which were training sets and radius of 40 km from one of the research centres. Self- data from these trials were discarded. Of the seven test trials, reported data were collected via touch screen questionnaires cards with matching symbols were presented on four occa- and interview. Information on the assessment procedure, sions selected at random. A demonstration of this test is avail- protocol and information on data access is available online able online (biobank.ctsu.ox.ac.uk/crystal/videos/snap.swf). (www.ukbiobank.ac.uk). For the purposes of estimating regression dilution, 20 346 individuals underwent a repeat Covariates assessment five years after their initial assessment. Data The effects of alcohol use on cognitive performance differ by from these respondents are used in the current longitudinal 24, 25 26 26 20 gender, education, past performance and age. These analysis. Individuals were omitted from the analysis if they variables were included as covariates in the present study disclosed a history of neurological disorder at either baseline or alongside deprivation as measured by the Townsend score, follow-up (Table S1), leaving 19 124 eligible participants. The physical activity assessed as walking activity, body mass index UK Biobank study was approved by the North West Multi- (BMI) and smoking status. Centre Research Ethics Committee (reference number 06/ MRE08/65). All participants gave written, informed consent. Data analysis Measures Alcohol consumption in grams per day was calculated by Alcohol use multiplying the average number of alcoholic drinks consumed Alcohol consumption was measured using the question each week by the average grams of alcohol contained in ‘about how often do you drink alcohol?’ Available responses each type of drink, determined using the UK Food Standard were ‘daily or almost daily’, ‘threeoffour times a week’, Agency’s guidelines. The total was then divided by seven to ‘once or twice a week’, ‘one to three times a month’, ‘special provide mean daily alcohol consumption. Alcohol consumption occasions only’, ‘never’ and ‘prefer not to answer’. Respondents was positively skew and log transformed. who drank alcohol once a week or more frequently were asked Consistent with established methods, RTs < 50 ms, to record how many alcoholic drinks they consumed on average indicating anticipation of the stimulus, were discarded as each week from a list of common alcoholic beverages (red and were RTs > 2 s, as the target stimulus had been withdrawn white wine, champagne, beer and cider, spirits and liquors, forti- at this point. RT was calculated as the arithmetic mean of fied wine, and other alcoholic drinks), or to respond ‘do not completed test trials. Intra-individual variation (IIV) was cal- know’ or ‘prefer not to answer’. Volumes were specified when culated as the standard deviation of each participant’sRTs referring to beverages (e.g. ‘there are six glasses in an average over the test trials. Participants with only one valid score at Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 306 JOURNAL OF PUBLIC HEALTH Table 1 Differences in mean reaction time (RT) and intra-individual baseline or follow-up were omitted. RT and IIV showed variation in reaction time (IIV) at follow-up according to socioeconomic log–normal distributions and were natural log transformed. and lifestyle factors. Values are means and (standard deviations) For the covariates, educational attainment was included as a binary variable (with or without a degree). BMI was Variable RT (ms) IIV included as two binary variables (normal <24.9 kg/m com- 2 2 pared to overweight 25–29.9 kg/m and obese ≥30 kg/m ) Age (years) as was smoking status (non-smoker compared to previous 40–52 512.76 (95.25)*** 68.53 (49.38)*** smoker and current smoker). Deprivation quintiles were 53–59 551.78 (102.18) 77.15 (55.23) included as a continuous variable, as were age and the time 60–63 571.90 (110.04) 82.27 (58.56) between baseline and follow-up assessments. Walking activ- 64+ 594.67 (114.15) 88.95 (66.03) ity was included as the number of days participants walked Gender for more than 10 min each week. Preliminary analyses found Females 562.51 (109.01)*** 79.80 (59.02) Males 548.80 (108.74) 77.40 (55.70) missing data was minimal (2.8%). Education Non-linearity in the alcohol–cognition relationship was No degree 564.44 (112.48)*** 81.25 (59.22)*** investigated using restricted cubic splines. A restricted cubic Degree 544.74 (103.67) 75.34 (54.98) spline is a cubic spline function with an additional constraint Deprivation quintile of linearity before the first knot and after the last knot. Least 553.77 (102.27)*** 78.54 (56.85)*** The number of knots was determined by examining the dis- 2 555.94 (109.68) 78.69 (57.23) tribution of average daily alcohol use in the sample, with the 3 555.49 (109.33) 79.72 (59.46) aim of locating boundaries between equal-sized categories 4 558.15 (109.76) 78.54 (57.59) and by comparing the Akaike Information Criteria (AIC) Most 558.80 (113.94) 77.65 (56.12) goodness of fit statistics across models. Alcohol intake Staged multivariable modelling, the regression of follow- Non-drinkers 574.79 (119.88)** 83.99 (67.08)** up RT and IIV on baseline alcohol consumption, began with Monthly 558.45 (112.07) 78.18 (56.34) Weekly 553.84 (107.35) 78.38 (56.98) an adjustment for age to establish the fundamental associ- Body mass index ation (model 1). Adjustment for social confounding was ≤Normal 555.57 (109.10)*** 78.72 (58.20)** made by further conditioning on lifestyle and background Overweight 553.86 (108.11) 77.45 (55.24) (model 2). The influence of baseline cognition was then Obese 561.73 (111.56) 81.26 (60.55) taken into account (model 3). Finally, interactions effects Walking tertiles (days/week) were included (model 4). All analyses were performed using 0–4 550.89 (106.82)* 76.87 (54.65)** Stata 14 (StataCorp. 2015. Stata Statistical Software: Release 5–6 555.28 (108.57) 78.35 (56.83) 14. College Station, TX: StataCorp LP). 7 559.50 (110.27) 80.02 (59.64) Smoking Non-smokers 552.91 (108.06)*** 78.22 (57.40) Results Previous smokers 561.07 (110.88) 79.48 (57.78) Current smokers 554.39 (107.61) 77.65 (55.23) Of the 19 124 individuals with follow-up data and no history of neurological disorder, there were 14 349 weekly drinkers. ***P < 0.001, **P < 0.01, *P < 0.05. Of these, complete data were available for 13 342 (93%). Weekly drinkers had lower levels of socioeconomic depriv- and 75th percentiles in the alcohol distribution provided the ation, were more likely to be male and to hold an undergraduate best fit for RT (AIC = −13 691, F(12, 13 329) = 475.54, degree or higher. Non-drinkers were older and reported worse P < 0.001) and IIV (AIC = 24,985, F(12, 13 329) = 53.28, cognitive scores across time (Table S2). RT varied by age, gen- P < 0.001). This curvilinear solution was superior to a linear der, education, BMI, walking activity, alcohol consumption and model (RT: AIC = −13 687.48; IIV: AIC = 24 987.20), and smoking status (Table 1). IIV varied by age, gender, education, models with two (RT: AIC = −13 687.48; IIV: AIC = 24 987.20) BMI, walking activity and alcohol consumption (Table 1). and four knots (RT: AIC = −13 690; IIV: AIC = 24 986.14). RT was associated with baseline RT, alcohol consumption, Curvilinear modelling age and years between assessments, gender, education Preliminary analyses aimed to identify the most parsimoni- and smoking status (Table 2). IIV was associated with ous curvilinear model. Models with knots at the 25th, 50th baseline IIV, alcohol consumption, age, years between Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 307 Table 2 Restricted cubic spline regression model results with baseline measures predicting mean reaction time (RT) and intra-individual variation in reaction time (IIV) at follow-up (N = 13 342) Predictors Outcomes β (95% CI) P value RT IIV Mean reaction time baseline 0.548 (0.529 to 0.561) – <0.001 Intra-individual variation at baseline – 0.178 (0.161 to 0.196) <0.001 Alcohol use: spline 1 (linear effect ≤ 10 g/day) −0.048 (−0.105 to −0.030) −0.055 (−0.125 to −0.034) 0.001 0.001 Alcohol use: spline 2 (slope effect) 0.035 (0.007 to 0.059) 0.034 (0.002 to 0.064) 0.013 0.039 Age in years (at repeat assessment) 0.135 (0.072 to 0.107) 0.085 (0.072 to 0.107) <0.001 <0.001 Gender (reference: female) −0.023 (−0.037 to −0.008) −0.005 (−0.023 to 0.013) 0.002 0.558 Education (reference: no degree) −0.027 (−0.047 to −0.014) −0.031 (−0.047 to −0.014) <0.001 <0.001 Previous tobacco use (reference: non-smoker) 0.020 (0.005 to 0.033) 0.008 (−0.009 to 0.025) 0.008 0.336 2 2 Fit R = 0.35; AIC = −13 691 R = 0.05; AIC = 24 985 B, unstandardized regression coefficient; SE B, standard error for the unstandardized regression coefficient; β (95% CI), standardized regression coefficient and 95% confidence intervals. Note: results for smoking, walking, BMI and deprivation omitted as did not approach statistical significance for either outcome assessments and education (Table 2). RT decreased by the non-linear effect, i.e. potential harm incurred above 0.102 SD units (0.048 ms) for every additional 1 g/day 10 g/day (β = −0.070, 95% CI: −0.093, −0.039) (Table S4). increase in alcohol consumption up to 10 g/day, meaning A similar effect was found for IIV. cognitive performance improved. Cognitive performance declined as alcohol consumption increased beyond 10 g/ Discussion day (Fig. 1). The limitations of the cubic spline method make it difficult to quantify the potential for harm, not Main finding of this study least due to the relatively small numbers of heavy drinkers In 13 342 weekly drinkers drawn from UK Biobank, 5-year in the UK Biobank sample. change in mean RT and IIV in RT were found to have IIV decreased by −0.055 units for every additional 1 g/day curvilinear associations with alcohol consumption. Cognitive increase in alcohol consumption up to 10 g/day, also indicat- performance improved as alcohol consumption increased up ing better performance at low alcohol levels but not at higher to 10 g/day and then deteriorated as alcohol consumption consumption levels. Multivariable modelling made no material increased beyond 10 g/day. As individuals age, this deleterious difference to these associations (Table 3). Although adjust- effect of alcohol on cognitive performance became more ment for social covariates and baseline cognition marginally pronounced. attenuated the association, statistical significance was retained. The model was refitted with interaction terms for age, gender, education, deprivation, smoking status, BMI and What is already known on this topic baseline cognition. For RT, age made little difference to the The long-term impact of alcohol use on cognition is contro- linear effect, i.e. potential benefit incurred below 10 g/day versial. Observational epidemiologic data of alcohol con- (β = 0.104, 95% CI: 0.101, 0.176) but moderately increased sumption and the incidence of cognitive impairment and 1 Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 308 JOURNAL OF PUBLIC HEALTH dementia show reduced risk with light to moderate alcohol randomization study in a Western population. Holmes 19, 32, 33 35 consumption. Studies of alcohol consumption and et al. failed to find evidence for a ‘J’ shaped association cognitive decline have reported a reduced rate of decline in between alcohol and cardiovascular risk, a condition that light and moderate drinkers compared to abstainers and hea- shares many if the mechanisms underlying cerebrovascular 11, 34 vy drinkers. risk. However, evidence against a ‘J’ shaped relationship accu- mulates. A Mendelian randomization instrumental variable What this study adds analysis in a Chinese population compared alcohol con- This study presents data on cognitive change at an individual sumption according to ADHD2 variants known to be asso- level across a wide range of alcohol consumption, in contrast ciated with alcohol consumption. A per-allele association to data on cognitive differences between alcohol consump- with cognitive performance between ADHD2 variants was tion groups. It also omits abstainers. Both design features not found. Although this study was underpowered, and hea- ameliorate the impact of reverse causation on the findings. vy drinkers (by Western standards) were absent from the The use of a RT task, conducted in a standard and con- sample, the finding is consistent with a large-scale Mendelian trolled environment, provided a precise and reliable measure of cognitive performance. Treating alcohol consumption as a continuous variable facilitated a dose–response analysis. These findings do not resolve the debate over whether bene- fit may be attributed to low level alcohol consumption. If there is no benefit, these findings demonstrate that adjusting adequately for confounding on this question is extremely difficult. If there is benefit, the mechanisms remain obscure. Given uncertainties concerning the shape of the asso- ciation there is a strong case for changing the focus of the debate to harm rather than benefit. There is little question that alcohol is neurotoxic and that no cognitive benefit derives from high consumption levels. The find- ings reported here indicate that harm becomes apparent at levels of alcohol consumption lower than previously reported. Zuccalà et al., for example, argue for a protective Fig. 1 Curvilinear association between average daily alcohol use at baseline effect of wine up to 40 g/day for women and up to 80 g/day and mean reaction time (RT) at follow-up for the full sample, with 99% for men. Britton et al. suggest that the beneficial effects of confidence intervals (N = 13 342). Estimates are adjusted for age, years alcohol among UK middle-aged adults occur up to 34 g/day, between assessments, gender, education, Townsend deprivation score, whilst UK department of Health guidelines are that drinkers smoking status, BMI, walking activity and RT at baseline. should not consume more than 16 g/day to minimize the Table 3 Restricted cubic regression of cognitive performance on daily alcohol consumption Cognition Alcohol consumption Model 1: Adjusted for age Model 2: Adjusted for age + Model 3: Adjusted for age + covariates + splines β (95% CI) covariates β (95% CI) baseline cognition β (95% CI) P-value P-value P-value RT Linear (Spline 1: Linear −0.102 (−0.187 to −0.099) −0.098 (−0.183 to −0.093) −0.048 (−0.105 to −0.030) effect up to 10 g/day) <0.001 <0.001 <0.001 Non-linear (Spline 2: Slope 0.056 (0.023–0.083) 0.055 (0.021–0.083) 0.035 (0.007–0.059) effect) 0.001 0.001 0.013 IIV Linear (Spline 1: Linear −0.064 (−0.137 to −0.046) −0.064 (−0.138 to −0.045) −0.055 (−0.125 to −0.034) effect up to 10 g/day) <0.001 <0.001 0.001 Non-linear (Spline 2: Slope 0.042 (0.009–0.072) 0.038 (0.005–0.068) 0.034 (0.002–0.064) effect) 0.011 0.024 0.039 RT, mean reaction time; IIV, intra-individual variability in reaction time; β (95% CI), standardized regression coefficient and 95% confidence intervals. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 309 risk of alcohol to health. Our findings suggest that to pre- alcohol consumption may deliver an additional cognitive serve cognitive performance 10 g/day is a more appropriate burden. Future studies are needed to test the differential or upper limit. This would translate into not more than one UK joint role of average volume versus drinking pattern in order standard unit of alcohol each day. Our findings are of par- to better understand the nature of the relationship between ticular relevance to older individuals who demonstrated a alcohol use and cognitive decline. A core methodological greater rate of decline as alcohol consumption increased. limitation is the use of self-report measures of alcohol con- sumption. Objective measures of alcohol consumption, such as metabolomic markers, are required to improve the rigour Limitations of the study of alcohol intake assessment. Statistical limitations require consideration. The restricted spline method enables the inflexion point in the curve to be identified but assumes linearity before and after the inflex- Conclusions ion. This assumption is unlikely to make much impact below Current advice from the UK Department of Health is for the inflexion point due to the limited scale range (the prox- men and women to not consume more than 16 g of pure imity of zero), but it is a strong assumption above the inflex- alcohol per day (two units) on average. Findings reported ion point. The wide confidence intervals on the curve above here suggest that daily alcohol consumption above one unit the inflexion (Fig. 1) indicate that further work is required to is may have an adverse cognitive impact. Recommendations reduce uncertainty in the functional relationship between should be sensitive to this, especially among middle-aged cognitive performance and alcohol consumption above and older members of the population. 10 g/day. The ‘J’ shaped association reported here should be con- sidered critically. To reduce the ‘sick quitter’ effect abstainers Supplementary data were omitted. However, participants who may have only Supplementary data are available at the Journal of Public reduced alcohol intake for health reasons rather than quit, Health online. remain in the analysis. Selection bias may also be operating at high levels of alcohol consumption in that ‘bright boo- zers’, those with high alcohol intake and high cognitive per- Conflicts of interest formance, may be over represented at recruitment and The authors have declared that no competing interests exist. follow-up, thus deflating estimates of harm at high levels of consumption. The extent to which this effect is ameliorated by heavy drinkers disproportionately under reporting con- Funding sumption levels is also unknown. The effect of these selec- This work was supported by funds from the Economic and tion and reporting biases is likely to be complex, but unlikely Social Research Council, the Medical Research Council and to materially affect the conclusion that alcohol consumption Alcohol Research UK to the ELAStiC Project (ES/L015471/1). deleteriously affects cognitive performance at lower intake The research was also supported by the MRC Dementias levels than previously thought. Platform UK (MR/ L023784/1 and MR/009076/1). Chinmoy Sarkar was funded by The University of Hong Kong’s Research Further work Assistant Professorship grant. The extent to which the association of alcohol with cogni- tion reported here may be generalized to other cognitive Data sharing domains is interesting. Due to its precision of measurement, RT is likely to be the most sensitive of the cognitive mea- This research has been conducted using the UK Biobank sures used in UK Biobank. Higher cognitive domains, such Resource under Application Number 14 935. The data as reasoning or memory, may provide greater opportunity reported in this article are available via application directly to for compensatory mechanisms to mask neurological impacts the UK Biobank. of alcohol consumption, particularly at low intakes. Nevertheless, the literature suggests the effect of alcohol on 19, 38, 39 References cognition is broad. The issue of alcohol consumption pattern is not 1 Crews FT, Nixon K. Mechanisms of neurodegeneration and regen- addressed in these data and single session heavy episodic eration in alcoholism. Alcohol Alcohol 2009;44(2):115–27. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 310 JOURNAL OF PUBLIC HEALTH 2 Crews FT, Collins MA, Dlugos C et al. Alcohol‐induced neurode- 19 Anstey KJ, Mack HA, Cherbuin N. Alcohol consumption as a risk generation: when, where and why? Alcohol Clin Exp Res 2004;28(2): factor for dementia and cognitive decline: meta-analysis of prospect- 350–64. ive studies. Am J Geriatr Psychiatry 2009;17(7):542–55. 3 Goodlett CR, Horn KH, Zhou FC. Alcohol teratogenesis: mechan- 20 Neafsey EJ, Collins MA. Moderate alcohol consumption and cogni- isms of damage and strategies for intervention. Exp Biol Med 2005; tive risk. Neuropsychiatr Dis Treat 2011;7:465–84. 230(6):394–406. 21 Peters R, Peters J, Warner J et al. Alcohol, dementia and cognitive 4 Rao R, Draper B. Alcohol-related brain damage in older people. decline in the elderly: a systematic review. Age Ageing 2008;37(5): Lancet Psychiatry 2015;2(8):674–5. 505–12. 5 Statistics OfN. Deaths Registered in England and Wales (Series DR): 22 Yeung SA, Jiang C, Cheng K et al. Evaluation of moderate alcohol 2015. 2016 https://www.ons.gov.uk/peoplepopulationandcommunity/ use and cognitive function among men using a Mendelian random- birthsdeathsandmarriages/deaths/bulletins/deathsregisteredinengland ization design in the Guangzhou biobank cohort study. Am J andwalesseriesdr/2015. Epidemiol 2012;175(10):1021–8. 6 Prince M, Bryce R, Albanese E et al. The global prevalence of 23 Sudlow C, Gallacher J, Allen N et al. UK biobank: an open access dementia: a systematic review and metaanalysis. Alzheimers Dement resource for identifying the causes of a wide range of complex dis- 2013;9(1):63–75 e2. eases of middle and old age. PLoS Med 2015;12(3):e1001779. 7 Winblad B, Amouyel P, Andrieu S et al. Defeating Alzheimer’s dis- 24 Ganguli M, Vander Bilt J, Saxton JA et al. Alcohol consumption and ease and other dementias: a priority for European science and soci- cognitive function in late life: a longitudinal community study. ety. Lancet Neurol 2016;15(5):455–532. Neurology 2005;65(8):1210–7. 8 Eurostat. Causes of Death Statistics. 2016 http://ec.europa.eu/ 25 Stampfer MJ, Kang JH, Chen J et al. Effects of moderate alcohol eurostat/statistics-explained/index.php/Causes_of_death_statistics. consumption on cognitive function in women. N Engl J Med 2005; 352(3):245–53. 9 Andreasson S. Alcohol and J-shaped curves. Alcohol Clin Exp Res 1998;22(7 Suppl):359S–64S. 26 KrahnD,FreeseJ,HauserR et al. Alcohol use and cognition at mid-life: the importance of adjusting for baseline cognitive ability 10 Anttila T, Helkala EL, Viitanen M et al. Alcohol drinking in middle and educational attainment. Alcohol Clin Exp Res 2003;27(7): age and subsequent risk of mild cognitive impairment and dementia 1162–6. in old age: a prospective population based study. Br Med J 2004;329 (7465):539. 27 Townsend P. Deprivation. J Soc Policy 1987;16(2):125–46. 11 Rodgers B, Windsor TD, Anstey KJ et al. Non-linear relationships 28 Finglas P, Roe M, Pinchen H et al. McCance and Widdowson’s The between cognitive function and alcohol consumption in young, Composition of Foods, 7th summary edn. Cambridge: Royal Society of middle-aged and older adults: the PATH Through Life Project. Chemistry, 2014. Addiction 2005;100(9):1280–90. 29 Lyall DM, Cullen B, Allerhand M et al. Cognitive test scores in UK 12 d’Orsi E, Xavier AJ, Steptoe A et al. Socioeconomic and lifestyle Biobank: data reduction in 480,416 participants and longitudinal sta- factors related to instrumental activity of daily living dynamics: bility in 20,346 participants. PLoS One 2016;11(4):e0154222. results from the English Longitudinal Study of Ageing. J Am Geriatr 30 Dykiert D, Der G, Starr JM et al. Age differences in intra-individual Soc 2014;62(9):1630–9. variability in simple and choice reaction time: systematic review and 13 Solfrizzi V, D’Introno A, Colacicco AM et al. Alcohol consumption, meta-analysis. PLoS One 2012;7(10):e45759. mild cognitive impairment, and progression to dementia. Neurology 31 Desquilbet L, Mariotti F. Dose-response analyses using restricted 2007;68(21):1790–9. cubic spline functions in public health research. Stat Med 2010;29(9): 14 Weyerer S, Schaufele M, Wiese B et al. Current alcohol consumption 1037–57. and its relationship to incident dementia: results from a 3-year fol- 32 Cao L, Tan L, Wang H-F et al. Dietary patterns and risk of demen- low-up study among primary care attenders aged 75 years and older. tia: a systematic review and meta-analysis of cohort studies. Mol Age Ageing 2011;40(4):456–63. Neurobiol 2016;53(9):6144–54. 15 Zanjani F, Downer BG, Kruger TM et al. Alcohol effects on cogni- 33 Hersi M, Irvine B, Gupta P et al. Risk factors associated with the tive change in middle-aged and older adults. Aging Ment Health 2013; onset and progression of Alzheimer’s disease: a systematic review of 17(1):12–23. the evidence. Neurotoxicology 2017;61:143–87. 16 Carrigan N, Barkus E. A systematic review of the relationship 34 Sabia S, Elbaz A, Britton A et al. Alcohol consumption and cogni- between psychological disorders or substance use and self-reported tive decline in early old age. Neurology 2014;82(4):332–9. cognitive failures. Cogn Neuropsychiatry 2016;21(6):539–64. 35 Holmes MV, Dale CE, Zuccolo L et al. Association between alcohol 17 Panza F, Capurso C, D’Introno A et al. Alcohol drinking, cognitive and cardiovascular disease: Mendelian randomisation analysis based functions in older age, predementia, and dementia syndromes. on individual participant data. Br Med J 2014;349:g4164. J Alzheimers Dis 2009;17(1):7–31. 36 Zuccala G, Onder G, Pedone C et al. Dose-related impact of 18 Sachdeva A, Chandra M, Choudhary M et al. Alcohol-related alcohol consumption on cognitive function in advanced age: dementia and neurocognitive impairment: a review study. Int J High results of a multicenter survey. Alcohol Clin Exp Res 2001;25(12): Risk Behav Addict 2016;5(3):e27976. 1743–8. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 311 37 Britton A, Singh-Manoux A, Marmot M. Alcohol consumption and 40 Zheng Y, Yu B, Alexander D et al. Metabolomic patterns and alco- cognitive function in the Whitehall II Study. Am J Epidemiol 2004; hol consumption in African Americans in the Atherosclerosis Risk 160(3):240–7. in Communities Study. Am J Clin Nutr 2014;99(6):1470–8. 38 Stavro K, Pelletier J, Potvin S. Widespread and sustained cognitive 41 Alcohol Policy Team DoH. How to Keep Health Risks From Drinking deficits in alcoholism: a meta-analysis. Addict Biol 2013;18(2): Alcohol to a Low Level Government Response to the Public Consultation. 203–13. Department of Health, 2016 https://www.gov.uk/government/uploads/ system/uploads/attachment_data/file/545911/GovResponse2. 39 Bernardin F, Maheut-Bosser A, Paille F. Cognitive impairments in pdf. alcohol-dependent subjects. Front Psychiatry 2014;5:78. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Public Health Oxford University Press

The relationship between alcohol use and long-term cognitive decline in middle and late life: a longitudinal analysis using UK Biobank

Download PDF
8 pages

Loading next page...
 
/lp/ou_press/the-relationship-between-alcohol-use-and-long-term-cognitive-decline-w3bUA2XvxF

References (85)

Publisher
Oxford University Press
Copyright
Copyright © 2022 Faculty of Public Health
ISSN
1741-3842
eISSN
1741-3850
DOI
10.1093/pubmed/fdx186
Publisher site
See Article on Publisher Site

Abstract

Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 Journal of Public Health | Vol. 40, No. 2, pp. 304–311 | doi:10.1093/pubmed/fdx186 | Advance Access Publication January 9, 2018 The relationship between alcohol use and long-term cognitive decline in middle and late life: a longitudinal analysis using UK Biobank 1,2 3 4 5 Giovanni Piumatti , Simon C. Moore , Damon M. Berridge , Chinmoy Sarkar , John Gallacher Department of Psychiatry, University of Oxford, Oxford, UK Unit of Development and Research in Medical Education (UDREM), Faculty of Medicine, University of Geneva, Geneva, Switzerland Violence & Society Research Group, School of Dentistry, Cardiff University, Cardiff, UK Farr Institute—CIPHER, Swansea University Medical School, Swansea UK Healthy High Density Cities Lab, HKUrbanLab, The University of Hong Kong, Hong Kong, China Address correspondence to Simon C. Moore, E-mail: MooreSC2@cardiff.ac.uk ABSTRACT Background Using UK Biobank data, this study sought to explain the causal relationship between alcohol intake and cognitive decline in middle and older aged populations. Methods Data from 13 342 men and women, aged between 40 and 73 years were used in regression analysis that tested the functional relationship and impact of alcohol on cognitive performance. Performance was measured using mean reaction time (RT) and intra-individual variation (IIV) in RT, collected in response to a perceptual matching task. Covariates included body mass index, physical activity, tobacco use, socioeconomic status, education and baseline cognitive function. Results A restricted cubic spline regression with three knots showed how the linear (β = −0.048, 95% CI: −0.105 to −0.030) and non-linear effects (β = 0.035, 95% CI: 0.007–0.059) of alcohol use on mean RT and IIV in RT (β = −0.055, 95% CI: −0.125 to −0.034; β = 0.034, 2 1 2 95% CI: 0.002–0.064) were significant adjusting for covariates. Cognitive function declined as alcohol use increased beyond 10 g/day. Decline was more apparent as age increased. Conclusions The relationship between alcohol use and cognitive function is non-linear. Consuming more than one UK standard unit of alcohol per day is detrimental to cognitive performance and is more pronounced in older populations. Keywords alcohol, alcohol consumption, public health The suggested curvilinear association between alcohol Introduction 16–18 and cognition, however, is controversial. Recent reviews The neurodegenerative effects of excessive alcohol con- 19–21 and meta-analyses indicate that there is little consensus 1–4 sumption are well documented. Alzheimer’s disease and on the level of alcohol consumption at which the harmful dementia have replaced ischaemic heart disease as the lead- effects of alcohol on cognition emerge. Furthermore, a ing cause of death in England and Wales, and death rates Mendelian Randomization study of alcohol and cognitive 6–8 for neurological disease are increasing worldwide. A lim- ited number of studies suggest a J- or U-shaped relationship between the volume of alcohol consumed and the long-term Giovanni Piumatti, Postdoctoral Researcher 9–11 cognitive decline, suggesting light to moderate alcohol Simon C. Moore, Professor of Public Health Research Damon M. Berridge, Professor of Applied Statistics consumption is a positive predictor of health status in older Chinmoy Sarkar, Assistant Professor adults, protects cognition and may reduce the risk of 13–15 John Gallacher, Professor of Cognitive Health dementia in later life. © The Author(s) 2018. Published by Oxford University Press on behalf of Faculty of Public Health. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), 304 which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 305 performance found evidence of benefit from reduced alco- bottle of wine’; ‘there are 25 standard measures in a normal hol intake at all levels of self-reported consumption. sized bottle’). Respondents who declared that they drank alco- The current study examined the shape of the association hol ‘one to three times a month’ or on ‘special occasions only’ between alcohol consumption and change in cognitive per- (henceforth monthly drinkers) were also asked to record how formance. Data were drawn from UK Biobank, a large many drinks they consumed on average each month. However, cohort of middle and older aged adults. Respondents who these questions were not included at baseline and therefore consumed alcohol at least once a week or more frequently more than half of the sample of monthly drinkers at baseline were eligible for inclusion to reduce selection and reporting were not assessed. Accordingly, only participants who declared biases. A reaction time (RT) task was used as a robust test that they drank at least once a week (henceforth, weekly drin- of central processing speed. Cognitive performance was kers) were included in primary analyses. measured using mean RT and IIV in RT. Cognitive performance Cognitive performance was assessed using a ‘stop-go’ RT task Methods in which participants were shown two cards simultaneously on Sample a computer screen. Each card had a symbol on it and partici- Between 2006 and 2010, a heterogeneous population sample pants were asked to respond as quickly as possible, using a of 502 649 adults aged 40–73 years participated in the UK button-box, if both symbols matched. RT, from the presenta- Biobank prospective cohort study at 22 research centres tion of the cards to their press of the button, was recorded in located across the UK. Participants were registered with milliseconds (ms). Each participant was presented with 12 the UK National Health Service (NHS) and lived within a pairs of cards, the first five of which were training sets and radius of 40 km from one of the research centres. Self- data from these trials were discarded. Of the seven test trials, reported data were collected via touch screen questionnaires cards with matching symbols were presented on four occa- and interview. Information on the assessment procedure, sions selected at random. A demonstration of this test is avail- protocol and information on data access is available online able online (biobank.ctsu.ox.ac.uk/crystal/videos/snap.swf). (www.ukbiobank.ac.uk). For the purposes of estimating regression dilution, 20 346 individuals underwent a repeat Covariates assessment five years after their initial assessment. Data The effects of alcohol use on cognitive performance differ by from these respondents are used in the current longitudinal 24, 25 26 26 20 gender, education, past performance and age. These analysis. Individuals were omitted from the analysis if they variables were included as covariates in the present study disclosed a history of neurological disorder at either baseline or alongside deprivation as measured by the Townsend score, follow-up (Table S1), leaving 19 124 eligible participants. The physical activity assessed as walking activity, body mass index UK Biobank study was approved by the North West Multi- (BMI) and smoking status. Centre Research Ethics Committee (reference number 06/ MRE08/65). All participants gave written, informed consent. Data analysis Measures Alcohol consumption in grams per day was calculated by Alcohol use multiplying the average number of alcoholic drinks consumed Alcohol consumption was measured using the question each week by the average grams of alcohol contained in ‘about how often do you drink alcohol?’ Available responses each type of drink, determined using the UK Food Standard were ‘daily or almost daily’, ‘threeoffour times a week’, Agency’s guidelines. The total was then divided by seven to ‘once or twice a week’, ‘one to three times a month’, ‘special provide mean daily alcohol consumption. Alcohol consumption occasions only’, ‘never’ and ‘prefer not to answer’. Respondents was positively skew and log transformed. who drank alcohol once a week or more frequently were asked Consistent with established methods, RTs < 50 ms, to record how many alcoholic drinks they consumed on average indicating anticipation of the stimulus, were discarded as each week from a list of common alcoholic beverages (red and were RTs > 2 s, as the target stimulus had been withdrawn white wine, champagne, beer and cider, spirits and liquors, forti- at this point. RT was calculated as the arithmetic mean of fied wine, and other alcoholic drinks), or to respond ‘do not completed test trials. Intra-individual variation (IIV) was cal- know’ or ‘prefer not to answer’. Volumes were specified when culated as the standard deviation of each participant’sRTs referring to beverages (e.g. ‘there are six glasses in an average over the test trials. Participants with only one valid score at Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 306 JOURNAL OF PUBLIC HEALTH Table 1 Differences in mean reaction time (RT) and intra-individual baseline or follow-up were omitted. RT and IIV showed variation in reaction time (IIV) at follow-up according to socioeconomic log–normal distributions and were natural log transformed. and lifestyle factors. Values are means and (standard deviations) For the covariates, educational attainment was included as a binary variable (with or without a degree). BMI was Variable RT (ms) IIV included as two binary variables (normal <24.9 kg/m com- 2 2 pared to overweight 25–29.9 kg/m and obese ≥30 kg/m ) Age (years) as was smoking status (non-smoker compared to previous 40–52 512.76 (95.25)*** 68.53 (49.38)*** smoker and current smoker). Deprivation quintiles were 53–59 551.78 (102.18) 77.15 (55.23) included as a continuous variable, as were age and the time 60–63 571.90 (110.04) 82.27 (58.56) between baseline and follow-up assessments. Walking activ- 64+ 594.67 (114.15) 88.95 (66.03) ity was included as the number of days participants walked Gender for more than 10 min each week. Preliminary analyses found Females 562.51 (109.01)*** 79.80 (59.02) Males 548.80 (108.74) 77.40 (55.70) missing data was minimal (2.8%). Education Non-linearity in the alcohol–cognition relationship was No degree 564.44 (112.48)*** 81.25 (59.22)*** investigated using restricted cubic splines. A restricted cubic Degree 544.74 (103.67) 75.34 (54.98) spline is a cubic spline function with an additional constraint Deprivation quintile of linearity before the first knot and after the last knot. Least 553.77 (102.27)*** 78.54 (56.85)*** The number of knots was determined by examining the dis- 2 555.94 (109.68) 78.69 (57.23) tribution of average daily alcohol use in the sample, with the 3 555.49 (109.33) 79.72 (59.46) aim of locating boundaries between equal-sized categories 4 558.15 (109.76) 78.54 (57.59) and by comparing the Akaike Information Criteria (AIC) Most 558.80 (113.94) 77.65 (56.12) goodness of fit statistics across models. Alcohol intake Staged multivariable modelling, the regression of follow- Non-drinkers 574.79 (119.88)** 83.99 (67.08)** up RT and IIV on baseline alcohol consumption, began with Monthly 558.45 (112.07) 78.18 (56.34) Weekly 553.84 (107.35) 78.38 (56.98) an adjustment for age to establish the fundamental associ- Body mass index ation (model 1). Adjustment for social confounding was ≤Normal 555.57 (109.10)*** 78.72 (58.20)** made by further conditioning on lifestyle and background Overweight 553.86 (108.11) 77.45 (55.24) (model 2). The influence of baseline cognition was then Obese 561.73 (111.56) 81.26 (60.55) taken into account (model 3). Finally, interactions effects Walking tertiles (days/week) were included (model 4). All analyses were performed using 0–4 550.89 (106.82)* 76.87 (54.65)** Stata 14 (StataCorp. 2015. Stata Statistical Software: Release 5–6 555.28 (108.57) 78.35 (56.83) 14. College Station, TX: StataCorp LP). 7 559.50 (110.27) 80.02 (59.64) Smoking Non-smokers 552.91 (108.06)*** 78.22 (57.40) Results Previous smokers 561.07 (110.88) 79.48 (57.78) Current smokers 554.39 (107.61) 77.65 (55.23) Of the 19 124 individuals with follow-up data and no history of neurological disorder, there were 14 349 weekly drinkers. ***P < 0.001, **P < 0.01, *P < 0.05. Of these, complete data were available for 13 342 (93%). Weekly drinkers had lower levels of socioeconomic depriv- and 75th percentiles in the alcohol distribution provided the ation, were more likely to be male and to hold an undergraduate best fit for RT (AIC = −13 691, F(12, 13 329) = 475.54, degree or higher. Non-drinkers were older and reported worse P < 0.001) and IIV (AIC = 24,985, F(12, 13 329) = 53.28, cognitive scores across time (Table S2). RT varied by age, gen- P < 0.001). This curvilinear solution was superior to a linear der, education, BMI, walking activity, alcohol consumption and model (RT: AIC = −13 687.48; IIV: AIC = 24 987.20), and smoking status (Table 1). IIV varied by age, gender, education, models with two (RT: AIC = −13 687.48; IIV: AIC = 24 987.20) BMI, walking activity and alcohol consumption (Table 1). and four knots (RT: AIC = −13 690; IIV: AIC = 24 986.14). RT was associated with baseline RT, alcohol consumption, Curvilinear modelling age and years between assessments, gender, education Preliminary analyses aimed to identify the most parsimoni- and smoking status (Table 2). IIV was associated with ous curvilinear model. Models with knots at the 25th, 50th baseline IIV, alcohol consumption, age, years between Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 307 Table 2 Restricted cubic spline regression model results with baseline measures predicting mean reaction time (RT) and intra-individual variation in reaction time (IIV) at follow-up (N = 13 342) Predictors Outcomes β (95% CI) P value RT IIV Mean reaction time baseline 0.548 (0.529 to 0.561) – <0.001 Intra-individual variation at baseline – 0.178 (0.161 to 0.196) <0.001 Alcohol use: spline 1 (linear effect ≤ 10 g/day) −0.048 (−0.105 to −0.030) −0.055 (−0.125 to −0.034) 0.001 0.001 Alcohol use: spline 2 (slope effect) 0.035 (0.007 to 0.059) 0.034 (0.002 to 0.064) 0.013 0.039 Age in years (at repeat assessment) 0.135 (0.072 to 0.107) 0.085 (0.072 to 0.107) <0.001 <0.001 Gender (reference: female) −0.023 (−0.037 to −0.008) −0.005 (−0.023 to 0.013) 0.002 0.558 Education (reference: no degree) −0.027 (−0.047 to −0.014) −0.031 (−0.047 to −0.014) <0.001 <0.001 Previous tobacco use (reference: non-smoker) 0.020 (0.005 to 0.033) 0.008 (−0.009 to 0.025) 0.008 0.336 2 2 Fit R = 0.35; AIC = −13 691 R = 0.05; AIC = 24 985 B, unstandardized regression coefficient; SE B, standard error for the unstandardized regression coefficient; β (95% CI), standardized regression coefficient and 95% confidence intervals. Note: results for smoking, walking, BMI and deprivation omitted as did not approach statistical significance for either outcome assessments and education (Table 2). RT decreased by the non-linear effect, i.e. potential harm incurred above 0.102 SD units (0.048 ms) for every additional 1 g/day 10 g/day (β = −0.070, 95% CI: −0.093, −0.039) (Table S4). increase in alcohol consumption up to 10 g/day, meaning A similar effect was found for IIV. cognitive performance improved. Cognitive performance declined as alcohol consumption increased beyond 10 g/ Discussion day (Fig. 1). The limitations of the cubic spline method make it difficult to quantify the potential for harm, not Main finding of this study least due to the relatively small numbers of heavy drinkers In 13 342 weekly drinkers drawn from UK Biobank, 5-year in the UK Biobank sample. change in mean RT and IIV in RT were found to have IIV decreased by −0.055 units for every additional 1 g/day curvilinear associations with alcohol consumption. Cognitive increase in alcohol consumption up to 10 g/day, also indicat- performance improved as alcohol consumption increased up ing better performance at low alcohol levels but not at higher to 10 g/day and then deteriorated as alcohol consumption consumption levels. Multivariable modelling made no material increased beyond 10 g/day. As individuals age, this deleterious difference to these associations (Table 3). Although adjust- effect of alcohol on cognitive performance became more ment for social covariates and baseline cognition marginally pronounced. attenuated the association, statistical significance was retained. The model was refitted with interaction terms for age, gender, education, deprivation, smoking status, BMI and What is already known on this topic baseline cognition. For RT, age made little difference to the The long-term impact of alcohol use on cognition is contro- linear effect, i.e. potential benefit incurred below 10 g/day versial. Observational epidemiologic data of alcohol con- (β = 0.104, 95% CI: 0.101, 0.176) but moderately increased sumption and the incidence of cognitive impairment and 1 Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 308 JOURNAL OF PUBLIC HEALTH dementia show reduced risk with light to moderate alcohol randomization study in a Western population. Holmes 19, 32, 33 35 consumption. Studies of alcohol consumption and et al. failed to find evidence for a ‘J’ shaped association cognitive decline have reported a reduced rate of decline in between alcohol and cardiovascular risk, a condition that light and moderate drinkers compared to abstainers and hea- shares many if the mechanisms underlying cerebrovascular 11, 34 vy drinkers. risk. However, evidence against a ‘J’ shaped relationship accu- mulates. A Mendelian randomization instrumental variable What this study adds analysis in a Chinese population compared alcohol con- This study presents data on cognitive change at an individual sumption according to ADHD2 variants known to be asso- level across a wide range of alcohol consumption, in contrast ciated with alcohol consumption. A per-allele association to data on cognitive differences between alcohol consump- with cognitive performance between ADHD2 variants was tion groups. It also omits abstainers. Both design features not found. Although this study was underpowered, and hea- ameliorate the impact of reverse causation on the findings. vy drinkers (by Western standards) were absent from the The use of a RT task, conducted in a standard and con- sample, the finding is consistent with a large-scale Mendelian trolled environment, provided a precise and reliable measure of cognitive performance. Treating alcohol consumption as a continuous variable facilitated a dose–response analysis. These findings do not resolve the debate over whether bene- fit may be attributed to low level alcohol consumption. If there is no benefit, these findings demonstrate that adjusting adequately for confounding on this question is extremely difficult. If there is benefit, the mechanisms remain obscure. Given uncertainties concerning the shape of the asso- ciation there is a strong case for changing the focus of the debate to harm rather than benefit. There is little question that alcohol is neurotoxic and that no cognitive benefit derives from high consumption levels. The find- ings reported here indicate that harm becomes apparent at levels of alcohol consumption lower than previously reported. Zuccalà et al., for example, argue for a protective Fig. 1 Curvilinear association between average daily alcohol use at baseline effect of wine up to 40 g/day for women and up to 80 g/day and mean reaction time (RT) at follow-up for the full sample, with 99% for men. Britton et al. suggest that the beneficial effects of confidence intervals (N = 13 342). Estimates are adjusted for age, years alcohol among UK middle-aged adults occur up to 34 g/day, between assessments, gender, education, Townsend deprivation score, whilst UK department of Health guidelines are that drinkers smoking status, BMI, walking activity and RT at baseline. should not consume more than 16 g/day to minimize the Table 3 Restricted cubic regression of cognitive performance on daily alcohol consumption Cognition Alcohol consumption Model 1: Adjusted for age Model 2: Adjusted for age + Model 3: Adjusted for age + covariates + splines β (95% CI) covariates β (95% CI) baseline cognition β (95% CI) P-value P-value P-value RT Linear (Spline 1: Linear −0.102 (−0.187 to −0.099) −0.098 (−0.183 to −0.093) −0.048 (−0.105 to −0.030) effect up to 10 g/day) <0.001 <0.001 <0.001 Non-linear (Spline 2: Slope 0.056 (0.023–0.083) 0.055 (0.021–0.083) 0.035 (0.007–0.059) effect) 0.001 0.001 0.013 IIV Linear (Spline 1: Linear −0.064 (−0.137 to −0.046) −0.064 (−0.138 to −0.045) −0.055 (−0.125 to −0.034) effect up to 10 g/day) <0.001 <0.001 0.001 Non-linear (Spline 2: Slope 0.042 (0.009–0.072) 0.038 (0.005–0.068) 0.034 (0.002–0.064) effect) 0.011 0.024 0.039 RT, mean reaction time; IIV, intra-individual variability in reaction time; β (95% CI), standardized regression coefficient and 95% confidence intervals. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 309 risk of alcohol to health. Our findings suggest that to pre- alcohol consumption may deliver an additional cognitive serve cognitive performance 10 g/day is a more appropriate burden. Future studies are needed to test the differential or upper limit. This would translate into not more than one UK joint role of average volume versus drinking pattern in order standard unit of alcohol each day. Our findings are of par- to better understand the nature of the relationship between ticular relevance to older individuals who demonstrated a alcohol use and cognitive decline. A core methodological greater rate of decline as alcohol consumption increased. limitation is the use of self-report measures of alcohol con- sumption. Objective measures of alcohol consumption, such as metabolomic markers, are required to improve the rigour Limitations of the study of alcohol intake assessment. Statistical limitations require consideration. The restricted spline method enables the inflexion point in the curve to be identified but assumes linearity before and after the inflex- Conclusions ion. This assumption is unlikely to make much impact below Current advice from the UK Department of Health is for the inflexion point due to the limited scale range (the prox- men and women to not consume more than 16 g of pure imity of zero), but it is a strong assumption above the inflex- alcohol per day (two units) on average. Findings reported ion point. The wide confidence intervals on the curve above here suggest that daily alcohol consumption above one unit the inflexion (Fig. 1) indicate that further work is required to is may have an adverse cognitive impact. Recommendations reduce uncertainty in the functional relationship between should be sensitive to this, especially among middle-aged cognitive performance and alcohol consumption above and older members of the population. 10 g/day. The ‘J’ shaped association reported here should be con- sidered critically. To reduce the ‘sick quitter’ effect abstainers Supplementary data were omitted. However, participants who may have only Supplementary data are available at the Journal of Public reduced alcohol intake for health reasons rather than quit, Health online. remain in the analysis. Selection bias may also be operating at high levels of alcohol consumption in that ‘bright boo- zers’, those with high alcohol intake and high cognitive per- Conflicts of interest formance, may be over represented at recruitment and The authors have declared that no competing interests exist. follow-up, thus deflating estimates of harm at high levels of consumption. The extent to which this effect is ameliorated by heavy drinkers disproportionately under reporting con- Funding sumption levels is also unknown. The effect of these selec- This work was supported by funds from the Economic and tion and reporting biases is likely to be complex, but unlikely Social Research Council, the Medical Research Council and to materially affect the conclusion that alcohol consumption Alcohol Research UK to the ELAStiC Project (ES/L015471/1). deleteriously affects cognitive performance at lower intake The research was also supported by the MRC Dementias levels than previously thought. Platform UK (MR/ L023784/1 and MR/009076/1). Chinmoy Sarkar was funded by The University of Hong Kong’s Research Further work Assistant Professorship grant. The extent to which the association of alcohol with cogni- tion reported here may be generalized to other cognitive Data sharing domains is interesting. Due to its precision of measurement, RT is likely to be the most sensitive of the cognitive mea- This research has been conducted using the UK Biobank sures used in UK Biobank. Higher cognitive domains, such Resource under Application Number 14 935. The data as reasoning or memory, may provide greater opportunity reported in this article are available via application directly to for compensatory mechanisms to mask neurological impacts the UK Biobank. of alcohol consumption, particularly at low intakes. Nevertheless, the literature suggests the effect of alcohol on 19, 38, 39 References cognition is broad. The issue of alcohol consumption pattern is not 1 Crews FT, Nixon K. Mechanisms of neurodegeneration and regen- addressed in these data and single session heavy episodic eration in alcoholism. Alcohol Alcohol 2009;44(2):115–27. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 310 JOURNAL OF PUBLIC HEALTH 2 Crews FT, Collins MA, Dlugos C et al. Alcohol‐induced neurode- 19 Anstey KJ, Mack HA, Cherbuin N. Alcohol consumption as a risk generation: when, where and why? Alcohol Clin Exp Res 2004;28(2): factor for dementia and cognitive decline: meta-analysis of prospect- 350–64. ive studies. Am J Geriatr Psychiatry 2009;17(7):542–55. 3 Goodlett CR, Horn KH, Zhou FC. Alcohol teratogenesis: mechan- 20 Neafsey EJ, Collins MA. Moderate alcohol consumption and cogni- isms of damage and strategies for intervention. Exp Biol Med 2005; tive risk. Neuropsychiatr Dis Treat 2011;7:465–84. 230(6):394–406. 21 Peters R, Peters J, Warner J et al. Alcohol, dementia and cognitive 4 Rao R, Draper B. Alcohol-related brain damage in older people. decline in the elderly: a systematic review. Age Ageing 2008;37(5): Lancet Psychiatry 2015;2(8):674–5. 505–12. 5 Statistics OfN. Deaths Registered in England and Wales (Series DR): 22 Yeung SA, Jiang C, Cheng K et al. Evaluation of moderate alcohol 2015. 2016 https://www.ons.gov.uk/peoplepopulationandcommunity/ use and cognitive function among men using a Mendelian random- birthsdeathsandmarriages/deaths/bulletins/deathsregisteredinengland ization design in the Guangzhou biobank cohort study. Am J andwalesseriesdr/2015. Epidemiol 2012;175(10):1021–8. 6 Prince M, Bryce R, Albanese E et al. The global prevalence of 23 Sudlow C, Gallacher J, Allen N et al. UK biobank: an open access dementia: a systematic review and metaanalysis. Alzheimers Dement resource for identifying the causes of a wide range of complex dis- 2013;9(1):63–75 e2. eases of middle and old age. PLoS Med 2015;12(3):e1001779. 7 Winblad B, Amouyel P, Andrieu S et al. Defeating Alzheimer’s dis- 24 Ganguli M, Vander Bilt J, Saxton JA et al. Alcohol consumption and ease and other dementias: a priority for European science and soci- cognitive function in late life: a longitudinal community study. ety. Lancet Neurol 2016;15(5):455–532. Neurology 2005;65(8):1210–7. 8 Eurostat. Causes of Death Statistics. 2016 http://ec.europa.eu/ 25 Stampfer MJ, Kang JH, Chen J et al. Effects of moderate alcohol eurostat/statistics-explained/index.php/Causes_of_death_statistics. consumption on cognitive function in women. N Engl J Med 2005; 352(3):245–53. 9 Andreasson S. Alcohol and J-shaped curves. Alcohol Clin Exp Res 1998;22(7 Suppl):359S–64S. 26 KrahnD,FreeseJ,HauserR et al. Alcohol use and cognition at mid-life: the importance of adjusting for baseline cognitive ability 10 Anttila T, Helkala EL, Viitanen M et al. Alcohol drinking in middle and educational attainment. Alcohol Clin Exp Res 2003;27(7): age and subsequent risk of mild cognitive impairment and dementia 1162–6. in old age: a prospective population based study. Br Med J 2004;329 (7465):539. 27 Townsend P. Deprivation. J Soc Policy 1987;16(2):125–46. 11 Rodgers B, Windsor TD, Anstey KJ et al. Non-linear relationships 28 Finglas P, Roe M, Pinchen H et al. McCance and Widdowson’s The between cognitive function and alcohol consumption in young, Composition of Foods, 7th summary edn. Cambridge: Royal Society of middle-aged and older adults: the PATH Through Life Project. Chemistry, 2014. Addiction 2005;100(9):1280–90. 29 Lyall DM, Cullen B, Allerhand M et al. Cognitive test scores in UK 12 d’Orsi E, Xavier AJ, Steptoe A et al. Socioeconomic and lifestyle Biobank: data reduction in 480,416 participants and longitudinal sta- factors related to instrumental activity of daily living dynamics: bility in 20,346 participants. PLoS One 2016;11(4):e0154222. results from the English Longitudinal Study of Ageing. J Am Geriatr 30 Dykiert D, Der G, Starr JM et al. Age differences in intra-individual Soc 2014;62(9):1630–9. variability in simple and choice reaction time: systematic review and 13 Solfrizzi V, D’Introno A, Colacicco AM et al. Alcohol consumption, meta-analysis. PLoS One 2012;7(10):e45759. mild cognitive impairment, and progression to dementia. Neurology 31 Desquilbet L, Mariotti F. Dose-response analyses using restricted 2007;68(21):1790–9. cubic spline functions in public health research. Stat Med 2010;29(9): 14 Weyerer S, Schaufele M, Wiese B et al. Current alcohol consumption 1037–57. and its relationship to incident dementia: results from a 3-year fol- 32 Cao L, Tan L, Wang H-F et al. Dietary patterns and risk of demen- low-up study among primary care attenders aged 75 years and older. tia: a systematic review and meta-analysis of cohort studies. Mol Age Ageing 2011;40(4):456–63. Neurobiol 2016;53(9):6144–54. 15 Zanjani F, Downer BG, Kruger TM et al. Alcohol effects on cogni- 33 Hersi M, Irvine B, Gupta P et al. Risk factors associated with the tive change in middle-aged and older adults. Aging Ment Health 2013; onset and progression of Alzheimer’s disease: a systematic review of 17(1):12–23. the evidence. Neurotoxicology 2017;61:143–87. 16 Carrigan N, Barkus E. A systematic review of the relationship 34 Sabia S, Elbaz A, Britton A et al. Alcohol consumption and cogni- between psychological disorders or substance use and self-reported tive decline in early old age. Neurology 2014;82(4):332–9. cognitive failures. Cogn Neuropsychiatry 2016;21(6):539–64. 35 Holmes MV, Dale CE, Zuccolo L et al. Association between alcohol 17 Panza F, Capurso C, D’Introno A et al. Alcohol drinking, cognitive and cardiovascular disease: Mendelian randomisation analysis based functions in older age, predementia, and dementia syndromes. on individual participant data. Br Med J 2014;349:g4164. J Alzheimers Dis 2009;17(1):7–31. 36 Zuccala G, Onder G, Pedone C et al. Dose-related impact of 18 Sachdeva A, Chandra M, Choudhary M et al. Alcohol-related alcohol consumption on cognitive function in advanced age: dementia and neurocognitive impairment: a review study. Int J High results of a multicenter survey. Alcohol Clin Exp Res 2001;25(12): Risk Behav Addict 2016;5(3):e27976. 1743–8. Downloaded from https://academic.oup.com/jpubhealth/article/40/2/304/4793394 by DeepDyve user on 19 July 2022 THE RELATIONSHIP BETWEEN ALCOHOL USE AND LONG-TERM COGNITIVE DECLINE 311 37 Britton A, Singh-Manoux A, Marmot M. Alcohol consumption and 40 Zheng Y, Yu B, Alexander D et al. Metabolomic patterns and alco- cognitive function in the Whitehall II Study. Am J Epidemiol 2004; hol consumption in African Americans in the Atherosclerosis Risk 160(3):240–7. in Communities Study. Am J Clin Nutr 2014;99(6):1470–8. 38 Stavro K, Pelletier J, Potvin S. Widespread and sustained cognitive 41 Alcohol Policy Team DoH. How to Keep Health Risks From Drinking deficits in alcoholism: a meta-analysis. Addict Biol 2013;18(2): Alcohol to a Low Level Government Response to the Public Consultation. 203–13. Department of Health, 2016 https://www.gov.uk/government/uploads/ system/uploads/attachment_data/file/545911/GovResponse2. 39 Bernardin F, Maheut-Bosser A, Paille F. Cognitive impairments in pdf. alcohol-dependent subjects. Front Psychiatry 2014;5:78.

Journal

Journal of Public Health – Oxford University Press

Published: Jun 1, 2018

Keywords: alcohol drinking; cognitive ability

There are no references for this article.