Critique 154: Effects of changes in alcohol consumption on risk of dementia among elderly women — 12 January 2015
Hoang TD, Byers AL, Barnes DE, Yaffe K. Alcohol Consumption Patterns and Cognitive Impairment in Older Women. Am J Geriatr Psychiatry 2014; 22:1663-1667.
Objective: Few studies have investigated changes in alcohol consumption and risk of cognitive impairment among oldest old adults.
Methods: In a prospective study of 1309 women ≥65 years old, alcohol use was assessed at repeated visits and used to estimate average change in alcohol consumption over 16 years. Clinically significant cognitive impairment (mild cognitive impairment and dementia) was assessed at year 20.
Results: Compared with the reference group (slight decrease in alcohol consumption by 0-0.5 drinks/week, 60.4%), increasing consumption over time (>0 drinks/week) was not associated with risk of cognitive impairment (5.0%, odds ratio [OR]: 1.00, 95% confidence interval [CI]: 0.54-1.85). Decreasing consumption by >0.5 drinks/week was associated with increased risk (34.5%, OR: 1.34, 95% CI: 1.05-1.70). Adjustment for age, education, diabetes, smoking, BMI, and physical activity attenuated the magnitude of the effect slightly and resulted in borderline statistical significance.
Conclusion: Women in their ninth and tenth decade of life who decrease alcohol use may be at risk of cognitive impairment.
A number of prospective studies have shown that moderate drinkers are at lower risk of developing dementia or cognitive impairment than are abstainers or heavy drinkers, but not all studies have confirmed such a relation (Anstey et al). Possible mechanisms of a putative beneficial effect are not known, but could relate to a reduction in the risk of cerebral atherosclerosis, a direct effect on cognition through the release of acetylcholine in the hippocampus, or the antioxidant effects of alcohol and polyphenols (Letenneur). Kim et al have also noted data suggesting that alcohol consumption could be neuroprotective from interactions between alcohol and protein kinase C, the adenosine receptor, and cardioprotective proteins (e.g., superoxide dismutase, nitric oxide synthase, and heat shock protein) or from the down-regulation of inducible nitric synthase and up-regulation of endothelial nitric oxide synthase.
The present study is based on 1,309 women who were aged 65 years or greater at the beginning of a 20-year follow up, during which they had at least two assessments of cognitive function. At baseline, 42.1% of the women reported that they were “light” drinkers (>0 – < 3 drinks/week, 13.8% reported “moderate” consumption (3-7 drinks/week), and 3.6% reported “heavy” consumption (> 7 drinks/week). By the 20th year of follow up, 17.5% of the previously normal women had developed dementia (by several standard tests, a previous diagnosis of dementia, or nursing home residence), and 22.7% had developed mild cognitive impairment (MCI). Thus, a total of 40.2% of the women had developed some degree of cognitive impairment, making this a good cohort in which to evaluate environmental factors associated with its development.
From repeated assessments of alcohol intake, the investigators stated that over 16 years, the women’s average change is alcohol use frequency was a decrease of about 0.5 drinks/week, with 60% falling into this category of change; women in this category made up the referent group. Only 5% reported an increase in drinking frequency. The investigators stated that baseline alcohol intake was not associated with the risk of developing cognitive impairment (43.1% for non-drinkers, 38.8% for light drinkers, 36.1% for moderate drinkers, and 38.3% for heavy drinkers).
In this paper, there was no suggestion that the small percentage of women who increased their intake changed their risk of cognitive impairment (adjusted OR=1.04, 95% CI 0.56, 1.95), when compared with women who did not change their intake or decreased it only slightly. However, for women who decreased their intake by > 0.5 drinks/week, the authors suggest that there was an increase in the risk of cognitive impairment, with unadjusted results showing an OR of 1.34, 95% CI: 1.05-1.70. However, with full adjustment for confounders, the effect was attenuated: OR = 1.26, 95% CI 0.98, 1.61; p = 0.07.
Some questions about the study and the authors’ implications: While the analyses were done appropriately, Forum members had concerns about whether the findings presented in this paper justify the wisdom of the authors in stating in their conclusion that “Women in their ninth and tenth decade of life who decrease alcohol use may be at risk of cognitive impairment.”
Alcohol assessments were made at baseline and at years 6, 8, 10, and 16, and “linear regression was used to estimate average change in alcohol consumption frequency between baseline and up to year 16, with random effects included for slope and intercept.” The investigators apparently counted alcohol intake up to the time of diagnosis of cognitive impairment, as they did not exclude alcohol data in the years shortly before a diagnosis was made; thus, developing dementia could have led to a decrease in intake, rather than the other way around.
Reviewer Ellison noted that “While the investigators adjusted for age, education, diabetes, BMI, physical activity, and smoking, the reason why some women decreased their alcohol intake or stopped drinking could not be ascertained. For example, having to move into a nursing home may have made it difficult or impossible to have access to alcohol (and being in a nursing home was one criterion for the diagnosis of cognitive impairment).” Reviewer Finkel also commented: “It seems nearly impossible to me for the authors to distinguish reduction of consumption as an a priore cause of impaired cognition from the reverse, factors incident to aging leading to reduced consumption.” Forum member Barrett-Connor added: “I agree with the comments of other Forum members, and I cannot tell in which direction the arrows go. I am not willing to assume causality from the data presented. Therefore it makes sense to question the authors’ conclusions that decreasing alcohol intake increases dementia risk.”
It was noted that here were very few heavy drinkers (3.6%), and only 5% of women increased their reported intake, so implications relating heavier drinking or an increase in amount of alcohol consumed to cognitive function are tenuous. Current smokers were more likely to increase their intake and obese women less likely to increase their intake, but the effects of baseline level of drinking (before the change in consumption) was apparently not taken into account. (It would have been helpful to know specifically the baseline level of drinking of subjects classified as showing an increase or a decrease in alcohol consumption.)
Forum reviewers had some concerns about how the authors dealt with mortality among their cohort, assuming a number of members of their original cohort had died during follow up. The women who were eligible for this study were apparently all survivors up to the 20-year examination; if so, the results would not relate to women who died earlier, whether they developed cognitive impairment or not. Reviewer Finkel commented: “I remain uneasy that in this study women consuming > 7 drinks/week were considered ‘heavy drinkers.’ (However, I have long objected to unsupported advocation of restricting alcohol simply because of advanced age.)”
Reviewer McEvoy stated: “I was impressed with the relative stability of drinking in this age group. However, I agree with other Forum members that there is concern about conclusions on the direction of causality. I am not sure it is valid to conclude that among older women, decreasing drinking may increase risk of cognitive impairment based on these results. Alzheimer’s dementia has a long prodromal period and latent pathology could precede/contribute to the decrease in drinking. (And even if not latent, there is no analysis here of the temporal relation of change in cognitive function with change in drinking).”
McEvoy added: “I sympathize with the authors that brief reports limit the amount of information that can be provided but I was also concerned about the lack of cognitive characterization of the cohort at baseline. (The authors simply state that at baseline, participants did not have dementia. There was no characterization of participants’ performance on any cognitive test; some could have performed below impairment cut-offs at baseline.) Also, a detailed demographic/clinical characterization of the cohort in terms of baseline drinking is needed.
“I do not know how to interpret the finding of a lack of association of baseline drinking with risk of cognitive impairment 20 years later given that we are not told whether the groups defined on baseline drinking differed from each other in age or other health risk factors. I agree that it would have been better if the authors somehow took baseline drinking into account when assessing effects of change in drinking. Decreasing drinking by > 0.5 glasses/week may be very different in someone drinking > 7 drinks/week than in someone drinking < 1 drink/week.”
Decreasing alcohol intake with ageing: Forum member Skovenborg commented on the usual effects of ageing on drinking patterns. “The authors do not acknowledge that a decreasing intake of alcohol is a normal pattern when you grow older. Several cross-sectional and longitudinal studies have shown a decline in alcohol intake with increasing age. The decline in percent of subjects consuming any alcohol over time has been around 2% per year (Adams et al)(Molgaard et al)(Moore et al). The reasons for drinking less could be fertile grounds for introduction of bias. In a study by Busby et al, the main reasons for decrease in alcohol intake were change in health and fewer social opportunities. A minority increased the intake of alcohol and the main reasons for increase in alcohol intake were more money and more time for drinking associated with social activities.”
Reviewer Keil commented: “In my opinion it is a general phenomenon in epidemiologic studies that people reduce their alcohol intake with age, as also shown in this study. (They also change their smoking behaviour with age by either quitting smoking or reducing the number of cigarettes; if the authors might have investigated this phenomenon they might have found that decreasing smoking is associated with more cognitive impairment.) The reasons for this behaviour are difficult to unravel from observational studies but a plausible explanation is that with increasing age people might not tolerate the previously consumed amount of alcohol or the development of some ill health might persuade them to drink less alcohol. The consumption of alcohol is generally a pleasurable and social endeavour and when people don`t feel well they might drink less or stop drinking.
“In light of this, the authors` conclusion that decreasing consumption of alcohol was associated with increased risk of cognitive impairment may just be an example of reverse causation. Thus, cognitive impairment might not have increased in the respective group over time because of a decrease of alcohol intake, but alcohol intake was decreased because some of the elderly women felt less well off. Knowing from other studies that light to moderate alcohol consumption is good for the heart and the head, I would not advise elderly people to stop drinking, but the conclusions drawn in this paper that decreasing alcohol consumption increases risk of cognitive impairment are not justified, because the design and analysis of the study do not permit such a statement.”
Reviewer Finkel had further remarks about drinking among the elderly: “In addition to the cogent reasons enumerated for elders drinking less, there must be something more. Even physically and psychologically healthy older people eat less, and do less. Although they still may, for example, drink wine with their dinners, their potable potion portion might be voluntarily reduced to half that of their previous intake, with equal satisfaction. However, I am not sure that they still get an equal pharmacological effect.”
Potential difficulties from “paradoxical” associations in epidemiologic analyses: Reviewer Zhang pointed out a potential problem of using a prevalent exposure, i. e., alcohol consumption reported at baseline, to assess effects of its change on the outcome, i.e., risk of cognitive function impairment, occurring during follow up. “The findings that baseline alcohol consumption was not associated with cognitive function are intriguing. However, in general, people do not start drinking alcohol at age 65 or later; thus, alcohol drinking assessed at baseline can be assumed to be a prevalent exposure: the drinkers at age 65 had probably been drinkers of similar amounts of alcohol for years.
“People with cognitive impairment before the baseline examination were excluded from this study. However, if alcohol is either neuroprotective or neurotoxic, subjects who were drinkers or non-drinkers entering the study (at baseline) could be quite different in many ways, and this could have led to bias in the results obtained. For example, if alcohol is neuroprotective, those who consumed alcohol may have been protected against previous cognitive impairment from the alcohol, thus were eligible for the study. Non-drinkers without previous cognitive impairment could have been protected from other reasons (unrelated to alcohol). If alcohol is neurotoxic, the opposite might be true: the drinkers who reached baseline without disease could have other protective mechanisms, whereas the non-drinkers would not have required such mechanisms to protect them. In either case, the drinkers and non-drinkers at baseline would have had different characteristics and may not have been comparable for an evaluation of baseline alcohol intake or change in alcohol consumption over the study period. This phenomenon has been described as the ‘prevalent user bias’ (Fleishmann et al)(Ray)(Danaei et al). The fact that previous alcohol use may affect the risk of later cognitive impairment can mean that the effect of changes in alcohol during follow up can be null, or even opposite to an earlier association, because of this type of bias.
“This phenomenon, i.e., ‘prevalent exposure bias,’ has been found in many studies, such as paradoxical effects of obesity on mortality among patients with heart failure (Curtis et al) and estrogen replacement therapy and risk of myocardial infarction from the Nurses’ Health Study (Grodstein et al). Also, in the British Male Doctors Study, the rate ratio of coronary death for cigarette smoking was increased among all age groups up to age 75, but was paradoxically decreased among participants aged 75-84 years (Doll et al). My associates and I have discussed this type of bias as it relates to rheumatic diseases (Choi et al).”
Zhang added: “To help judge the potential relevance of such a bias in the present study, the authors should at least have presented the results of association between baseline alcohol consumption and change of alcohol consumption over the study period. If change of alcohol consumption was related to baseline alcohol consumption, it may confound the association between such a change and cognitive impairment.”
Reviewer Thelle agreed with the opinions of other Forum members regarding potential bias in the study, but stated that this is not an unusual problem: “Actually, bias or their sources are what keeps us epidemiologists busy!” Reviewer Skovenborg had one concluding statement: “It seems to me that the present report is an example of a study where observational epidemiology faces its limits: the exposure of alcohol is very small, the longitudinal changes of exposure are minuscule, and the associations are weak, making the conclusions presented questionable to say the least (as described by Taubes & Mann).”
References from Forum review
Adams WL et al. Alcohol intake in the healthy elderly. Changes with age in a cross-sectional and longitudinal study. J Am Geriatr Soc 1990;38:211-216.
Anstey KJ, Mack HA, Cherbuin N: Alcohol consumption as a risk factor for dementia and cognitive decline: meta-analysis of prospective studies. Ame J Geriatr Psychiatry 2009; 17:542e555.
Busby WJ, et al. Alcohol use in a community-based sample of subjects aged 70 years and older. J Am Geriatr Soc 1988;36:301-305.
Choi HK, Nguyen U-S, Niu J, Danaei G, Zhang Y. Selection bias in rheumatic disease research. Nat Rev Rheumatol 2014; advance online publication; doi:10.1038/nrrheum.2014.36
Curtis JP, Selter JG, Wang Y, et al. The Obesity Paradox: Body Mass Index and Outcomes in Patients With Heart Failure. Arch Intern Med 2005;165:55-61.
Danaei G, Tavakkoli M, Hernan MA. Bias in Observational Studies of Prevalent Users: Lessons for Comparative Effectiveness Research From a Meta-Analysis of Statins. Am J Epidemiol 2012;175:250-62.
Doll R, Hill AB. Mortality of British doctors in relation to smoking; observations on coronary thrombosis. In: Haenszel W, ed. Epidemiological approaches to the study of cancer and other chronic diseases. Monog Natl Cancer Inst. 1966;19:205-68.
Fleischmann E, Teal N, Dudley J, May W, Bower J, Salahudeen A. Influence of excess weight on mortality and hospital stay in 1346 hemodialysis patients. Kidney Int 1999;55:1560–1567.
Grodstein F, Stampfer MJ, Manson JE, et al. Postmenopausal estrogen and progestin use and the risk of cardiovascular disease. N Engl J Med. 1996;335:453–461.
Kim JW, Lee DY, Lee BC, et al. Alcohol and Cognition in the Elderly: A Review. Psychiatry Investig 2012;9:8-16.
Letenneur L. Moderate alcohol consumption and risk of developing dementia in the elderly: the contribution of prospective studies. Ann Epidemiol. 2007;17(Suppl):S43-S45.
Molgaard CA et al. Prevalence of alcohol consumption among older persons. Journal of Community Health 1990;15:239-251
Moore AA et al. Longitudinal patterns and predictors of alcohol consumption in the United States. Am J Public Health 2005;95:458-465.
Ray WA. Evaluating Medication Effects Outside of Clinical Trials: New-User Designs. Am J Epidemiol 2003;158:915-920.
Taubes G, Mann CC. Epidemiology faces its limits. Science 1995;269:164-169.
A number of prospective studies have shown that moderate drinkers are at lower risk of developing dementia or cognitive impairment than abstainers or heavy drinkers; possible mechanisms of a putative beneficial effect could relate to a reduction in the risk of cerebral atherosclerosis, a direct effect on cognition through the release of acetylcholine in the hippocampus, the antioxidant effects of alcohol and polyphenols, or from the down-regulation of inducible nitric synthase and up-regulation of endothelial nitric oxide synthase.
The present study is based on 1,309 women who were aged 65 years or greater at the beginning of a 20-year follow up, during which they had at least two assessments of cognitive function. At baseline, 42.1% of the women reported that they were “light” drinkers (>0 – < 3 drinks/week, 13.8% reported “moderate” consumption (3-7 drinks/week), and 3.6% reported “heavy” consumption (> 7 drinks/week). By the 20th year of follow up, 17.5% of the previously normal women had developed dementia (by several standard tests, a previous diagnosis of dementia, or nursing home residence), and 22.7% had developed mild cognitive impairment (MCI). Thus, a total of 40.2% of the women had developed some degree of cognitive impairment.
In this paper, there was no suggestion that the small percentage of women (5%) who increased their intake changed their risk of cognitive impairment (fully adjusted OR=1.04, 95% CI 0.56, 1.95), when compared with women who did not change their intake or decreased it only slightly. However, for women who decreased their intake by > 0.5 drinks/week, the authors suggest that there was an increase in risk of cognitive impairment, with unadjusted results showing an OR of 1.34, 95% CI: 1.05-1.70. However, with full adjustment for confounders, the effect was attenuated: OR = 1.26, 95% CI 0.98, 1.61; p = 0.07.
Forum members noted that the reasons that some women changed their drinking habits could not be ascertained (as is usually the case in observational studies). Alzheimer’s dementia has a long prodromal period and latent pathology could precede/contribute to a decrease in drinking. They were also concerned about the lack of cognitive characterization of the cohort at baseline. Thus, Forum members had some difficulty in knowing how to interpret the finding of a lack of association of baseline drinking with risk of cognitive impairment 20 years later given that we are not told whether the groups defined on baseline drinking differed from each other in age or other health risk factors. For those who changed their intake, decreasing drinking by > 0.5 glasses/week may be very different for someone drinking > 7 drinks/week than in someone drinking < 1 drink/week. It would have been better if the authors somehow took baseline drinking into account when assessing effects of change in drinking.
There was also the concern that any observational study that relates the effects of changes in alcohol intake on the risk of a disease without somehow taking earlier drinking into account may give paradoxical results. (An example is the relation of obesity to cardiovascular disease; overall, there is an increase in risk for obese individuals but when changes in intake among older people are evaluated, the result is often no effect or even a protective effect: the “obesity paradox.”) The same has been shown for the relation of usual alcohol intake and changes in intake to other diseases.
Thus, while the analyses were done appropriately, Forum members had concerns about whether the findings presented in this paper justify the wisdom of the authors in stating in their conclusion that “Women in their ninth and tenth decade of life who decrease alcohol use may be at risk of cognitive impairment.” As one reviewer stated: “Knowing from other studies that light to moderate alcohol consumption is good for the heart and the head, I would not advise elderly people to stop drinking. However, the conclusions of the authors that this paper shows that decreasing alcohol consumption increases the risk of cognitive impairment in elderly women may be true, but are not justified because the design, analysis, and results of the study do not permit such a statement.”
* * *
Comments for this critique by the International Scientific Forum on Alcohol Research were provided by the following members:
Elizabeth Barrett-Connor, MD, Chief of the Division of Epidemiology, Distinguished Professor in the Departments of Family and Preventive Medicine & Medicine, University of California, San Diego, La Jolla, CA, USA
Giovanni de Gaetano, MD, PhD, Research Laboratories, Catholic University, Campobasso, Italy
R. Curtis Ellison, MD. Section of Preventive Medicine & Epidemiology, Department of Medicine, Boston University School of Medicine, Boston, MA, USA
Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA
Ulrich Keil, MD, Institute of Epidemiology & Social Medicine, University of Muenster, Germany
Linda McEvoy, PhD, Department of Radiology, University of California at San Diego (UCSD), La Jolla, CA, USA
Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark
Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway
Dag S. Thelle, MD, PhD, Senior Professor of Cardiovascular Epidemiology and Prevention, University of Gothenburg, Sweden; Senior Professor of Quantitative Medicine at the University of Oslo, Norway
Marianne van den Bree, PhD. MRC Centre for Neuropsychiatric Genetics and Genomics, Institute of Psychological Medicine and Clinical Neurosciences, Cardiff University, UK
Yuqing Zhang, MD, DSc, Clinical Epidemiology, Boston University School of Medicine, Boston, MA, USA