Wright CB, Elkind MS, Luo X, Paik MC, Sacco RL. Reported alcohol consumption and cognitive decline: the northern Manhattan study. Neuroepidemiology. 2006; 27: 201-207.

• Measure the effect of alcohol intake on cognitive performance over time in a younger, multi-ethnic, community-based sample
• Assess the effect of having an apolipoprotein E ε4 (APOE-4) allele.

• Resident of Northern Manhattan for three or more months
• English- or Spanish-speaking
• Resident of a household with a telephone
• Age 40 or more years
• No stroke diagnosis at enrollment or before the first cognitive assessment in 2001
• No history of alcohol-related hospitalization
• Data on reported alcohol intake
• Two or more cognitive function scores available
• Subjects provided written informed consent; the Columbia University Medical Center ethics committee approved the study.

• Non-resident of Northern Manhattan, or resident for less than three months
• Not English- or Spanish-speaking
• Resident of a household without a telephone
• Age less than 40 years
• Stroke diagnosis either before enrollment or prior to the first cognitive assessment in 2001
• History of alcohol-related hospitalization
• No data on reported alcohol intake
• Less than two cognitive function scores available.

Recruitment

Dual-frame, random digit dialing sample of subjects residing in Northern Manhattan.

Design

• Prospective cohort study with annual follow-ups
• At enrollment (sometime between 1993 and 2001), participants completed a telephone interview, physical examination and medical records review
• Participants completed subsequent annual telephone interviews
• Cognitive assessments were included beginning in 2001.

Dietary Intake/Dietary Assessment Methodology

Alcohol intake was measured in a structured interview adapted from food-frequency questionnaires (FFQ). It included responses for beer, wine and liquor, with nine frequency responses (none to seven or more drinks per day). At enrollment, subjects were asked about average consumption in the past year and average lifetime consumption. At the time of the cognitive assessment, subjects were asked about average consumption in the past six months.

Statistical Analysis

• Subjects who had a stroke sometime after the first cognitive assessment had all subsequent cognitive scores censored
• Generalized estimating equations were used to measure an association between reported alcohol intake at the first cognitive assessment and changes in cognitive scores over time, controlling for potential confounders, such as time between cognitive assessments, age, education, gender, race or ethnicity, health insurance status, high-density lipoprotein (HDL-C) level, body mass index (BMI), history of hypertension, diabetes, cardiac disease, current smoking, depression and physical inactivity
• Stratified analyses were repeated in a subset of the sample for whom APOE-4 allele status was known
• Potential bias due to dropouts was assessed in a logistic regression model. Drop-out after the first cognitive assessment was predicted by the first assessment score, reported alcohol intake, the interaction between the two and relevant covariates.

Timing of Measurements

• Subjects were enrolled sometime between 1993 and 2001. At enrollment, subjects completed a telephone interview, physical examination including lab assessment and medical records review.
• A subset (N=600) had APOE-4 allele status determined
• Subjects completed annual telephone interviews after enrollment. Cognitive assessment was included beginning in 2001.

Dependent Variables

Annual cognitive functioning (based on the modified Telephone Interview for Cognitive Status [TICS-m]):

• TICS-m assesses orientation, attention, immediate recall of a 10-word list, calculations, judgment, language, finger tapping, antonyms and delayed recall of the 10-word list
• Scored out of 51 points; normally distributed without a ceiling effect and good validity and reliability in multi-ethnic community samples.

Independent Variables

Alcohol consumption in the previous six months months (five categories: never, past, less than one drink per week, one drink per week to two drinks per day, more than two drinks per day):

• Those who reported alcohol intake at enrollment, but not at the time of the first cognitive assessment were considered past drinkers
• Those who reported no intake at either enrollment or the first cognitive assessment were considered never drinkers
• Those whose intake at follow-up differed by two categories from enrollment were excluded (N=10).

Control Variables

• APOE-4 allele status (baseline fasting blood sample; determined by H-haI digestion and PCR amplification of genomic DNA)
• Time between cognitive assessments
• Sociodemographics:
  • Age (years)
  • Education
  • Gender
  • Race/ethnicity (self-reported as Hispanic, black, white or other groups)
• Health insurance status
• HDL-C level (baseline fasting blood sample, determined with an automated spectrometer)
• BMI
• Current smoking
• Physical inactivity
• History of medical conditions:
  • Hypertension (systolic blood pressure (SBP) higher than 140mmHg or diastolic blood pressure (DBP) higher than 90mmHg on the mean of two measures, self-reported diagnosis or medical treatment for hypertension)
  • Diabetes (fasting blood glucose 127mg per dL or higher, self-reported diagnosis or insulin or oral hypoglycemic use)
  • Cardiac disease (history of coronary artery disease, atrial fibrillation, or myocardial infarction)
  • Depression (Hamilton Depression Rating Scale score higher than 10 or history of antidepressant use).

• Initial N: 3,298 enrolled between 1993 and 2001 (68% response rate)
• Attrition (final N):
  • 1,428 analyzed (43% of initial N); 6,913 person-years of follow-up (mean 2.2 years; range 0.5 to 4.4 years); 600 had APOE-4 allele status tested
  • Subjects were excluded or lost to follow-up for history of alcohol-related hospitalization (N=31), death (N=812), stroke (N=128), insufficient data on alcohol intake or fewer than two cognitive assessment scores (N=851). 48 subjects were not accounted for
  • Compared to those who were excluded, the sample had more Hispanics and diabetics and fewer blacks. There was no difference in cognitive scores or in the interaction between cognitive scores and alcohol intake among those who dropped out after the first cognitive assessment
• Age: Mean age 71 years
• Ethnicity: 62% Hispanic, 19% black, 19% white
• Other relevant demographics:
  • 12% uninsured, 45% on Medicaid, 43% privately insured
  • 82% had hypertension, 25% were diabetic, 19% had cardiac disease, 3% had a history of depression
  • 13% were current smokers
  • 21% were never drinkers, 44% were past drinkers
  • Of those tested, 155 were APOE-4 allele carriers and 445 were non-carriers
• Anthropometrics: BMI: (kg/m²)
  • 27% less than 25
  • 43% 25 to 30
  • 30% higher than 30
• Location: Northern Manhattan.

Key Findings

• Mean TICS-m scores declined an average of 0.4 points from the first to the last assessment
• Amount of current (but not past) alcohol intake was positively associated with TICS-m scores over time
  • The association was diminished when adjusted for sociodemographic variables, but was still significant
  • Including vascular risk factors (Model three) further diminished the association, but more than one drink per week was still associated with higher cognitive function scores
• Neither baseline TICS-m scores nor changes over time differed significantly by APOE-4 allele status. Allele status did not modify the effect of intake on cognitive function.

Model one is adjusted for age and education; Model two is Model one plus gender, race/ethnicity and insurance status; Model three is Model two plus history of hypertension, diabetes, cardiac disease, physical inactivity, depression, current smoking, HDL-C level and BMI.

Other Findings

In a sensitivity analysis, those with a history of alcohol-related hospitalization (N=31) further attenuated the effect in the more than two drinks per day category, but was still significant.

• Current drinkers had less cognitive decline than never drinkers, adjusting for sociodemographic and vascular risk factors. The findings are in general agreement with previous research
• The past drinking group did not differ from non-drinkers, which may partially be due to heterogeneity among past drinkers in terms of amount of past drinking and amount of time abstinent
• APOE-4 allele status did not alter the association.

• Author-identified strengths:
  • Prospective design
  • Community-based sample including Hispanics and blacks (groups with higher risk of dementia)
• Author-identified limitations:
  • Current drinkers may be healthier than non-drinkers or past drinkers because the latter may be abstaining because of health problems. This potential bias was addressed by studying change scores
  • Non-drinkers tended to have lower sociodemographic characteristics than drinkers, which may have biased the findings toward the null
  • Cognitive function was assessed over the phone, which may not be as valid as a full neuropsychiatric exam
• Additional concerns with the article:
  • 48 subjects were not accounted for from the initial 3,298 at enrollment to final analysis sample of 1,428
  • There was no description of how a sub-sample was selected for APOE-4 allele status determination
  • Table one stated that the mean age was 71; the text said mean age was 66 (range 40 to 98)
  • The initial response rate was 68%, and only 43% of initial cases were analyzed. Comparisons to population demographics were not discussed, so it is difficult to determine how representative the final sample was
  • Although alcohol type was included in the intake measure, any potential differences among beer, wine and liquor were not explored.