Associations of alcohol use in early and middle adulthood with mid- and late-life cognition - a synthetic cohort approach

Using data from the NLSY79, HRS, and a synthetic cohort, this study found no significant association between early-to-middle adulthood alcohol consumption (including abstention, light/moderate, or heavy drinking) and mid-to-late life memory performance, suggesting that alcohol's effects on cognition may not manifest until later in life.

The Gist

Imagine you are trying to solve a mystery: Does the amount of alcohol you drink when you're young (in your 20s) affect how sharp your memory is when you get older (in your 50s, 60s, and beyond)?

This is a tricky puzzle because most studies only have one piece of the picture:

• Study A knows what people drank when they were young, but stops tracking them before they get old enough to show serious memory issues.
• Study B knows exactly how old people's memories are getting, but didn't start tracking them until they were already 50, so it has no idea what they drank in their youth.

To solve this, the researchers in this paper used a clever trick called a "Synthetic Cohort." Think of this like Frankenstein's monster, but for data. They took the "young body" (the drinking history) from Study A and stitched it onto the "old brain" (the memory test results) from Study B. This created a "super-participant" that had a full life story from age 18 to age 80, allowing them to see the long-term effects.

Here is the breakdown of what they found, using simple analogies:

1. The Three Groups of Drinkers

The researchers looked at three types of people:

• The Light/Moderate Drinkers: People who had a glass of wine or a beer now and then (the "social sippers").
• The Abstainers: People who never drank a drop.
• The Heavy Drinkers: People who drank a lot or had "binge" episodes.

2. The Findings: A Tale of Two Times

The "Young Adult" Test (The NLSY79 Data)
When they looked at people in their 20s and 30s, they found no clear link between drinking and memory.

• The Analogy: Imagine a car engine. Whether you put premium gas, regular gas, or no gas in it when the car is brand new doesn't seem to change how the engine runs right now. The engine is too new to show wear and tear yet.
• Result: Whether you were a teetotaler, a social sipper, or a heavy drinker in your 20s, your memory score in middle age looked about the same.

The "Mid-Life" Test (The HRS Data)
When they looked at people who were already in their 50s and 60s, the story changed slightly.

• The Analogy: Now the car has 30,000 miles on it. The researchers noticed that the cars that had no gas (abstainers) and the cars that were flooded with bad fuel (heavy drinkers) were running slightly worse than the cars with the "Goldilocks" amount of fuel (light/moderate drinkers).
• Result: In middle age, both never drinking and drinking too much were linked to slightly lower memory scores compared to moderate drinking.

The "Frankenstein" Test (The Synthetic Cohort)
This is the most important part. They combined the "young drinking history" with the "old memory scores."

• The Analogy: They asked, "If we take the drinking habits of a 25-year-old and fast-forward them to see how that specific person's memory looks at age 65, what happens?"
• Result: Nothing significant. The drinking habits from 30 or 40 years ago did not predict memory problems later in life. The "sips" or "binges" of your youth didn't seem to leave a scar on your brain decades later.

3. Why Did the Results Change? (The "Sick Quitter" Mystery)

The researchers noticed something interesting about the "Abstainers" (people who never drank) in the older group. They performed worse on memory tests.

• The Theory: This is often called the "Sick Quitter" effect. Imagine a group of people who stopped drinking because they got sick, had a heart attack, or were already feeling unwell. They aren't "healthy non-drinkers"; they are "sick people who quit."
• The Twist: In the "Young Adult" group, this didn't happen because young people rarely quit drinking due to serious illness. But in the older group, the "never drinkers" included many people who had health issues that made them stop (or never start) drinking, which skewed the results.

The Bottom Line

Does your drinking in your 20s ruin your memory in your 70s?
According to this study, probably not.

• Heavy drinking in your youth didn't seem to cause a faster memory decline later on.
• Never drinking in your youth didn't seem to give you a "super-memory" advantage later on.
• The only time drinking habits seemed to matter for memory was right now (in middle age), where both heavy drinking and stopping drinking (often due to illness) were linked to slightly lower scores.

The Takeaway:
Think of your brain like a garden. This study suggests that the weeds you planted (or didn't plant) when you were a teenager don't seem to be the ones choking the garden 40 years later. The health of the garden seems to depend more on what you are doing today and what is happening to your body right now, rather than what you did decades ago.

Note: This is a preprint (a draft study), so it hasn't been fully checked by other scientists yet, but it offers a fascinating new way of looking at old data.

Technical Summary

1. Problem Statement and Research Gap

Alcohol use is a known modifiable risk factor for dementia, yet existing literature lacks clarity on lifecourse timing: whether the effects of alcohol vary depending on when it is consumed (e.g., early adulthood vs. middle age) and whether effects are limited to heavy drinking.

• Data Limitation: No single existing dataset captures both long-term prospective alcohol use data from early adulthood and longitudinal cognitive aging data extending into late life.
  • The National Longitudinal Study of Youth 1979 (NLSY79) has prospective alcohol data from ages 18–48 but only limited cognitive assessments in middle age.
  • The Health and Retirement Study (HRS) has extensive longitudinal memory data (ages 50+) but lacks early adulthood alcohol history.
• Objective: To estimate the effects of prospectively reported alcohol use in early and middle adulthood on memory outcomes in mid-to-late life by utilizing a synthetic cohort approach to bridge these data gaps.

2. Methodology

Study Design and Data Sources

The study utilized a three-pronged approach:

1. NLSY79 ($n=7,540$): Analyzed alcohol use (ages 18–48) against midlife memory scores.
2. HRS ($n=13,090$): Analyzed midlife alcohol use against mid-to-late life memory scores.
3. Synthetic Cohort ($n \approx 8,710$ matched pairs): Created by matching NLSY79 participants (providing early-life exposure) to HRS participants (providing late-life outcomes) to simulate a single cohort with full lifecourse data.

Exposure and Outcome Definitions

• Alcohol Use: Harmonized across waves into three categories:
  • Abstention: No alcohol.
  • Light/Moderate: $\le$7 drinks/week (women) or $\le$14 (men) without heavy episodic drinking.
  • Heavy: $>$7 drinks/week (women) or $>$14 (men), or any heavy episodic drinking ($\ge$6 drinks in 30 days).
• Cognitive Outcome: Memory score derived from the word list recall test (immediate + delayed recall of 10 words).
  • Scores were pre-processed to adjust for practice effects (repeated testing) using regression residuals.
  • Outcomes included: Memory Level (intercept at age 50) and Memory Decline (slope over time).

Synthetic Cohort Construction

• Matching Algorithm: HRS participants were matched to NLSY79 participants based on time-invariant covariates (sex, race/ethnicity, nativity, parental education, age, birth year) and time-varying mediators (education, marital status, health conditions, smoking, depression, physical activity, income).
• Distance Metric: A weighted Euclidean distance was used for continuous variables (z-scored).
• Optimization: The study tested various matching calipers (0.25 to 1.5 SD) and variable sets to balance match fidelity with statistical power, selecting the set that minimized bias without reducing power.
• Analysis Strategy: Each HRS participant was matched to up to 10 NLSY79 participants. The data was split into 10 analytic sets, analyzed via linear mixed-effects models, and pooled using Rubin's rules.

Statistical Analysis

• Models: Covariate-adjusted linear mixed models.
• Adjustments: Controlled for time-invariant confounders (e.g., AFQT scores, parental education) and time-varying confounders (e.g., income, health status) measured prior to or contemporaneously with exposure.
• Key Parameters: Main effects of alcohol on memory level and interaction terms (Alcohol $\times$ Age) to assess memory decline.

3. Key Results

A. NLSY79 (Early/Mid-life Alcohol $\to$ Mid-life Memory)

• Abstention: No significant association with midlife memory scores compared to light/moderate drinking (e.g., at ages 18–26: $\beta = -0.09$, 95% CI: -0.30, 0.11).
• Heavy Drinking: Generally associated with worse memory, though significance varied by wave. Notably, heavy drinking at ages 18–26 was associated with lower scores ($\beta = -0.26$, 95% CI: -0.48, -0.04).
• Decline: No significant association found between alcohol use and the rate of memory decline in middle adulthood.

B. HRS (Mid-life Alcohol $\to$ Mid-to-Late Life Memory)

• Abstention: Significantly associated with lower memory scores compared to light/moderate drinkers ($\beta = -0.14$, 95% CI: -0.25, -0.02).
• Heavy Drinking: Significantly associated with lower memory scores ($\beta = -0.25$, 95% CI: -0.42, -0.07).
• Decline: Neither abstention nor heavy drinking showed a statistically significant association with the rate of memory decline, though point estimates suggested faster decline.

C. Synthetic Cohort (Early-life Alcohol $\to$ Late-life Memory)

• Abstention: No significant association with mid-to-late life memory scores ($\beta = 0.13$, 95% CI: -0.10, 0.36 for ages 18–26).
• Heavy Drinking: No significant association with memory scores ($\beta = 0.02$, 95% CI: -0.25, 0.28 for ages 18–26).
• Decline: No significant association between early-life alcohol use and the rate of memory decline in later life.

4. Key Contributions

1. Methodological Innovation: Successfully applied a synthetic cohort approach to link early-life exposure data (NLSY79) with late-life outcome data (HRS), overcoming the limitation of no single dataset covering the full lifecourse.
2. Lifecourse Timing: Provided evidence that the association between alcohol and cognition may be time-dependent. The "sick-quitter" phenomenon (where abstainers have poor health) appears more relevant in midlife (HRS data) than in early adulthood (NLSY79 data).
3. Null Findings on Early Exposure: Demonstrated that prospectively reported alcohol use in early adulthood (ages 18–48) does not appear to have a significant direct effect on memory levels or decline rates in mid-to-late life, challenging the notion that early heavy drinking inevitably leads to accelerated cognitive aging.

5. Significance and Discussion

• Interpretation of Abstention: The study suggests that the negative association between abstention and cognition observed in older adults (HRS) may be driven by residual confounding (e.g., former heavy drinkers quitting due to illness) rather than a causal protective effect of light/moderate drinking. This association was absent in the NLSY79 data where participants were younger and less likely to have "sick-quitter" dynamics.
• Clinical/Public Health Implications:
  • There is little evidence that alcohol use within the ranges reported in these cohorts substantially increases the rate of memory decline.
  • The benefits of complete alcohol abstention for cognitive health remain uncertain; the data does not support a strong causal link between early-life drinking patterns and late-life dementia risk.
• Limitations:
  • Observational Nature: Cannot definitively prove causality due to potential unmeasured confounding.
  • Measurement Error: Alcohol reporting is subject to desirability bias; different measurement scales across decades may introduce artifacts.
  • Synthetic Assumptions: The validity of the synthetic cohort relies on the assumption that the matching algorithm successfully controls for all confounders, which is difficult to guarantee.

Conclusion: The study concludes that while midlife abstention and heavy drinking are associated with lower memory scores (likely due to health selection bias), early-to-midlife alcohol use does not appear to have a significant causal effect on memory decline in later life. Future research should focus on changes in drinking patterns over time rather than static lifetime categories.