Etiology and correlates of alcohol misuse in early midlife

Abstract

Background

Early midlife individuals (ages 30–40) experience demographic shifts that may influence the remainder of adult life. Although new or persistent alcohol misuse is common during this period, early midlife is understudied in alcohol use literature. We examined the heritability of alcohol misuse; the associations between alcohol misuse and sociodemographic factors, physical health, and well‐being; and whether these associations were robust in cotwin comparisons.

Methods

Participants were 1446 Finnish twin pairs and 748 nonpaired Finnish twins with mean age 34 years. The alcohol misuse index was a composite measure of frequency of use, intoxication, heavy episodic drinking, and alcohol problems assessed with the Malmö‐modified Michigan Alcoholism Screening Test and the Rutgers Alcohol Problem Index. Early midlife correlates included relationship status and length, family formation, unemployment status, education level, self‐rated health, pain, sleeping difficulties, life satisfaction, psychological health, and other substance use. We employed a sex‐limitation model to estimate early midlife heritability. Linear and fixed effects regression models were used for individual and cotwin comparison analyses, respectively.

Results

Additive genetic (A) and unique environmental (E) components of alcohol misuse variance differed across sex (Females: A = 62%, E = 38%; Males: A = 49%, E = 51%). In individual‐based analyses, higher scores on the alcohol misuse index were associated with lower relationship stability, financial situation, education level, self‐rated health, physical fitness, life satisfaction and psychological health, and higher self‐reported pain, sleep difficulties, unemployment rates and other substance use (R ² = 0.008–0.12). Associations remained significant in cotwin comparison analyses (R ²  = 0.004–0.10) except for financial situation and education level.

Conclusions

There is evidence of sex differences in the etiological factors that influence early midlife drinking. After controlling for confounding familial factors, associations between alcohol misuse and poorer early midlife functioning largely remained, suggesting that alcohol misuse may play a role in poorer functioning across several outcomes.

We found evidence of sex differences in the heritability of early midlife alcohol misuse, with higher genetic influences on female (vs. male) alcohol misuse. We also found associations between early midlife alcohol misuse and functioning domains (i.e., sociodemographic, physical health, and well‐being) that were largely robust to controls of confounding familial factors. These findings provide insight into lifespan alcohol heritability patterns and identify functioning domains that may require additional support as a result of early midlife alcohol misuse.

INTRODUCTION

Early midlife, also known as established adulthood, is a period associated with changes in many functional domains (Lachman, 2015; Mehta et al., 2020). Individuals in this age range (ages 30–40 years old) experience demographic shifts such as changes in relationship status, parenthood, advanced education, and financial independence (Allred, 2019; Bell, 2009; Mathews & Hamilton, 2016; OECD, 2014; US Census Bureau, 2019, 2022; Wong, 2020). Alongside these demographic changes, alcohol misuse is common in early midlife: 25% of early midlife adults report recent heavy episodic drinking (5+ drinks in a session) and 16% meet criteria for past year alcohol use disorder (Braudt, 2018; Grant et al., 2015; Schulenberg et al., 2017). Epidemiological evidence further suggests that early midlife individuals may be particularly susceptible to the onset of AUD or the formation of a persistent course of alcohol misuse (Jester et al., 2016; Meier et al., 2013; Vergés et al., 2012). The early midlife period is a pivotal time for AUD treatment seeking as well (Blanco et al., 2015; Grant et al., 2015) as these individuals, compared with younger demographics, are at particularly high risk for AUD relapse (Dawson et al., 2007).

Despite the prevalence of alcohol misuse in this age range, the field's understanding of the etiology and correlates of early midlife alcohol misuse is limited. Twin and family studies indicate that AUD is approximately 50% heritable (Verhulst et al., 2015), and there is evidence that the heritability of AUD may increase across development (Kendler et al., 2008; van Beek et al., 2012), underscoring the dynamic nature of genetic influences on AUD (Dick, 2011; Stephenson et al., 2024). There is mixed evidence for sex‐specific differences in the etiology of alcohol misuse (Hardie et al., 2008; Kendler et al., 2016; Salvatore et al., 2017; Seglem et al., 2016; Verhulst et al., 2015), with some evidence for differences between males and females in other developmental periods (e.g., young adulthood) (Hardie et al., 2008; Salvatore et al., 2017; Seglem et al., 2016). Yet, the relative contribution of genetic and environmental influences on alcohol misuse phenotypes in early midlife, and whether there are sex differences in heritability, has not been examined directly.

Additionally, although many studies have examined the correlates of drinking in adolescence through young adulthood (i.e., the 20s) (Capaldi et al., 2015; Schulenberg et al., 2015), the literature on the correlates of alcohol misuse in midlife is far more limited. The largest study of early midlife drinking comes from Monitoring the Future (Schulenberg et al., 2015), which found that individuals who misuse alcohol in early midlife are also more likely to use other substances, and have poorer overall health, sleep and life and marital satisfaction. Interestingly, employed individuals (defined as individuals with full‐ or part‐time jobs) as compared to nonemployed individuals (defined as individuals who were unemployed, homemakers, or without paid employment) and individuals with higher educational attainment (grouped by individuals with a bachelor's degree or higher and those with less than a bachelor's degree) were also more likely to meet criteria for AUD in early midlife (Schulenberg et al., 2015). Marital/cohabitation status, parenthood status, and financial independence were not significantly associated with alcohol misuse in early midlife (Schulenberg et al., 2015).

In this study, we built on the modest extant literature and used a genetically informed approach to understand the etiology and correlates of early midlife alcohol misuse using the FinnTwin16 cohort, a longitudinal, population‐based sample of Finnish twins. First, we used standard genetic partitioning methods (Neale & Cardon, 1992) to estimate the degree to which additive genetic (A), shared environmental (C), and unique environmental (E) factors contributed to variation in early midlife alcohol misuse. Second, we leveraged the twin design to elucidate associations between alcohol misuse in early midlife and functioning across key domains (relationship status and length, family formation, education level, employment status, financial situation, physical health, well‐being, and other substance use). A notable limitation of prior early midlife alcohol misuse research is the use of samples of unrelated individuals (Schulenberg et al., 2015). As with all correlational research in unrelated samples, unmeasured factors that vary between families (e.g., genetic makeup, income, socioeconomic status, race/ethnicity, and parental substance use) may confound observed associations. Prevention and intervention approaches may be predicated on previously reported associations in unrelated samples that are not robust to familial (genetic and environmental) controls.

A cotwin comparison design can overcome this limitation by controlling for genetic and environmental factors shared by cotwins through the comparison between exposures and outcomes within families (D'Onofrio et al., 2013; Gustavson et al., 2024). Cotwin comparison designs account for such confounds by evaluating whether differences in a predictor of interest (e.g., early midlife alcohol use) are associated with differences in outcomes (e.g., early midlife functioning domains) within members of a twin pair who share genetic variation as well as environmental factors. Assuming there is sufficient power, if associations are not robust within a cotwin comparison design, this suggests that familial factors (genetic and shared environmental) are likely confounding the results.

Current study

In summary, we used the FinnTwin16 cohort to:

1. estimate the relative contribution of genetic and environmental factors on early midlife alcohol misuse and evaluate whether there were any sex differences in those etiological factors;

2. examine the associations between early midlife alcohol misuse and functioning in early midlife domains (relationship status and length, family formation, education level, employment status, financial situation, physical health, well‐being, and other substance use) in individual‐level analyses; and

3. examine associations between early midlife alcohol misuse and functioning in early midlife domains using a cotwin comparison design, an approach that strengthens inferences beyond individual‐level analyses by controlling for familial confounders.

In light of prior work (Lachman, 2015; Meier et al., 2013; Schulenberg et al., 2015; Verhulst et al., 2015), we hypothesized that additive genetic, shared environmental, and unique environmental factors would account for approximately 50%, 10%, and 40% of the variance in early midlife alcohol misuse and that higher early midlife alcohol misuse would be associated with poorer functioning in multiple early midlife domains. We did not advance directional hypotheses regarding association robustness using the cotwin comparison design in view of the limited prior research in this area.

METHODS

Sample

Participants came from the FinnTwin16 sample, a longitudinal, population‐based sample of Finnish twins (Kaprio, 2013). Monozygotic (MZ) and same‐sex (SS) and opposite‐sex (OS) dizygotic (Plomin et al., 2022) twins born between 1975 and 1979 were recruited through Finland's Central Population Registry. Twin pairs were mailed surveys to complete at age 16. Participants were invited to complete follow up assessments at ages 17 (97% retention), 18.5 (97% retention), 25 (88% retention), and 34 (79% retention). This study used data from the early midlife (age 34) assessment, which included a total of 4248 participants (Kaidesoja et al., 2019). After removing participants with incomplete biological sex, zygosity ratings, and alcohol misuse measures, the final analytic sample size was 3640 participants (57% female), with an average age of 34 years at time of assessment (SD = 1.1, range = 32–37). Participants were 1446 complete twin pairs (469 SS‐DZ, 471 OS‐DZ, 506 MZ) and 748 nonpaired twins (564 DZ, 184 MZ). We conducted attrition analyses to determine whether participants in the analytic sample significantly differed from excluded participants on key indicators across early midlife domains of interest (i.e., sociodemographic, physical, and psychological health). Participants in the analytic sample reported higher education levels (OR = 1.89, 95% CI [1.47, 2.44]) and better self‐rated health (OR = 1.48, 95% CI [1.01, 2.15]) when compared to participants excluded from the analytic sample, while participant's current smoking status and psychological health did not significantly differ across inclusion status.

Measures

Detailed information on the measures examined in this report is presented in Table 1 and described briefly here. Following the methods from Vachon et al. (2017) and leveraged in prior FinnTwin analyses (Pascale et al., 2022), we created an early midlife alcohol use index (AUI) including five alcohol‐related measures (frequency of use, heavy episodic drinking, frequency of intoxication, and alcohol problems as assessed with the Malmö‐modified Michigan Alcoholism Screening Test (MmMAST) (Kristenson & Trell, 1982) and the Rutgers Alcohol Problem Index (RAPI) (White & Labouvie, 1989)). All AUI measures were significantly correlated (all r ≥ 0.26, p < 0.01) and demonstrated strong internal consistency (α = 0.86). We first converted the response options for each measure to a scale of 0–10 and then averaged to create a composite score with a range of 0–10. Prior validation of this analytic approach reported high correlation across development (r = 0.98 and above) between the AUI and a latent alcohol use measure of the same variables derived from data‐drive techniques (i.e., factor analysis) (Vachon et al., 2017).

TABLE 1.

Early midlife functional domain measures.

Domain	Question at early midlife assessment
Alcohol misuse
Frequency of alcohol use	“How often do you drink alcohol on the whole? Try to take into account also the times you have drunk very small amounts, e.g. half a bottle of medium strength beer or a drop of wine.” Recoded such that 1 = I don't drink alcohol—9 = daily
Heavy episodic drinking	“At present how often do you drink more than 5 bottles of beer, more than a bottle of wine or more than half a bottle of liquor (or an equal amount of alcohol) at one and same occasion?” Recoded such that 1 = never drinkers—9 = daily
Frequency of Intoxication	“How often do you drink alcohol to the amount that you get drunk?” Recoded such that 1 = never drinkers—9 = daily
MmMAST	Assesses habits and experiences related to current and former alcohol use with answer options (0) = “no,” and (1) = “yes.” Scores were summed to create a composite score between 0 and 11. Participants with fewer than nine complete items were coded as missing
RAPI	22 questions assessing negative consequences of alcohol use with (1) = “never”—(4) = “10 or more times.” Scores were summed, and a pro‐rated mean between 1 and 4 was calculated. Participants with fewer than 18 complete items were coded as missing
Demographic characteristics
Relationship status	Participants were asked to report if they were currently in a relationship with (0) = “no,” and (1) = “yes”
Relationship length	“For how long has your current relationship lasted”
Number of cohabitations	Participants reported the number of current or prior partners they cohabitated with. Response options ranged from (1) = “none” to (4) = “three or more”
Pregnancy status	Participants were asked whether they or their partner was currently pregnant with (0) = “no,” and (1) = “yes”
Parenthood Status	Participants were asked whether they had biological children with (0) = “no,” and (1) = “yes.” Participants were also asked if they had non‐biological children living with them with (0) = “no,” and (1) = “yes”
Number of biological children	Participants were asked the number of biological children they had
Education level	“What schools/degrees have you completed.” Response options were (1) = “junior high school,” (2) = “vocational school or corresponding school,” (3) = “college level or corresponding level,” (4) = “senior high school,” (5) = “university of applied sciences,” and (6) = “college or university”
Unemployment status	Participants were asked to report whether they were primarily working outside the home, working at home, a student, unemployed, in the military, or doing something else. Responses were recoded such that (0) = not in workforce or employed part/full time, and (1) = unemployed
Financial situation	Participants were asked to indicate their current living and financial situation. Response options ranged from (1) = “very bad,” to (5) = “very good”
Physical health
Self‐rated health	“What do you think about your current health, is it…?” Response options ranged from (1) = “very poor,” to (5) = “very good”
Physical fitness	“What do you think about your current physical state, is it…?” Response options ranged from (1) = “very poor,” to (5) = “very good”
Somatic pain symptoms	Participants were asked to rate the frequency of their “stomachaches,” “headaches,” “low back pain,” and “neck and shoulder pain” during the past 6 months. Response options ranged from (1) = “seldom or never,” to (4) = “nearly every day”
Sleeping difficulties	“During the past 6 months, have you had any of the following symptoms, and if so how often; difficulty getting to sleep or waking up at night?” Response options ranged from (1) = “seldom or never,” to (4) = “nearly every day”
Well‐being
Life satisfaction	“Are you satisfied with your life?” Response options ranged from (1) = “strongly disagree,” to (7) = “strongly agree”
Psychological health	The General Health Questionnaire (GHQ‐12) asked participants to rate their health over the past month. Response options ranged from (1) = “better than usual,” to (4) = “much less than usual.” Recoded such that 1 = much less than usual—4 = better than usual. Scores were summed and a pro‐rated mean of 1–4 was calculated. Participants with fewer than nine complete items were coded as missing
Other substance use
Current smoking status	“Which of the following best describes your smoking habits?” Response options ranged from (1) = “I smoke daily,” to (5) = “I have never smoked.” Participants also reported if they smoked cigars, cigarillos, or pipes. Response options ranged from (1) = “never,” to (3) = “regularly”
Illicit drug use	“Have you ever used hash, marijuana, or other drugs e.g., sniffed glue?” Response options ranged from (1) = “not once,” to (5) = “20 times or more”

Abbreviations: MmMAST, Malmö‐modified Michigan Alcoholism Screening Test (α = 0.64 for males, 0.73 for females); RAPI, Rutgers Alcohol Problem Index (α = 0.88).

Early midlife sociodemographic factors included relationship status and length, parenthood status (biological and non‐biological), number of biological children, education level, unemployment status, and financial situation. Early midlife physical health correlates included self‐rated health, physical fitness, somatic pain symptoms (stomach, head, low back, and neck and shoulder), and sleeping difficulties. Early midlife well‐being correlates included life satisfaction and psychological health using the 12‐item General Health Questionnaire (GHQ) (Goldberg et al., 1988; Penninkilampi‐Kerola et al., 2006). Other early midlife substance use correlates included current smoking status and illicit drug use. Current smoking status included two product types of nicotine use, (1) cigarette use and (2) cigar, cigarillo, or pipe use. If participants reported current use of either product, they were considered current smokers (coded as 1). If they reported no use or past use only in both product types, they were categorized as current nonsmokers (coded as 0).

Statistical analysis

The analytic plan was preregistered (https://doi.org/10.17605/OSF.IO/XD3KF). As a prelude to the formal twin modeling of the etiological influences on early midlife AUI, we ran twin pair correlations stratified by zygosity. Using the R package OpenMx, version 4.1.1 (Boker et al., 2023; Neale & Cardon, 1992), we employed a univariate twin model, which compares MZ‐pair correlations (who share 100% of their genetic variation) to DZ‐pair correlations (who share approximately 50% of their genetic variation), to calculate the relative contribution of additive genetic (A), shared environmental (C), and unique environmental (E) influences on early midlife alcohol misuse. Additive genetic influences are indicated when MZ‐pair correlations are higher than DZ‐pair correlations, shared environmental influences are indicated when DZ‐pair correlations are more than half of the MZ‐pair correlation, and unique environmental influences are indicated when MZ‐pair correlations are less than 1. By including DZ opposite‐sex pairs in the univariate model, we simultaneously tested for sex differences in the source (i.e., qualitative) and relative (i.e., quantitative) estimates of ACE influences on early midlife alcohol misuse. Qualitative sex differences are present if opposite‐sex DZ‐pair correlations are significantly lower than same‐sex DZ‐pair correlations, while quantitative sex differences are present if comparisons between MZ‐pair correlations and DZ‐pair correlations differ by sex (Salvatore et al., 2017).

We also examined the associations between early midlife functioning domains (sociodemographic factors, physical health correlates, well‐being correlates, and other substance use) and AUI in individual‐level linear regression analyses that accounted for familial nesting. We then tested the robustness of the individual‐level associations between the early midlife functioning domains and AUI by controlling for familial confounds (i.e., genetic and shared environmental influences) using a within‐family, twin fixed effects model. The cotwin comparison analysis was separated into two tests (Gustavson et al., 2024). We ran a cotwin comparison model with both MZ and DZ twin pairs together, then limited the analyses to MZ twin pairs only—who share all their genetic variation identical by descent—which corresponds to the most conservative test of the correlates of early midlife alcohol misuse within families. Sex was used as a covariate in each analysis. Individual‐level and twin fixed effects regressions were run using R {plm} package version 1.6‐6 (Croissant & Millo, 2008).

RESULTS

Descriptive statistics for the early midlife AUI (frequency of use, intoxication, heavy episodic drinking, and alcohol problems) and early midlife functioning domains can be found in Table 2. In this population‐based sample, males had significantly higher AUI scores than females. Males were also more likely than females to be current smokers and have used illicit drugs in their lifetime. Females were more likely than males to be in a relationship, have biological and nonbiological children, and experience somatic pain symptoms (stomach pain, headaches, and neck and shoulder pain) and sleeping difficulties. Females also reported longer relationship length, higher education level, financial situation, and life satisfaction, but lower levels of psychological health than males.

TABLE 2.

Descriptive statistics of study variables for analytic sample and by sex.

Variable	Full sample mean (SD)/N (%)	Females mean (SD)/N (%)	Males mean (SD)/N (%)	Sex differences (t/$\mathcal{x}$ 2)
Age at assessment	33.96 (1.14)	33.95 (1.15)	33.98 (1.14)	t = −0.86
Alcohol Use Index (AUI)	3.37 (1.79)	2.77 (1.57)	4.16 (1.76)	t = −24.65***
Demographic predictors
Relationship status				$\mathcal{x}$ x 2   = 5.08*
In relationship	2958 (82)	1705 (83)	1253 (80)
Not in relationship	666 (18)	352 (17)	314 (20)
Relationship length (years)	8.94 (4.87)	9.34 (4.99)	8.40 (4.65)	t = 5.24***
Number of cohabitations	2.30 (0.80)	2.33 (0.78)	2.25 (0.82)	t = 2.97**
Pregnancy status				$\mathcal{x}$x2 = 0.004
Participant/Partner pregnant	281 (8)	160 (8)	121 (8)
Participant/Partner not pregnant	3343 (92)	1897 (92)	1446 (92)
Biological children				$\mathcal{x}$ x 2   = 26.00***
Yes	2171 (60)	1305 (64)	866 (55)
No	1466 (40)	756 (36)	710 (45)
Number of biological children	1.14 (1.14)	1.22 (1.14)	1.03 (1.12)	t = 5.24***
Non‐biological children				$\mathcal{x}$ x 2   = 22.15***
Yes	256 (7)	109 (5)	147 (9)
No	3371 (93)	1945 (95)	1426 (91)
Education level				$\mathcal{x}$ x 2   = 66.94***
Junior High School	881 (24)	413 (20)	468 (30)
Vocational School/Senior High School	660 (18)	342 (17)	318 (20)
College Level/University of Applied Sciences	1070 (29)	669 (32)	401 (25)
University	1028 (28)	636 (31)	392 (25)
Unemployment status				$\mathcal{x}$ x 2   = 13.38***
Yes	115 (3)	46 (2)	69 (4)
No	3519 (97)	2012 (98)	1507 (96)
Financial situation				$\mathcal{x}$ x 2   = 52.9***
Very bad	75 (2)	39 (2)	36 (2)
Fairly bad	339 (9)	230 (11)	109 (7)
Average	1287 (35)	793 (39)	494 (31)
Fairly good	1523 (42)	794 (39)	729 (46)
Very good	413 (11)	203 (10)	210 (13)
Physical health predictors
Stomach pain				$\mathcal{x}$ x 2   = 209.51***
Seldom or never	2445 (68)	1189 (58)	1256 (81)
Approximately once a month	874 (24)	649 (32)	225 (14)
Approximately once a week	214 (6)	158 (8)	56 (4)
Nearly every day	62 (2)	47 (2)	15 (1)
Headaches				$\mathcal{x}$ x 2   = 139.00***
Seldom or never	1597 (44)	743 (36)	854 (54)
Approximately once a month	1377 (38)	854 (42)	523 (34)
Approximately once a week	562 (16)	399 (19)	163 (10)
Nearly every day	76 (2)	56 (3)	20 (1)
Low back pain				$\mathcal{x}$x2 = 5.23
Seldom or never	1890 (52)	1067 (52)	823 (53)
Approximately once a month	1030 (28)	571 (28)	459 (29)
Approximately once a week	459 (13)	260 (13)	199 (13)
Nearly every day	236 (7)	150 (7)	86 (5)
Neck and shoulder pain				$\mathcal{x}$ x 2   = 124.00***
Seldom or never	1364 (38)	638 (31)	726 (46)
Approximately once a month	1193 (33)	678 (33)	515 (33)
Approximately once a week	715 (20)	484 (24)	231 (15)
Nearly every day	346 (10)	250 (12)	96 (6)
Self‐rated health				$\mathcal{x}$x2 = 3.11
Very poor	9 (<1)	6 (<1)	3 (<1)
Fairly poor	88 (2)	56 (3)	32 (2)
Moderate	574 (16)	319 (15)	255 (16)
Fairly good	2162 (40)	1214 (59)	948 (60)
Very good	798 (22)	462 (22)	336 (21)
Physical fitness				$\mathcal{x}$ x 2   = 11.22*
Very poor	26 (1)	18 (1)	8 (1)
Fairly poor	269 (7)	172 (8)	97 (6)
Moderate	1208 (33)	675 (33)	533 (34)
Fairly good	1741 (48)	989 (48)	752 (48)
Very good	394 (11)	205 (10)	189 (12)
Sleeping difficulties				$\mathcal{x}$ x 2   = 23.07***
Seldom or never	1539 (43)	816 (40)	723 (46)
Approximately once a month	949 (26)	542 (26)	407 (26)
Approximately once a week	776 (21)	459 (22)	317 (20)
Nearly every day	351 (10)	232 (11)	119 (8)
Well‐being predictors
Life satisfaction				$\mathcal{x}$ x 2   = 27.89***
Strongly disagree	52 (1)	23 (1)	29 (2)
Disagree	160 (4)	81 (4)	79 (5)
Slightly disagree	354 (10)	195 (10)	159 (10)
Neither agree nor disagree	153 (4)	65 (3)	88 (6)
Slightly agree	942 (26)	523 (26)	419 (27)
Agree	1431 (40)	831 (41)	600 (38)
Strongly agree	530 (15)	332 (16)	198 (13)
Psychological health	3.10 (0.46)	3.07 (0.48)	3.14 (0.44)	t = −4.63***
Other substance use
Current smoking status				$\mathcal{x}$ x 2   = 211.50***
Yes	243 (7)	29 (1)	214 (14)
No	3393 (93)	2029 (99)	1364 (86)
Illicit drug use				$\mathcal{x}$ x 2   = 78.14***
Not Once	2707 (75)	1633 (79)	1074 (68)
1–3 times	569 (16)	277 (13)	292 (19)
4–9 times	148 (4)	75 (4)	73 (5)
10–19 times	68 (2)	31 (2)	37 (2)
20 or more times	138 (4)	40 (2)	98 (6)

Note: Bolded values indicate significant sex difference, *p < 0.05, **p < 0.01, ***p < 0.001; Available N's for each measure range from 2945 to 3640; Female N range: 1692–2061; Male N range: 1253–1579. x ²  = Pearson's chi‐squared test, t = Student's t‐tests.

Genetic and environmental influences on AUI

Results for the twin pair correlations of the early midlife AUI variable can be found in Table 3. As expected, point estimates for the correlations indicated that MZ twins resembled each other more than DZ twins. Results for the univariate sex‐limitation models testing for qualitative and quantitative sex differences can be found in Table S1. According to AIC model fit index and likelihood‐ratio test, the best fitting model was the AE model (additive genetic and unique environmental influences only) allowing for quantitative sex differences between males and females. Results for the quantitative sex model of early midlife AUI variance can be found in Table 4. The quantitative AE model indicated that, for females, 62% (95% CI [0.55, 0.67]) of alcohol misuse in early midlife was attributable to additive genetic influences and 38% (95% CI [0.33, 0.45]) to unshared environmental influences or error. For males, 49% (95% CI [0.39, 0.58]) of alcohol misuse in early midlife was attributable to additive genetic influences and 51% (95% CI [0.42, 0.61]) to unshared environmental influences or error. Some have suggested using the AIC fit index to determine the best‐fitting model (e.g., AE, CE, or E sub‐models) introduces bias (Sullivan & Eaves, 2002). Therefore, for completeness, results for the full ACE model of early midlife alcohol misuse in males and females is shown in Table S2.

TABLE 3.

Early Midlife Alcohol Use Index (AUI) twin pair correlations.

Zygosity group	N	r	95% CI
MZ female	326	0.65	[0.58, 0.71]
MZ male	170	0.52	[0.40, 0.62]
DZ female	245	0.23	[0.10, 0.34]
DZ male	208	0.32	[0.19, 0.43]
DZ opposite sex	459	0.20	[0.12, 0.29]

Abbreviations: DZ, dizygotic; MZ, monozygotic; N, number of twin pairs.

TABLE 4.

Quantitative sex limitation AE model of early midlife AUI variance.

AE decomposition by sex	Estimate	95% CI
Female
Additive genetic (A)	0.62	[0.55, 0.67]
Unique environment (E)	0.38	[0.33, 0.45]
Male
Additive genetic (A)	0.49	[0.39, 0.58]
Unique environment (E)	0.51	[0.42, 0.61]

Abbreviation: CI, confidence interval.

Individual‐level associations

Results for individual‐level associations can be found in Table 5 and Figure S1. With respect to early midlife sociodemographic factors, higher early midlife AUI was associated with a lower likelihood of being in a relationship, shorter relationship length, a higher number of relationships involving cohabitation, a lower likelihood of being pregnant, fewer biological children, lower education level, a higher likelihood of unemployment, and worse financial situation (R ² = 0.009–0.039). For early midlife physical health correlates, higher AUI was associated with lower self‐rated health and physical fitness, higher levels of pain in the stomach, low back, and neck and shoulders, and more sleeping difficulties (R ² = 0.008–0.052). For early midlife well‐being correlates, higher AUI was associated with lower life satisfaction and psychological health (R ² = 0.033–0.040). For other early midlife substance use, higher AUI was associated with a higher likelihood of being a current smoker and lifetime illicit drug use (R ² = 0.089–0.119). No significant associations were found between AUI and headaches nor likelihood of having nonbiological children.

TABLE 5.

Individual‐level, Cotwin (all pairs), and MZ‐only associations with alcohol misuse in linear regression analyses.

Variable	Analysis type	β (95% CI)	p	R 2
Demographic characteristics
Relationship status	Individual level	−0.03 (−0.03, −0.02)	<0.001	0.020
	Cotwin	−0.05 (−0.06, −0.03)	<0.001	0.032
	Cotwin MZ	−0.06 (−0.09, −0.04)	<0.001	0.040
Relationship length	Individual level	−0.41 (−0.52, −0.30)	<0.001	0.024
	Cotwin	−0.48 (−0.69, −0.26)	<0.001	0.022
	Cotwin MZ	−0.24 (−0.68, 0.20)	0.28	0.004
Number of cohabitations	Individual level	0.08 (0.06, 0.09)	<0.001	0.036
	Cotwin	0.06 (0.03, 0.08)	<0.001	0.015
	Cotwin MZ	0.03 (−0.02, 0.08)	0.26	0.002
Pregnancy status	Individual level	−0.02 (−0.02, −0.01)	<0.001	0.010
	Cotwin	−0.03 (−0.03, −0.01)	<0.001	0.015
	Cotwin MZ	−0.02 (−0.04, −0.00) a	0.04	0.009
Biological children (Y/N)	Individual level	−0.05 (−0.06, −0.04)	<0.001	0.039
	Cotwin	−0.08 (−0.09, −0.06)	<0.001	0.067
	Cotwin MZ	−0.06 (−0.09, −0.03)	<0.001	0.026
Number of biological children	Individual level	−0.10 (−0.12, −0.08)	<0.001	0.034
	Cotwin	−0.13 (−0.17, −0.10)	<0.001	0.043
	Cotwin MZ	−0.08 (−0.15, −0.01)	0.02	0.011
Non‐biological children (Y/N)	Individual level	0.00 (−0.01, 0.00)	0.73	0.005
	Cotwin	0.00 (−0.01, 0.01)	0.62	0.001
	Cotwin MZ	0.00 (−0.02, 0.01)	0.66	<0.001
Education level	Individual level	−0.04 (−0.06, −0.02)	<0.001	0.033
	Cotwin	−0.03 (−0.06, 0.002)	0.07	0.037
	Cotwin MZ	0.00 (−0.05, 0.05)	0.98	<0.001
Unemployment status	Individual level	0.01 (0.00, 0.01) a	<0.001	0.009
	Cotwin	0.01 (0.00, 0.01) a	0.006	0.009
	Cotwin MZ	0.02 (0.01, 0.03)	<0.001	0.024
Financial situation	Individual level	−0.03 (−0.04, −0.01)	0.003	0.022
	Cotwin	−0.02 (−0.05, 0.01)	0.26	0.006
	Cotwin MZ	−0.03 (−0.09, 0.03)	0.33	0.002
Physical health
Self‐rated health	Individual level	−0.04 (−0.05, −0.03)	<0.001	0.042
	Cotwin	−0.05 (−0.08, −0.03)	<0.001	0.013
	Cotwin MZ	−0.02 (−0.07, 0.03)	0.45	0.001
Physical fitness	Individual level	−0.03 (−0.05, −0.01)	<0.001	0.019
	Cotwin	−0.05 (−0.07, −0.02)	0.002	0.008
	Cotwin MZ	−0.01 (−0.06, 0.05)	0.79	<0.001
Stomach pain	Individual level	0.02 (0.01, 0.04)	<0.001	0.052
	Cotwin	0.04 (0.01, 0.06)	0.005	0.035
	Cotwin MZ	0.04 (−0.01, 0.09)	0.12	0.005
Low back pain	Individual level	0.04 (0.02, 0.06)	<0.001	0.008
	Cotwin	0.04 (0.01, 0.08)	0.020	0.004
	Cotwin MZ	0.03 (−0.03, 0.10)	0.35	0.002
Neck and shoulder pain	Individual level	0.05 (0.03, 0.07)	<0.001	0.045
	Cotwin	0.06 (0.02, 0.09)	<0.001	0.047
	Cotwin MZ	0.05 (−0.02, 0.13)	0.16	0.004
Headaches	Individual level	0.00 (−0.01, 0.02)	0.73	0.042
	Cotwin	0.01 (−0.02, 0.04)	0.53	0.035
	Cotwin MZ	0.02 (−0.04, 0.08)	0.43	0.001
Sleeping difficulties	Individual level	0.08 (0.06, 0.10)	<0.001	0.026
	Cotwin	0.08 (0.05, 0.12)	<0.001	0.017
	Cotwin MZ	0.09 (0.02, 0.16)	0.02	0.012
Well‐being
Life satisfaction	Individual level	−0.13 (−0.16, −0.11)	<0.001	0.040
	Cotwin	−0.17 (−0.22, −0.12)	<0.001	0.035
	Cotwin MZ	−0.19 (−0.29, −0.09)	<0.001	0.027
Psychological distress	Individual level	−0.04 (−0.05, −0.03)	<0.001	0.033
	Cotwin	−0.05 (−0.06, −0.03)	<0.001	0.026
	Cotwin MZ	−0.05 (−0.08, −0.01)	0.006	0.015
Other substance use
Current smoking status	Individual level	0.03 (0.02, 0.03)	<0.001	0.089
	Cotwin	0.03 (0.02, 0.04)	<0.001	0.069
	Cotwin MZ	0.04 (0.02, 0.05)	<0.001	0.038
Illicit drug use	Individual level	0.17 (0.15, 0.19)	<0.001	0.119
	Cotwin	0.13 (0.11, 0.16)	<0.001	0.084
	Cotwin MZ	0.05 (0.01, 0.10)	0.01	0.013

Note: Bolded values indicate significant association, p < 0.05.

Abbreviation: CI, confidence interval.

^a CI reported as 0.00 due to rounding, but β remains significant.

Cotwin comparison

A within‐pair correlation test (intraclass correlation [ICC] = 0.35; 95% CI [0.31, 0.40]) determined there was sufficient alcohol use variation within pairs to conduct a cotwin comparison analysis. Table 5 and Figure S1 provide the results for cotwin comparison analyses. Associations between early midlife AUI and relationship status, current relationship length, number of cohabitations, pregnancy status, parenthood status (biological children), number of biological children, unemployment status, self‐rated health, physical fitness, stomach, low back, and neck and shoulder pain, sleep difficulties, life satisfaction, psychological health, current smoker status, and illicit drug use remained significant. However, the associations between early midlife AUI and education level and financial situation were no longer statistically significant. When only MZ twins were included, the most stringent cotwin comparison of twins who share all their genetic sequence variation, associations between AUI and relationship status, pregnancy status, parenthood status (biological children), number of biological children, unemployment status, sleep difficulties, life satisfaction, psychological health, current smoker status, and illicit drug use remained statistically significant. Associations between early midlife AUI and relationship length, number of cohabitations, self‐rated health, physical fitness, and stomach, low back, and neck and shoulder pain were no longer significant in the MZ‐only analysis.

DISCUSSION

In a sample of Finnish twins, we found evidence for quantitative sex differences in the relative influence of genetic and environmental factors on early midlife alcohol misuse, such that there were greater additive genetic influences in females than in males. We also found statistically significant associations between early midlife alcohol misuse and sociodemographic factors, and physical health and well‐being indicators at the individual‐level, the magnitude of which largely remained comparable within cotwin comparison analyses that controlled for genetic and environmental familial confounds.

Heritability of early midlife alcohol misuse

The quantitative sex‐limitation model provided the best fit for the early midlife alcohol misuse data and identified differences in the magnitude (but not source) of genetic and environmental variance across sex. The AE submodel fit the data best, suggesting a minimal role for shared environmental influences on early midlife alcohol misuse. This pattern is consistent with evidence from prior studies of adolescent and young adult drinking in the same (Rose et al., 2001; Viken et al., 1999) and similar (Baker et al., 2011; van Beek et al., 2012) cohorts, and aligns with evidence of higher genetic influences on AUD in adulthood (including participants in early midlife) when compared to younger developmental periods (Kremen et al., 2006; Verhulst et al., 2015). A recent longitudinal analysis leveraging genome‐wide polygenic scoring methods reported increases in SNP‐based heritability from adolescence to adulthood (i.e., age 26 and above) consistent with the findings from twin and family designs (Thomas et al., 2024). However, these results conflict with findings from a sample of male–male twin pairs from the US‐based Virginia Adult Twin Study of Psychiatric and Substance Use Disorders (VATSPSUD), for which there was a brief spike in shared environmental influences on alcohol consumption in the mid‐30s (Kendler et al., 2008). The discrepancy between our findings and this earlier report may be attributable to methodological and sample differences, as the VATSPSUD analyses were limited to male–male twin pairs, and the alcohol phenotype was a consumption measure (vs. alcohol misuse) measured using retrospectively reported life‐history calendar methods (Freedman et al., 1988).

We also found evidence of sex differences, unlike other studies (Baker et al., 2011; Salvatore et al., 2017; Verhulst et al., 2015), such that additive genetic influences played a larger role in females' alcohol misuse compared to males. Further exploration of sex differences in additional samples is warranted given the mixed evidence for sex differences in the heritability of alcohol misuse (Hardie et al., 2008; Kendler et al., 2016; Salvatore et al., 2017; Seglem et al., 2016; Verhulst et al., 2015). Secular changes in drinking patterns as a function of sex may offer a partial explanation for the sex differences in genetic and environmental influences on early midlife alcohol misuse. Adult women in western European countries have steadily increased drinking behaviors in recent years, while men have decreased or maintained their drinking rates in the same time frame (Tigerstedt et al., 2020; Virtanen et al., 2019). Thus, there is a convergence of drinking prevalence among men and women where men previously exhibited significantly higher drinking behaviors compared to women (Keyes et al., 2019). These loosening social norms surrounding women's drinking may contribute to the higher genetic influences on early midlife alcohol misuse in women compared to men.

Alcohol misuse and early midlife functioning

We found that early midlife alcohol misuse was correlated with several indicators across sociodemographic, physical, and well‐being domains. Higher alcohol misuse was associated with lower likelihood of being in a relationship, shorter relationship length, and lower education level, financial situation, self‐rated health, physical fitness, life satisfaction, and psychological health. Higher alcohol misuse was also associated with higher unemployment rates, pain levels, sleep difficulties, and likelihood of being a current smoker. These results expand upon prior studies of the correlates of early midlife alcohol misuse, which documented associations with adolescent substance use, low academic achievement, a family and adolescent history of psychiatric disorders, lower levels of constraint, and higher levels of negative emotions (Meier et al., 2013; Schulenberg et al., 2015).

Our results contrast previously reported associations between early midlife drinking and employment (vs. non‐employment) and higher education in the US‐based Monitoring the Future (MTF) study (Schulenberg et al., 2015). Previous studies have shown conflicting associations between alcohol misuse and socioeconomic factors in the United States (Collins, 2016; Walia et al., 2021) and Finland (Barr et al., 2016; Peña et al., 2018). Potential explanations for the discrepancy in early midlife are (1) our measure of unemployment excluded individuals not in the workforce, while the MTF sample included homemaker and no‐paid employment, (2) our alcohol misuse index included alcohol problems in addition to frequency of use, intoxication, and heavy episodic drinking measures versus alcohol problems alone in the MTF sample, and (3) the MTF sample controlled for earlier risk factors (e.g., adolescent alcohol use) while we did not control for alcohol use in other developmental periods. Future research should explore these associations in additional samples with controls for longitudinal risk factors to determine the influence of prior developmental exposures on associations between early midlife alcohol misuse and functioning.

As noted by others (D'Onofrio et al., 2013), directly addressing genetic and environmental confounds is useful for strengthening inferences about the nature of associations from observational studies. After controlling for potentially confounding familial factors in the cotwin comparison analyses, the associations between early midlife alcohol misuse and poorer functioning across these domains largely remained statistically significant with minimal attenuation in effect sizes. This suggests that these associations observed at the population level are not more simply explained by a shared genetic or environmental vulnerability. There were two exceptions to this trend, financial situation and education level. For financial situation, the magnitude of effect sizes remained the same across the individual, cotwin, and MZ‐only analyses, suggesting the nonsignificant association in the cotwin analyses was likely due to the lack of statistical power and imprecision in the estimate as indicated by wide confidence intervals. However, for education level, the effect size was attenuated in the cotwin comparison analysis, and further reduced in the MZ‐only analysis. This pattern of effects is consistent with prior evidence that an underlying genetic factor contributes to both alcohol misuse and educational attainment in early midlife (Groen‐Blokhuis et al., 2011; Piirtola et al., 2018).

The pattern of results from individual‐level to cotwin comparison analyses demonstrates that changes in early midlife alcohol misuse may impact functioning in sociodemographic, physical health, and well‐being domains. These findings may have implications for popular and effective interventions for alcohol use disorders such as motivational interviewing (Lundahl et al., 2010). Motivational interviewing allows the client to resolve conflict by identifying intrinsic motivators that align their behaviors with their goals, values, and identities (Rollnick & Miller, 1995). To the extent that patients may find it motivating to know how changes in alcohol misuse might affect functioning in other domains, and with the caveat that cotwin comparisons do not account for potential confounding due to factors that differ between twins our findings provide some of the strongest nonexperimental evidence that reductions in alcohol use may improve a range of social, physical, and well‐being indicators.

Limitations

The current findings should be considered in light of the following limitations. This sample is drawn from a population‐based, longitudinal study of Finnish twins. According to the Organization for Economic Cooperation and Development, Finland outperforms other countries on most social, physical, and mental well‐being measures such as educational attainment, financial stability, and life satisfaction (OECD, n.d.). Therefore, these results may not be generalizable to non‐Scandinavian countries. Moreover, the analytic sample reported higher educational attainment and better self‐rated health when compared to early midlife Finnish Twins not in the analytic sample, which may further reduce generalizability. Additionally, our index of early midlife alcohol misuse included a measure of consumption in addition to measures of problematic alcohol use. Evidence suggests that alcohol consumption and problems may represent distinct genetic etiology and correlations with sociodemographic, physical, and psychological health variables (Dick et al., 2011; Johnson et al., 2021). We removed frequency of alcohol consumption from the alcohol use index to determine whether early midlife alcohol misuse heritability and correlates of early midlife functioning changed. Results from these sensitivity analyses can be found in Tables S3, S4, and S5. Consistent with heritability estimates derived from the full AUI, the best‐fitting model allowed for quantitative sex differences between males and females (Table S3) and the AE submodel fit the data best (Table S4). In females, additive genetic and unique environmental estimates in the AUI without frequency of alcohol use (A = 65%; E = 35%) were comparable to estimates derived from the full AUI (A = 62%; E = 38%). In males, after removing the frequency of alcohol use measure from the AUI, additive genetic influences were nominally higher (A = 59%) than estimates derived from the full AUI (A = 49%) and unique environmental influences were nominally lower (E = 41%) than estimates derived from the full AUI (E = 51%). These findings suggest that, in males, there may be greater genetic influences on problematic alcohol use when compared to genetic influences on alcohol consumption. Therefore, further research into the etiology of alcohol use and problems in early midlife, specifically in males, is warranted. In addition, removing frequency of alcohol use from the index resulted in minimal differences in the cotwin comparison associations between early midlife functioning and alcohol misuse (Table S5). These results suggest that alcohol problems (vs. alcohol consumption) were driving the robust associations with early midlife sociodemographic, physical, and psychological health.

Although cotwin comparison analyses attempt to control for the majority of genetic and environmental confounding, there are instances in which control between twins cannot be accounted for, such as nonshared environmental exposures (e.g., unmeasured traumatic life events that occurred to one twin and not the other (Dixon et al., 2009)). Due to a compounding of measurement error, cotwin analyses also involve an increase in Type 2 error when compared to individual‐level analyses (McGue et al., 2010). Cotwin comparisons also reduce sample sizes (Boardman & Fletcher, 2015) compared with individual‐level analyses. This corresponds to a reduction in the statistical power to detect effects, which is especially evident in the smaller MZ‐only sample. Lastly, although the cotwin comparison design offers advantages for strengthening inferences in observational research, we are unable to conclude the direction of associations discussed due to the cross‐sectional study design. Future genetically informed, longitudinal analyses leveraging phenotypic data from earlier developmental periods may prove beneficial in disentangling causality.

CONCLUSIONS

In a sample of Finnish twins, we found quantitative sex differences in alcohol misuse heritability estimates, with a higher additive genetic influence in females than in males. We also found that higher alcohol misuse in early midlife was associated with lower functioning in sociodemographic, physical health, and well‐being domains. Many of these associations remained robust with minimal effect size attenuation after controlling for familial confounds (genetic and environmental) using a cotwin comparison analysis. These novel epidemiological associations describe a relatively unaddressed age period in alcohol misuse literature, allowing for enhanced insight into life span alcohol misuse heritability patterns and identification of sociodemographic, physical, and well‐being domains that may require additional support as a result of early midlife alcohol misuse.

FUNDING INFORMATION

This work was primarily supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health under award number R01AA015416.

CONFLICT OF INTEREST STATEMENT

None to declare.

Supporting information

Figure S1.

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Lumpe, E. , Pascale, A. , Stephenson, M. , Barr, P. , Cooke, M.E. , Latvala, A. et al. (2025) Etiology and correlates of alcohol misuse in early midlife. Alcohol: Clinical and Experimental Research, 49, 301–314. Available from: 10.1111/acer.15513

DATA AVAILABILITY STATEMENT

Research data are not shared owing to Finnish data privacy laws.