Genetic and environmental influences on the progression from alcohol use disorder to alcohol-related medical conditions

Abstract

Background:

Medical conditions related to alcohol use disorders (AUD) represent a substantial public health concern. However, only a subset of individuals with AUD develop these conditions, and the extent to which genetic and environmental factors shared with, versus distinct from, AUD contribute to this progression has not yet been determined.

Methods:

Using data from Swedish national registries for a cohort born 1932-1970 (N=1,319,214, 48.9% women), we conducted twin-sibling biometric model fitting to examine the genetic and environmental sources of variance that contribute to liability to alcohol-related medical conditions (AMC). Progression to AMC, determined using medical registry data, was contingent on an AUD registration, which was determined using medical and criminal registry data.

Results:

We identified AUD registrations in 3.2% of women and 9.2% of men. Among those with an AUD registration, 14.4% of women and 15.4% of men had an AMC registration. In the final models, we constrained the beta pathway from AUD to AMC and the genetic and unique environmental paths to be equal across sexes. The beta path was estimated at 0.59. AMC was modestly heritable in women (A=0.32) and men (A=0.30). The proportion of total heritability unique to AMC was 39.6% among women and 41.3% among men. A higher proportion of total environmental variance was unique to AMC: 76.7% for women and 77.2% for men. In a sensitivity analysis limited to liver-related AMC, we observed similar results, with a slightly lower beta path from AUD to AMC (0.46) and higher proportions of AMC-specific genetic (70.0% in women; 71.7% in men) and environmental (84.5% in both sexes) variance.

Conclusions:

A moderate to substantial proportion of genetic and environmental variance contributing to AMC risk is not shared with AUD, underscoring the need for additional gene identification efforts for AMC. Furthermore, the prominent influence of environmental factors specific to AMC provides a promising area for the identification of potential prevention targets. We did not observe significant sex differences in the etiology of AMC, though follow-up is warranted in other well-powered studies.

Introduction

Alcohol use disorders (AUDs) represent a significant public health burden: In the United States, excessive drinking was estimated to cost $249 billion in 2010 alone (Sacks et al., 2015), and the World Health Organization estimated that alcohol accounted for 5.1% of the global burden of disease and injury in 2012 (World Health Organization, 2014). Alcohol misuse can result in medical consequences, including alcoholic liver disease and pancreatitis. However, only a fraction of those with AUD experience medical issues as a result of their drinking: It is estimated that about 20% of AUD cases develop fatty liver disease, 10-35% develop alcoholic hepatitis (McCullough and O’Connor, 1998), and 10-15% develop cirrhosis (Mann et al., 2003). The factors that contribute to variation in the liability to alcohol-related medical conditions (AMC) are complex and not yet fully characterized.

The heritability of AUD is moderate, estimated at 0.49 in a meta-analysis of twin studies (Verhulst et al., 2015), and ranging from 0.22-0.57 in a large sample of Swedish twins and siblings (Kendler et al., 2016). However, the extent to which liability to AMC is genetically influenced is unclear. Studies of white male US twins found higher degrees of concordance for cirrhosis among monozygotic (MZ) than dizygotic (DZ) twins, providing evidence of the importance of genetic factors (Hrubec and Omenn, 1981, Reed et al., 1996). Reed et al. (1996) reported a heritability of 0.47 for cirrhosis, which was largely shared with AUD. Given modest rates of progression to AMC among those with AUD, and the pathophysiology of AMC (Boccuto and Abenavoli, 2017, Herreros-Villanueva et al., 2013, Jupp et al., 2010), the genetic etiology of various medical conditions is almost certainly at least partially independent of genes that impact AUD. Indeed, molecular genetic analyses have identified common variants associated with specific AMC that are not associated with AUD (Derikx et al., 2015, Innes et al., 2020, Schwantes-An et al., 2021), though such studies are limited in number and scale, and further research is warranted.

Some evidence suggests that women develop AMC at lower levels of consumption than men. In a population-based Danish cohort, Becker et al. (1996) found that women were at higher risk of alcoholic liver disease than men at all levels of consumption. There are potential biological and behavioral explanations for the observed sex differences (Mann et al., 2003, Saunders et al., 1981, Biswas and Ghose, 2018), including drinking patterns (e.g., frequency vs. quantity of consumption), diet, bioavailability of alcohol (Frezza et al., 1990), and the impact of estrogen on alcohol metabolism (Eagon, 2010, Zakhari and Li, 2007). However, sex differences are not uniformly observed (Tao et al., 2003); confirmation of these differences, and clarification of underlying mechanisms, could be of clinical utility given their potential to inform treatment decisions.

In the current study, our primary research aim was to determine the genetic and environmental structure of AMC in the context of a causal-contingent-common pathway (CCC) model (Kendler et al., 1999). We leveraged Swedish national registries, which provide a substantial sample size, access to objective medical and legal records of AUD and AMC, and absence of reporting bias. A previous twin analysis of cirrhosis utilized traditional multivariate modeling (Reed et al., 1996); by using the CCC model, we account for the contingency of AMC on AUD and may therefore improve the partitioning of variance.

Materials and Methods

Sample

We linked nationwide Swedish registers via the unique 10-digit identification number assigned at birth or immigration to all Swedish residents. The identification number was replaced by a serial number to ensure confidentiality. The following sources were used to create our dataset: Total Population Register, containing information about year of birth, sex, family and marital status; Multi-Generation Register, linking individuals born after 1932 to their parents and grandparents; The Swedish Twin Register, the Hospital Discharge Register, containing hospitalizations for Swedish inhabitants from 1964-2017; Outpatient Care Register, containing information from all outpatient clinics from 2001 to 2017; and regional Primary Health Care Registers (data available for different years depending on region, from 1992-2018). In addition, we used the Crime Register, which included national complete data on all convictions in lower court from 1973-2017; the Swedish Suspicion Register that included national data on individuals strongly suspected of crime from 1998-2015; and the Mortality Register with dates and causes of death from 1952 until 2016.

We included monozygotic (MZ) and dizygotic (DZ) twins, full siblings and maternal half siblings born up to five years apart from the birth cohort 1950 to 1990. Both same-sex and opposite-sex pairs were included.

Alcohol use disorder

AUD was defined from Swedish medical registries by the following ICD codes: ICD-8: 571.0, 291, 303; ICD-9: 305A, 357F, 571A, 571B, 571C, 571D, 425F, 535D, 291, 303; and ICD-10 codes: E24.4, G31.2, G62.1, G72.1, I42.6, K29.2, K70.0, K70.1, K70.2, K70.3, K70.4, K70.9, K85.2, K86.0, O35.4, F10.1, F10.2, F10.3, F10.4, F10.5, F10.6, F10.7, F10.8, F10.9. In addition, we identified AUD among individuals convicted for or suspected of at least two alcohol-related crimes (e.g. drunk driving and drunk boating) according to law 1951:649, paragraph 4 and 4A and law 1994:1009, Chapter 20, paragraph 4 and 5 from the Swedish Crime Register, and code 3005 and 3201 in the Suspicion register.

Alcohol-related medical conditions

We identified registrations for AMC using ICD-10 codes that corresponded to the following medical consequences of alcohol misuse: liver diseases caused by alcohol (K70), alcohol-induced pseudo-Cushing’s syndrome (E24.4), degeneration of nervous system due to alcohol (G31.2), alcoholic polyneuropathy (G62.1), alcoholic myopathy (G72.1), alcoholic cardiomyopathy (I42.6), alcoholic gastritis (K29.2), alcohol-induced acute pancreatitis (K85.2), alcohol-induced chronic pancreatitis (K86.0), and maternal care for (suspected) damage to fetus from alcohol (O35.4); from ICD-9 we used V79B, 357F, 425F, 535; and from ICD-8, we used 291. In sensitivity analyses, we conducted analyses using only liver-related registrations for AMC: ICD8: 571.0, ICD9: 571A, 571B, 571C, 571D; and ICD10: K700, K701, K702, K703, K704, K709.

Statistical methods

We modelled AUD and the possible subsequent AMC utilizing a Causal Contingent Common (CCC) pathway model (Kendler et al., 1993). In this model there is a direct path (beta, or β) from AUD to AMC, and consequently AMC can be found only in subjects with AUD.

We assume that the liability to AUD and AMC can be represented by three sources of variance: additive genetic (A), common environment (C), and unique environment (E). We denote the three variance components for AUD by A₁, C₁, and E₁, which we have investigated previously (Kendler et al., 2016). In that report, we found evidence of quantitative and qualitative gender differences. In the current analyses we therefore make the corresponding assumptions for AUD while focusing on investigating variance components for AMC. The proportion of variance of AMC that is explained by the same factors that are involved in the development of AUD is transmitted via the β path. The extreme β = 0 implies that no factors involved in the development of AMC are shared with those involved in the development of AUD, while the other extreme β = 1 represents the situation where there are no additional factors involved in the development of AMC. The three sources of variance that are unique to AMC are denoted A₂, C₂, and E₂. In total, the variance components representing the three sources of variance for AMC, A_c, C_c, and E_c are represented by: A_c = β² × A₁+ A₂, C_c = β² × C₁+ C₂, and E_c = β² × E₁+ E₂. For this aspect of the model, we have sparse data and will assume there are no qualitative sex differences; in other words, the genetic correlation (r_g) and the shared environmental correlation (r_c) are constrained to unity across the sexes.

Different prevalences for twins, full siblings and maternal half siblings are represented by different sex-specific thresholds. To account for the increase in diagnosis by age we allow the prevalence to depend on birth year by including a regression term where the thresholds for AUD and AMC depend linearly on birth year. These regression parameters are sex-specific but constrained among relative types.

Estimates from a full model would in general result in more accurate estimates; however, due to sparse data on the AMC component, we elected to run submodels to investigate support for and stability of the estimates. We tested a limited set of submodels and report both the likelihood ratio test (compared with the full model) and Akaike Information Criterion (AIC) values (Akaike, 1987) for each model. The latter balances explanatory value and parsimony and allows us to compare models that are not nested.

Descriptive analyses were conducted using SAS software Version 9.4 of the SAS system for Windows. Twin and family model fitting was conducted using OpenMx (Neale et al., 2016).

Results

Descriptive statistics.

The total sample size was N=1,319,214 (N=645,222 [48.9%] women). Of the total sample, 3.2% (N=22,698) of women and 9.2% (N=63,473) of men had an AUD registration. Among those with an AUD registration, 14.4% (N=3272) of women and 15.4% (N=9778) of men had an AMC registration, a significant sex difference (OR=1.08 [95% confidence intervals 1.04; 1.13]). Because of left truncation of some of the registry data as a function of when records became available in different municipalities, we were unable to obtain reliable estimates for age of AUD and AMC onset. However, the mean birth year for women with AUD was 1953.6 (SD=9.1), and for men was 1953.0 (SD=9.1). The mean birth year for women with AMC was 1950.3 (SD=8.4) compared to 1954.1 (SD=9.1) in the subgroup without AMC, a mean difference of 3.8 years (95% confidence intervals [CI] 3.4, 4.2; p<0.0001). For men, the mean birth year among those with AMC was 1950.0 (SD=8.4) compared to 1953.6 (SD=9.2) in the subgroup without AMC, a mean difference of 3.6 years (95% CI 3.3, 3.8; p<0.0001).

Of the N=20,782 women with AUD registrations only from medical registers, 10.7% (N=2,222) were registered with AMC; 8.9% (N=188) of the N=2102 whose AUD was identified from criminal records were registered for AMC. The risk of AMC was slightly higher among the N=12,315 women defined from both medical diagnosis and criminal records: 13.9% (N=1217) had AMC. The pattern was similar for men. Among N=49,267 defined as AUD from medical registers, 15.3% (N=7527) had AMC, and of N=24,544 defined as AUD based on criminal records, 11.2% (N=2744) had AMC. For the N=12,315 men defined from both medical registration and criminal records, 20.1% (N=2470) had AMC.

Table 1 presents sample sizes as a function of relative pair type (e.g., monozygotic twin, full sibling, etc.). Cross-sibling correlations are presented in Table 2. Correlations generally, though not uniformly, decreased as genetic relatedness decreased, suggesting a role of genetic factors. We noted substantially lower sibling concordance for AMC among full (non-twin) and maternal half-siblings relative to twins. Due to low AMC prevalence, correlation estimates were imprecise.

Table 1.

Sample size information for relative pairs, including the prevalence of alcohol use disorder (AUD) and alcohol-related medical conditions (AMC).

	Women			Men
	N pairs	With AUD registration N (%)	With AMC registration N (% of AUD)	N pairs	With AUD registration N (%)	With AMC registration N (% of AUD)
Same-sex pairs
MZ	3,869	263 (3.40%)	34 (12.93%)	3,193	464 (7.27%)	60 (12.93%)
DZ	5,060	348 (3.44%)	48 (13.79%)	4,702	792 (8.42%)	115 (14.52%)
FS	355,349	24,115 (3.39%)	3,582 (14.85%)	373,963	69,270 (9.26%)	10,926 (15.77%)
MHS	13,329	1,680 (6.30%)	205 (12.20%)	13,633	4,448 (16.31%)	597 (13.42%)
Opposite sex pairs
DZ	11,074	450 (4.06%)	59 (13.11%)	11,074	1,109 (10.01%)	177 (15.96%)
FS	726,050	23,568 (3.25%)	3,531 (14.98%)	726,050	69,352 (9.55%)	11,069 (15.96%)
MHS	26,973	1,718 (6.60%)	233 (13.08%)	26,973	4,501 (16.69%)	616 (13.69%)

MZ=monozygotic twin; DZ=dizygotic twin; FS=full sibling; MHS=maternal half sibling

Table 2.

Cross-sibling correlations (SE) for alcohol use disorder (AUD) and alcohol-related medical conditions (AMC).

	N pairs	AUD	AMC
Women
MZ	3,869	0.58 (0.05)	0.698 (0.25)
DZ	5,060	0.29 (0.06)	0.394 (0.44)
FS	355,349	0.27 (0.01)	0.043 (0.07)
MHS	13,329	0.16 (0.03)	0.005 (0.25)
Men
MZ	3,193	0.60 (0.04)	0.518 (0.22)
DZ	4,702	0.29 (0.04)	0.747 (0.13)
FS	373,963	0.30 (<0.01)	0.112 (0.03)
MHS	13,633	0.21 (0.02)	−0.098 (0.11)
Opposite Sex Pairs
DZ	11,074	0.22 (0.03)	0.130 (0.26)
FS	726,050	0.23 (<0.01)	0.153 (0.03)
MHS	26,973	0.12 (0.02)	0.083 (0.12)

MZ=monozygotic twin; DZ=dizygotic twin; FS=full sibling; MHS=maternal half sibling

Twin-sibling model fitting.

We previously reported model fitting for AUD (Kendler et al., 2016) and therefore focused our current efforts on AMC. Due to sparse data and correspondingly modest statistical power, the cross-sex genetic and shared environmental correlations for AMC were fixed to 1 across (i.e., we did not allow for quantitative sex differences). We considered this as our “full” model (Table 3, Model 1).

Table 3.

Model descriptions and fit statistics.

Model #	Description	−2logL	AIC	Number of parameters	Δ 2logL	p-value
1	Full model, including age regressions	1,596,781	−4,956,219	31
2	No age regression on AMC	1,600,828	−4,952,176	29	4047.531	<0.001
3	No age regression on AUD	1,597,215	−4,955,798	29	433.82	<0.001
4	Drop AMC-specific C paths	1,596,782	−4,956,222	29	1.45	0.485
5	Drop AMC-specific A paths	1,596,788	−4,956,216	29	6.73	0.035
6	Equate AMC AE paths + betas across sexes; drop AMC-specific C path	1,596,785	−4,956,223	27	4.50	0.343

AMC=alcohol-related medical conditions; AUD=alcohol use disorder; A=additive genetic variance; E=unique environmental variance; C=shared environmental variance; −2logL=−2 times the log likelihood; AIC=Akaike Information Criterion

From this structure, we next tested whether the age regression could be dropped from either the AMC component (Model 2) or the AUD component (Model 3). Both reduced model fit. We subsequently tested whether we could remove AMC-specific C variance (Model 4) or AMC-specific A variance (Model 5) from the full model. The former resulted in a model improvement in AIC, while the latter reduced fit. Finally, we tested whether we could remove AMC-specific C and equate AMC-specific A, E, and beta paths across sexes (Model 6). Model 6 provided the lowest AIC value of all models tested. We therefore selected this as the final model. We present parameter estimates for both the full and final models in Supplementary Table 1 for the sake of comparison. The Figure provides a graphic depiction of the final model, and Table 4 provides variance components for the final model. We were unable to obtain reliable standard errors for some estimates. Exploratory analyses suggested that this might be due to the inclusion of opposite sex sibling pairs; however, point estimates were quite comparable with or without these individuals, and we elected to retain them in the model due to their substantial contribution to the total sample size.

Figure.

Parameter estimates for women (in black text) and men (in grey text) from the final model (Table 3, Model 6). We were unable to obtain reliable standard errors for some paths.

A=additive genetics; C=shared environment; E=unique environment; AUD=alcohol use disorder; AMC=alcohol-related medical conditions

Table 4.

Variance components estimates from the final model.

Variance Components	A	C	E
Females
AUD	0.55	<0.01	0.45
AMC	0.32	<0.01	0.68
Males
AUD	0.51	0.05	0.44
AMC	0.30	0.02	0.68

In the final model, the beta path was estimated at 0.59. The total heritability of AMC was 0.32 for women and 0.30 for men. Of AMC’s total heritability, 39.6% was unique to AMC among women, while 41.3% was unique to AMC among men. C (conveyed via the beta path from its influence on AUD) accounted for ≤5% of the total variance in AMC for either sex. In contrast, the majority of environmental variance contributing to AMC liability was specific to AMC (76.7% for women and 77.2% for men).

Sensitivity analysis.

Given concerns about the heterogeneity of our AMC variable, we conducted a sensitivity analysis including only individuals with a liver-related AMC (AMC_liver). Nearly two-thirds of women with AMC (64.6%, N=2715) had a liver-related condition, compared to slightly more than half (53.7%; N=6738) of men with an AMC. Details on sample sizes, model fit statistics, and variance components are available in the Supplemental Tables 2–4. In the final model, the total heritability of AMC_liver was 0.38 in women and 0.37 in men; C variance conveyed through the beta path from AUD accounted for ≤5% of total variance. In contrast to the more inclusive definition of AMC, the heritability of AMC_liver was primarily due to AMC_liver-specific variance: among women, 70.0% of the heritability was unique to AMC_liver, with a corresponding value of 71.7% among men. Consistent with results for AMC overall, the vast majority of environmental variance contributing to AMC_liver liability was unique to AMC_liver in both sexes (84.5%).

Discussion

We examined the genetic and environmental structure of liability to alcohol-related medical conditions among individuals with an alcohol use disorder in a large, genetically informative Swedish cohort. Our methodological approach accounted for genetic and environmental influences on AUD, and thereby enabled us to quantify risk attributable to distinct sources. We found that AMC is moderately heritable (30-32%), and that a non-trivial proportion (~40%) of total heritability is independent of genetic factors influencing AUD. AMC liability was predominantly due to environmental influences, which were largely (77%) specific to AMC. When restricted to liver-related AMC, the specificity of genetic and environmental factors increased, to ~70% and 85%, respectively. These findings provide previously unavailable insight to the etiology of AMC. Most notably, genetic and environmental liability to AMC is only modestly correlated with the corresponding influences on AUD.

Genetic contributions to AUD have long been recognized (Cloninger et al., 1988, Cloninger et al., 1981); however, far fewer studies have focused on whether liability to the medical sequelae of AUD is due in part to genetic factors. Reed and colleagues (Reed et al., 1996) employed biometric models in their study of a US male veteran cohort assessed for alcoholism and its organ-specific medical sequelae and reported little to no genetic or environmental variance specific to medical outcomes. The current study differs in important ways: i) we have access to a far larger sample; ii) we included both sexes; and iii) our model specifies a contingency pathway from AUD to AMC. Our findings contrast previous studies, most notably in our identification of AMC-specific influences, underscoring the need for very large samples when examining low-prevalence outcomes.

More recently, AMC have been the target of gene identification efforts. These efforts have largely been restricted to candidate genes (Israelsen et al., 2020, Beaudoin et al., 2021), which are not reflective of the likely polygenicity of AMC. However, several genomewide association studies, which employed different methods to account for the contingency of AMC on alcohol problems, identified a limited number of loci associated with alcoholic chronic pancreatitis (Derikx et al., 2015) and alcohol-related cirrhosis (Innes et al., 2020, Schwantes-An et al., 2021). Furthermore, (Emdin et al., 2021) found that individuals with high polygenic liability for cirrhosis (alcohol-related or otherwise) were at increased risk if they were heavy alcohol users. These analyses did not report single nucleotide polymorphism-based heritability estimates, but their findings are consistent with the current results in their support for genetic risk factors for AMC that act independently of AUD.

Despite prior evidence that women with AUD are at higher risk for developing AMC than their male counterparts, we did not observe striking sex differences. First, men with AUD were slightly more likely to develop AMC than were women, though this effect was not strong. Second, we observed very similar parameter estimates across sexes in the full model, prior to constraining paths (see Table 4). Third, we were able to equate genetic and environmental paths, and the beta path from AUD to AMC, across women and men, without a detriment to model fit. It remains possible that we lacked sufficient power to detect quantitative sex differences; however, the current sample size is considerably larger than most if not all prior studies. Identifying a sample better suited to detect sex differences may be a challenge, but the increasing availability of large biobanks holds promise for further interrogation of genetic variation by sex.

The moderate beta path from AUD to AMC, which was reduced in our sensitivity analysis of AMC_liver (from beta=0.59 to beta=0.46), in conjunction with AMC-specific genetic and environmental variance, provides insight to the observation that only a minority of individuals with AUD suffer medical consequences. In particular, environmental factors account for a substantial fraction of risk – 62-68% in our sensitivity and primary analyses. These exposures may include other medical comorbidities, lower health care utilization (e.g., lack of preventative care), diet, or other factors. Future studies aimed at identifying features that distinguish individuals with AUD who do versus do not develop AMC are therefore critical to the identification of prevention targets.

The current findings have implications for gene identification efforts and for clinicians. First, they confirm the relevance of AMC-specific genetic factors. Molecular studies will be necessary to implicate individual loci and further dissect AMC-specific etiology (e.g., via pathway analysis). The availability of novel statistical genetic approaches such as GenomicSEM (Grotzinger et al., 2019) and GWAS-by-subtraction (Demange et al., 2021) may be useful in this regard. In the near term, while personalized genomic medicine is unlikely to be more informative than family history for use in a clinical setting for AMC outcomes, the use of molecular genetic approaches has the potential to extend our findings and elucidate the biological mechanisms underlying risk. As molecular genetic studies of alcohol consumption continue to improve in statistical power, their results could be used in subsequent analyses pertaining to AMC. For example, variants that are reliably implicated in genetic studies of consumption could potentially be used as instruments in Mendelian randomization analyses to assess the causal impact of consumption (distinct from full-fledged AUD) on AMC. Environmental exposures may also prove to be a fruitful area for future study: Not only do environmental factors account for a large proportion of AMC liability, they also have the advantage of being potentially modifiable. Finally, the moderate causal pathway from AUD to AMC underscores the fact that ongoing efforts to reduce the prevalence of AUD will have an overall positive impact, reducing the personal and economic toll of AMC as well.

Our study must be considered in the context of several methodological limitations. First, our primary analyses collapsed a wide range of AMC into a single outcome, potentially obscuring important etiologic differences, for example across different organs impacted by heavy alcohol use. We attempted to attenuate this issue through our sensitivity analysis of liver-related AMC, which comprised the largest somewhat homogeneous subgroup of medical sequelae. We observed only modest changes in variance components between AMC overall and AMC_liver, but the proportion of AMC-specific genetic and environmental variance increased in the AMC_liver group. Future studies should distinguish among various AMC outcomes if statistical power allows for it; this may prove challenging, since as demonstrated in the current study, rare phenotypes may contribute to fluctuations in concordance and imprecise estimates.

Second, risk of AMC is likely a function of alcohol consumption, at least to some degree. While consumption and alcohol problems are positively correlated, both phenotypically (Dick et al., 2011) and genetically (Sanchez-Roige et al., 2019, Kranzler et al., 2019), we do not have data on alcohol consumption due to our reliance on registry resources. We are thus unable to address whether, for example, the beta path would differ if AUD were replaced with alcohol consumption, or whether we might detect more pronounced sex differences if AUD were replaced with alcohol consumption. These issues should be considered high priority in future studies.

Third, the coverage for Swedish national registries varies across years and municipalities. It is possible that some cases of AUD and AMC were not detected, particularly for early-onset problems among older cohorts, where coverage was sparser. However, the investigation of younger cohorts may be more relevant as environmental influences may change over time. Our findings therefore warrant replication in other large, representative samples. Finally, while the availability of socialized medical care in Sweden reduces concerns about representativeness as function of accessibility within Sweden, it is possible that the relationship between AUD and AMC might differ in countries without socialized medicine, such as the United States. Reduced access to AUD treatment, for example among socioeconomically disadvantaged individuals, could impact the AUD→AMC association and lead to different conclusions on shared versus distinct etiologic factors.

In summary, the current study clearly indicates that genetic and environmental factors independent of those underlying AUD liability contribute to risk of AMC. Further studies are necessary to identify specific genetic loci and determine how they contribute to etiology, potentially through metabolic pathways or physiological response to toxins or tissue damage. The impact of AMC-specific environmental factors offers an important opportunity for follow-up studies to identify exposures that distinguish individuals with AUD who do versus do not develop AMC and determine whether these may be viable prevention targets.