Orbitofrontal cortex thickness and substance use disorders in emerging adulthood: Causal inferences from a cotwin control/discordant twin study

Abstract

Background/aims.

Research linking orbitofrontal cortex (OFC) structure and substance use disorders (SUDs) is largely correlational and often implies a causal effect of addiction/substance exposure on the brain, but familial risk factors (e.g., genetic liability) may confound these associations. We tested whether associations between alcohol, cannabis, and tobacco use disorders and OFC thickness reflected the potential causal effects of familial risk or SUDs-related consequences (e.g., substance exposure).

Design.

A cotwin control/discordant twin design separated familial risk confounding from SUDs-related consequences.

Setting/participants.

A population-based sample of 436 24-year-old twins (62% monozygotic) from the Minnesota Twin Family Study, USA.

Measurements.

Alcohol, cannabis, and tobacco use disorders were assessed using the Composite International Diagnostic Interview – Substance Abuse Module. Cortical thickness of the medial and lateral OFC (mOFC and lOFC, respectively) was assessed using MRI.

Findings.

Lower mOFC (p-values ≤0.006) but not lOFC (p-values ≥0.190) thickness was observed in diagnosed individuals (n = 185) relative to non-SUDs controls (n = 251). Cotwin control analyses offered evidence that mOFC associations were consistent with familial risk across SUDs (between-pair effect: p-values ≤0.047) and the independent consequences of having an alcohol or cannabis use disorder (within-pair effect: p-values ≤0.024). That is, within alcohol/cannabis discordant twin pairs, affected twins had significantly lower mOFC thickness compared with their unaffected cotwins.

Conclusions.

A confounder adjusted analysis of the Minnesota Twin Family Study appeared to indicate that, beyond a substance use disorders general familial risk effect, the experience of an alcohol or cannabis use disorder in emerging adulthood reduces the thickness of the medial orbitofrontal cortex, a region associated with value-guided decision making.

Introduction

Addiction and substance use disorder (SUDs) are leading public health concerns for young adults. Emerging adulthood (ages 18 to 25) represents an important developmental period (1) involving continued prefrontal cortex development (2,3) alongside peak lifetime rates of SUDs (4–6), which may create a vulnerable period where the still-developing young adult brain is particularly sensitive to the potential deleterious consequences of SUDs.

Two core psychological attributes of SUDs are the disruption of value-guided decision making and contingent learning processes (7–11), likely mediated by the medial and lateral orbitofrontal cortex (OFC), respectively (12–14). Prominent models suggest a fundamental involvement of the OFC (and its related cognitive functions) across forms of addiction and compulsive drug seeking behaviors given its crucial role in modulating the value of reinforcers (10). Evidence from cross-sectional studies has established a reliable association between SUDs and OFC structural deviations (15–19). Lower medial OFC (mOFC) thickness is observed across alcohol, tobacco, cannabis, and other forms of illicit SUDs, possibly reflecting a biological index of addiction in general, whereas lower lateral OFC (lOFC) thickness is primarily found in alcohol use disorder (16).

The vast majority of prior research in this area has been conducted using cross-sectional, case-control research designs that cannot make strong causal inferences, and yet models often implicitly or explicitly assume a deleterious substance exposure effect on the brain. Individuals with SUDs typically differ from those without on important familial characteristics (e.g., genes, rearing environment) that may confound the relationship between SUDs and brain outcomes (20) – association but not causation. Despite significant public health implications of understanding these potential causal sources (e.g., targeted preventions, policy decisions), the causal nature of SUDs-OFC associations remain unclear (as discussed in (10)).

Does lower OFC thickness reflect a consequence of addiction (e.g., chronic substance exposure, increased stress response (7)), an endophenotype of SUDs-related genetic liability (20), or a mixture of environmental and familial factors? In psychopathology research with humans, one cannot conduct a randomized controlled trial to observe the consequences of SUDs on OFC thickness by randomly assigning individuals to either a SUDs or no-SUDs (or substance use) condition. This is clearly not feasible, but as discussed by Thapar and Rutter (21), there are alternative strategies to testing causal hypotheses in psychopathology research. “Natural” quasi-experiments approximate true experiments by attempting to address the problem of confounding between exposure, familial risk, and outcome in an a priori manner through research design to maximize causal inference within observational/correlational frameworks (22,23).

One such design is the cotwin control (CTC) analysis or discordant-twin-pair design (24,25), which uses twins as ideal genetic and shared environmental controls to more stringently test for SUDs exposure effects (unconfounded by familial influences) than is possible with cross-sectional or longitudinal studies of genetically unrelated individuals. In this design, outcomes of SUD unaffected twins provide a close approximation of the expected outcomes for the SUDs affected cotwins had they been unaffected (analogous to observing the counterfactual; (22)). For example, if lower OFC thickness reflects a deleterious consequence of alcohol use disorder (e.g., heavy alcohol exposure), then within discordant twin pairs, the affected twin is expected to have thinner cortex relative to their unaffected cotwin. In contrast, if familial risk accounts for the observed effects, then cortical thickness should be comparable even between discordant cotwins.

To our knowledge, no study to date has tested the degree to which the relationship between SUDs and the OFC during emerging adulthood reflects a SUDs consequence effect or a preexisting liability. To address this important question, we tested the effects of alcohol, cannabis, and tobacco use disorders, representing three of the most common addictive substances used by emerging adults in the United States (6), on mOFC and lOFC thickness in a large population-based, etiologically-informative sample of 24-year-old twins. Guided by previous work (16), we hypothesized that individuals with a SUD would show lower OFC thickness (primarily in the mOFC) relative to participants with no SUDs. Given suggestive evidence that substance use may have sex-specific effects on the brain (26,27) and the need for more well-powered neuroimaging studies on this topic (28), we also evaluated whether SUD effects differed between women and men. Of primary interest was the use of the CTC analysis to assess the relative contribution of familial risk and potential consequences of SUDs on OFC thickness in twins discordant or concordant for SUDs.

Methods

Sample

Participants were same-sex twins assessed at the target age of 24 from the population-based Minnesota Twin Family Study Enrichment Study (29). By design (participants were able to complete in-person MRI assessments; met standard MRI safety exclusions), 441 individuals underwent structural MRI scans. Data from individuals with clinically significant brain anomalies (determined by a clinical radiologist) and one with coil failure during scanning were excluded from analysis. The final sample included 436 individuals (age: mean [SD] = 24.3 [0.8] years; 254 women; racial composition: 92.2% White/Caucasian, 2.8% Black/African American; 2.5% Hispanic; 1.4% mixed/other; 0.7% Native American; 0.5% Asian/Pacific Islander), with 270 monozygotic (120 complete pairs) and 166 dizygotic (66 complete pairs) twins.

Substance use disorder assessment

An expanded version of the Substance Abuse Module of the Composite International Diagnostic Interview (30) was administered by trained interviewers to assess Diagnostic and Statistical Manual of Mental Disorders (DSM), 5th edition (31) SUDs criteria. Symptom presence was assigned using a consensus approach by pairs of individuals with advanced clinical training. DSM-5 alcohol, cannabis, or tobacco use disorder diagnoses were defined by endorsement of two or more symptoms (clustering within the same 12-month period in the past seven years to cover the emerging adulthood period and interval since prior assessment) of the respective criteria (cannabis data was missing for one individual). A control group was formed of those who were not diagnosed with any alcohol, cannabis, or tobacco use disorder. Three (non-mutually exclusive) SUD groups were formed for those diagnosed with an alcohol, cannabis, or tobacco use disorder; participants could be included in more than one SUD group. For convenience, these are collectively referred to as SUDs in this report.

Neuroimaging acquisition and processing

Structural MRI data were collected on 3T Siemens Trio (n = 100) and Prisma (n = 336) MRI scanners (32-channel array head coil) at the Center for Magnetic Resonance Research, University of Minnesota. A software upgrade occurred during the study (n: pre-upgrade = 306, post-upgrade = 130). Three-dimensional T₁-weighted sagittal plane anatomical images were acquired using the following magnetization prepared rapid gradient echo sequence: TR = 2530 ms; TE = 3.65 ms; flip angle = 7°; matrix size = 256 × 256; FOV = 256 mm; GRAPPA = 2; 240 coronal slices with 1-mm isotropic voxels; single shot; interleaved acquisition. All images were normalized and manually reviewed for artifacts/structural anomalies, processed using the standard Freesurfer pipeline (version 5.3.0) (32,33), and segmented using the Desikan-Killiany atlas (34) to calculate cortical thickness of the mOFC and lOFC.

Statistical analyses

Linear mixed models (LMMs; lme4 package (35)) were fit in R (36) with family-level random intercepts to adjust for within-twin-pair correlations in dependent measures and Kenward-Roger approximated denominator degrees of freedom (lmerTest package (37)). Residual-bootstrapped 95% basic confidence intervals (2000 draws; clustered by family) were calculated (bootmlm package (38)). All models included sex, age, zygosity, scanner, and acquisition software as covariates. As we had no a priori hypothesis regarding hemisphere effects, to reduce Type I error likelihood, total (summing across left/right hemispheres) mOFC and lOFC thickness scores were used in the primary analyses. Results by hemisphere are reported in the Supplement for completeness. Analyses were not pre‐registered; results should be considered exploratory. The a priori alpha rate equaled 0.05.

First, we compared the phenotypic (individual-level) differences in mOFC or lOFC total thickness between participants with either an alcohol, cannabis, or tobacco use disorder (n = 122, 94, and 104, respectively) and non-SUD controls (n = 251). Separate LMMs were computed for each pairwise combination of OFC thickness (medial, lateral) and SUD; thickness was the dependent measure and the binary SUD variable (e.g., 0 = control, 1 = alcohol) was the fixed effects predictor. Unstandardized betas and Cohen’s d (derived from the LMM t-statistic to adjust for covariates) are reported as effect size estimates. Further analyses were performed to test for sex by SUD interactions.

For each significant individual-level association, follow-up cotwin control analyses (using the 186 complete twin pairs) evaluated causal consequence and familial risk effects of having a SUD on OFC thickness by utilizing twins as ideal genetic and shared environmental controls to adjust for all measured and unmeasured sources of familial confounding. Outcomes (e.g., mOFC thickness) were compared between members of a twin pair; for example, if twins were discordant for alcohol use disorder, the outcome of the unaffected twin provides a close approximation of the expected outcome (unobserved counterfactual (22)) for the affected twin had she/he been unaffected.

The CTC for a binary measure is a variant of the discordant-twin-pair design (39). The individual-level predictor (binary diagnosis variable [0 = unaffected, 1 = affected] for each SUD) was decomposed into two orthogonal components: 1) the twin-pair mean score (between-pair effect), indexing all familial vulnerability influences (genetic; shared environment), whether measured or unmeasured; and 2) an individual’s deviation from their respective twin-pair mean (within-pair effect), reflecting the nonshared environmental consequence effects of having a SUD. The between-pair effect is the average of the two dummy variables coding for presence (X_ij = 1) or absence (X_ij = 0) of a SUD (e.g., alcohol use disorder) in each twin, i, in twin pair j, which is an analog to the mean “rate” of disorder presence in each twin pair. There are only three possible values of the twin-pair mean score: 0 = neither twin experienced the disorder (concordant unaffected); 0.5 = twins were discordant for the disorder; and 1 = both twins experienced the disorder (concordant affected). The within-pair effect is the value of each twin’s deviation from his/her respective twin-pair mean, which also has three possible values: 0 = both twins in concordant pairs; 0.5 = the affected twin in discordant pairs; and −0.5 = the unaffected twin in discordant pairs. Only discordant twin pairs affect estimates of the within-pair effect in the CTC models. The distributions of these scores for all three SUDs are shown in Supplemental Figure S1. LMMs were fit with thickness as the dependent variable and the within-pair and between-pair terms as independent variables. A significant between-pair effect would be consistent with familial risk influencing both the SUD and cortical thickness. A significant within-pair effect would be consistent with variations in thickness reflecting the potential consequence of having the SUD (e.g., within discordant pairs, the affected twins exhibiting decreased thickness relative to their unaffected cotwins), unconfounded by all familial factors influencing SUD risk (40). We compared the magnitude of significant within-pair effects between MZ (100% genetic control) and DZ (50% genetic control) twin pairs with a zygosity by within-pair interaction; statistically comparable MZ/DZ effects would be strongly consistent with a SUD consequence effect unconfounded by familial influence (24).

Results

Descriptive statistics

Descriptive statistics for OFC thickness and alcohol, cannabis, and tobacco use disorder prevalence rates are presented in Table 1. In terms of SUDs, 53 (29%) and 41 (22%) individuals met criteria for two or all three SUDs, respectively, and 251 did not meet criteria for any (i.e., the control group). As shown in Table 1 (and Supplementary Table S1), there were moderate to large familial influences (with tobacco being the most heritable) and appreciable within-pair differences in SUD prevalence, which supported the use of the cotwin control analysis in this sample.

Table 1.

Descriptive statistics, twin correlations, and standardized biometric estimates.

	Mean (SD)	Twin correlations (95% CI)		Biometric estimates (95% CI)
		MZ	DZ	A	E
Medial OFC thickness	4.49 (0.27)	0.54 (0.41, 0.64)	0.24 (0.04, 0.41)	0.51 (0.37, 0.62)	0.49 (0.38, 0.63)
Lateral OFC thickness	5.01 (0.22)	0.54 (0.42, 0.65)	0.38 (0.20, 0.54)	0.56 (0.43, 0.66)	0.44 (0.34, 0.57)
	Total N
Alcohol Use Disorder
Cases	122	0.58 (0.28, 0.77)	0.37 (−0.06, 0.69)	0.59 (0.36, 0.82)	0.41 (0.18, 0.64)
Controls	314
Cannabis Use Disorder
Cases	94	0.76 (0.51, 0.90)	0.14 (−0.41, 0.61)	0.74 (0.55, 0.95)	0.26 (0.06, 0.46)
Controls	341
Tobacco Use Disorder
Cases	104	0.81 (0.58, 0.92)	0.57 (0.18, 0.81)	0.82 (0.68, 0.96)	0.18 (0.04, 0.32)
Controls	332

Notes: Medial and lateral OFC thickness are the sum of left and right hemisphere thickness. Prevalence rates for the disorders refer to the number of individuals that met criteria (cases) or did not meet criteria (controls) for a given diagnosis. See Supplementary Table S1 for details on biometric model fit statistics. For the twin correlations and biometric estimates, OFC measures were adjusted for sex, age, scanner, and acquisition software, and substance use disorder measures (0 = control, 1 = case) were adjusted for sex and age.

Abbreviations: OFC, orbitofrontal cortex; MZ, monozygotic; DZ, dizygotic; CI, confidence interval; A, additive genetic variance; E, nonshared environmental variance.

Individual-level associations between substance use disorder diagnosis and OFC thickness

As reported in Table 2 and shown in Figure 1, individuals with an alcohol, cannabis, or tobacco use disorder exhibited significantly lower total mOFC cortical thickness relative to non-SUD controls. In contrast, no significant differences were found for total lOFC thickness. Results did not statistically differ by hemisphere and the same pattern was observed in the right and left hemispheres (Supplemental Table S2).

Table 2.

Individual-level associations between alcohol, cannabis, or tobacco use disorders and medial or lateral orbitofrontal cortex (OFC) thickness.

	Medial OFC					Lateral OFC
Model	Beta (95% CI)	SE	t(df)	p	Cohen’s d	Beta (95% CI)	SE	t(df)	p	Cohen’s d
Alcohol	−0.107 (−0.169, −0.043)	0.032	−3.339(362)	<0.001	−0.374	−0.035 (−0.085, 0.017)	0.027	−1.312(366)	0.190	−0.146
Cannabis	−0.103 (−0.171, −0.036)	0.035	−2.928(332)	0.004	−0.361	−0.020 (−0.078, 0.038)	0.029	−0.673(333)	0.502	−0.083
Tobacco	−0.097 (−0.161, −0.026)	0.035	−2.758(321)	0.006	−0.338	−0.015 (−0.069, 0.044)	0.030	−0.507(362)	0.613	−0.062

Note: Results reflect the main effect comparing the respective substance use disorder group (coded as 1) to the control group (no alcohol, cannabis, or tobacco use disorder diagnoses; coded as 0). Separate linear mixed models were constructed for each pairwise combination of orbitofrontal cortex thickness (medial or lateral) and substance use disorder (alcohol, cannabis, or tobacco). All statistical models included scanner, acquisition software, sex, age, and zygosity as covariates. Medial and lateral regions are the sum of the respective left and right hemisphere thickness scores. Significant effects (p < 0.05) are in bold.

Figure 1.

Right panel. Group differences in total medial orbitofrontal cortex (OFC) thickness between control participants (no substance use disorders [SUDs]) and participants with an alcohol, cannabis, or tobacco use disorder. The violin plots show the overall distribution (kernel density plot) of medial OFC thickness for each group. The white circles represent the conditional mean (from the linear mixed models, adjusting for covariates) and solid bars represent ± 1 standard error of the mean for each group; the dashed lines represent the error bar boundaries for the No SUD control group. Medial OFC thickness was lower in each of the alcohol, cannabis, and tobacco use disorder groups relative to control participants. Left panel. Same as the right panel, but for total lateral OFC thickness. Lateral OFC thickness was comparable between the alcohol, cannabis, and tobacco groups and control participants.

The approximately equal number of women and men in this sample allowed us to test for sex differences in the OFC-SUD associations. As expected, men were more likely than women to have an alcohol, cannabis, or tobacco use disorder (χ²₍₁₎ ≥ 18.885, p-values < 0.001). However, a sex by diagnosis interaction was not significant for mOFC or lOFC (p-values range: 0.140–0.987), indicating that the SUDs-OFC effects did not statistically differ by sex.

Cotwin control analysis of SUD diagnosis on medial OFC thickness

The CTC analysis was used to separate familial risk influences from deleterious environmental consequence effects on the negative associations between mOFC thickness and each SUD. Results of the CTC analysis are reported in Table 3.

Table 3.

Cotwin control analysis of alcohol, cannabis, or tobacco use disorder on medial orbitofrontal cortex (OFC) thickness.

	Within-pair effect				Between-pair effect
Model	Beta (95% CI)	SE	t(df)	p	Beta (95% CI)	SE	t(df)	p
Alcohol
Medial OFC	−0.095 (−0.171, −0.018)	0.041	−2.335(185)	0.021	−0.105 (−0.198, −0.016)	0.046	−2.276(179)	0.024
Cannabis
Medial OFC	−0.108 (−0.199, −0.015)	0.047	−2.284(184)	0.024	−0.103 (−0.206, −0.001)	0.052	−2.005(178)	0.047
Tobacco
Medial OFC	−0.055 (−0.151, 0.040)	0.049	−1.120(185)	0.264	−0.171 (−0.257, −0.084)	0.044	−3.914(179)	<0.001

Notes: Separate cotwin control models were computed for each substance use disorder. All statistical models included scanner, acquisition software, sex, age, and zygosity as covariates. Medial OFC thickness is the sum of left and right hemisphere thickness. Significant effects (p < 0.05) are in bold.

The between-pair effect (reflecting genetic and environmental influences shared by members of a twin pair) had a significant negative association with mOFC thickness for all three disorders. The within-pair effect (reflecting nonshared environmental effects unconfounded by familial influence) had a significant negative association with mOFC thickness for both alcohol and cannabis use disorders. While in the expected negative direction, mOFC thickness was statistically comparable between cotwins discordant for tobacco use disorder. Consistent with expectations for a within-pair effect, the MZ and DZ within-pair effects were statistically equivalent for both the alcohol and cannabis CTC models; adding a zygosity by within-pair effect interaction did not improve model fits (Δχ²₍₁₎ ≤ 0.729, p-values ≥ 0.393) and the interaction terms were non-significant (p-values ≥ 0.397).

Figure 2 depicts mOFC thickness as a function of twin discordance. This illustrates the within-pair CTC finding that, after accounting for all familial confounding, the mOFC thickness of alcohol or cannabis use disorder affected twins was significantly thinner than that of their unaffected cotwins, consistent with a deleterious environmental consequence effect.

Figure 2.

Total medial orbitofrontal cortex (OFC) thickness as a function of twin discordance for alcohol, cannabis, and tobacco use disorder diagnosis. To illustrate the within-pair cotwin control analysis effects, the conditional mean (and ±1 standard error of the mean bars) for medial OFC thickness is plotted for unaffected and affected twins from twin pairs that were discordant for an alcohol, cannabis, or tobacco use disorder. The p-values correspond to the within-twin-pair effect from the cotwin control analyses. Within twin pairs discordant for an alcohol (n = 47 pairs) or cannabis (n = 35 pairs) use disorder, the medial OFC thickness of affected cotwins was significantly thinner relative to their unaffected cotwins. This suggests that after accounting for all familial risk confounding shared by members of a twin pair, the presence of an alcohol or cannabis use disorder confers a deleterious environmental consequence effect on medial OFC thickness. The direction of the effect for twin pair discordance in tobacco use disorder (n = 33 pairs) was negative but not significant, suggesting that medial OFC thickness was statistically equivalent between the unaffected and affected cotwins.

Cotwin control follow-up analyses

We recomputed the alcohol and cannabis CTC models to test whether the significant alcohol and cannabis within-pair effects on mOFC thickness held after accounting for several factors that may potentially confound within-pair effects, namely subthreshold cases, unshared confounding from twin differences in comorbid SUDs, and recent substance use. Specifically, for alcohol discordant twin pairs, twin pairs were excluded from the CTC if the twin who did not meet diagnostic threshold for alcohol use disorder did endorse one symptom (i.e., subthreshold case); excluding these pairs resulted in twin pairs for the CTC that were ‘truly’ discordant. The same subthreshold exclusion criteria were applied for the cannabis CTC. In addition, while the CTC controls all shared confounders, it cannot account for unshared confounders (influences unique to one member of a twin pair) that may in turn confound within-pair effects. For example, the within-pair effect for alcohol use disorder on mOFC thickness could potentially be influenced by twin discordance in other SUDs. To address this, we used a within-pair mediation approach (41) to test whether twin differences in comorbid SUDs accounted for the observed alcohol or cannabis within-pair effects. Specifically, the within-pair scores for the other two SUDs were included as joint predictors of mOFC thickness in the alcohol and cannabis CTC models to test whether the observed within-pair effects remain significant after adjusting for the potential individual-specific confounding from co-occurring SUDs. Finally, past week alcohol and cannabis use were included in the CTC models to account for any effect of recent substance use (see Supplement for descriptives). In all, the follow-up CTC models were computed using 157 complete twin pairs (28 discordant pairs) for the alcohol CTC, and 161 complete twin pairs (24 discordant pairs) for the cannabis CTC.

After removing subthreshold cases and adjusting the alcohol CTC model for the aforementioned confounders, the alcohol within-pair effect on mOFC thickness remained significant (Beta [95% CI] = −0.121 [−0.231, −0.013], t(161) = −2.208, p = 0.029). The cannabis within-pair effect also remained significant (Beta [95% CI] = −0.145 [−0.274, −0.019], t(165) = −2.239, p = 0.027) after excluding subthreshold cases and adjusting the cannabis CTC model for confounders. This suggests that the within-pair effects of an alcohol or cannabis disorder were robust, independent, and could not be attributed to potential unshared confounding of comorbid SUDs or recent alcohol/cannabis use.

Discussion

The current report tested the associations between alcohol, cannabis, and tobacco use disorders during emerging adulthood and medial and lateral OFC thickness in a large population-based sample of 24-year-old twins. Structural differences in the OFC have been frequently linked to SUDs (15–18), yet the nature of this association has been unclear as disentangling familial risk from environmental effects is difficult in observational research without genetically-informative samples and quasi-experimental designs (21). Using a “natural” quasi-experimental CTC/discordant twin design to help draw causal inferences, results of this study provide new evidence that lower mOFC thickness likely reflects both a brain-based expression of the familial risk for all three SUDs and the deleterious environmental consequence of alcohol and cannabis use disorder on the brain.

Individual differences in the structure and function of the OFC have been frequently linked to SUDs (15–18,42). Replicating these findings, we found that alcohol, cannabis, and tobacco use disorders were all associated with lower thickness of the mOFC. In contrast, we found no significant effects for the lOFC, which has been related to credit assignment/contingent learning processes (12). This is in line with some prior work suggesting a null relationship between most forms of SUDs and lOFC structure (15–17), although there is suggestive evidence that lOFC thickness may be lower in individuals with only alcohol dependency (16). Human lesion studies offer evidence that the mOFC is involved in value-guided decision-making processes and filtering distracting/irrelevant salient options to make the optimal or appropriate choice (12). Individual differences in mOFC structure may relate to clinical characteristics of SUDs such as disadvantageous decision making/anticipatory planning when faced with a choice between several valued options (e.g., relapse risk) and increased salience attribution/motivation (e.g., craving) toward substances and substance-related cues despite negative long-term consequences (7,11). This is consistent with studies reporting poor neuropsychological performance (43) and lower mOFC/reward network fMRI activity (28) during decision making tasks (e.g., Iowa gambling task) in SUDs. However, it is important to note that the relationship between structural variation and function is far from straightforward; further work is needed to fully understand the complex relationship between cortical thickness, decision making, and real-world substance use problems.

Results from the CTC analysis, which utilizes twins as ideal genetic and shared environmental controls, suggested a deleterious consequence effect of alcohol and cannabis use disorder on lower mOFC thickness (within-pair effect). When comparing members of twin pairs discordant for an alcohol or cannabis use disorder, which controls for all sources of familial confounding shared by members of a twin pair, affected twins had significantly lower mOFC thickness relative to their unaffected cotwins. In contrast, while the tobacco use disorder within-pair effect was in the expected direction (affected twin mOFC thickness < unaffected twin), this finding was not statistically significant. Importantly, the alcohol and cannabis within-pair effects held when accounting for twin differences in comorbid SUDs and recent alcohol and cannabis use, suggesting that they cannot be readily attributed to these potentially confounding individual-specific factors. Findings suggest that experiencing an alcohol or cannabis use disorder in emerging adulthood confers a deleterious environmental effect on mOFC thickness as early as age 24, offering support for emerging adulthood as a vulnerable developmental period that warrants increased attention in addiction research.

While the cotwin control results offer evidence that alcohol or cannabis use disorders may lead to lower mOFC thickness, they do not necessarily implicate a specific causal mechanism. The within-discordant-pair effects may reflect the causal effects of alcohol or cannabis exposure; alcohol is a neurotoxin at high doses, and chronic Δ9-THC exposure may affect the cannabinoid receptor modulation of the OFC, itself a relatively cannabinoid receptor-rich area (44) that influences decision making and goal directed behaviors through the endocannabinoid system (in rodents) (45). Or rather than reflecting a direct neurotoxic consequence, alcohol or cannabis use disorders may cause increased stress and cortisol levels, risky behaviors, or other physical/emotional psychosocial effects that may in turn affect mOFC structure. The lack of statistically significant within-pair tobacco use disorder effect may be attributed to several factors. Preclinical rodent work suggests that the adult brain may be less sensitive to possible nicotine-related neurotoxicity than the adolescent brain (46–49); future work using etiologically-informative adolescent human samples is needed. Nicotine has a high affinity with α4β2 nicotinic acetylcholine receptors (nAChR); if nicotine does confer a neurotoxic effect, it may be more readily seen in α4β2 nAChR enriched regions, such as the cingulo-insular network, thalamus, and hippocampus, compared to those with relatively low density, such as the OFC (50,51).

While the CTC offers evidence in support of an environmental exposure effect, we caution that the CTC is not without limitations and cannot definitively establish causality or rule out reverse causation (for a discussion, see (24)). The current study is cross-sectional, and longitudinal genetically-informative studies can help make further causal claims by establishing the temporal sequence of changes in mOFC as a function of alcohol or cannabis use disorder presence. Nevertheless, if confirmed through further research, the finding that alcohol and cannabis use disorders during emerging adulthood confer deleterious effects on the mOFC has potentially significant public health implications given the alarmingly high rates of these disorders during this developmental period (6).

The cotwin control analysis also provided evidence for mOFC effects that are likely not due to the consequences of having a SUD, but instead reflect the brain-based expression of the vulnerability toward alcohol, cannabis, and tobacco use disorders (between-pair effect). Specifically, mOFC thickness decreased as the amount of “risk” within a twin pair (number of diagnosed twins) for an alcohol, cannabis, or tobacco use disorder increased. That is, for each SUD, mOFC thickness was greatest for concordant unaffected twin pairs (lowest risk) followed by discordant affected twin pairs (intermediate risk) and concordant affected twin pairs (highest risk). The between-pair effect was significant for all three SUDs, suggesting that lower mOFC thickness may be a premorbid characteristic (endophenotype; (52,53)) that confers a general familial risk (e.g., genetic liability) for SUDs. These effects are consistent with prior work in adolescents suggesting that lower OFC volume indexes the familial risk for alcohol use (54), is observed prior to cannabis use initiation (55), and is associated with polygenic risk for tobacco smoking initiation (56). Given the high degree of comorbidity between SUDs and the broad externalizing spectrum (20,57), individuals with this brain-based risk factor are likely at heightened risk for other negative outcomes (e.g., impulsive/antisocial behavior; internalizing psychopathology; psychosocial impairment) (58–60).

A major strength of this study is its potential generalizability to the community at large by use of a population-based representative sample of 24-year-old young adults rather than the clinical/treatment-seeking samples often used in SUDs research. There are also limitations to the current report. As noted above, future work using prospective etiologically-informative research designs and complementary imaging analyses can help assess whether mOFC deviations predate or follow SUDs onset/development (e.g., do OFC gyrification patterns, which are set early in development, predict later SUDs?). More work is needed to determine if mOFC structure is associated with SUDs-related genetic variants (e.g., polygenic risk scores) identified by genome-wide association studies. The CTC accounts for all confounding influence shared by twins but does not control for unshared factors that may relate to the exposure and outcome (24). While the present sample was designed to reflect the demographics of Minnesota in the target birth years, it is predominantly White/Caucasian; replication of these findings is needed in large, diverse samples. As noted above, the functional significance of the mOFC thickness findings is not known, and future work is needed to link SUD-related cortical thickness variations to personality and neurocognitive measures known to be associated with SUDs (e.g., impulsivity, risk taking). Despite both the large sample and fact that these twins were assessed at the end of emerging adulthood, a developmental period associated with peak levels of lifetime substance use/SUDs (5,6), it is possible that a tobacco use disorder within-pair effect may be detected in larger samples or chronic cases.

The current report, utilizing within- and between-pair differences in alcohol, cannabis, and tobacco use disorders as a “natural” quasi-experiment to isolate sources of genetic and environmental influence, provides strong and novel evidence that lower mOFC thickness appears to reflect both the familial risk toward SUDs (20,61) and the potential deleterious consequence of alcohol/cannabis use disorders. These complementary causal sources have important implications regarding etiological/developmental models of SUDs and public health, such as targeting individuals with this brain-based predisposition to help identify high-risk youth for prevention effects before substance use initiation/escalation and focusing public health messaging on the risk for insult to the still developing young adult brain associated with alcohol and cannabis use disorders. While replication is necessary, these findings represent an important contribution to our understanding of the effects of addiction on the brain.