shows that the regular smokers were significantly older and had significantly higher rates of all types of DSM-IV psychopathology compared to the nonregular smokers. The group means for each motive (computed using observed scores) also are shown. Regular smokers had significantly higher observed mean values for the coping and enhancement motives and significantly lower values for the conformity motives.
Descriptive Statistics for Regular Smoker and Nonregular Smoker Groups and Results of Statistical Tests for Between-Group Differences
Validity of the 4-Factor DMQ-R Structure
The 4-factor structure (with correlated factors) adequately fit the DMQ-R data in the regular and nonregular smoker groups. In regular smokers, CFI = 0.954, TLI = 0.987, RMSEA = 0.077, and in the nonregular smokers, CFI = 0.950, TLI = 0.985, and RMSEA = 0.075.
Tests for invariance of thresholds and factor loadings indicated that MI was not obtained, χ2
(adjusted df, 48) = 127.98, p
< 0.001. However, noninvariance was attributed to threshold differences between just 3 items: (i) How often do you drink to get high? (enhancement); (ii) How often do you drink because you feel more self-confident or sure of yourself? (coping); and (iii) How often do you drink to be sociable? (social). This indicated that the regular and nonregular smokers tended to use the response categories for these 3 items differently (Millsap and Yun-Tein, 2004
). The CFA model assumed the response categories were cut-points on an underlying continuous latent response distribution (Lubke and Muthén, 2004
; Muthén, 1984
). Higher threshold values for the regular smoker group indicated that compared to nonregular smokers, regular smokers had higher values on this underlying distribution before they chose a higher response category (e.g., almost never) over a lower response category (e.g., never).
When the thresholds for these 3 items were freely estimated across the groups, partial MI was obtained, χ2 (adjusted df, 43) = 56.65, p = 0.08. Notably, when we deleted these items from the model and again tested for MI, the thresholds and factor loadings for the remaining 17 items were fully invariant across the groups, χ2 (adjusted df, 43) = 57.14, p = 0.07. Here, we report results from the model with partial MI, because follow-up CFAs (that used only the 17 invariant items) yielded identical results.
Interactions of AD and Covariates With Smoking Status
Compared to the model where regression coefficients associated with zygosity, age, mood disorder, and externalizing problems and AD were freely estimated across the groups, constraining the coefficients did not result in a significant decrease in model fit, χ2 (adjusted df, 12) = 17.245, p = 0.1406. This indicated the size of the coefficients in regular smokers did not significantly differ from the size of the coefficients in the nonregular smokers; we did not find evidence for interactions of smoking status with AD or the covariates (i.e., there was no evidence that the relationships between the drinking motives and alcohol dependence differed between regular and nonregular smokers). The unstandardized coefficients (i.e., partial regression coefficients) for each variable and for each latent drinking motive factor are shown in .
Results From the Regression of the Drinking Motive Factors onto Zygosity, Age, Mood Disorder Problems, Externalizing Problems, and Alcohol Dependence. Unstandardized Partial Regression Coefficients Are Shown
Relationships of AD With the Drinking Motives
After adjusting for zygosity, age, and mood disorder and externalizing problems, AD diagnosis was significantly associated with higher scores on enhancement, coping, conformity, and social motives. Additionally, mood disorder problems were associated with higher scores on coping and conformity motives and lower scores on enhancement and social motives. Also, a 1-year increase in age was associated with a 0.035 decrease in enhancement scores and 0.019 decrease in coping scores. Finally, externalizing problems were associated with higher coping scores. In a follow-up regression analysis, we examined the specificity of each drinking motive as a predictor of AD. After adjusting for enhancement, social, and conformity motives, only coping was significantly associated with AD (z = 6.76, p < 0.001) in both groups.
Differences in Drinking Motive Scores Between Regular and Nonregular Smokers
The covariate-adjusted differences between group means are shown in . Positive values indicate larger mean values for the regular versus the nonregular smoker group. After adjusting for zygosity, age, mood disorder and externalizing problems, and AD, the means for enhancement, coping, and social motives in the regular smokers were larger than those in the nonregular smokers. Conversely, the mean for conformity in the regular smokers was significantly lower than that in the nonregular smokers.
Covariate-Adjusted Group Mean Differences Between Regular and Nonregular Smokers, Standard Errors (SE), z-Test Results
Genetic analyses are based on 837 twin pairs. Of these, 217 MZ and 152 DZ pairs were concordant for regular smoking; 170 MZ and 80 DZ pairs were concordant nonregular smoking; and 107 MZ and 111 DZ pairs were discordant for smoking status. shows the standardized estimates (A, C, and E) from the univariate twin models for each drinking motive as estimated for the entire sample (Panel 1), the MZ and DZ correlations for twin pairs concordant and discordant for smoking status (Panel 2) and the standardized estimates for each drinking motive from the univariate twin models stratified by smoking status (Panel 3).
Standardized Variance Estimates from Univariate Twin Models of Drinking Motives and MZ and DZ Correlations for Twin Pairs Concordant and Discordant for Smoking Status
Results of the univariate twin models for the entire sample (which included twin pairs discordant for smoking status) suggest statistically significant additive genetic effects on the scores for coping (30%) and social (25%) motives, but not on the scores for enhancement and conformity. Shared environment effects accounted for a significant portion of the variance in enhancement scores, but not in the scores for any other motives. For all 4 of the motives, the remaining variation in the scores was attributable to nonshared environment.
Examination of the MZ and DZ twin pair correlations for each drinking motive in the twins concordant for smoking status suggests an interesting trend. Specifically, in twins concordant for regular smoking, the size of the DZ twin pair correlation for each drinking motive is about half the size of the MZ twin pair correlation. In contrast, the MZ and DZ correlations are about equal in the twin pairs concordant for nonregular smoking. This suggests the influence of additive genetic effects on the drinking motive scores in regular smokers but not in nonregular smokers.
In the univariate twin models stratified by smoking status, there were no statistically significant differences in the additive genetic influences on the drinking motives between the regular and nonregular smokers. Specifically, the chi-square difference tests indicated that constraining the additive genetic (a), shared environmental (c), and nonshared environmental coefficients (e) to be equal across the groups did not result in a significant change in the fit of the model for enhancement χ2 (3) = 2.69, p = 0.443, for coping, χ2 (3) = 2.32, p = 0.501, for conformity, χ2 (3) = 3.12, p = 0.374, or for social motives χ2 (3) = 3.30, p = 0.348. As shown in Panel 3 in , the confidence intervals for the standardized estimates of the additive genetic influences on the drinking motives in the regular and in nonregular smokers all overlap. However, it is notable that the point estimates for the additive genetic influences on all of the drinking motives are consistently larger in the regular versus the nonregular smokers. Further, additive genetic influences on enhancement, coping, and social motive scores reached statistical significance in regular smokers, but did not in nonregular smokers.
To investigate this further, we computed another series of models for each drinking motive in the regular and nonregular smoker groups. Here, we dropped A, C, or A and C from the models and tested for significant deterioration in model fit using chi-square difference tests. For the regular smokers, dropping A from the models for enhancement and coping resulted in a significant deterioration in model fit, χ2 (1) = 5.35, p < 0.05 and χ2 (1) = 3.96, p < 0.05, respectively. In contrast, dropping C from these models did not result in a significant deterioration in model fit (p > 0.05 for both). For conformity and social motives, dropping either A or C did not result in a significant deterioration in model fit, whereas the models that dropped both A and C yielded a significantly poorer fit to the data, χ2 (2) = 9.27, p < 0.05 and χ2 (2) = 52.55, p < 0.05, respectively.
In the nonregular smoker group, dropping C resulted in significantly poorer fit for enhancement, χ2 (1) = 7.51, p < 0.05, and social motives χ2 (1) = 3.86, p < 0.05. For coping and conformity motives, A or C could be dropped without significantly reducing model fit. However, dropping both A and C yielded a significant deterioration in the fit of the model for coping, χ2 (2) = 19.50, p < 0.05 and for conformity, χ2 (2) = 6.47, p < 0.05.
These results are summarized in Panel 4 in , where the standardized variance estimates for the best-fitting models (either A–E, C–E, or A–C–E) are shown. Taken together, our results suggest a trend where drinking motives in regular smokers are genetically influenced more so than in nonregular smokers.