The 373 subjects included in these analyses were San Diego Prospective Study probands who participated in the ongoing 25-year followup to date and who had complete data for all of the required analyses. They represent an estimated 94% of those who had been scheduled for interview. These men were an average age of 46.5 (3.29) years at their 25-year followup, all were Caucasian (including White Hispanic), and 53.1% had an alcohol-dependent father. Over the 25 years, the rate of AUDs in the FHP probands was 44.9% and in FHNs was 29.1% (χ2 = 9.90, p < .002). At baseline 18.5% were married, 26.5% reported having no religious preference, while the values for Novelty Seeking were a mean of 15.6 (4.78) with a range of 3 to 32 and for Reward Dependence were 17.7 (4.41) with a range of 6 to 28.
Among the 373 probands, 140 men (37.5%) met criteria for alcohol abuse or dependence during the 25-year followup, with an average age of onset of an AUD of 27.2 (6.60) years. The proportion of alcoholics (Groups 2 through 4) who ever met criteria for alcohol dependence was 71.4% overall, with the remainder only meeting criteria for alcohol abuse. Over the 25 years of the study, 10.7% of the 140 alcoholics had ever received treatment for their AUD from a healthcare provider, and an additional 12.1% had participated in a self-help group such as Alcoholics Anonymous, but had never received formal treatment. Therefore, a total of 22.8% of the 140 had received either or both of these sources of help.
As shown in , the 4 groups included 233 drinkers (62.5%) in Group 1 who had never developed an AUD during the relevant follow-up interval; 64 men in Group 2 (17.2% of the total and 45.7% of the 140 probands with an AUD) who developed their disorder during the first 10-year followup and maintained an AUD for all subsequent evaluations; 25 individuals in Group 3 (6.7% of the total and 17.9% of the alcoholics) who had an AUD onset after the T10 followup and a subsequent chronic course; and 51 men in Group 4 (13.7% of the total sample and 36.4% of those with an AUD) who were alcoholic by T10 and who subsequently recovered from their AUD by T20. Across Groups 2, 3, and 4, the ages of onset of the AUD were 24.5 (3.33), 37.9 (6.83), and 25.2 (3.80) years, respectively (F = 94.8, p < .001). While the Tukey HSD post-hoc analysis (the approach most appropriate across the three groups) for ages of onset for Groups 2 versus 4 was not significant (p = .67), Groups 2 and 4 were each different from Group 3 (p < .001). For Group 3, 8 men (32.0% of Group 3 and 5.7% of the 140 alcoholics) developed their disorder at age 40 or above.
Comparisons of T1 and T10 Variables across 4 Alcohol Diagnosis Groups with Planned Comparisons for 373 Men*
presents the distribution across the 4 groups for T1 and T10 potential predictors of the onset and course of AUDs. At the bottom of the table we also report the percent who met criteria for alcohol dependence in each AUD group, along with a summary of treatment received for AUDs and for 12-step program participation for relevant groups through T25. As all variables in were significantly different across the 4 groups, we carried out a series of planned comparisons, with results presented in the four last columns in . First, men in the combined Groups 2 through 4 who had developed an AUD were compared to the non-diagnostic Group 1. Almost all baseline variables were significantly different across the two broad diagnostic categories, including an association for AUDs with lower LR values, an FHP designation, higher Novelty Seeking, and higher T1 alcohol and substance use patterns. The only baseline variable that did not significantly differentiate between Groups 1 versus Groups 2–4 was T1 Reward Dependence. While the comparisons across non-alcoholic and alcoholic groups for T10 variables reflected the fact that all but 25 of the 140 alcoholics had manifested their diagnosis by the 10-year follow-up, the lower T10 figures for alcoholics for ever having been married and for practicing a religion, and the higher percent who smoked and used drugs during the first decade of follow up may be worth noting.
Additional information regarding the prediction of AUDs can be gathered by comparing the later onset subjects in Group 3 (who did not have an AUD at T10) with men in Group 1. Here, compared to those in Group 1 with no diagnosis, Group 3 men with later onset AUDs had significantly lower T1 LR, fewer years of schooling, and higher usual T1 drinking quantities. However, at T1 Groups 1 and 3 men were similar on FH, the percent with non-diagnostic alcohol problems, smoking histories, Novelty Seeking, and Reward Dependence. At T10, a time frame before their AUD onset, Group 3 men had higher drinking frequencies and quantities and demonstrated a trend for a higher proportion with any T10 alcohol problems (p = .07).
The third planned evaluation of results from involved comparing subjects with an AUD onset after T10 (Group 3) to the alcoholics who evidenced onsets earlier than the T10 followup (combined Groups 2 and 4). As shown in the second to last column in , at T1 the men who would experience a later onset of their AUD were less likely than other alcoholics to have reported any of the 9 non-diagnostic alcohol problems and to smoke, and had lower scores than other alcoholics for Novelty Seeking. Once again, while T10 alcohol items reflect the differences between groups in their AUD status at the 10-year followup, the higher proportion of ever married men by T10 in Group 3 may be worth noting, as might their lower proportion of those who used an illicit drug between T1 and T10. Later-onset subjects were also more likely than those with AUDs by T10 to have developed only alcohol abuse, not dependence. Interestingly, the frequencies of drinking were similar at T10 across earlier versus later onset alcoholics.
The final planned comparison evaluates predictors of remission (Group 4) versus a chronic course (Group 2) for men with an AUD onset by T10. At T1, compared to men who would become alcoholics with a chronic course, those with subsequent remission had lower drinking frequencies and a trend for higher scores on Reward Dependence (p = .09). At T10, a time when both groups were already demonstrating their AUD, those who subsequently remitted were more likely to be divorced or separated and reported lower usual and maximum drinking frequencies, although drinking quantities were more similar. While not shown in the table, the direction of the relationship to remission was the same for those divorced as for those separated at T10.
For each of the planned comparisons, all items that differentiated across the relevant groups were then entered into regression analyses predicting the relevant outcomes, as shown in . For the predictors of developing an AUD (Groups 2, 3, and 4) among all 373 men, the regression revealed that 7 of the 10 baseline (T1) items predicting an AUD contributed significantly to the equation. Note only items from that were significant in any comparison are listed in , with “NA” indicating when a variable was not relevant to a specific column of that table, either because it was not significant for that outcome in or was not appropriate to test in the regression for that outcome. An example of the latter is that T10 alcohol frequency could not be used to predict an AUD by T10, and is “NA” for data column 1, but was relevant for the evaluation of predictors of remission among those with an AUD at T10 in data column 4. The T1 items that contributed to an estimated 41% of the variance for development of an AUD (the pseudo R2), included all 3 key variables of lower LR, an FHP status, and higher Novelty Seeking scores. Also significant were higher T1 alcohol use and problem histories, and lower education.
Significant Items in Regressions Predicting Each of Four Outcomes: Wald χ2 (Odds Ratio)*
Outcomes for the 140 alcoholics (Groups 2–4) reflected a combination of alcohol dependence (71% of Groups 2 – 4) and abuse (29%). To better understand the implications of this approach, the major analyses for Groups 1 versus 2 through 4 were repeated to evaluate the two diagnoses separately. A regression predicting dependence revealed significant contributions for most T1 items significant in the first data column of , including a lower LR (OR = 0.64); the FHP status (OR = 2.14); fewer years of education (OR = 0.71); higher T1 alcohol frequency (OR = 1.08), quantity (OR = 1.46), and problems (OR = 2.59); as well as higher Novelty Seeking (OR = 1.07), with R2 = .50. When the regression was repeated for abuse outcomes, the results revealed significant contributions to the prediction of abuse for a lower LR (OR = 0.50); fewer years of education (OR = 0.78); and higher T1 alcohol frequencies (OR = 1.06) and quantities (OR = 1.06), with R2 = .65.
The second regression in focused on the prediction of later onset AUDs (Group 3) for men in Groups 1 and 3. Here, while significant on its own in , LR did not add to the equation that generated an R2 of .57. Of the significant items in , only T1 education and T10 drinking frequencies were significant in the regression.
The regression analysis in the third data column of describes the performance of the 3 significant T1 items from regarding the prediction of later onset AUDs among alcoholics. Two T1 variables contributed significantly to the regression predicting later onset among alcoholics, including the absence of non-diagnostic alcohol problems at T1 and lower scores for Novelty Seeking, combining for a pseudo R2 of .55. Adding the presence or absence of a dependence (versus abuse only) diagnosis to the regression did not contribute significantly to the regression.
The data in column 4 of evaluates predictors of recovery for these men who demonstrated an AUD by T30. The regression demonstrates that the predictors of remission included lower drinking frequencies at both T1 and T10, higher Reward Dependence, and a T10 marital status of having been divorced or separated (R2 = .30). To evaluate the impact of abuse versus dependence diagnosis, the regressions were re-evaluated after adding the abuse/dependence designation, but the results did not change from those listed in . To test the possible impact of having received alcohol-related treatment or participating in AA for this recovery-related outcome, these items were also added to the regression, but did not contribute significantly to the prediction of outcome.