In a recent review of drug abuse research it has been stated that some of the advancements parallel concepts that are also part of the developmental psychopathology approach. Among various research areas, this perspective includes antecedent and co-occurring psychopathological conditions and other problem behaviour [1]. To date, there is only a small number of longitudinal studies that have taken this perspective by studying the association of adolescent substance use with adult outcome [e.g. [2–5]] or the developmental antecedents, patterns, and correlates of substance use in adolescence [6–10].

Within the developmental perspective of adolescent substance use, various risk factors including life stress, personality features, parental behaviour, peer influences, school characteristics, and other environmental features have been studied as can be delineated from both major reviews and empirical studies [6, 11–17]. The frequent co-occurring mental disorders and behaviour problems have been addressed particularly in another series of recent reviews with a strong emphasis on co-morbid conduct disorders or externalizing problems [13, 18–20].

The present study attempted to shed some more light on the developmental trajectories of substance use from adolescence to young adulthood. Based on data from a longitudinal study, the main focus was on problematic substance use rather than on substance use disorders. In accordance with various studies, weekly or daily consumption of tobacco and/or alcohol, problem and heavy drinking aiming at drunkenness, and cannabis use more frequently than three times in the past month was considered to reflect problematic substance use [21]. The major aim of the present study was the analysis and predictive power of selected psychosocial correlates across time in subjects with problematic substance use (PSU). The psychosocial correlates of the longitudinal study had been selected on the basis of a theoretical model that will be outlined with the methods below.

Three main questions were addressed in the analyses. First, the continuity of PSU across time was studied. The second series of cross-sectional analyses compared psychosocial correlates in the PSU groups at each time with the rest of the sample representing the controls. The third question asked for preceding psychosocial markers during middle and late adolescence of later PSU in young adulthood and was, thus, using the longitudinal data.

The size of the groups with PSU steadily increased from 5.1% risk subjects at time 1 to 26.1% subjects at time 2 and 48.9% subjects at time 3. There was significant continuity from time1 to time 2 with 25/30 (83%) of persistent subjects with PSU compared to 130/563 (23%) of subjects who newly developed PSU (McNemar Test p < .001). In the same way, the PSU group at time 1 showed significantly more PSU also at time 3 (28/30 subjects; 93%) than the control group from time 1 (262/563 subjects; 47%; McNemar Test p < .001). Continuity from time 2 to time 3 was also highly significant with 131/155 (85%) of the subjects with previous PSU showing ongoing PSU whereas only 159/438 (36%) developed PSU de novo (McNemar p < .001). There were also significant gender effects. At time 1, boys were less frequent than girls in the PSU group (13/30 vs. 17/30) and the controls (271/563 vs. 292/563) (McNemar p < .001) whereas males were more frequent among the adolescents at time 2 in the PSU group (79/155 vs. 76/155) and less frequent in the control group (205/437 vs. 233/437) (McNemar p < .001). At time 3 there were no significant gender effects. Furthermore, the PSU group was significantly older than the controls at time 1 (Mean = 15.1, SD = 1.0 vs. Mean = 13.5, SD = 1.6, t = -7.99, df = 36.9, p < .001) and also at time 2 (Mean = 17.0, SD = 1.5 vs. Mean = 16.4, SD = 1.6, t = -4.32, df = 299, p < .001). There were no significant age differences between the two groups at time 3.

The present study compared PSU groups and control groups in the community rather than clearly defined clinical groups fulfilling criteria for manifest substance abuse. The emphasis on PSU was taken because these subjects should be potentially more apt and open for preventive measures. If there are cheap, reliable, and valid assessment tools for the identification of PSU subjects, identification in the community would be the first step in order to start interventions. These tools should not only be simple to administer but also allow composite definitions of substance use. Rather than focussing only on a single substance the present study is based on a composite definition of PSU reflecting the fact that most adolescent and young adult users consume various substances. Furthermore, the present longitudinal study allowed both for the repeated analysis of cross-sectional correlates and the predictive power of these variables on substance use at risk.

Starting from the identification of PSU by a combination of items asking for the frequency of alcohol, tobacco, and cannabis use, a steadily increasing number of subjects across the time span from early adolescence to young adulthood was identified in the present study. A more than nine-fold increase of PSU was observed in this period with a slight majority of females in a rather small risk group in early adolescence and a male predominance in males at later times. The increasing number and significant continuity of PSU is very much in line with what can be expected from recent reviews of international trends in use of a range of substances [13, 20].

In the next step of the analyses it was shown that PSU at each point in time of development was significantly associated with a number of behavioural and other psychosocial features. These correlates were not only important by themselves but also served as a validation of the definition of PSU. There was a clear association with externalizing problems across all three times which is also very much in accordance with other studies on developmental patterns of substance use and abuse [1, 5, 7–9, 13, 18–20]. However, at time 2 in late adolescence PSU subjects also showed higher scores for internalizing problems. This finding matches other studies that found mood and anxiety disorders to be frequently associated with substance use disorders in adolescents [20].

Among the various psychosocial correlates the most consistent finding of PSU was the persistent association with life events from early adolescence to young adulthood which is in line with the finding that negative life events contribute to escalated substance use during adolescence [6]. Also in line with other studies [6, 17] is the consistent finding that low parental support in terms of lower perceived parental acceptance and higher rejection at both times during adolescence is a marker of PSU. Avoidant coping as an additional correlate of substance use at risk was identified only at time 2 in late adolescence, which has been observed in terms of non-adaptive coping as another association with escalating substance use in adolescence [6]. Finally, among the various correlates of perceived school environment during adolescence, PSU was most consistently characterized across two times by the more controlling behaviour of the teacher, and in late adolescence also by the lack of possibility to participate and a higher amount of perceived peer acceptance.

The joint consideration of the correlated variables in logistic regression analyses allowed a more definite evaluation of the most relevant associations with PSU. At times 1 and 2 during adolescence, age was strongly contributing to the prediction of risk status indicating that there is a developmental pattern with increasing age contributing to PSU. The second strongest and most persistent variable across all three times contributing to PSU are externalizing problems. At younger age in early adolescence, negative perceived parenting contributes significantly to PSU whereas with greater autonomy in late adolescence these associations are replaced by deficits in active problem solving as indicated by avoidant coping, a feeling of having a lack of possibilities to participate at school, and peer influences as reflected by high peer acceptance. Later in young adulthood, none of the psychosocial variables besides externalizing problems (and the reverse, i.e. low scores on the internalizing dimension) had any predictive power. However, it should be taken into account that there were less potential correlates assessed at this time compared to the two previous times in the study.

The longitudinal analyses showed that at each time during adolescence it is predominantly the preceding PSU status that predicts PSU at outcome in young adulthood. Thus, the self-perpetuating character and the high rate of persisting PSU were underlined again by the data in this study. Furthermore, age was a predictor only at time 1 and 2. Together with the increased age of the PSU groups at these times, this finding points to the rather trivial fact that PSU is age-dependent during adolescence. Other major and persistent risk factors were externalizing problems and male gender. Despite the trend for a closing gender gap with increasing age in a previous cross-sectional analysis of substance use in an even larger sample [16], the present approach with a focus on specifically defined subjects emphasizes the particular risk of males for remaining PSU subjects across time. Among the other psychosocial variables only a few contributed to the prediction and included some variables that were assessed in early adolescence, namely, low scores on the internalizing dimension which is just the correlate of the externalizing problems, heightened self-awareness, and a feeling of lacking possibilities to participate at school. In late adolescence, only perceived paternal acceptance contributed to the prediction which may be due to the fact that the father was seen as a role model particularly for the males in the PSU group.

In conclusion, the present study points to a rather persistent pattern of problematic substance use with a number of associated and predictive psychosocial features that both can be assessed in the community as a first step for the identification of individuals who are in danger of developing long-term risk behaviours in adulthood. The strength of the present study lies in the large community sample and the longitudinal approach. Limitations include the emphasis on PSU, rather than, substance use disorders according to the major schemes of diagnostic classification, and the reliance on self-reports of the subjects. However, it should be kept in mind that at this age in adolescence and early adulthood parents and caretakers are less reliable informants as to substance use in youths.