Bodin & Strandberg  reported an evaluation of the Örebro prevention programme (ÖPP), targeting youth drinking, initial scientific reporting of which was published in this journal in 2008 by Koutakis & Stattin . In their effectiveness study they concluded that ÖPP does not work as intended. In the Abstract they stated: ‘In contrast to the original study but consistent with a recent Dutch trial, our results do not support the hypothesis that ÖPP reduces youth drinking’, and in the Discussion they claimed that: ‘Our results do not support the hypothesis that ÖPP as currently delivered in Sweden reduces youth drinking’. Despite the rigorous design and robust findings reported by Bodin & Strandberg, we believe that their conclusions should be re-examined.
There are three reasons for re-examining Bodin & Strandberg's findings. First, according to the authors, the implementation did not, for the most part, reach the recommended number of sessions. Only a small proportion of the parents received the full dose of the programme. With this type of poor fidelity, preventive interventions are less likely to be effective, and this might be a strong reason why the programme did not work as intended. A second reason for re-examining the findings is that Bodin & Strandberg reduced variance in the outcome measures by dichotomizing their measures of youth drinking. Thus, their measures might not have been sensitive enough to detect differences between the groups. Finally, Bodin & Strandberg focused only on the programme effect, without considering the theory behind it. The theory is important, because the effectiveness of any preventive intervention is dependent upon the extent to which the programme is successful in changing the mediating mechanisms through which it is theorized to work . Therefore, although examining differences in youth drinking at group level (between the intervention and the control schools) is important, these analyses should be complemented with direct tests of the programme theory. These three issues—poor fidelity, reduced variance in the outcomes and not testing mediating mechanisms—might be addressed through re-analyses of Bodin & Strandberg's data. Bodin kindly granted us access to the data and permission to publish our re-analyses. Thus, we carried out the re-analyses as described below.
We were able to address the second and third reasons of the three mentioned for re-examining the findings. Concerning the first—poor implementation—the implementation was not measured at the individual level; therefore, it was not possible to estimate the extent to which the implementation affected the outcomes. Thus, we re-analysed the data to address the reduced variance in the outcomes and to test mediating mechanisms. We performed these analyses using a latent growth modelling approach, which is suggested to be a strong analytical approach for answering questions regarding individual change over time and individual differences in change process [4, 5], and has been demonstrated to be suitable for testing intervention effects [6, 7]. In our analyses, we used the original metric of the variables, without reducing variations, and we took into account the clustering of the data to achieve robust estimates of the effects.
To examine the extent to which reduced variance in the outcomes affected Bodin & Strandberg's results, we tested the effect of the programme on the four undichotomized outcomes measures: life-time drunkenness, drunkenness in past month, frequency of drinking and amount of drinking on each occasion. We followed the procedure outlined by Cheong et al. . The results showed that the programme had a significant effect on life-time drunkenness (est. = −0.37, P = 0.034) and a marginally significant effect on past-month drunkenness (est. = −0.352, P = 0.054). The two models yielded a perfect fit, with a non-significant χ2 model fit: χ2 (2) = 1.855, P = 0.39 for life-time drunkenness, and χ2 (2) = 0.824, P = 0.66 for past-month drunkenness. In sum, youths in the intervention group increased in drunkenness at a significantly lower rate over time than the youths in the comparison condition. However, we found no direct effect of the programme on the frequency and amount of drinking. We also tested the same models with dichotomized variables following Bodin & Strandberg's strategy. None of the models suggested a significant programme effect. Thus, we conclude that Bodin & Strandberg's dichotomization of the outcomes seems to have had some effect in suppressing significant results.
The state-of-the-art method of testing programme theory is fitting mediation models that elucidate the effect of the programme on the targeted outcomes through mediating mechanisms . With this purpose in mind, we first tested if the programme had a significant effect on parental attitudes towards youth drinking; it did (est. = 0.19, P < 0.001). Next, for a test of mediating mechanisms, we tested whether the programme influenced youth drinking through its effect on parental attitudes. A separate model was fitted for each of the outcome measures. The results were clear across all models: the ÖPP programme had significant indirect effects on life-time drunkenness (est. = −0.057, P = 0.016), past-month drunkenness (est. = −0.072, P = 0.017), frequency of drinking (est. = −0.05, P = 0.003) and amount of drinking (est. = −0.06, P = 0.006) through its effect on parents' attitudes toward youth drinking. The indirect effects were robust, as bootstrapped confidence intervals confirmed the systematic indirect effect of the programme on youth outcomes. Thus, these findings show that when the programme influenced parents' attitudes, according to all the outcome measures, it succeeded in reducing youth drinking.
Overall, our conclusions suggest the opposite of what Bodin & Strandberg concluded. The ÖPP programme appears to influence changes in youth drunkenness, which is the aim of the programme. In addition, the effect of the programme is seemingly explained by changes in parental attitudes to youth drinking, providing strong support for the validity of the programme theory. The effects in these re-analyses were smaller than was reported originally in the efficacy study by Koutakis et al. . We agree with Bodin & Strandberg that this might be due potentially to the media campaigns and public health enterprises to increase parents' restrictive attitudes to youth drinking which were ongoing simultaneously with Bodin & Strandberg's evaluation, resulting in higher parental restrictiveness today than was the case when the original study by Koutakis et al. was published.
Declarations of interest
Koutakis, Stattin & Kerr published the original evaluation of the ÖPP programme with funding from the Swedish National Institute of Public Health. ÖPP can be used by anyone without cost. Neither Stattin or Özdemir have any financial interests to declare. Bodin & Strandberg's  study was also funded by the Swedish National Institute of Public Health. No funding was received for the present re-analysis.
Metİn Özdemİr & Håkan Stattin
Center for Developmental Research, Örebro University, Fakultetsgatan 1, Örebro 70182, Sweden. E-mail: firstname.lastname@example.org