Premature conclusions about depression prevention programmes

In my opinion, the meta-analysis by Jané-Llopis et al (2003) suffers from some methodological flaws that misguided the authors to draw premature conclusions on predictors of prevention in depression prevention programmes.

First, many of the selected studies did not target the prevention of depression but examined therapeutic or preventive strategies for other primary disorders and used depression scores as secondary outcome measures. For example, Bisson et al (1997) studied the efficacy of psychological debriefing on the development of post-traumatic stress disorder (PTSD) in victims of acute burn traumas. They showed that psychological debriefing may even worsen the long-term course of burn victims. But while psychological debriefing may have been mistakenly considered helpful for preventing PTSD in the past, no reasonable therapist or researcher has ever claimed that massive emotional confrontation would represent a promising strategy for depression or depression prevention.

Second, the coding of respective methods looks rather inconsistent, and I wonder how the authors were able to reach such a high interrater reliability across codes. For example, the psychological debriefing method used by Bisson et al (1997) was coded as ‘behavioural, cognitive and educational’ (p. 389), while the code ‘cognitive’ was missing for Seligman et al's (1999) intervention based on cognitive therapy. Similarly, four research groups using similar variants of the Coping with Depression Course by Lewinsohn et al (1984) were coded differently (e.g. ‘cognitive and competence’, ‘behavioural, cognitive, educational and social support’, ‘cognitive’, and ‘behavioural, cognitive, competence and educational’ (pp. 386–391)). Finally, the coding category ‘behavioural methods’ incorporates very heterogeneous strategies. For example, behavioural strategies found to be helpful in cognitive–behavioural therapy for depression focus on increasing pleasant activities and social skills training (Lewinsohn et al, 1984), whereas the delivery of peer support telephone dyads by lay persons, as used in the studies by Heller et al (1991), may be regarded as a very specific behavioural strategy which has so far not been recommended as a helpful intervention by the research community. In Jané-Llopis et al's meta-analysis, respective interventions from the studies by Heller et al (1991) had negative effect sizes and therefore may have substantially accounted for the missing or even negative effect of the ‘behavioural’ component of preventive measures.