Mead et al ^{Reference Mead, Lester, Chew-Graham, Gask and Bower1} recently meta-analysed data on the effectiveness of befriending interventions on reducing depressive symptoms. Befriending was defined as a non-professional intervention that provides clients with non-directive, emotionally focused support by one or more individuals; was not psychoeducational or mentoring in nature; and did not constitute formal psychotherapy. Mead et al found that befriending interventions had a modest, statistically significant effect on depressive symptoms within 12 months of randomisation (standardised mean difference 0.27, 95% CI 0.06 to 0.48, nine studies) and a slightly smaller effect on longer-term outcomes (standardised mean difference 0.18, 95% CI 0.05 to 0.32, five studies).

As the authors noted, the effect sizes for befriending were essentially equivalent to effect sizes from collaborative care depression interventions in primary care. In a 2006 meta-analysis, Gilbody et al ^{Reference Gilbody, Bower, Fletcher, Richards and Sutton2} reported a short-term (within 6 months) standardised mean difference effect size for symptom reduction from collaborative or enhanced depression care of 0.25 (95% CI 0.18 to 0.32, 35 studies) and longer-term effect sizes of 0.15 at 2 years post-randomisation (95% CI –0.03 to 0.32, 9 studies) and 0.15 at 5 years post-randomisation (95% CI 0.001 to 0.30, 2 studies). As Mead and colleagues note, the implications of this are important. Befriending or social support interventions could provide a less expensive and potentially ‘less medicalised’ option of care for patients with mild to moderate symptoms of depression in primary care. Indeed, collaborative care is a complex, multifaceted, expensive organisational intervention that can be difficult to implement outside of research settings. ^{Reference Katon and Seelig3,Reference Katon, Unützer, Wells and Jones4}

There are caveats, however. As noted by Mead et al, only a small set of heterogeneous studies have examined the effects of befriending interventions on depressive symptoms. Furthermore, funnel plot asymmetry suggested that publication bias may have influenced the estimate of the degree to which befriending may affect depressive symptoms. The authors did not assess the degree to which publication bias may have influenced the results of the meta-analysis. However, if only studies with statistical power of at least 0.70 among the studies with short-term outcomes evaluated by Mead et al are analysed, the resulting synthesised effect estimate is 0.08 (95% CI –0.06 to 0.21, four studies), a substantially smaller estimate than that produced by all nine studies (0.27, 95% CI 0.06 to 0.48). Thus, as noted by Mead et al, more high-quality research is needed on befriending in order to determine the likely benefit to patients in clinical practice.

Meanwhile, the results of the meta-analysis suggest that future research on collaborative care should use a befriending or attention control group. Up to now, collaborative care interventions have been compared with usual care, and it is not known to what degree the effects that have been reported are due to specific effects of the collaborative care intervention versus effects that may come from the substantially increased attention and support received by patients in collaborative care.