Does baseline severity interact with the effects of psychotherapy for depression? A meta-analytic review

Introduction: It is not yet clear whether baseline severity is associated with the effects of psychotherapies. We examined baseline severity at the study level in a large sample of randomized controlled trials comparing psychotherapies against a control condition for the treatment of depression.
Methods: We used an existing large database of randomized trials comparing psychotherapies for depression with control groups (www.metapsy.org). We converted baseline severity scores across different depression measures into a common metric. We ran bivariable and multivariable meta-regression analyses to examine the association of effect sizes with baseline severity. We also examined response rates in treatment and control conditions.
Results: We included 387 randomized trials (463 comparisons; 47,315 patients). The pooled effect size of the psychotherapies was g = 0.77 (95 % CI, 0.70; 0.84). In the main analyses, we found a highly significant association between the effect size and baseline severity (bivariable coefficient: 0.024 (SE = 0.006; p < 0.0001), multivariable coefficient: 0.022 (SE = 0.007; p = 0.002)). This was confirmed in some but not all sensitivity analyses. Absolute response rates in the control conditions remained stable across different levels of baseline severity (bivariable metaregression analyses: p = 0.545), or showed a negative association (multivariable analyses: p = 0.002). In the therapy conditions the response rates were significantly larger with increasing levels of baseline severity (bivariable: p ≤0.0001; multivariable: p = 0.006).
Conclusion: The effects of psychotherapies are probably associated with baseline severity. Response rates in control conditions remained relatively stable across different levels of baseline severity, while in the treatment conditions the response rates increased with increasing levels of baseline severity.