Thirty-five RCTs were included in the review (n=2,243), of which 19 were eligible for meta-analysis (n=1,431). One study was classed as grade A for risk of bias, eight as B and 26 as C. Only four studies reported using intention-to-treat analyses.
Glycaemic control: Pooled effect size for HbA1c was -0.32 (95% CI -0.47 to -0.16, I2=44%; 19 RCTs, n=1,431). The treatment effect in absolute units was a decrease of 0.54% (-0.83 to -0.23), which indicated an improvement in glycaemic control following treatment with a psychological intervention (for the nine studies only of interventions delivered by generalist clinicians the result was 0.51% and for the nine studies only of interventions by psychological specialists the result was a 0.57% reduction; both analyses displayed moderate statistical heterogeneity, with I2 values of 52% for the former and 46% for the latter). There was no evidence of publication bias. There was no statistically significant association between improvements in HbA1c and duration of follow-up or duration of therapy, but there was evidence of an association between improvements in HbA1c and increased number of sessions (p=0.001).
Psychological status: Psychological interventions resulted in improvements in psychological status (SMD -0.56, 95% CI -1.00 to -0.13; 13 RCTs); there was no evidence of publication bias, but there was evidence of statistically significant heterogeneity (I2 value not reported). There was evidence of an association between improvements in psychological status and increased duration of therapy, increased follow-up length and increased number of sessions (all p<0.001).