Forty-one trials set in North America, Europe and Asia were included in the review. The reported number of participants differed between the text and tables, but appeared to be more than 7,745. Drop-out rates ranged from zero to 39.8% and the drop-out rate difference between study arms ranged from zero to 26.1%. Allocation concealment was adequate in 15 studies.
The intervention improved glycaemic control (SMD -0.38, 95% CI -0.47 to -0.29) corresponding to an absolute mean difference in HbA1c of 0.51% between the intervention and control groups. There was evidence of statistical heterogeneity between studies (Ι²=66%).
Univariate meta-regression showed that the effect of the intervention was greater in studies where the mean HbA1c was 8% or more compared with less than 8% (SMD -0.45, 95% CI -0.56 to -0.34 compared to SMD -0.14, 95% CI -0.25 to -0.05).
Programmes in which the disease manager could start or modify treatment were associated with a greater improvement in HbA1c compared to those that could not (SMD -0.60, 95% CI -0.73 to -0.47 compared to SMD -0.28, 95% CI -0.37 to -0.18).
Programmes with a high frequency of contact led to a significantly greater reduction in HbA1c compared to those with low frequency (SMD -0.56, 95% CI -0.72 to -0.40 compared to SMD -0.30, 95% CI -0.54 to 0.06). There was no difference in the effect of intervention between moderate and low frequency programmes.
No multivariable meta-regression results were reported. Sensitivity analyses showed that the results were robust to exclusion of studies defined as low quality.
Six out of nine studies that reported data on hypoglycaemic events found no difference in rate between intervention and control groups; two of the remaining three found a higher rate in the control compared to intervention group. No overall difference between groups was found in 20 studies that reported mortality. Data on hospital admissions were not clearly or systematically reported.