Thirteen studies (n=2,160) were included in the review; 12 studies (n=2,011) were included in the meta-analysis. Three studies compared SMBG with SMUG, nine compared SMUG with an intervention without self-monitoring, and two compared SMBG with SMBG and feedback programmes.
The median internal validity score was 9 (range: 4 to 10) out of a possible 13. No studies reported adequate allocation concealment or patient blinding, and only one reported blinding of the outcome assessors.
Pooled estimates of change from baseline in HbA1c.
The pooled changes from baseline in HbA1c were -0.47% (95% CrI: -0.66, -0.28) for interventions without self-monitoring, -0.61% (95% CrI: -1.20, -0.05) for interventions with SMUG, -0.87% (95% CrI: -1.14, -0.58) for interventions with SMBG, and -1.48% (95% CrI: -2.06, -0.89) for interventions with SMBG and feedback.
Relative effects based on MTC meta-analysis.
It was estimated that there was a 69% probability that an intervention with SMUG results in a larger reduction in HbA1c than an intervention without self-monitoring, and a 99% probability that an intervention with SMBG results in a larger reduction in HbA1c than an intervention without self-monitoring. The probability that interventions with SMBG are more effective than interventions with SMUG was estimated as 84%.
Adjustment for baseline glycaemic control and weighting by quality score.
When estimated relative effects were adjusted for the baseline HbA1c, the differences between interventions increased, except for the difference between SMBG and SMUG which was reduced. When studies were additionally weighted by quality score, the probability that SMUG was more effective than interventions without self-monitoring reduced to 54%, and the probability that interventions with SMBG are more effective than SMUG increased to 88%. The results were similar when analyses were restricted to studies of non-insulin requiring patients.