Nine studies (1,822 participants) were included in the review. Two studies were graded as high quality, five as moderate quality and two as low quality. Three studies did not describe the process of randomisation; five did not mention allocation concealment; seven studies presented drop-out rates. Loss to follow-up ranged from 7.1% to 23.8%. Four studies used intention-to-treat analysis and three used per-protocol analysis.
Glycosylated haemoglobin was lower in patients who received the web-based intervention, the mean difference was -0.71% (95% CI -1.00, -0.43) after three months, -0.52% (95% CI -0.75, -0.29) after six months and -0.55% (95% CI -0.70, -0.39) after 12 months. At 12 months there was evidence of statistical heterogeneity. subgroup analysis indicated that this was driven by two studies which used mobile phone as well as internet-based interventions, which had a greater effect on glycosylated haemoglobin than internet-only interventions.
Among the three studies which reported the effect of web-based tools on low-density lipoprotein cholesterol, the intervention was beneficial, mean difference -0.23mmol/L (95% CI -0.28, -0.19).
Web-based tools were not effective in reducing fasting plasma glucose (four studies), high-density lipoprotein cholesterol or total cholesterol (nine studies each).
There was no statistical evidence of significant publication bias.