Seven RCTs were included (28,065 participants; range 110 to 11,140). Median follow up periods ranged from two to 11.1 years. Most studies had a low risk of bias rating for all domains of the Cochrane risk of bias tool, except for blinding of participants and personnel, where all but two studies were subject to a high risk of bias.
Compared with conventional treatment, intensive glucose control statistically significantly reduced the risk of microalbuminuria (RR 0.86, 95% CI 0.76 to 0.96, seven trials, Ι²=64%) and macroalbuminuria (RR 0.74, 95% CI 0.65 to 0.85, six trials, Ι²=13%), but there were no significant differences for the clinical renal outcomes: doubling of the serum creatinine level (four trials), end-stage renal disease (five trials) and death from renal disease (three trials). The result for macroalbuminuria was not significant when only the two studies with adequate participant/personnel blinding were used in the analysis. Other risk of bias sensitivity analyses had little effect on the results.
Meta-regression indicated that larger differences in haemoglobin A1c between intensive and conventional therapy were associated with greater benefit for both microalbuminuria and macroalbuminuria. The median year of enrolment, the years since diabetes diagnosis and the duration of therapy were only associated with the doubling of the serum creatinine level outcome.
Funnels plots suggested there was evidence of publication bias.