Twenty to sixty percent of diabetic patients are affected by hypertension and antihypertensive agents are used to treat this condition. These agents are also used to prevent the onset of kidney disease both in normotensive and hypertensive diabetics. To evaluate the comparative effects of antihypertensive agents in patients with diabetes and normoalbuminuria. MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials, conference proceedings, and contact with investigators were used to identify relevant trials. Randomised controlled trials (RCTs) comparing any antihypertensive agent with placebo or another agent in hypertensive or normotensive patients with diabetes and no kidney disease (albumin excretion rate < 30 mg/d) were included. Two investigators independently extracted data on renal outcomes and other patient relevant outcomes (all-cause mortality, serious cardiovascular events), and assessed quality of trials. Analysis was by a random effects model and results expressed as relative risk (RR) and 95% confidence intervals (CI). Sixteen trials (7603 patients) were identified, six of angiotensin converting enzyme inhibitors (ACEi) versus placebo, six of ACEi versus calcium channel blockers (CCBs), one of ACEi versus CCBs or combined ACEi and CCBs and three of ACEi versus other agents. Compared to placebo, ACEi significantly reduced the development of microalbuminuria (six trials, 3840 patients: RR 0.60, 95% CI 0.43 to 0.84) but not doubling of creatinine (three trials, 2683 patients: RR 0.81, 95% CI 0.24 to 2.71) or all-cause mortality (four trials, 3284 patients: RR 0.81, 95% CI 0.64 to 1.03). Compared to CCBs, ACEi significantly reduced progression to microalbuminuria (four trials, 1210 patients: RR 0.58, 95% CI 0.40 to 0.84). A significant reduction in the risk of developing microalbuminuria in normoalbuminuric patients with diabetes has been demonstrated for ACEi only. It appears that the effect of ACEi is independent of baseline blood pressure, renal function and type of diabetes, but data is too sparse to be confident that these are not important effect modifiers and an individual patient data meta-analysis is required.