We determined plasma metalloproteinase-9 (MMP-9) concentrations in 30 patients with non-insulin-dependent diabetes mellitus (NIDDM) at an initial examination (baseline) and on three separate occasions during a 48-month follow-up period. All patients had normal urinary albumin excretion (<20 microg/min) at the first three examinations. At the fourth examination (48 months after the first examination), 22 patients had normal urinary albumin excretion and eight had microalbuminuria (median, 36.4 microg/min; range, 20.2 to 46.6 microg/min). Compared with patients with normal urinary albumin excretion, patients with microalbuminuria had significantly higher plasma levels of MMP-9 at the second (56+/-14 microg/L v36+/-12 microg/L; P < 0.05), third (88+/-23 microg/L v 39+/-14 microg/L; P < 0.01), and fourth (117+/-30 microg/L v 44+/-16 microg/L; P < 0.01) examinations, but not at the first examination (34+/-12 microg/L v 33+/-14 microg/L; P=NS). An increase in plasma MMP-9 concentrations preceded the occurrence of microalbuminuria within 4 years. The groups did not differ with regard to age, sex, duration of NIDDM, blood pressure, or mean glycated hemoglobin. In addition, the eight patients with microalbuminuria were treated with an angiotensin-converting enzyme inhibitor (cilazapril; 0.5 mg once daily) for 6 months. Microalbuminuria was reduced to within the normal range, and plasma MMP-9 concentrations were significantly decreased with the cilazapril treatment (52+/-18 microg/L; P < 0.01). However, serum MMP-1 and tissue inhibitor of MMP-1 showed no change during the study period. These data suggest that plasma MMP-9 concentrations preceded and may predict the development of microalbuminuria in NIDDM.