Background. The risk of microalbuminuria in patients with insulin-dependent diabetes mellitus (IDDM) is thought to depend on the degree of hyperglycemia, but the relation between the degree of hyperglycemia and urinary albumin excretion has not been defined. Methods. We measured urinary albumin excretion in three random urine samples obtained at least one month apart from 1613 patients with IDDM. Microalbuminuria or overt albuminuria was considered to be present if the ratio of albumin (in micrograms) to creatinine (in milligrams) was 17 to 299 or greater than or equal to 300, respectively, for men and 25 to 299 or greater than or equal to 300, respectively, for women. Measurements of glycosylated hemoglobin (hemoglobin A(1)) obtained up to four years before the urine testing were used as an index of hyperglycemia. Twelve percent of the patients had overt albuminuria and were excluded from subsequent analyses. Results. The prevalence of microalbuminuria was 18 percent in patients with IDDM. It increased with increasing postpubertal duration of diabetes and, within each six-year interval of disease duration, it increased nonlinearly with the hemoglobin A(1) value. For hemoglobin A(1) values below 10.1 percent, the slope of the relation was almost flat, whereas for values above 10.1 percent, the prevalence of microalbuminuria rose steeply (P<0.001). For example, as the hemoglobin A(1) value increased from 8.1 to 10.1 percent, the odds of microalbuminuria increased by a factor of 1.3, but as the value increased from 10.1 to 12.1 percent, the odds were increased by a factor of 2.4. Conclusions. The risk of microalbuminuria in patients with IDDM increases abruptly above a hemoglobin A(1) value of 10.1 percent (equivalent to a hemoglobin A(1c) value of 8.1 percent), suggesting that efforts to reduce the frequency of diabetic nephropathy should be focused on reducing hemoglobin A(1) values that are above this threshold.