Aims/hypothesis: Anaemia has been suggested to be an independent risk factor for subsequent progression of advanced diabetic nephropathy; however, the relationship between haemoglobin levels and progression of nephropathy in patients without clinical albuminuria is unknown. Methods: We conducted this prospective hospital-based cohort study of 464 type 2 diabetic patients (149 women and 315 men, 55±13 [mean±SD] years of age) with serum creatinine <177 μmol/l (2.00 mg/dl) and urinary albumin : creatinine ratio <300 mg/g creatinine. GFR was estimated using the equation formulated by the Modification of Diet in Renal Disease Study group, refitted for Japanese individuals. Most patients had haemoglobin concentrations in the normal range (144±15 g/l), only modest renal impairment (GFR: 74.8±14.5 ml min −1 1.73 m −2 ), and normal urinary albumin levels (81.5/18.5% with normo-/microalbuminuria). The primary outcome measurement was the rate of change in GFR determined by regression analysis with GFR as a function of time. Patients were followed up for a mean observation period of 5.0±0.9 (range: 2.5 to 6.2) years. Results: Univariate and multiple regression analyses yielded a significant association between the rate of change in GFR and baseline haemoglobin concentration. After adjusting for covariates, the rate of decline in GFR was significantly greater in patients in the lowest haemoglobin quartile (−3.27 ml min −1 1.73 m −2 year −1 ) than in the third (−2.71 ml min −1 1.73 m −2 year −1 , p=0.024) and highest quartiles (−2.78 ml min −1 1.73 m −2 year −1 , p=0.046). Conclusions/interpretation: Lower haemoglobin concentrations in type 2 diabetic patients without clinical albuminuria may be a significant predictor of subsequent decline in GFR.