OBJECTIVES: The association between HbA1c variability and diabetic nephropathy risk over the short term remains uninvestigated. The objective of this study was to evaluate the association between HbA1c variability and new-onset albuminuria in patients with type 2 diabetes (T2DM) over a 3-year period. METHODS: A retrospective cohort study was carried out at Khoo Teck Puat Hospital, Singapore, between June 2010 and June 2014. Patients who already had albuminuria at baseline were excluded. HbA1c variability was defined as the coefficient of variation (HbA1c-CV), calculated as the ratio of standard deviation to mean HbA1c values. Patients were defined as albuminuric if their urine albumin levels were above 30 mg/day at any point. Multivariate logistic regression was performed to adjust for the effects of known risk factors of albuminuria. Model discrimination was ascertained using the receiver operating characteristic (ROC) curve.  RESULTS: A total of 1,714 patients were included. Subjects had a median age of 57 (IQR 49–65) and a median baseline HbA1c of 8.0% (IQR 7.2% – 9.4%). During the 3 year follow-up, 842 out of 1,539 (54.7%) albuminuria-naïve patients developed albuminuria. HbA1c-CV was associated with new-onset albuminuria, independent of mean HbA1c and other known risk factors including age and presence of hypertension. The odds ratios for patients with moderate and high HbA1c-CV relative to those with low HbA1c-CV were 1.52 (95% CI 1.19–1.95) and 1.89 (95% CI 1.37–2.61), respectively. The inclusion of HbA1c-CV as a predictor significantly improves the area under the ROC for predicting 3 year albuminuria risk (0.658 vs. 0.669, p=0.03). CONCLUSIONS: Higher HbA1c-CV is independently associated with an increased risk of new-onset albuminuria within a 3-year period in T2DM patients. HbA1c-CV also improves short-term albuminuria prognostication which may be useful for communicating short-term risks of nephropathy to patients to effect favorable behavioral modification.