To evaluate the association of both mean HbA1c and HbA1c variability with DR development in patients with type 2 diabetes. Patients with type 2 diabetes who received dilated funduscopic examination annually and who underwent at least 2-year follow-up were included in this longitudinal study. Subjects were excluded if they took less than five HbA1c measurements during the follow-up period. HbA1C variability was expressed as A1c-SD, and the mean of HbA1c (A1c-Mean) was calculated. In addition, medical history and clinical data of all subjects were collected and analyzed. According to A1c-Mean above or below the value 7% and A1c-SD above or below the population mean value 0.76%, subjects were divided into four quartiles: Q1(A1c-Mean < 7%, A1c-SD < 0.76%); Q2(A1c-Mean < 7%, A1c-SD ≥ 0.76%); Q3(A1c-Mean ≥ 7%, A1c-SD < 0.76%); Q4(A1c-Mean ≥ 7%, A1c-SD ≥ 0.76%). 3152 participants were included in the study analysis with a median follow-up period of 3.95 years (2-5 years), 17.6% (n = 556) were found to have DR, and these patients also had higher HbA1c levels (P < 0.001). Linear mixed-effect models were performed after adjusting for the characteristics of participants and the results showed that HbA1c variability is an independent risk factor for DR. Cox regression revealed that patients in Q4 group had the highest DR prevalence (HR = 1.624, P < 0.001) while Q1 group had the lowest. In addition, patients in Q2 group (HR = 1.429, P = 0.006) had a higher risk of DR than those in Q3 group (HR = 1.334, P < 0.001). HbA1c variability is an independent predictor of DR in patients with type 2 diabetes in Asia. It may play a greater role in DR development than mean HbA1c does when the mean value of HbA1c variability index is above 0.75%, indicating that aggressive A1c lowering strategies may, in fact, contribute excessively to risk of DR in patients with type 2 diabetes; steady decline of A1c should be taken into consideration.