The present study aimed to explore the early predictive marker of diabetic retinopathy (DR) and to elucidate the associated demographic, metabolic, and ocular factors. We enrolled 43 type 2 diabetic subjects with mild non-proliferative retinopathy (MNPDR), 30 diabetic subjects with no retinopathy (DNR), and 35 healthy controls (HC). The study groups showed no significant alteration in central macular thickness (CMT) and visual acuity (VA). The contrast sensitivity (CS) score was found to be significantly lower among DNR and MNPDR subjects compared to HCs (p < 0.0001). Between MNPDR and DNR subjects, the CS score was significantly lower in the former (p = 0.0036). CS score discriminated DNR subjects from HC, with 74% accuracy for the optimal threshold 0.71. The associated area under the ROC curve (AUC) is 0.82 (p < 0.0001) while the discrimination rule has 66% sensitivity and 80% specificity. The CS score also discriminated MNPDR subjects from DNR with 64% accuracy for the optimal threshold 0.53. The associated AUC is 0.65 (p < 0.023) and the rule has 86% sensitivity and 33% specificity. According to best subset regression analysis, not only glycaemic parameters but also lipid parameters [low-density lipoprotein cholesterol (LDL-C) (p = 0.045) and triglycerides (TG) (p = 0.0005)] were found to be significant predictors of CS. CMT (p = 0.058) was another marginally significant predictor of CS. CS may be used as an early predictive marker for DR. So, not only hyperglycemia, but also hyperlipidemia seems to significantly affect retinal CS function in diabetes.