Purpose: To examine the capability of optical coherence tomography-derived retinal thickness measures in detecting 4-year incident diabetic peripheral neuropathy (DPN). Methods: 145 eyes of 145 participants with diabetes but no DPN at baseline were examined for incident DPN. HbA1c levels, nephropathy, neuropathy (DPN), cardiovascular measures, and various retinal thickness measures were examined at baseline and after 4 years. Incidence of DPN was defined as newly developed DPN at follow-up. Baseline factors were assessed by univariate and a step-wise multiple logistic regression, and the predictors were examined for diagnostic capabilities. Results: Of the 145 participants without DPN at baseline, 51 had developed DPN when examined after 4 years (35% incidence). Of the ophthalmic variables, the mean (S.D.) of the overall thickness in the parafovea at baseline was 315 (18) μm in the no DPN group and 306 (18) μm in the ‘incidence’ group, and the differences were significant, p = 0.005. The superior hemisphere parafovea (mean (S.D.): 318 (17) μm vs 310 (20) μm, p = 0.02) and inferior hemisphere parafovea (313 (19) μm vs 302 (18) μm, p = 0.002) were different in the incident DPN group compared with the no DPN group. When adjusted for age, retinal thickness in the parafovea (AUC = 0.65, p = 0.003, 86% sensitivity and 44% specificity at 321 μm criterion), and body mass index or BMI (AUC = 0.65, p = 0.003, 49% sensitivity and 83% specificity at 29.3 kg m−2 criterion) at baseline were significant predictors for 4-year incident DPN. Conclusions: A lower retinal thickness at the parafovea and a higher BMI can predict 4-year incident neuropathy in patients with diabetes, with acceptable diagnostic accuracies. This OCT-derived measure may serve as a potential ophthalmic marker in the screening of patients at risk of developing DPN.