AIM: To investigate whether heart rate variability (HRV) is a predictor for the incidence of diabetes in a 4-year follow-up. MATERIALS AND METHODS: The HRV of 9192 participants free of diabetes was analysed in time and frequency domains and stratified based on the reference values presented in the literature. The participants were then allocated to one of three groups, according to age-specific value distributions for each HRV domain: lower than the 25(th) percentile, between the 25(th) and 75(th) percentiles, and higher than the 75(th) percentile. The association between HRV and diabetes incidence at 4-year follow-up was analysed using Poisson regression models with robust estimator. RESULTS: Six hundred thirty-four participants (6.90%) developed diabetes within 4 years and five out of six HRV analysed indices showed increased relative risk of developing diabetes associated with low HRV: SDNN (RR = 1.29; 95% CI, 1.09-1.52; .003), pNN50 (RR = 1.33; 95% CI, 1.11-1.58; .001), RMSSD (RR = 1.29; 95% CI, 1.09-1.53; .004), LF (RR = 1.25; 95% CI, 1.05-1.48; .012), and HF (RR = 1.39; 95% CI, 1.16-1.63; .001). CONCLUSIONS: This study suggests that both overall variability and changes in parasympathetic modulation precede the incidence of diabetes. For four HRV indices below the 25(th) percentile, the risk for incident diabetes was 68% higher than for those participants who presented none. We concluded that HRV is an independent risk predictor of diabetes in a 4-year period.