Aims/hypothesis: Cardiac autonomic nervous dysfunction (CAND) raises the risk of mortality, but the glycaemic threshold at which it develops is unclear. We aimed to determine the prevalence of, risk factors for and impact of CAND in glucose intolerance and diabetes. Methods: Among 1,332 eligible participants aged 55–74 years in the population-based cross-sectional KORA S4 study, 130 had known diabetes mellitus (k-DM), and the remaining 1,202 underwent an OGTT. Heart rate variability (HRV) and QT variability were computed from supine 5 min ECGs. Results: In all, 565 individuals had normal glucose tolerance (NGT), 336 had isolated impaired fasting glucose (i-IFG), 72 had isolated impaired glucose tolerance (i-IGT), 151 had combined IFG–IGT (IFG–IGT) and 78 had newly detected diabetes mellitus (n-DM). Adjusted normal HRV limits were defined in the NGT population (5th and 95th percentiles). Three HRV measures were more frequently abnormal in those with k-DM, n-DM, IFG–IGT and i-IFG than in those with NGT (p < 0.05). The rates of CAND (≥2 of 4 HRV indices abnormal) were: NGT, 4.5%; i-IFG, 8.1%; i-IGT, 5.9%; IFG–IGT, 11.4%; n-DM, 11.7%; and k-DM, 17.5% (p < 0.05 vs NGT, except for i-IGT). Reduced HRV was associated with cardiovascular risk factors used to construct a simple screening score for CAND. Mortality was higher in participants with reduced HRV (p < 0.05 vs normal HRV). Conclusions/interpretation: In the general population aged 55–74 years, the prevalence of CAND is increased not only in individuals with diabetes, but also in those with IFG–IGT and, to a lesser degree, in those with i-IFG. It is associated with mortality and modifiable cardiovascular risk factors which may be used to screen for diminished HRV in clinical practice.