BACKGROUND: Despite the emerging clinical relevance of heart rate variability (HRV) as a potential biomarker of cognitive decline and as a candidate target for intervention, there is a dearth of research on the prospective relationship between HRV and cognitive change. In particular, no study has addressed this issue in subjects with a diagnosis of cognitive status including cognitive impairment. OBJECTIVE: To investigate HRV as a predictor of cognitive decline in subjects with normal cognition (NC) or Mild Cognitive Impairment (MCI). Specifically, we tested the literature-based hypothesis that the HRV response to different physical challenges would predict decline in different cognitive domains. METHODS: This longitudinal study represents the approximately 3-year follow-up of a previous cross-sectional study enrolling 80 older outpatients (aged >/= 65). At baseline, power spectral analysis of HRV was performed on five-minute electrocardiographic recordings at rest and during a sympathetic (active standing) and a parasympathetic (paced breathing) challenge. We focused on normalized HRV measures [normalized low frequency power (LFn) and the low frequency to high frequency power ratio (LF/HF)] and on their dynamic response from rest to challenge (Delta HRV). Extensive neuropsychological testing was used to diagnose cognitive status at baseline and to evaluate cognitive change over the follow-up via annualized changes in cognitive Z-scores. The association between Delta HRV and cognitive change was explored by means of linear regression, unadjusted and adjusted for potential confounders. RESULTS: In subjects diagnosed with MCI at baseline a greater response to a sympathetic challenge predicted a greater decline in episodic memory [adjusted model: Delta LFn, standardized regression coefficient (beta) = -0.528, p = 0.019; Delta LF/HF, beta = -0.643, p = 0.001] whereas a greater response to a parasympathetic challenge predicted a lesser decline in executive functioning (adjusted model: Delta LFn, beta = -0.716, p < 0.001; Delta LF/HF, beta = -0.935, p < 0.001). CONCLUSION: Our findings provide novel insight into the link between HRV and cognition in MCI. They contribute to a better understanding of the heart-brain connection, but will require replication in larger cohorts.