The present study aimed to investigate the potential causal pathways and temporal relationships of glucose metabolism and diabetes with heart rate variability (HRV). This cohort study was conducted among a sample of 3858 Chinese adults. At baseline and 6 years follow-up, participants underwent HRV measurement (low frequency [LF], high frequency [HF], total power [TP], standard deviation of all normal-to-normal intervals [SDNN], and square root of the mean squared difference between adjacent normal-to-normal intervals [r-MSSD]) and determination of glucose homeostasis (fasting plasma glucose [FPG] and insulin [FPI], homeostatic model assessment for insulin resistance [HOMA-IR]). The temporal relationships of glucose metabolism and diabetes with HRV were evaluated using cross-lagged panel analysis. FPG, FPI, HOMA-IR, and diabetes were cross-sectionally negatively associated with HRV indices at baseline and follow-up (P < 0.05). Cross-lagged panel analyses demonstrated significant unidirectional paths from baseline FPG to follow-up SDNN (beta = -0.06), and baseline diabetes to follow-up low TP group (beta = 0.08), low SDNN group (beta = 0.05), and low r-MSSD group (beta = 0.10) (P < 0.05). No significant path coefficients were observed from baseline HRV to follow-up impaired glucose homeostasis or diabetes. These significant findings persisted even after excluding participants who were taking antidiabetic medication. The results support that elevated FPG and the presence of diabetes may be the causes rather than the consequences of HRV reduction over time.