Abstract Background DNA methylation-derived epigenetic clocks and frailty are well-established biological age measures capturing different aging processes. However, whether they are dynamically linked to each other across chronological age remains poorly understood. Methods This analysis included 1 309 repeated measurements in 524 individuals aged 50–90 years from the Swedish Adoption/Twin Study of Aging. Frailty was measured using a validated 42-item frailty index (FI). Five epigenetic clocks were calculated, including 4 principal component (PC)-based clocks trained on chronological age (PCHorvathAge and PCHannumAge) and aging-related physiological conditions (PCPhenoAge and PCGrimAge), and a pace of aging clock (DunedinPACE). Using dual change score models, we examined the dynamic, bidirectional associations between each of the epigenetic clocks and the FI over age to test for potential causal associations. Results The FI exhibited a nonlinear, accelerated increase across the older adulthood, whereas the epigenetic clocks mostly increased linearly with age. For PCHorvathAge, PCHannumAge, PCPhenoAge, and PCGrimAge, their associations with the FI were primarily due to correlated levels at age 50 but with no evidence of a dynamic longitudinal association. In contrast, we observed a unidirectional association between DunedinPACE and the FI, where a higher DunedinPACE predicted a subsequent increase in the FI, but not vice versa. Conclusions Our results highlight a temporal order between epigenetic aging and frailty such that changes in DunedinPACE precede changes in the FI. This potentially suggests that the pace of aging clock can be used as an early marker of the overall physiological decline at system level.