INTRODUCTION: Serial changes in cardiac troponin in hemodialysis (HD) patients have uncertain clinical implications. We evaluated associations of adverse outcomes in HD patients with reference change value (RCV) data and tertile concentrations for cardiac troponin I (cTnI) and cTnT measured by high-sensitivity (hs) assays. METHODS: RCV data and tertiles for hs-cTnI and hs-cTnT were determined from plasma samples collected 3 months apart in 677 stable outpatient HD patients and assessed for their associations with adverse outcomes using adjusted Cox models. Primary outcomes were all-cause mortality and sudden cardiac death (SCD). RESULTS: During a median follow-up of 23 months, 18.6% of patients died. RCVs were: hs-cTnI +37% and - 30%; hs-cTnT +25% and -20%. Patients with serial hs-cTnI and hs-cTnT changes >RCV (increase or decrease) had all-cause mortality of 25.2% and 23.8% respectively, compared to 15.0% and 16.5% with changes ≤RCV [adjusted hazard ratios (aHRs): 1.9, P = 0.0003 and 1.7, P = 0.0066), respectively]. Only hs-cTnI changes >RCV were predictive of SCD (aHR 2.6, P = 0.005). hs-Cardiac troponin changes >RCVimproved all-cause mortality prognostication compared to changes ≤RCV in tertile 2: hs-cTnI aHR, 2.70 (P = 0.003); hs-cTnT aHR, 1.98 (P = 0.043). The aHR of changes in hs-cTnI in tertile 2 >RCV for SCD was 5.62 (P = 0.039). CONCLUSIONS: Changes over 3 months in hs-cTnI and hs-cTnT of >RCV identified patients at greater risk of all-cause mortality, and for hs-cTnI were also predictive of SCD. Among patients with middle tertile cardiac troponin concentrations, hs-cTnI changes >RCV provided additive prognostic value for both SCD and all-cause mortality, whereas those for hs-cTnT provided additive prognostic value only for all-cause mortality.