Central sleep apnea (CSA) is characterized by recurring cycles of crescendo-decrescendo ventilation during sleep, and enhances sympathetic nerve activity. Thus CSA has a prognostic impact in patients with chronic heart failure (CHF). Although nocturnal oxygen (O2) therapy decreases frequency of CSA and improves functional exercise capacity, it is also known that some non-responders to the therapy exist. We thus aimed to identify predictors of responders to nocturnal O2 therapy in CHF patients with CSA. In 12 CHF patients with CSA hospitalized at our department, sleep study was performed at 2 consecutive nights. Patients nasally inhaled O2 at either the first or second night in a randomized manner. To predict the percentage reduction in apnea-hypopnea index (%ΔAHI) in response to the nocturnal O2 therapy, we performed multiple regression analysis with a stepwise method with variables including age, brain-natriuretic peptide, circulation time, baseline AHI, hypercapnic ventilatory response and end-tidal carbon dioxide tension (PETCO2). Nocturnal O2 therapy significantly decreased AHI (from 32 ± 13 /h to 12 ± 10 /h, P < 0.0001). Among the possible predictors, PETCO2 was the only variable that is predictive of % changes in AHI. Receiver operating characteristics analysis determined 4.25% as the optimal cutoff PETCO2 level to identify responder to nocturnal O2 therapy (> 50% reduction of AHI), with 88.9% of sensitivity and 66.7% of specificity. In conclusion, PETCO2 is useful to predict the efficacy of O2 therapy in CHF patients with CSA, providing important information to the current nocturnal O2 therapy.