OBJECTIVE: To investigate the trends in mortality, as well as in the timing and cause of death, among extremely preterm infants at the limit of viability, and thus to identify the clinical factors that contribute to decreased mortality. METHODS: We retrospectively reviewed the medical records of 382 infants born at 23-26 weeks' gestation; 124 of the infants were born between 2001 and 2005 (period I) and 258 were born between 2006 and 2011 (period II). We stratified the infants into two subgroups-"23-24 weeks" and "25-26 weeks"-and retrospectively analyzed the clinical characteristics and mortality in each group, as well as the timing and cause of death. Univariate and multivariate logistic regression analyses were done to identify the clinical factors associated with mortality. RESULTS: The overall mortality rate in period II was 16.7% (43/258), which was significantly lower than that in period I (30.6%; 38/124). For overall cause of death, there were significantly fewer deaths due to sepsis (2.4% [6/258] vs. 8.1% [10/124], respectively) and air-leak syndrome (0.8% [2/258] vs. 4.8% (6/124), respectively) during period II than during period I. Among the clinical factors of time period, 1-and 5-min Apgar score, antenatal steroid identified significant by univariate analyses. 5-min Apgar score and antenatal steroid use were significantly associated with mortality in multivariate analyses. CONCLUSION: Improved mortality rate attributable to fewer deaths due to sepsis and air leak syndrome in the infants with 23-26 weeks' gestation was associated with higher 5-minute Apgar score and more antenatal steroid use.