Association of caffeine citrate plus vitamin A/D drops with bronchopulmonary dysplasia and respiratory outcomes in preterm infants born at <32 weeks' gestation: a 72-h landmark retrospective cohort study.

OBJECTIVE: The aim of this work is to examine whether the early addition of vitamin A/D drops to caffeine citrate was associated with bronchopulmonary dysplasia (BPD) and short-term respiratory outcomes in preterm infants born at <32 weeks' gestation. METHODS: This single-center, 72-h landmark retrospective cohort study included 126 infants admitted to the hospital between October 2022 and November 2024 with a gestational age <32 weeks and birth weight ≤1.5 kg. Exposure was determined at 72 h as either caffeine alone or caffeine plus vitamin A/D drops. Complete-case analyses were performed. Among the 119 infants who survived to 36 weeks' postmenstrual age, BPD, longitudinal outcomes, and clinical course were evaluated. Multivariable regression and stabilized inverse probability of treatment weighting (IPTW) were applied. RESULTS: BPD occurred in 20 of 55 survivors (36.4%) receiving caffeine alone and 11 of 64 (17.2%) receiving combination therapy. After adjustment for gestational age, birth weight, and grade III-IV respiratory distress syndrome, combination therapy was associated with a lower risk of BPD [adjusted odds ratio (aOR), 0.36; 95% confidence interval (CI), 0.14-0.90]. The composite outcome of death or BPD occurred in 25 of 60 infants (41.7%) in the caffeine group and 13 of 66 (19.7%) in the combination group (extended-model aOR, 0.33; 95% CI, 0.14-0.80); IPTW results were consistent. Combination therapy was also associated with more favorable changes in blood gas and inflammatory markers, along with shorter durations of respiratory support and hospitalization. No serious adverse event required permanent treatment discontinuation. CONCLUSION: Among infants who survived to 72 h, early initiation of vitamin A/D drops as an adjunct to caffeine was associated with lower risks of BPD and death or BPD, as well as more favorable short-term respiratory outcomes. Non-randomized treatment allocation, complete-case selection, and residual confounding preclude causal inference.