Results

A total of 729 patients were recruited: 319 in the Theophylline group and 410 in the non-Theophylline group.
Tables 1 and 2 summarize the two groups’ baseline characteristics. No differences were found in sociodemographic variables, but significant clinical differences were identified, being infants in Theophylline group more fragile (less GA at birth and at KMCP entry, less weight at birth and at entry, more hospital stay, and less non-invasive ventilation).
Table 3 describes the follow-up results. The Theophylline group had more BPD defined as OD >28 days and GA ≧ 36 weeks, but when BPD was defined as OD >28 days and GA ≧40 weeks, there was no difference between groups18. Infants who received Theophylline had a higher frequency of apneas during the hospital stay and ambulatory follow-up. Additionally, chronological age at oxygen weaning and weight at 40 weeks were higher than in the non-Theophylline group. Although the feeding pattern at oxygen weaning was similar between the two groups, the non-Theophylline group had a higher proportion of exclusive breast milk (EBM) at 40 weeks (46 vs 36%).
No differences were found in the number of readmissions up to 40 weeks. Neither in the frequency of intraventricular hemorrhage (IVH), retinopathy of prematurity (ROP) or neurodevelopmental problems.
Regarding the possible side effects associated with Theophylline treatment, we found no differences in GER or gastrointestinal disorders. However, more tachycardia episodes were recorded in the Theophylline group (1.3% vs 0).
Finally, patients in the Theophylline group had a higher percentage of losses (3.5% vs 1%), although in general, these were minor for the two cohorts.
In order to adjust for baseline differences between groups (GA, birth weight, GA at study recruitment, frequencies of IUGR and apnea, days in NICU and days with oxygen at study recruitment), we performed propensity score matching. Then, multiple linear regression with total days of oxygen dependency as the outcome variable and having received Theophylline as the exposure variable. The control variables were frequency of EBF at O2 weaning, days of mechanical ventilation, weight at study recruitment, weight gain between study recruitment and oxygen weaning and weight reached at oxygen weaning. Probable interactions were entered, and model diagnostics were performed (after individual evaluation, 6 outliers were excluded).
Tables 4 and 5 show the propensity score matching and multiple linear regression analysis. After adjusting for baseline differences between groups with propensity score matching, it was found that controlling for days of mechanical ventilation, weight at study recruitment and weight gain to reach the oxygen weaning weight (Figure 1), EBF could reduce oxygen dependency in 9 days; on the other hand, for each day with mechanical ventilation, oxygen dependency increases in 2 days. No effect of Theophylline administration was found between the two groups (682 observations, R2 0.63, 95% CI (-1.01, 4.66) p=0.21).