**Mann Whitney Test
†Fisher’s exact test
¶Chi-square test
Gestational Age (GA), birthweight (BW), Intravenous antibiotics (IVABx), parental nutrition (PN), Patent Ductus Arteriosus (PDA), High Frequency Oscillation ventilation (HFOV), Mean Airway Pressure (MAP), inspired oxygen (FiO2), oxygen saturations(SpO2), partial pressure of carbon dioxide on blood gas (pCO2), Haemoglobin (Hb), Pulmonary interstitial emphysema (PIE), chest xray (CXR)
Table 3 shows the results of the univariate and multivariate binary logistic regression analyses with unadjusted and adjusted odds ratios (OR). Those variables included in the multivariable analysis (p <0.2 on univariate) were the GA at birth, birthweight, presence of chorioamnionitis, number of failed extubations, age in days at starting dexamethasone, current weight, ventilation mode at time of dexamethasone, FiO2 24hrs prior to dexamethasone, MAP, PN use, feed type, last pH and pCO2 on gas prior, PDA status, episodes of sepsis, airleak and CXR grade. Corrected GA at the time of starting dexamethasone was not included in the model because it was linearly correlated with age at stating dexamethasone so only one of these variables was included (i.e., age at stating dexamethasone).
The univariate analysis suggests that GA at birth was strongly protective with each additional gestational week leading to a 1.8 increase in the odds of successful extubation by 14 days (p<0.001). This was confirmed when adjusting for the other patient variables (OR=1.533; 95% CI: 1.122,2.096; p<0.01). BW, age in days and current weight at the time of dexamethasone were all strongly predictive of successful extubation in univariate analysis however after mutual adjustment these factors were no longer significant.
Several potentially important prognostic factors identified by the univariate analysis, could no longer be shown to be significant predictors when we adjusted for other factors. These include HFOV, PDA status and CXR grading of BPD.
Multivariable modelling revealed two other variables that remained significantly different once adjusting for all other variables. The predictive effects of FiO2 and MAP remained significant even after adjusting for other important prognostic indicators. For every additional unit of average FiO2 requirement in preceding 24 hours, there was an associated 6% decrease in the odds pf successful extubation by day 14 (OR=0.94, 95%CI: 0.921, 0.997; p<0.05) and higher MAP resulting in 0.76 times the odds in extubation (p<0.01).
Table 3. Unadjusted and adjusted odds ratios (OR) for odds of being successfully extubated by day 14 after starting dexamethasone.