Predictors and Outcomes of Extubation Failure in Preterm Neonates: A Systematic Review.

CONTEXT Extubation failure (EF) is common in preterm neonates and may be associated with adverse outcomes. OBJECTIVE To systematically review and meta-analyze the existing literature on predictors and outcomes of EF in preterm neonates. DATA SOURCES MEDLINE, Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Embase (OvidSP), CINAHL (EBSCOHost), and Cochrane Library (Wiley) from 1995 onward. The search strategy was developed by a reference librarian. STUDY SELECTION Experimental or observational studies reporting on predictors and/or outcomes related to EF (defined as reintubation within 7 days) in preterm neonates less than 37 weeks were eligible. Predictors included machine learning (ML) algorithms and lung ultrasound (LUS). Main outcome of interest was association of EF with mortality and/or bronchopulmonary dysplasia (BPD). DATA EXTRACTION Studies identified by the search strategy were screened based on title and abstract. Data from included studies were extracted independently by 2 authors, along with adjudication of risk of bias. RevMan Web was used to conduct meta-analyses. RESULTS Out of 8336 studies screened, 120 were included. Neonates with lower gestational age at birth, birthweight, postmenstrual age, and weight at extubation were more likely to experience EF. Higher level of pre-extubation respiratory support, indicated by lower pre-extubation pH and higher pre-extubation mean airway pressure, fraction of inspired oxygen, and Pco2 were associated with EF risk. ML models showed variable accuracy and lower external validity. LUS may be a promising predictor, though scoring systems varied. EF was associated with higher odds of mortality and/or BPD (pooled odds ratio [OR], 4.7; 95% CI, 2.84-7.76) as well as the individual components of the composite: mortality (pooled OR, 3.87; 95% CI, 2.35-6.36) and BPD (pooled OR, 3.27; 95% CI, 2.54-4.21). LIMITATIONS Associations were derived from unadjusted data, precluding a definitive causal relationship between EF and predictors/outcomes. CONCLUSIONS Lower gestational and chronological age and higher levels of pre-extubation ventilation support were associated with EF. ML models and LUS scores require further validation in larger studies. EF was associated with mortality and/or BPD.