Objective: To determine whether the neonatal chest radiograph (CXR) at 28 days in very low birthweight (VLBW) infants who develop chronic neonatal lung disease (CNLD) predicts oxygen therapy duration or CXR abnormalities in early childhood. Also, to assess the inter-observer reliability of the radiologists scoring the CXR.
Methodology: Clinically well survivors of CNLD (n = 46) had neonatal CXR scored (mean age 28.5 days) and compared with current CXR (mean age 40 months). The CXR were scored independently and 'blindly' by two paediatric radiologists using a standardized scoring system (range 0-10).
Results: There was no correlation between neonatal CXR scores and current CXR scores for either radiologist. There was no association between CXR severity scores and duration of oxygen therapy for either neonatal or current CXR. Radiologist A scored the current CXR significantly more abnormal than radiologist B [medians (range): 3 (1-6) vs 1 (0-5), P < 0.001] with reasonable correlation (r = 0.593, P < 0.005) but worse than chance agreement (kappa = -0.034). The median scores for the neonatal CXR were similar [1.5 (0-8) vs 2 (0-8), P = 0.789] and again there was good correlation (r = 0.760, P < 0.0005) although poor individual agreement (kappa = 0.243) between radiologists.
Conclusions: Follow-up CXR abnormalities in VLBW infants with CNLD are usually minor and are not predictive of the duration of oxygen therapy that will be required nor of the CXR appearance in early childhood. Considerable inter-observer variation exists in the interpretation of the CXR in CNLD.