Diagnostic Accuracy of Preoperative CT Imaging in Pleuropulmonary Blastomas and Congenital Lung Malformations: Results from a Multi-Institutional Operative Registry
Introduction: The ability of computed tomography (CT) imaging to accurately discriminate between pleuropulmonary blastoma (PPB) and benign congenital lung malformations (CLM) is controversial. Therefore, we performed a blinded study to determine whether pediatric academic radiologists could adequately discern PPB from CLM using cases from a multicenter research collaborative.
Methods: After central reliance IRB approval, pathologically confirmed cases of PPB were reviewed from a multicenter operative registry of lung lesions (n=521). Preoperative CT scans of PPB lesions were age matched with scans from children with pathologically confirmed CLMs and interpreted independently by blinded, board-certified pediatric radiologists. Diagnostic confidence and suspicion for malignancy scores were captured using 5-point Likert scales. Statistical analysis included Cohen’s kappa statistics and logistic regression as appropriate.
Results: Forty patients with new, postnatally detected lung lesions were selected. Pathologic diagnoses included cystic PPB (n=9, 22.5%) and congenital pulmonary airway malformation (n=20, 50%). There were no significant differences in age at CT [PPB: 8.2 mos, interquartile range (IQR), 4.0-70.1 vs. non-PPB: 7.3 mos, IQR, 2.0-22.3; p=0.75] and age at resection (PPB: 10.2 mos, IQR, 4.2-71.2 vs. non-PPB: 8.8 mos, IQR, 5.1-24.4; p=0.83). The interrater reliability (kappa score) was 41.18.4% (p<0.001). The mean diagnostic confidence score was 3.80.3, and the accuracy rate of CT for differentiating benign versus malignant lesions was 75.5%. The positive predictive value and negative predictive value for a CT diagnosis of cancer were 52.8% and 90.1%, respectively. However, logistic regression showed that increasing suspicion for malignancy (mean score, 2.20.5) was significantly associated with PPB pathology (odds ratio, 13.6; 95% confidence, 4.1-44.8; p<0.001, Figure).
Conclusion: This multi-institutional study suggests that there is relatively low agreement and reliability in distinguishing CLM and PPB lesions amongst experienced radiologists. A low threshold for operative management in any child with a newly diagnosed lung lesion is warranted.