Comparison of Haller index values calculated with chest radiographs versus CT for pectus excavatum evaluation

Abstract

Pectus excavatum is a common chest wall anomaly in children. Pre-operative imaging for pectus excavatum is performed with CT, which is used to calculate the Haller index to determine the severity of pectus excavatum. To determine the correlation between Haller index values calculated with two-view chest radiographs and those calculated with CT and to determine, with CT as the reference standard, the diagnostic performance of radiographic Haller index for identifying cases that meet imaging criteria for surgical correction of pectus excavatum. For the period 2001-2009, our radiology information system was searched to identify all children who had undergone CT for Haller index calculation. Children who had also undergone two-view chest radiography (CXR) within 6 months of the CT were included in this retrospective study. Two radiologists independently calculated CT Haller index and radiographic Haller index. Data distributions were tested for normality with the Shapiro-Wilk W test. The associations between CT Haller index and radiographic Haller index were determined with the Spearman coefficient of rank correlation. Differences between CT Haller index and radiographic Haller index were tested with the Wilcoxon signed rank test. Haller index values were dichotomized into positive (>3.2) and negative (≤3.2) cases. Using CT as the reference standard, the sensitivity, specificity, and accuracy of radiographic Haller index in identifying children who meet imaging criteria for surgery were calculated. CT and CXR for evaluation of pectus excavatum were available for 32 children (25 male; median age 14.5 years). With CT, the median Haller indices for observers 1 and 2 were 3.4 and 3.5 and with CXR 3.5 and 3.5. There were statistically significant correlations between the radiographic Haller index and CT Haller index estimated by the two observers [Spearman correlation coefficient (95% confidence interval) for observer 1 = 0.71 (0.48-0.85, P < 0.01) and for observer 2 = 0.77 (0.52-0.88, P < 0.01)]. A statistically significant correlation was found between the radiographic Haller index calculated by the two observers [Spearman correlation coefficient = 0.98 (0.95-0.99, P < 0.01)]. Using CT Haller index as the reference standard, radiographic Haller index had a sensitivity of 0.95 (0.75-0.99), specificity of 0.75 (0.43-0.94), and accuracy of 0.88 (0.72-0.95) for observer 1, and a sensitivity of 0.95 (0.75-0.99), specificity of 0.67 (0.35-0.90), and accuracy of 0.84 (0.68-0.93) for observer 2. Radiographic Haller index correlates strongly with CT Haller index, has good interobserver correlation, and has a high diagnostic accuracy for pre-operative evaluation of pectus excavatum. We suggest that a CT of the chest is not required for pre-operative evaluation of pectus excavatum, and a two-view chest radiograph is sufficient for preoperative imaging of pectus excavatum.