The reliability and value of determining mediastinal involvement and width on chest radiographs in patients with Hodgkin's disease

Over the years several methods for evaluating mediastinal involvement in Hodgkin's disease have been applied to chest radiographs and conflicting results have been reported. In a retrospective study of 104 patients we evaluated interobserver variability inassessing mediastinal involvement and investigated various cut-off points for mediastinal size as to their ability to identify patients with high- and low-risk for recurrence. For mediastinal involvement the concordance rate for two reviewing radiologists was 94% ( 98 104 ) and compared with prior assessment by outside radiologists the concordance rates were 90% ( 94 104 ) and 88% ( 92 104 ), respectively. A good correlation between the reviewing radiologists was found for the quantitative evaluation of mediastinal diameter and thoracic ratios. ROC curves and relative risk figures were used to investigate the various cut-off points for mediastinal width and for the ratios of the maximal mediastinal diameter to the chest diameter at Th 5–6 (M1) and to the chest diameter at the widest thoracic level (M2). Neither the ROC curve analysis nor the use of relative risk figures revealed a cut-off point clearly more accurate in predicting recurrence. In conclusion, our results do not suggest that interobserver variability in mediastinal assessment, differences in the method of mediastinal measurement, or the cut-off points applied to mediastinal width can explain the discrepancies in the reported data on the prognostic value of mediastinal width in Hodgkin's disease, but rather factors such as patient selection and differences in treatment given may be responsible.