Background
Standard imaging modalities are inaccurate in staging malignant pleural mesothelioma (MPM). Single institution studies suggest that volumetric computed tomography (VolCT) is more accurate but labor intensive. We established a multicenter network to test interobserver variability, accuracy (relative to pathologic stage) and prognostic significance of semi-automated VolCT.
Methods
Six institutions electronically submitted clinical and pathologic data to an established multicenter database on patients with MPM who had surgery. Institutional radiologists reviewed preoperative CT scans for quality then submitted via electronic network (AG mednet) to biostatistical center (BC). Two reference radiologists, blinded to clinical data, performed semi-automated tumor volume calculations using commercially available software (Vitrea Enterprise 6.0), then submitted readings to BC. Study endpoints included: feasibility of network; interobserver variability for VolCT; correlation of tumor volume to pTN stages, and overall survival (OS).
Results
Of 164 cases, 129 were analyzable and read by reference radiologists. Most tumors were <500cm3. A small bias was observed between readers, as one provided consistently larger measurements than the other (mean difference=47.9, p=.0027), but for 80% of cases, the absolute difference was ≤ 200cm3. Spearman correlation between readers was 0.822. Volume correlated with pTN stages and OS, best defined by 3 groups with average volumes of: 91.2, 245.3, 511.3cm3, associated with median OS of 37, 18, 8 months respectively.
Conclusions
For the first time, a multicenter network was established and initial correlations of tumor volume to pTN stages and OS shown. A larger multicenter international study is planned to confirm results and refine correlations.