Pathologic grading of malignant pleural mesothelioma : an evidence-based proposal

Introduction: A pathologic grading system (PGS) for malignant pleural mesothelioma (MPM) is warranted to better identify different risk categories of patients, plan therapeutic options, and activate clinical trials. Methods: A series of 940 patients with MPM (328 in a training set and 612 in a validation set) that was diagnosed between October 1980 and June 2015 at the participant institutions was retrospectively assembled. A PGS was constructed by attributing to each histologic parameter, independent at multivariate analysis with excellent reproducibility (κ > 0.75), different scores based on the increase in corresponding hazard ratios. The relevant PGS score thus ranged from 0 to 8 points for individual patients with MPM. Conclusions: The PGS was constructed by taking into consideration the histological subtyping of MPM (epithelioid/biphasic = 0 pointssarcomatoid = 2 points), necrosis (absent = 0 points versus present = 1 point), mitotic count per 1 mm2 (cutoffs as follows: 1-2 = 0 points, 3-5 = 1 point, 6-9 = 2 points, or ≥10 = 4 points), and Ki-67 labeling index based on 2000 cells (<30% = 0 points versus ≥30 = 1 point), all of which are independent factors in both patient sets after adjustment for stage and age at diagnosis. No heterogeneity was seen across the validation centers (p = 0.19). Epithelioid/biphasic MPM patterning and biopsy versus resection did not affect survival, whereas the PGS outperformed mitotic count and Ki-67 LI in both the training (area under the curve receiver operating characteristic = 0.76) and validation sets (area under the curve receiver operating characteristic = 0.73) (p < 0.01). Patient survival progressively deteriorated from a score of 0 (median times of 26.3 and 26.9 months) to a score 1 to 3 (median times of 12.8 and 14.4 months) and a score of 4 to 8 (median times of 3.7 and 7.7 months) in both sets of patients, with the hazard ratio for a 1-point increase in score being 1.46 (95% confidence interval: 1.36-1.56) in the training set and 1.28 (95% confidence interval: 1.22-1.34) in the validation set (after adjustment for age and [when available] tumor stage). The PGS was effective even in subgroup analysis (epithelioid, biphasic, and sarcomatoid tumors). Discussion: A simple and reproducible multiparametric PGS effectively predicted survival in patients with MPM