A Comparison of 2 Disease Burden Assessment Methods (3D Volume Versus the Number of Lesions) for Prognostication of Survival in Metastatic Melanoma: Implications for the Characterization of Oligometastatic Disease

Abstract

Purpose: Oligometastatic disease (OMD), generally defined by the presence of ≤5 metastatic lesions, represents an intermediate state between localized and widespread metastatic disease. This study aimed to question the conventional definition of OMD and assess the significance of the total volume and loci of metastases in characterizing OMD using an unselected metastatic melanoma cohort.

Methods and materials: We identified 86 consecutive patients with metastatic melanoma who received pembrolizumab monotherapy from 2015 to 2020. We retrospectively contoured the gross tumor volumes of all metastatic lesions on baseline and follow-up imaging. The number, total volume, and loci information of metastases was collected. The primary endpoint was overall survival. A density histogram plot was used for tumor characteristic descriptions, and classification analysis using the decision tree and random forest methods was performed to determine the optimal combination of prognostic factors in the clinical setting.

Results: A total of 2728 gross tumor volumes were delineated. On baseline imaging, the median number and total volume of metastases was 7 (interquartile range, 3-17) and 28.4 cc (interquartile range, 8.4-88.78), respectively. The lymph node was the most common metastatic site (n = 46, 54%), followed by the lungs (n = 32, 37%), liver (n = 23, 27%), and bones (n = 21, 24%). Two-year overall survival rates of patients with 1 to 5, 6 to 10, 11 to 20, and >20 metastases were 58%, 47%, 31%, and 14%, respectively, and with ≤10, 11 to 30, 31 to 130, and >130 cc of metastatic volume were 64%, 43%, 33%, and 25%, respectively. K-adaptive partitioning revealed that the optimal cutoff was 20 and 37.9 cc. Decision tree and random forest analyses revealed that volume and loci (brain and liver metastases) were the most important factors (Harrell's C-index, 0.78).

Conclusions: The OMD state could represent a continuous spectrum of disease burden instead of a binary phenomenon. We propose integrating the volumetric and spatial information of metastases into the characterization of OMD and the stratification tool of clinical trials in the metastatic setting, although external validation studies are needed.