Importance: Accuracy of the predicted metastasis-free survival (MFS) by a commercially available gene expression profiling (GEP) test is not known.
Objective: To compare the predicted MFS with the observed MFS in patients in this cohort and with those in published studies (published MFS, meta-analysis).
Design, setting, and participants: This cohort study included consecutive patients from the University of Iowa and Cleveland Clinic who were diagnosed with uveal melanoma who underwent prognostic fine-needle aspiration biopsy at the time of primary treatment. Patients were recruited from December 2012 to December 2020. The predicted MFS for patients was extracted from the GEP report. The observed MFS was defined as time to metastasis. Cox proportional hazards models were fit to identify tumor variables impacting MFS in patients with class 2 tumors. The overall estimate of the published MFS was obtained by performing meta-analysis of data from published series. Analysis took place in August 2021.
Main outcomes and measures: MFS.
Results: There were 92 patients from the University of Iowa and 255 patients from the Cleveland Clinic. The mean (SD) age at diagnosis was 59.4 (13.0) years. The median (IQR) follow-up interval was 38.0 (19.0-57.0) months. The observed MFS for patients with class 2 tumor in this cohort (3 years: 67% [95% CI, 59%-77%]; 5 years: 47% [95% CI, 37%-61%]) and in published studies (3 years: 62% [95% CI, 57%-66%]; 5 years: 40% [95% CI, 34%-46%]) were better than those predicted (50% and 28% for 3 and 5 years, respectively). Within patients with class 2 tumor, those with metastasis had larger tumors compared with nonmetastatic tumors (mean largest basal diameter difference, 1.7 [95% CI, 0.5-3.0] mm; P = .01; mean thickness ratio, 1.3 [95% CI, 1.04-1.5]; P = .01, respectively). An increasing tumor size was significantly associated with increased hazard ratio (1.16 [95% CI, 1.06-1.27]; P < .001) of metastasis.
Conclusions and relevance: These findings suggest the predicted MFS for metastatic tumors (class 2) appears to be worse than that observed here and reported by others. Incorporation of tumor size in the prediction model may enhance its accuracy. Adjuvant therapy trials may not be able to rely on predicted MFS to calculate efficacy with a high degree of confidence.