Improving Risk Estimates of Metastasis From Uveal Melanoma Through Mutation Profiling

The possibility of predicting metastasis from uveal melanoma was revolutionized when cytogenetic and gene-expression profiling could tell with unprecedented accuracy tumors that disseminate from those unlikely to do so. ¹ Subsequently, typical mutations were identified: BAP1 (a regulatory enzyme) is mutated in the high-risk type, and EIF1AX (a translation initiation factor) and SF3B1 (a splicing factor) in the low-risk type. ¹  

Ewens and colleagues ² show that combining genetic predictors with chromosome 3 status broadens the spread of risk estimates and increases prognostic accuracy. For example, when adjusting for tumor size and location, patients with monosomy 3 and mutated BAP1 but wild-type EIF1AX were 37.5 times more likely to die of metastases within 2 years than patients blessed with the opposite, low-risk phenotype. The confidence intervals of this pilot study are wide, ranging from not larger than 4.3 times higher risk in the above example. Also, it consisted of comparable numbers of selected patients who developed metastasis within 2 years and controls who remained metastasis-free at that time. Such a nonconsecutive, nonrandom design precludes Cox proportional hazard regression. ³ Logistic regression was appropriate but could not use follow-up data beyond 2 years. Thus, the reported estimates preferentially refer to fast-growing metastases, and additional patients will develop metastases later on. 

With due allowance to the limitations of their study, its design is clinically attractive. The authors acknowledge and confirm that time-honored indicators of prognosis ⁴ —tumor dimensions and location—also captured in the American Joint Committee on Cancer staging, ⁵ modify the risk related to genetic markers. For example, taking these factors into account weakened somewhat the estimated effect of wild-type EIF1AX in the face of monosomy 3 and mutated BAP1, but considerably strengthened the effect of E1F1AX in the presence of disomy 3 and wild-type BAP1. Their data indicate that both clinicopathologic and molecular markers are essential for prognostication.