Blood Plasma Metabolomics to Support Uveal Melanoma Diagnosis

ImportanceUveal Melanomas (UM) micro-metastasis can be present prior to diagnosis and relapse after treatment. Earlier detection resulted in an increased incidence of small (T1 and T2) tumors allowing for novel eye-preserving treatment strategies but reducing available tumor tissue needed for prognostic genomic profiling, creating the need for minimal-invasive detection and novel prognostication methods. ObjectiveTo determine whether tumor presence can be confirmed using metabolite patterns in blood plasma and to evaluate if these patterns differ between high risk (BRCA1-associated protein-1, BAP1), intermediate risk (Splicing Factor 3b Subunit 1, SF3B1) and low risk (Eukaryotic Translation Initiation Factor 1A X-Linked, EIF1AX) mutated tumors. DesignRetrospective observational study including discovery (n=53) and replication (n=42) convenience sample sets compared to unaffected control-participants (n=46) as well as across mutation-based subgroups. SettingPatients from two tertiary referral centers specialized in ocular oncology: The Rotterdam Eye Hospital and the Erasmus MC Cancer Institute were included. ParticipantsSex-matched controls and patients were included based on their prognostic relevant secondary driver mutations. Peripheral blood plasma was collected at diagnosis, prior to treatment. Exclusion criteria were the presence of other malignancies or co-occurrence of systemic diseases at time of diagnosis. Main outcome and measureMetabolite profiles of patients and control-participants were generated as mass/charge (m/z) features using ultra-high performance liquid chromatography mass-spectrometry. After normalization, discriminatory feature patterns were determined using a random forest classifier and leave-one-out cross-validation. ResultsWe detected differential metabolic patterns with a sensitivity of 0.95 and 0.90 and a specificity of 0.98 and 0.98 in the positive and negative ion modes, respectively. The accuracy of the model for classifying the subgroups was insufficient for the discovery (0.600 and 0.614 in the positive and negative ion modes, respectively) and replication cohort (0.544 and 0.672 in the positive and negative ion modes, respectively). Conclusion and relevanceMinimally invasive metabolomics does not discriminate between the prognostic relevant BAP1, SF3B1 and EIF1AX mutated UM-subgroups. However, this technique has the potential to allow for minimal invasive screening as it distinguishes metabolite patterns in peripheral blood derived plasma of UM-patients from control-participants. Key pointsO_ST_ABSQuestionC_ST_ABSCan we discriminate uveal melanoma patients and mutation subgroups from unaffected control-participants using the metabolome of peripheral blood plasma taken at time of diagnosis? FindingsIn this retrospective observational study, we find a low sensitivity and specificity to detect subgroups but a high sensitivity and specificity to discriminate patients from control-participants by measuring metabolite abundancy in plasma using ultra-high performance liquid chromatography mass-spectrometry and reach a receiver operating characteristic area under the curve of 0.993. MeaningThese results suggest that surveying the metabolome of uveal melanoma patients could aid in the minimal invasive detection of uveal melanoma.