Reassessment of Published Comparative Performance Model for 31-GEP and CP-GEP in Sentinel Lymph Node Biopsy Prediction in Cutaneous Melanoma

Background A recent large study of the CP-GEP for prediction of sentinel lymph node biopsy (SLNB) results in cutaneous melanoma (CM) failed to meet its primary endpoint of attaining a negative predictive value (NPV) >95%, indicating the test alone cannot be safely used to guide decisions to forgo SLNB under current clinical standards. In a separate analysis, CP-GEP performance from the study was compared to another SLNB predictive test, i31-GEP/i31-SLNB, by reanalyzing the largest i31-SLNB validation cohort. However, the reanalyzed data used in the analysis does not reflect the validation cohort, misrepresenting i31-SLNB performance. Objective To detail calculation errors in the comparative model of the i31-SLNB and CP-GEP for predicting SLNB positivity, and re-evaluate the tests’ performances by reconstructing the analyses using the datasets from the original publications. Methods From patients who underwent SLNB with T1-T3 tumors, NPVs were calculated and false omission rate (FOR=1−NPV) modeled as a function of prevalence for each of the i31-SLNB and CP-GEP. NPVs among all SLN-assessed patients in other published United States-based cohorts of varying SLN positivity prevalences were analyzed. Results In each respective cohort, 96.1% NPV for the i31-SLNB (n=1,152) and 92.9% NPV for the CP-GEP (n=1,761) were observed. Modeled FOR indicated less frequent false negatives for the i31-SLNB than CP-GEP across SLN positivity prevalences. Further, in reported United States-based cohorts, the i31-SLNB had 100% NPV in two additional cohorts with 20.5% and 6.4% SLN positivity prevalence, while the CP-GEP had 93.8% NPV in one additional cohort with 21.2% SLN positivity prevalence. Conclusion Although both tests were designed to attain >95% NPV in prediction of CM SLN metastasis in SLNB-eligible patients, only the i31-SLNB met this benchmark, supporting its utility as the only GEP to identify patients at low risk (95%, suggesting that SLNB should still be considered rather than omitted based on this result alone.