External validation of a Dutch predictive nomogram for complete response to T-VEC in an independent American patient cohort.

Abstract

9563 Background: Talimogene Laherparepvec (T-VEC) is a genetically modified herpes simplex type 1 virus and known as an effective oncolytic immunotherapy for injectable cutaneous, subcutaneous, and nodal melanoma lesions in stage IIIB-IVM1a patients. Recently, Stahlie et al. published (Cancer Immunol Immunother '21) a model for predicting a complete response (CR) to T-VEC based on 3 easily accessible tumor characteristics identified using univariable and multivariable logistic regression analysis. The aim of this study was to externally validate this model in an independent, American patient cohort. Methods: A total of 76 patients with stage IIIB-IVM1a melanoma treated with T-VEC at Moffitt Cancer Center were included. A second nomogram was built incorporating the same predictive factors: tumor size (diameter of largest metastasis in mm), type of metastases (cutaneous, subcutaneous and nodal) and number of metastases (cut-off: <20 and >20). Predictive accuracy was assessed through calculation of overall performance, discriminative ability, and calibration. Outcomes and previously published outcomes were compared. Statistical analyses were done using R software. Results: Overall performance of the validation dataset nomogram was calculated with the Brier score and found to be 0.195, demonstrating good overall performance and similar to the original model Brier score of 0.182. Discriminative power, assessed by calculating the area under the receiver operating characteristic (ROC) curve was similar for both models, 0.767 and 0.755 for the NKI and Moffitt, respectively, resulting in a fair discriminative ability. The calibration curve showed mostly slight underestimation for predicated probabilities >0.37 and slight overestimation <0.37. Conclusions: An independent dataset externally validated a recently published predictive nomogram for CR to T-VEC in stage IIIB-IVM1a melanoma, with both models resulting in overall performances that were comparable and good. The second model reinforces the conclusion that for the best response to T-VEC, it should be used early on in the course of the disease, when the patient’s tumor burden is cutaneous with smaller diameter and fewer of metastases.[Table: see text]