33P - Nomogram to predict survival of patients with unresectable melanoma receiving immune checkpoint inhibitors

Background

40-60% of patients with advanced melanoma receiving immune checkpoint inhibitors (ICI) do not derive benefit. Baseline predictive and prognostic biomarkers for ICI are currently missing. Here we aim to identify predictive clinicopathologic features using a random survival forest (RSF) and develop a nomogram for PFS in this collective.

Methods

We included patients with stage IV unresectable melanoma receiving first-line ICI and documented clinicopathologic features and blood markers at the time of therapy start. The variable importance was calculated to screen for predictors of progression-free survival (PFS) using an RSF. We developed a Cox regression prediction model based on the RFS results and established a prognostic nomogram for PFS. Performance of the nomogram was evaluated using calibration plots, Brier score, and discrimination by the Harrel C-Index and AUC. Performance was also verified by internal validation. Patients were divided into four subgroups based on quartiles of the total points (TP) of the nomogram. Kaplan Meier survival analysis was performed.

Results

We included 296 patients with a median PFS of 25 months. Median age was 67 (IQR:57, 77) years. The RSF model identified nine variables associated with PFS [LDH, type of melanoma, neutrophile/lymphocyte rate (NLR), age, S100, number of metastatic organs, presence of liver or brain metastases, and BMI] that were used to construct the prediction model. The C-index was 0.66, and calibration curves showed a good correlation between the predicted and actual progression risks. Patients with total points below 63.88 (below first quartile of TP, n=73) had a median PFS of 63 months (95% CI: 48-76), while the group with points above 121.45 (above third quartile of TP, n=75) had a median PFS of 11 months (95% CI 7-19); HR=3.95 (95% CI:2.63-5.94; p<0.001).

Conclusions

We identified predictors of PFS in advanced unresectable melanoma patients. We also established a nomogram integrating clinicopathologic features and blood markers, with good accuracy for predicting PFS that can be used for risk stratification at the time of therapy start.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

T. Sinnberg, H. Niessner: Financial Interests, Institutional, Funding: Novartis, Pierre Fabre. A. Forschner: Financial Interests, Personal, Advisory Role: Roche, Novartis, MSD, Pierre Fabre; Financial Interests, Personal, Invited Speaker: Roche, Novartis, BMS, MSD, CeGaT; Financial Interests, Institutional, Research Grant: BMS Stiftung Immunonkologie; Financial Interests, Personal, Other, travel support: Roche, Novartis, BMS, Pierre Fabre. U. Leiter-Stoppke: Financial Interests, Personal, Funding: MSD; Financial Interests, Personal, Advisory Role: Sun Pharma, Sun Pharma, Sanofi, Novartis, MSD, Roche, Almirall Hermal; Financial Interests, Personal, Invited Speaker: Sun Pharma, Sanofi, MSD, Novartis, Roche, Almirall Hermal; Financial Interests, Institutional, Invited Speaker: Sanofi, MSD. L. Flatz: Financial Interests, Personal, Research Grant: Hookipa Pharma, SAKK/Immunophotonics, Deutsche Forschungsgemeinschaft, Philogen, Mundipharma; Financial Interests, Personal, Advisory Role: Philogen, Sanofi, Novartis, BMS, Data Safety Board University; Financial Interests, Personal, Stocks/Shares: Hookipa Pharma. T.M.S. Amaral: Financial Interests, Personal, Invited Speaker: CeCaVa, BMS, Novartis, Pierre Fabre; Financial Interests, Institutional, Funding: Novartis, Neracare, Sanofi, Skyline-Dx; Financial Interests, Institutional, Research Grant: Novartis, iFIT; Non-Financial Interests, Personal, Member: Portuguese Society for Medical Oncology, Portuguese Society of Medical Oncology - Young Oncologists Group, ASCO; Other, Personal, Other, Clinical expert: Infarmed. All other authors have declared no conflicts of interest.