An Evaluation of Unreported Subgroup in Head and Neck Squamous Cell Carcinoma (HNSCCS) Study Based on the Bipartite Graph

Speaker(s)

ABSTRACT WITHDRAWN

OBJECTIVES: The KEYNOTE-048 (NCT02358031) study designed two Combined Positive Score(CPS) subgroups (CPS ≥ 1 and CPS ≥ 20) to compare Pembrolizumab with or without chemotherapy and Cetuximab-chemotherapy. 1 ≤ CPS < 20 subgroups (or intermediate CPS subgroup) accounting for 42.24% of the total HNSSC patients, however, was unreported. This study aims to determine its clinical value.

METHODS: In a bipartite graph, matching is a set of edges that connect elements on the both side. A matching contains a greater number of edges than any other matchings called the maximal matching of this bipartite graph, which also represents the most matched part of two sides. Calculating the maximal matching of CPS ≥20 and CPS ≥1 subgroup, the unmatched part in the CPS ≥1 set represented the unreported subgroup. The Hungarian algorithm was used to find the maximal matching, which included a new edge in each iteration by reversing the augmenting path. A KMsubstraction package was applied to achieve the matching and retrieving. The Bland-Altman plots and Empirical Cumulative distribution were utilized to evaluate the quality of matching. The whole calculation was finished in R.

RESULTS: In terms of Overall Survival of the intermediate CPS group, Pembrolizumab with or without chemotherapy is more preferred with a Hazard Ratio (HR) of 0.70 (0.53-0.92, 95%CI) and 0.84 (0.65-1.01, 95%CI). Bland-Altman plots showed an absolute difference of 0.27 and the Kolmogorov-Smirnov test was used to compare the matched part of CPS ≥1 and CPS ≥20, the P-value was 0.997 and indicated the matching was successful.

CONCLUSIONS: Using the Hungarian Algorithm in the bipartite graph, the Overall Survival IPD of the intermediate CPS group was reconstructed. Align with the findings of other subgroups, Pembrolizumab is more favorable than Cetuximab. The following analysis of the intermediate CPS group and its cost-effectiveness analysis will be disclosed in the future.

Code

SA125

Topic

Clinical Outcomes, Methodological & Statistical Research

Topic Subcategory

Clinical Outcomes Assessment, Confounding, Selection Bias Correction, Causal Inference, Missing Data

Disease

Oncology, Sensory System Disorders (Ear, Eye, Dental, Skin)