Abstract
BACKGROUND: Pembrolizumab is a promising treatment option for platinum-failed thymic carcinoma; however, the lack of established predictive biomarkers remains a challenge. Therefore, this study aimed to assess the predictive value of artificial intelligence (AI)-powered tumor-infiltrating lymphocyte (TIL) analysis of pembrolizumab for thymic carcinoma. METHODS: Patients with platinum-failed, advanced thymic carcinoma treated with pembrolizumab between January 2016 and December 2021 were included. Hematoxylin and eosin-stained sections from the samples closest to the time before pembrolizumab treatment were analyzed using an AI-powered TIL analyzer. Intratumoral TIL (iTILs) and stromal TIL (sTILs) were quantified, and their immune phenotypes (IP) were identified. RESULTS: In total, 10 patients were included in this study. The best response was complete response in 1 patient (10%) and partial response in 1 patient (10%). The median progression-free survival (PFS) was 5.0 months. Patients with higher iTIL (>27.23/mm(2)) exhibited longer PFS (median, 9.5 vs. 1.5 months, P=0.03) and overall survival (OS) (median, not determined vs. 4 months, P=0.03). Patients with higher sTIL (>252.54/mm(2)) exhibited longer PFS (median, 10 vs. 1 month, P=0.006) and OS (median, not determined vs. 9 months, P=0.01). Patients with inflamed IP exhibited longer PFS than those with non-inflamed IP (median, 10 vs. 3 months, P=0.046). CONCLUSIONS: Increased infiltration of both iTIL and sTIL is associated with longer PFS and OS. Additionally, an inflamed IP is associated with longer PFS. Thus, TIL density and IP may be promising predictive biomarkers for pembrolizumab in patients with platinum-failed thymic carcinoma.