Tumor-associated immune cells and progression-free survival in advanced endometrial cancer (EC), results from the PHAEDRA trial (ANZGOG 1601).

Abstract

5584 Background: Activity of durvalumab in patients with deficient mismatch repair (dMMR) advanced endometrial carcinoma (EC) was confirmed in the PHAEDRA trial (ANZGOG 1601). This study investigated the association between immune biomarkers and clinical outcomes in PHAEDRA. Methods: Formalin-fixed paraffin embedded sections immunohistochemically stained for PD-L1 using the Ventana platform, were with matched H&E slides scored independently by two pathologists according to the Ventana PD-L1 (SP263) algorithm for urothelial carcinoma (UC). Immune biomarkers assessed were PD-L1 staining of tumor cells (TCP) and immune cells (IC), and presence of tumor-associated immune cells (ICP). Results: Sixty-seven of the 71 patients had sufficient tumor for PD-L1 testing. AUC were 0.667, 0.726 and 0.644 for TCP, ICP and IC, respectively for predicting tumor response. Optimal cutpoints were TCP≥1%, ICP≥10% and IC≥35%. ICP≥10% achieved the highest sensitivity (53%) and specificity (82%) of the individual cutpoints. The optimal cutpoint algorithm was able to identify patients who would not respond, (sensitivity 88%, negative predictive value 92%), but had low specificity (48%) and positive predictive value (37%). Differences in PFS were found using ICP≥10% (logrank p = 0.01), compared to TCP (p = 0.25), IC (p = 0.48) and the UC algorithm (p = 0.08) (Figure 1). PFS was shorter in patients with pMMR than dMMR after adjusting for ICP (HR 2.99, 95%CI: 1.61-5.57, p < 0.001). Adjustment for MMR reduced the prognostic significance of ICP≥10% for PFS (HR 0.59, 95% CI: 0.28-1.23, p = 0.16). For OS, differences were seen for the UC algorithm (p = 0.02), but not ICP (p = 0.07), TCP (p = 0.18) or IC (p = 0.23). Similarly to PFS, adjustment for MMR reduced the prognostic significance of the UC algorithm for OS (HR: 0.53, 95% CI: 0.25-1.12, p = 0.10). Conclusions: In this exploratory analysis, ICP was more closely associated with tumor response and PFS than TCP or IC. ICP alone was better than the UC algorithm for predicting PFS. The optimum cutpoint algorithm was promising for identifying non-responders, but requires external validation. Clinical trial information: ACTRN12617000106336.