Clinical, radiographic, and genomic predictors of response to neoadjuvant chemotherapy for high-risk localized upper tract urothelial carcinoma.

Abstract

861 Background: Recent data support a pathologic response in 60-65%, including 10-20% with a pathologic complete response (CR) for patients treated with neoadjuvant chemotherapy (NAC, 4-cycles of cisplatin-based treatment) at definitive extirpative surgery; nearly 35% of patients do not respond. We evaluated clinical, radiographic, and genomic predictors of pathologic response to NAC. Methods: Patients undergoing definitive surgery for high-risk UTUC within 16-weeks of completing NAC between 2000-2024 were identified from an institutional database. Response to NAC was defined as < ypT2 N0 and non-response as ≥ypT2 N0/+. The association between clinical predictors and response was evaluated using Welch Two Sample t-tests, two sample tests of equality of proportions with continuity correction, and Fisher’s Exact tests (N = 144). The correspondence between post-NAC radiographic response and pathological response was tested using an Exact test (N = 127). Tumor genomic profiles were sequenced by MSK-IMPACT and alterations were filtered for oncogenic variants using OncoKB (N = 65). Results: Overall, 90% of patients were clinically stage as ≥cT2 at NAC initiation, with final surgical pathology identifying 65% (N = 94) responders and 35% non-responders. Nearly all patients (97%) received gemcitabine-cisplatin NAC, and 89% completed adequate chemotherapy (≥3-cycles). Demographics were similar between responders and non-responders. Of disease characteristics evaluated, only variant histology at pre-NAC biopsy was associated with non-response to NAC (p = 0.049). Non-responders more frequently harbored variant histologies at final pathology than responders (34% vs 2%, p = 0.003). Radiographic response was associated with pathologic response (p = 0.004), however, 9.3% patients with a radiographic CR were pathological non-responders. The mutational landscape of this cohort was enriched for TERT (60%), FGFR3 (43%), and histone methylation/chromatin remodeling gene (25-30%) mutations. Somatic mutation frequencies differed between responders and non-responders for TSC1 (0% vs 23%), EP300 (0% vs 17%), and HRAS (17% vs 0%), respectively. Conclusions: Limited clinical factors predict response to NAC; variant histology may predict for non-response to NAC. While radiographic response predicts pathologic response to NAC, it is not a sufficient surrogate for pathological CR. Genomic alterations in TSC1, EP300, and HRAS were associated with response to NAC and warrant further study.