Assessment of the Prognostic and Predictive Values of the Deficient Mismatch Repair Phenotype in Patients Treated with Radical Nephroureterectomy for Upper Tract Urothelial Carcinoma.

Abstract

Background and objective: Given the conflicting evidence currently available in the literature, our aim was to assess the prognostic and predictive values of the deficient mismatch repair (dMMR) phenotype in a large cohort of upper tract urothelial carcinoma (UTUC) patients.

Methods: Based on our national network, we performed a retrospective multicenter study including 281 UTUC patients treated with radical nephroureterectomy between 2000 and 2015 at ten French hospitals. The dMMR phenotype as well as PD-L1 and PD-1 expression were determined using immunohistochemistry analyses based on 2-mm-core tissue microarrays. Multivariable Cox regression models were fitted to assess the impact of the dMMR phenotype on recurrence-free (RFS), cancer-specific (CSS), and overall (OS) survival using interaction terms to test the heterogeneity of the treatment effect of adjuvant chemotherapy (AC). Multivariable logistic regression models were also fitted to assess the impact of the dMMR phenotype on PD-L1 and PD-1 expression.

Key findings and limitations: Overall, 76 (27.0%) patients had a dMMR phenotype, which was an independent predictor of prolonged RFS (hazard ratio [HR] = 0.41; 95% confidence interval [CI] = [0.21-0.83]; p = 0.01), CSS (HR = 0.38; 95% CI = [0.18-0.83]; p = 0.02), and OS (HR = 0.44; 95% CI = [0.22-0.89]; p = 0.02), with a significant interaction with the use of AC in multivariable Cox regression models (all p_interaction < 0.05). Subgroup analyses showed that the use of AC was significantly associated with prolonged RFS (HR = 0.14; 95% CI = [0.06-0.30]; p < 0.001), CSS (HR = 0.10; 95% CI = [0.03-0.29]; p < 0.001), and OS (HR = 0.23; 95% CI = [0.10-0.54]; p = 0.001) in non-dMMR patients only, without any significant benefit in dMMR patients (all p > 0.05). In multivariable logistic regression analyses, the dMMR phenotype was significantly associated with inverse PD-L1 (OR = 0.20; 95% CI = [0.10-0.80]; p = 0.001) and PD-1 (OR = 0.36; 95% CI = [0.16-0.79]; p = 0.01) expression.

Conclusions and clinical implications: We observed that the dMMR phenotype was associated with favorable pathological characteristics and prognosis in UTUC patients, despite conferring decreased sensitivity to AC and lower PD-L1 or PD-1 expression.