Outcomes of Patients with Advanced Urothelial Carcinoma after Anti–programmed Death-(ligand) 1 Therapy by Fibroblast Growth Factor Receptor Gene Alteration Status: An Observational Study

Take Home Message This retrospective study found some evidence of potential lower response rates and shortened overall survival following anti–programmed death-(ligand) 1 therapy in patients with metastatic urothelial carcinoma with prespecified fibroblast growth factor receptor alterations (FGFRa) than in those without FGFRa who received prior immunotherapy between May 2018 and July 2019.

p = 0.054). Limitations include this study's retrospective nature and a potential selection bias from small sample size. Conclusions: Some evidence of lower response rates and shortened OS following anti-PD-(L)1 therapy was observed in FGFRa+ patients. The phase 3 THOR study (NCT03390504) will prospectively compare FGFRa+ patients with advanced mUC treated with erdafitinib versus pembrolizumab. Patient summary: Patients with metastatic urothelial carcinoma and prespecified fibroblast growth factor receptor alterations (FGFRa) potentially have worse clinical outcomes when treated with anti-PD-(L)1 therapy than those without FGFRa.

Introduction
In recent years, insights into the potential role of immunotherapies for bladder cancer have led to the approval of checkpoint inhibitors, such as atezolizumab (first-line treatment of platinum-ineligible patients regardless of programmed death ligand-1 [PD-L1] status and those with PD-L1+ [!5%] tumors), avelumab (first-line maintenance irrespective of cisplatin eligibility), nivolumab (adjuvant treatment for those at a high risk of recurrence after radical resection and second-line treatment after platinum-based chemotherapy), and pembrolizumab (first-line treatment of platinum-ineligible patients or second-line treatment after platinum-based chemotherapy) for patients with locally advanced or metastatic urothelial carcinoma [1][2][3][4]. While these immunotherapies have improved survival in patients with locally advanced or metastatic urothelial carcinoma [5][6][7], clinical benefit may vary depending on the molecular subtype and underlying immune landscape [8]. More specifically, response to checkpoint inhibitors may be dependent on T-cell infiltration of the tumor and T-cell function in the tumor microenvironment [8,9], as improved outcomes have been observed in patients with programmed death-(ligand) 1 (PD-[L]1)positive tumors [10]; however, as demonstrated in anti-PD-(L)1 clinical trials [6,[11][12][13], many patients with advanced urothelial carcinoma do not have PD-(L)1positive tumors.
Fibroblast growth factor receptor (FGFR) alterations (FGFRa; mutations or fusions) are detected in approximately 15-20% of patients with locally advanced or metastatic urothelial carcinoma [14,15]. Previous studies have shown that FGFR3 mutations are encountered more frequently in luminal tumors, which are known to be comparatively less responsive to checkpoint inhibition, and that FGFR3mutated bladder tumors are associated with decreased Tcell infiltration and low PD-L1 expression [15][16][17].
Several recent studies have reported the clinical outcomes of patients with FGFRa (FGFRa+) following anti-PD-(L)1 therapy, with differing outcomes [18][19][20][21]. Only one of 22 patients enrolled in BLC2001 who had received prior immunotherapy was reported as having responded to immunotherapy, highlighting the need for additional treatment options [21]. First-line anti-PD-(L)1 treatment in patients with FGFRa+ may be associated with poorer overall survival (OS); however, poorer OS was not observed in patients with FGFRa+ treated with any-line or second-line anti-PD-(L)1 therapy [18]. Similarly, the JAVE-LIN Bladder 100 study reported poorer survival outcomes in patients with high versus low FGFR3 gene expression who received first-line anti-PD-(L)1 therapy [20]. It was also shown that patients with FGFRa+ who received anti-PD-(L)1 alone as first-line therapy had an adjusted risk of progression two times higher than that of patients with wild-type FGFR [22]. However, data from cohorts 1 and 2 of the IMVigor 210 study demonstrated no statistically significant difference in response rates in patients with mutant versus wild-type FGFR3 with urothelial carcinoma treated with atezolizumab [19]. While patients from the PURE-01 study with high FGFR3 gene expression showed a lower complete response rate versus those with low FGFR3 gene expression following neoadjuvant pembrolizumab, the correlation between FGFR3 activity or mutation/fusion and complete response was not established [23]. Real-world data from patients with advanced urothelial carcinoma treated with anti-PD-(L)1 therapy also demonstrated that FGFR3-altered and wild-type tumors have equivalent Tcell receptor diversity, with comparable objective response rates (ORRs), progression-free survival, and OS [24].
Recent data from cisplatin-ineligible patients with locally advanced or metastatic urothelial carcinoma showed that the majority of platinum-naïve patients who progressed to anti-PD-(L)1 therapy responded to enfortumab vedotin [25,26]. Preliminary data from the NORSE study (NCT03473743) demonstrated improved efficacy with erdafitinib (a pan-FGFR inhibitor approved for the treatment of adult patients with locally advanced and metastatic urothelial carcinoma, and susceptible FGFR3 or FGFR2 genetic alterations, who have progressed during or following one or more prior lines of platinum-based chemotherapy) and the anti-PD-1 monoclonal antibody cetrelimab compared with erdafitinib alone (68% ORR [13/19] vs 33% ORR [6/18]) in patients with newly diagnosed locally advanced or metastatic urothelial carcinoma and FGFRa who were ineligible for cisplatin-based therapy, suggesting the potential value of combining therapies to overcome treatment resistance [27]. Therefore, treatment sequencing strategies should be considered carefully in light of emerging evidence on biomarker-directed therapies, including pan-FGFR inhibitors.
To build on this existing evidence, we conducted a retrospective analysis of the effects of any FGFRa in patients with locally advanced or metastatic urothelial carcinoma who received anti-PD-(L)1 therapy.

Study design
This was a noninterventional, retrospective, multicenter study conducted at five sites in the USA and three sites in Europe (Fig. 1). Clinical practice data were collected from patients at selected BLC2001 study sites (NCT02365597) between May 2018 and July 2019 [21]. These patients were not enrolled in the BLC2001 study because of screening failure (either they did not meet the molecular eligibility criteria or they elected not to enroll in the trial), and were required to have previously been treated or treated subsequently with an anti-PD-(L)1 agent.
Investigator determined ORR, investigator determined disease control rate (DCR), and OS per multivariate and unadjusted analyses were assessed for this study.

Study population
Eligible patients were diagnosed with urothelial carcinoma, received an anti-PD-(L)1 agent, and were either positive or negative for FGFR molecular alterations (any FGFR mutation or gene fusion, and copy number alterations/gene amplifications were not eligible; Supplementary  When required by the study site, patients or their legally acceptable representatives provided written consent before participation. The study protocol and its amendments were approved by review boards at all participating institutions.   All patients in the FGFRa+ cohort and 54 patients in the FGFRa-cohort received anti-PD-(L)1 therapy (Fig. 1); two patients were excluded for not meeting the study eligibility ECOG PS = Eastern Cooperative Oncology Group performance status; FGFR = fibroblast growth factor receptor; FGFRa+/-= fibroblast growth factor receptor alteration positive/negative; PD-(L)1 = programmed death (ligand) 1. a Nine FGFRa+ patients received treatment with FGFR inhibitors, but none of these patients received this treatment before receiving anti-PD-(L)1 therapy after the advanced diagnosis date. b Before advanced diagnosis date, defined as the date of first diagnosis of urothelial carcinoma (when available) or the date of first diagnosis of metastatic disease. c Includes patients who received multiple lines of immunotherapy.  Table 1). The proportion of patients receiving an immunotherapy/immunotherapy combination was higher in the FGFRa+ group than in the FGFRaÀ group (18% vs 6%).

Outcomes by FGFR status
The median follow-up duration was 31.1 (range, 5.7-299.9) mo. There was some evidence of lower ORRs and DCRs to anti-PD-(L)1 therapy in FGFRa+ versus FGFRa-patients regardless of the number of prior lines of therapy; however, the difference in rates between groups did not reach conventional levels of statistical significance ( Table 2). Among the 92 patients who received any line of anti-PD-(L)1 therapy, ORRs in those with FGFRa+ and FGFRa-were 16% and 26%, respectively (relative risk: 1.14 [95% CI, 0.92-1.40]; p = 0.3).
Although some evidence of shorter median OS was also observed in the univariate analysis for patients with FGFRa + versus those with FGFRa-, irrespective of the sequence number in which prior immunotherapy was used, the difference in OS between groups did not reach conventional levels of statistical significance (Table 2  Among the 24 patients who received first-line immunotherapy, the median OS was 18.3 mo in those who were FGFRa+ (n = 14) and 25.3 mo in those who were FGFRa-(n = 10; HR: 1.12 [95% CI, 0.33-3.84]; p = 0.9). Among the 36 patients who received second-line immunotherapy treatment, the median OS was 7.69 mo in those who were FGFRa+ (n = 11) and 11.0 mo in those who were FGFRa-(n = 25; HR: 1.47 [95% CI, 0.60-3.60]; p = 0.4).
OS was shorter in FGFRa+ patients than in FGFRapatients who received prior platinum chemotherapy and subsequent anti-PD-(L)1 therapy; however, the difference was not statistically significant ( Table 2 and Fig. 2). A multivariate analysis provided some evidence for shorter OS in FGFRa+ than in FGFRa-patients, with an HR of 1.81 (95% CI, 0.99-3.31) in those who had any line of anti-PD-(L)1 therapy (p = 0.054), 5.92 (95% CI, 0.40-87.54) in those who received first-line anti-PD-(L)1 treatment (p = 0.2), and 2.46 (95% CI, 0.47-12.80) in those who had second-line anti-PD-(L)1 therapy (p = 0.3); however, the difference in OS between groups did not reach conventional levels of statistical significance (Fig. 3).

Discussion
In this retrospective analysis of patients with locally advanced/metastatic urothelial carcinoma, some evidence of poorer outcomes was observed in those with FGFR+ alter-   [29], and 11.1 mo for atezolizumab) [6]. It is worth noting that this study was carried out prior to the approval of FGFR inhibitors for any indication. Importantly, recent promising data on the use of enfortumab vedotin in cisplatin-ineligible patients with locally advanced or metastatic urothelial carcinoma who progressed after anti-PD-(L)1 therapy [25,26] suggest that appropriate treatment sequencing strategies should be considered as clinical evidence with biomarker-directed thera-pies, including FGFR inhibitors, continues to emerge. Other clinical studies evaluating FGFR inhibition in patients with advanced urothelial carcinoma whose tumors expressed FGFRa also found a poor response to prior immunotherapy. While it may not be surprising to see a lower response rate to anti-PD-(L)1 in a relapsed/refractory population, it is interesting that 59% of patients in the BLC2001 primary analysis responded to erdafitinib following anti-PD-(L)1 therapy [21]. Likewise, in a phase 1 study of rogaratinib in patients with advanced cancers selected according to FGFR mRNA expression, approximately 30% of patients with urothelial carcinoma who received prior immunotherapy responded to rogaratinib [30]. However, these results are not conclusive since it was also demonstrated that FGFR3 alterations do not preclude a response to nivolumab in metastatic urothelial cancer [31], suggesting that further studies are needed in this setting to clarify the potential effects of FGFRa on clinical outcomes.
The current study was limited by its retrospective nature, the relatively small number of patients, potential selec- Patients who terminated study participation or were lost to follow-up were censored at the date they were last known to be alive. CI = confidence interval; FGFRa+/-= fibroblast growth factor receptor alteration positive/negative; HR = hazard ratio; PD-(L)1 = programmed death (ligand) 1.
tion bias, and nonstatistically significant results. Patients were selected for their suitability to receive an FGFR inhibitor; FGFRa-patients who were included in this analysis failed screening for the BLC2001 study because they did not meet the molecular eligibility criteria. Likewise, FGFRa + patients who were not enrolled in the BLC2001 study because of screening failure (or elected not to enroll in the trial) may not be representative of FGFRa+ patients. Therefore, patients included in this analysis do not represent a randomly selected population, which is a limitation of this study. However, the baseline data from the two cohorts (FGFRa+ vs FGFRa-patients) were generally similar and prognostically comparable (based on Bellmunt scores), supporting the assessment of anti-PD-(L)1 therapy outcomes between these groups. Another potential source of selection bias is that, owing to small numbers of patients in each cohort, patients who were permitted to receive an anti-PD-(L)1 agent alone or in combination with chemotherapy or other treatments, any number of prior lines of therapy, and treatment with an anti-PD-(L)1 agent in either a clinical study or treatment setting were pooled together. Furthermore, patients with copy number alterations and gene amplifications were not considered, as this study was designed to investigate mutations and fusions that were more reflective of the population that are clinically targeted by FGFR inhibitors. In addition, it was not possible to ascertain the dynamics of FGFRa positivity throughout patients' treatment course, highlighting the potential value for using circulating tumor DNA testing to monitor genomic alterations over time, as an alternative to tumor tissue testing [32]. Of the 38 FGFRa+ patients, nine received FGFR inhibition prior to receiving immunotherapy; this additional targeted treatment for FGFRa+ patients represents a source of a potential bias as one could expect different outcomes from these patients. However, the evidence toward worse outcomes in FGFRa+ patients despite this additional treatment shows a clinical need in this patient population.
The findings of this study contribute to the emerging data on the predictive value of FGFRa on outcomes of patients with advanced or metastatic urothelial carcinoma following anti-PD-(L)1 therapy and the unmet medical need in this targetable patient population. Further studies are needed to confirm these results in a larger patient cohort and to clarify whether other underlying concomitant genomic alterations dictate the treatment response.

Conclusions
In this retrospective study, there was some evidence of lower ORRs and DCRs in patients with FGFRa+ versus those with FGFRa-and advanced or metastatic urothelial carcinoma who had received anti-PD-(L)1 therapy. A multivariate analysis showed some evidence toward shorter median OS in patients with FGFRa+ versus those with FGFRa-in this cohort of patients treated with immunotherapy. These data provide some evidence toward the hypothesis that patients with FGFR gene alterations have poor outcomes with anti-PD-(L)1 agents and contribute to the emerging data on outcomes of FGFRa+ patients with available therapies.
This work was previously (virtual) presented at the European Society for Medical Oncology Congress, September 19-21, 2020 (abstract 757P).
Author contributions: Arash Rezazadeh Kalebasty had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.