JAVELIN Solid Tumor (NCT01772004) was a phase I, open-label, multicohort trial assessing avelumab monotherapy for various solid tumors. The study design and full methodology in UC cohorts have been reported previously.10 Briefly, patients with locally advanced or metastatic urothelial carcinoma, disease progression following platinum-based chemotherapy or platinum ineligible, and no prior immunotherapy, received avelumab 10 mg/kg by 1 hour intravenous infusion every 2 weeks until confirmed disease progression, unacceptable toxicity or other protocol-specified criteria for withdrawal occurred. Premedication with an antihistamine (diphenhydramine or equivalent) and acetaminophen was administered 30–60 min prior to each avelumab infusion. Dose modifications of avelumab (reductions or increases) were not permitted. Tumor response and progression were assessed per Response Evaluation Criteria in Solid Tumors V.1.1 by a blinded independent review committee. In exploratory post hoc analyses, clinical activity was assessed in subgroups defined by standard baseline characteristics or characteristics associated with high-risk disease or poor response to chemotherapy. PD-L1 expression in tumor samples was assessed using a research-only assay (Dako PD-L1 IHC 73-10 pharmDx), and PD-L1+ status was defined prospectively as expression on ≥5% of tumor cells, consistent with previous analyses.9 10 Adverse events (AEs) were graded according to National Cancer Institute Common Terminology Criteria for Adverse Events V.4.0. Immune-related AEs (irAEs) were identified using a prespecified list of Medical Dictionary for Regulatory Activities preferred terms, followed by comprehensive medical review using predefined criteria. Infusion-related reactions (IRRs) and related events were identified via an expanded definition that included prespecified Medical Dictionary for Regulatory Activities terms (IRR, drug hypersensitivity, anaphylactic reaction, hypersensitivity and type 1 hypersensitivity) occurring on the day of or the day after the infusion, in addition to signs and symptoms of IRR that occurred on the day of infusion and resolved within 2 days.

The primary end point for this study was the occurrence of dose-limiting toxicities during the first 3 weeks of avelumab treatment in the dose-escalation part of the study (reported previously),26 and best overall response per RECIST V.1.1, adjudicated by independent end point review committee, in efficacy expansion cohorts. Secondary end points included duration of response (DOR), progression-free survival (PFS), OS, tumor PD-L1 expression and safety.

Time-to-event end points (DOR, PFS and OS) were analyzed using the Kaplan-Meier method. For post hoc subgroup analyses, ORR (proportion of patients with a complete response (CR) or partial response (PR)) and disease control rate (DCR; proportion of patients with CR, PR, stable disease (SD) or non-CR/non-progressive disease (PD)) were analyzed using event proportions, and DCR and PFS were assessed as sensitivity analyses. Additionally, a sensitivity analysis was performed for post hoc subgroups based on continuous covariates, including creatinine clearance and albumin concentration; an automated cut-off search was applied using recursive partitioning to test the robustness of cut-offs selected.27 This data-mining approach conditionally partitions the covariate space that best predicts the end point of interest. DCR was selected to avoid redundancy and use classification tree models. The partition algorithm selected the input variable that had the strongest association with disease control and implemented a binary split in the selected input variable; this process was repeated recursively. Post hoc analysis of the association between time to response (TTR) and DOR or OS was performed using landmark Kaplan-Meier analysis. Associations were assessed visually using a scatter plot matrix for TTR, DOR, PFS and OS (data not shown). Subgroups were defined based on whether response occurred by week 7 (first tumor assessment ±5 days; DOR analysis) or by 3 months (12 weeks; second tumor assessment ±5 days; OS analysis), and analyses included only those patients who remained on study at the defined time point. Univariate and multivariate Cox regression model analyses were also performed (data not shown). Following methodology based on previously published studies, further OS analyses were carried out that defined separate categories for patients with early response versus SD/late response versus other patients, and for patients who continued treatment beyond PD versus patients who discontinued treatment when PD occurred.28 OS analyses were confirmed by supplementary analyses, including one using an alternative landmark definition (week 7, as above) and another that excluded patients who were not evaluable by blinded IRC.

Between September 3, 2014, and March 15, 2016, 329 patients with advanced/metastatic UC were screened for enrollment into two study cohorts; of those, 249 met eligibility criteria and received avelumab (table 1; online additional file 1). Detailed patient characteristics for the pooled population have been reported previously.10 Briefly, median age was 68 years (range 30–89 years), most patients were male (178 (71.5%)), and 124 (49.8%) had received ≥2 prior lines of therapy for advanced disease. At data cut-off on April 10, 2018, median follow-up was 31.9 months (range 24–43 months). Patients received a median of six doses of avelumab (range 1–93 doses) for a median of 2.8 months (range 0.5–42.8 months). Twelve patients (4.8%) remained on treatment at last follow-up. The most common reason for treatment discontinuation was disease progression (155 (62.2%)); other reasons were AEs (39 (15.7%)), death (13 (5.2%)), withdrawal of consent (13 (5.2%)), protocol non-compliance (2 (0.8%)) and other reasons (15 (6.0%)). Consistent with the approved label of avelumab and a previous report,10 seven platinum-naïve patients were excluded from efficacy analyses but were included in safety analyses.

Of 242 postplatinum patients, 40 had a confirmed response (CR in 10 (4.1%); PR in 30 (12.4%)), resulting in an ORR of 16.5% (95% CI 12.1% to 21.8%; online additional file 2). Data for best overall response were missing or not assessable in 43 patients (17.8%; no postbaseline tumor assessment in 35, postbaseline assessment but response not evaluable in 2, stable disease of insufficient duration in 5 and non-evaluable with PD >12 weeks after study assessment in 1). Median TTR was 2.7 months (range 1.3–11.0 months). At last follow-up, response was ongoing or considered SD in 20 of 40 responding patients (50.0%; online additional file 3A). Median DOR was 20.5 months (95% CI 9.7 months to not estimable), and the estimated proportion of responses lasting ≥12 months was 65.4% (95% CI 47.0% to 78.8%). Of 198 evaluable patients with a baseline and on-study tumor assessment, 54 (27.3%) had tumor shrinkage of ≥30% (online additional file 3B). In patients with ≤1 (n=118), 2 (n=73) or ≥3 (n=51) previous lines of therapy for advanced disease, ORRs (95% CIs) were 19.5% (12.8% to 27.8%), 16.4% (8.8%% to 27.0%) and 9.8% (3.3% to 21.4%), respectively. Median PFS was 1.6 months (95% CI 1.4 to 2.7 months), and PFS rates at 6 and 12 months were 26.5% (95% CI 20.9% to 32.5%) and 16.8% (95% CI 11.9% to 22.4%), respectively (figure 1A). Median OS was 7.0 months (95% CI 5.9 to 8.5 months), and OS rates at 12 and 24 months were 35.9% (95% CI 29.9% to 42.0%) and 20.1% (95% CI 15.2% to 25.4%), respectively (figure 1B).

For PD-L1 expression, 214 patients (88.4%) were evaluable, of whom 84 (39.3%) had PD-L1+ tumors and 130 (60.7%) had PD-L1− tumors. Efficacy findings in PD-L1 subgroups are shown in online additional files 3B and 4.

Responses to avelumab occurred in all subgroups defined by individual risk factors (figure 2). Factors associated with a higher ORR compared with respective comparator subgroups were absence of baseline visceral or liver metastases, older age (≥75 years) and high albumin (≥35 g/L). ORRs were similar within subgroups defined by renal function (creatinine clearance <60 or ≥60 mL/min) and tumor location (upper vs lower tract). In patients with a Bellmunt prognostic (or risk) score of 0, 1, 2 or 3,1 ORRs (95% CI) were 22.2% (12.0% to 35.6%), 21.4% (14.2%% to 30.2%), 6.9% (1.9% to 16.7%) and 0% (0% to 18.5%), respectively. DCR and PFS were analyzed in selected subgroups (online additional files 5 and 6). Similar to ORR analyses, no difference in DCR or PFS was found between subgroups defined by renal function or tumor location, and patients without baseline liver metastases had a higher DCR and longer PFS than those with liver metastases. In contrast to ORR analyses, subgroups defined by age or albumin level showed no differences in DCR or PFS. Data-driven cut-off searches yielded no new relevant subgroups, and cut-offs found using recursive partitioning were similar to predefined cut-offs (data not shown).

In a post hoc exploratory analysis comparing DOR in patients who had achieved an objective response (CR or PR) either by week 7 (first tumor assessment; n=15) or after that time point (n=25), the median DOR was 25.9 months (95% CI 9.0 months to not estimable) vs 20.5 months (95% CI 8.1 months to not estimable), respectively. CIs were large due to the small sample size. Post hoc landmark analyses were also performed to assess the association between response by 3 months of treatment (second tumor assessment) and OS. In patients who had an objective response by this time point, the median duration of OS beyond 3 months was not reached (95% CI 18.9 months to not estimable) vs 7.1 months (95% CI 5.2 to 9.0 months) in patients who did not respond by 3 months (figure 3). To investigate this observation in more detail, patients without response by 3 months were further categorized according to whether they had SD by 3 months and/or achieved a later response, or whether their best response was PD or not evaluable; median duration of OS beyond 3 months in these subgroups was 9.5 months (95% CI 6.5 to 14.4 months) and 4.3 months (95% CI 2.8 to 7.1 months), respectively (figure 3). Further analyses found that OS was similar between patients with PD who continued treatment beyond investigator-confirmed PD and those who did not (figure 3), although this analysis might be biased because the decision to continue treatment would have also been based on patient characteristics at the time of PD. OS findings were confirmed by a sensitivity analysis that used an earlier landmark time point (7 weeks).

Safety findings were consistent with the 6-month analysis of this population (N=249).10 Treatment-related AEs (TRAEs) of any grade occurred in 177 patients (71.1%), which included 11 additional patients compared with the 6-month analysis (online additional file 7). Grade ≥3 TRAEs occurred in 29 patients (11.6%; 8 additional patients compared with the 6-month analysis). The most common grade ≥3 TRAE was fatigue (four (1.6%)); all other grade ≥3 TRAEs occurred in one to two patients only. irAEs occurred in 51 patients (20.5%; 17 additional patients compared with the 6-month analysis), including 12 (4.8%) who had grade ≥3 events (online additional file 8). The most commonly occurring categories of irAEs (any grade) were immune-related rash (28 (11.2%), including various rash/pruritus AEs) and immune-related thyroid disorders (13 (5.2%), including hypothyroidism, hyperthyroidism and blood thyroid-stimulating hormone increased). IRRs and related events (based on an expanded definition) occurred in 78 patients (31.3%), including grade ≥3 events in three (1.2%). Treatment-related death occurred in one patient (0.4%), due to pneumonitis in a patient with ongoing, treatment-unrelated Clostridium difficile colitis and diverticulitis, as reported previously.

In an exploratory post hoc analysis, frequencies of the TRAEs examined were generally similar in high-risk patient subgroups compared with the overall population (table 2). In particular, patients with renal insufficiency or upper tract disease did not experience significant increases in serum creatinine, patients with liver metastases did not have an increased risk of hepatic events, and incidences of pneumonitis, gastrointestinal events (diarrhea and colitis), endocrine events (hypothyroidism and adrenal insufficiency) and IRRs were not elevated in any subgroup examined.