Efficacy and Safety of Durvalumab in Locally Advanced or Metastatic Urothelial Carcinoma

Study Design and Participants

For the overall study population, patients with solid tumors ages 18 years or older with histologically and/or cytologically confirmed disease were eligible for inclusion. Eligible patients could have disease that progressed on prior therapy or be treatment-naïve, had Eastern Cooperative Oncology Group performance status 0/1, and had adequate organ and bone marrow function. Patients were not eligible if they had received any immunotherapy or investigational anticancer therapy within the past 4 weeks (6 weeks for monoclonal antibodies) or if they were receiving any concurrent chemotherapy, immunotherapy, biologic, or hormonal therapy for cancer. Patients who met these criteria and had locally advanced/metastatic UC and whose disease had progressed while they were receiving prior therapy or were ineligible for or refused any number of prior therapies were eligible for inclusion in the UC cohort.

This study was conducted according to the Declaration of Helsinki and approved by the independent ethics committee or institutional review board at each participating center, with written informed consent obtained from all patients. Study participants were not compensated. See Supplement 1 for the trial protocol and the statistical analysis plan.

Procedures

Durvalumab was administered by intravenous infusion, 10 mg/kg every 2 weeks (Q2W), for up to 12 months, or until confirmed progressive disease, initiation of another anticancer therapy, unacceptable toxic effects, consent withdrawal, or other reasons for discontinuation. For patients with disease progression during follow-up who had not received another anticancer therapy and had not met criteria for discontinuing study treatment, a 12-month course of durvalumab retreatment was allowed. Safety was assessed from start of study with monitoring continued through 90 days after the last durvalumab dose or until initiation of another anticancer therapy. Toxic effects were graded according to National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.03. Tumor assessments occurred at 6, 12, and 16 weeks, and every 8 weeks thereafter during treatment. Following discontinuation, tumor assessments were performed every 2 months for 1 year and then every 3 months thereafter.

Outcomes

Primary safety end points included AEs, serious AEs, laboratory evaluations, vital signs, and physical examinations. Also assessed were AEs of special interest (AESIs) and immune-mediated adverse events (imAEs [AESIs requiring systemic steroids, endocrine therapy, or other immunosuppressants within 30 days of onset and prior to resolution that were consistent with an immune-mediated mechanism and had no clear alternate etiology]). The primary efficacy end point was objective response rate (ORR; proportion with confirmed complete response or partial response by blinded independent central review using Response Evaluation Criteria in Solid Tumors [RECIST], version 1.1). Key secondary end points included duration of response, time to response, change in target lesion size, disease control rate (confirmed complete response or partial response or stable disease for ≥6 weeks), PFS (time from first dose until documented disease progression per RECIST, version 1.1, or death), and OS (time from first dose until death).

PD-L1 expression was evaluated by immunohistochemical analysis in tumor tissue obtained prior to treatment, using the SP-263 anti-PD-L1 antibody assay (Ventana Medical Systems), as described previously. With respect to patient eligibility for the trial, the first 20 patients were enrolled regardless of PD-L1 expression; however, preliminary data suggested that PD-L1 may be expressed more commonly on immune cells than on TCs. Therefore, to ensure assessment of the contribution of PD-L1–expressing TCs to response with durvalumab, subsequently enrolled patients (n = 43) were required to have PD-L1 expression of at least 5% on TCs. However, an interim analysis showed that ORRs in patients with less than 5% PD-L1 expression (n = 86) were similar to ORRs in all patients. Therefore, a protocol amendment removed this requirement. For purposes of biomarker analyses, a 25% cutoff for defining TC- or immune cell–dependent expression status was chosen, as previously described, because this cutoff seemed to enrich for response, based on review of PD-L1 expression in the first 20 enrolled patients who were followed for a minimum of 12 weeks. This exploratory analysis also suggested the optimal scoring algorithm to be a unique combined assessment of PD-L1 staining of TCs and immune cells (PD-L1 “high,”≥25% of either TCs or immune cells staining for PD-L1, and PD-L1 “low or negative,”<25% of both TCs and immune cells staining for PD-L1).

Statistical Analysis

This planned study update is based on more than 3 times as many patients and increased follow-up compared with the initial report of this study, and its timing was predicated by regulatory interactions. In addition, we examined time to onset of treatment-related AEs and AESIs, which was defined as the time from the first dose of study treatment to the onset date of the AE. All patients who received their first dose of durvalumab at least 30 days before July 24, 2016 (data cutoff for the interim analysis), were included in the safety and efficacy analyses (“as-treated” population). A total of 191 patients with UC would provide a width between the observed ORR and its lower limit of the exact 2-sided 95% CI ranging from 6% to 7%, when the ORR was expected to be in the 20% to 30% range.

The ORR, disease control rate, and their respective exact 2-sided 95% CIs were estimated using the Clopper-Pearson method and by PD-L1 expression status (although the study was not designed to perform formal statistical comparisons of PD-L1 subgroups). Time-to-event end points, such as duration of response (in patients with objective response), PFS, OS, time to response, landmark PFS and OS rates, and cumulative incidences of AEs, were estimated by Kaplan-Meier method (relevant censoring times are defined in the eMethods in Supplement 2) with 2-sided 95% CIs provided by Brookmeyer and Crowley method. SAS statistical software (version 9.3 or higher) was used for all statistical analyses.

Patients in the UC Cohort

A total of 191 patients with locally advanced/metastatic UC from 60 sites across 9 countries had received treatment (as-treated population) (Figure 1). The median exposure duration in the as-treated UC population was 2.8 months (range, 0.4-12.5 months), and the median follow-up was 5.78 months (range, 0.4-25.9 months). As of the data cutoff of October 24, 2016 (90-day update), 44 patients (23.0%) were still receiving durvalumab.

Demographic and baseline characteristics of the patients with UC are summarized in Table 1 (eTable 1 in Supplement 2 summarizes baseline characteristics for all patients who received durvalumab, 10 mg/kg Q2W [n = 970]). The median age of patients with UC was 67.0 years and most patients were male (136 [71.2%]) and white (123 [71.1%]). All patients with UC had stage 4 disease, and 190 (99.5%) had received prior anticancer therapy; 64 (33.5%) had received at least 2 prior regimens, and 188 (98.4%) had received prior platinum-based treatment. Overall, 182 of 191 patients (95.3%) whose disease had progressed while receiving or after receiving a platinum-based therapy or within 12 months of receiving therapy in a neoadjuvant or adjuvant setting (hereinafter referred to as second-line or greater postplatinum subgroup). The other 9 patients (4.7%) were either treatment naïve or designated as first-line patients who had received platinum-based therapy in the neoadjuvant/adjuvant setting and whose disease progressed more than 12 months after the last dose of therapy. At study entry, 177 of 191 patients (92.7%) had visceral metastases; 82 of 191 (42.9%), liver metastases; and 14 of 191 (7.3%), lymph node–only disease; PD-L1 expression was high in 98 of 191 (51.3%), low or negative in 79 of 191 (41.4%), and unknown in 14 of 191 (7.3%).

Antitumor Activity in the UC Cohort

In the as-treated population (n = 191), the ORR was 17.8% (34 of 191; 95% CI, 12.7%-24.0%), including 7 complete responses (3.7%; see the eResults and eTable 4 in Supplement 2 for additional details of the complete responses), and was consistent with that seen in the second-line or greater postplatinum subgroup (32 of 182 [17.6%] [95% CI, 12.3%-23.9%]); ORRs were 27.6% (n = 27; 95% CI, 19.0%-37.5%) in PD-L1 high patients and 5.1% (n = 4; 95% CI, 1.4%-12.5%) in PD-L1 low or negative patients (Table 2). The complete response rates did not numerically differ by PD-L1 status, and responses in both subgroups were durable (Table 2). Objective responses were observed across all subgroups, including subsets with poor prognosis: ORRs in patients with visceral metastases and liver metastases were 15.3% (27 of 177; 95% CI, 10.3%-21.4%) and 7.3% (6 of 82; 95% CI, 2.7%-15.2%), respectively; the ORR in patients with baseline lymph node–only disease was 50.0% (7 of 14; 95% CI, 23.0%-77.0%).

The disease control rate by blinded independent central review assessment was 36.6% (95% CI, 29.8%-43.9%) in the as-treated population and 36.3% (95% CI, 29.3%-43.7%) in the second-line or greater postplatinum subgroup. The disease control rates were numerically greater in PD-L1 high vs low or negative subgroups (44.9% vs 21.5% in the as-treated population; 44.2% vs 20.5% in the second-line or greater postplatinum population) (Table 2).

Responses occurred early and were durable (Figure 2A). Median time to response was 1.41 months (range, 1.2-7.2), coinciding with the first protocol-specified imaging assessment. Time to response did not numerically differ by PD-L1 expression status and was similar in the second-line or greater postplatinum subgroup (data not shown). Median duration of response in the as-treated population had not been reached at data cutoff (range, ≥0.9 to ≥19.9 months). Seventeen of 34 responders (50.0%) had a response lasting at least 6 months (Figure 2A, Table 2), and 26 (76.5%) had an ongoing response at data cutoff (see the eResults in Supplement 2 for additional details regarding patients without an ongoing response at data cutoff).

Among the 159 patients with target lesions at baseline and at least 1 postbaseline scan, 51 (32.1%) experienced a target lesion reduction of at least 30% from baseline (Figure 2B). Tumor shrinkage and deep durable changes were seen in both PD-L1 subgroups (Figure 2B; eFigure 1 in Supplement 2).

Survival in the UC Cohort

Given limited follow-up, OS data were considered immature at data cutoff. The median PFS was 1.5 months (95% CI, 1.4-1.9) in the as-treated population and 2.1 months (95% CI, 1.4-2.8) and 1.4 months (95% CI, 1.3-1.5) in PD-L1 high and low or negative patients, respectively. The PFS rates at 6, 9, and 12 months were 22% (95% CI, 16%-28%), 18% (95% CI, 12%-25%), and 16% (95% CI, 10%-23%), respectively, in the as-treated population (eFigure 2 in Supplement 2). The median OS was 18.2 months (95% CI, 8.1 to not estimable) in the as-treated population and 20.0 months (95% CI, 11.6 to not estimable) and 8.1 months (95% CI, 3.1 to not estimable) in PD-L1 high and low or negative patients, respectively (Figure 2C). The OS rates at 6, 9, and 12 months were 64% (95% CI, 56%-71%), 57% (95% CI, 47%-66%), and 55% (95% CI, 44%- 65%), respectively, in the as-treated population (Figure 2C).

Safety in the UC Cohort

Any-grade treatment-related AEs were reported in 60.7% (Table 3). With a median time to onset of 6.1 weeks (range, 4.1-10.1 weeks), the cumulative incidence of treatment-related AEs occurred early, was highest during the first 8 weeks, and subsequently seemed to plateau at around 32 weeks (eFigure 3 in Supplement 2). The most common AEs were fatigue (19.4%), decreased appetite (9.4%), diarrhea (8.4%), and rash (7.3%). Grade 3 or 4 treatment-related AEs occurred in 6.8% (Table 3).

Treatment-related AEs leading to death occurred in 2 of 191 patients (1.0%), 1 each due to autoimmune hepatitis (in a man in his 70s with no history of autoimmune disease, liver disease, or liver abnormalities, but who had elevated liver enzyme levels at baseline prior to dosing) and pneumonitis (in a man in his 50s with a history of untreated grade 1 dyspnea and remote tobacco use); see the eResults in Supplement 2 for additional details regarding the treatments and events that led to both deaths. There were no clinically meaningful differences in AE rates between PD-L1 high and low or negative subgroups (data not shown). Serious treatment-related AEs were reported in 4.7%. Autoimmune hepatitis occurred in 2 patients, whereas other serious AEs only occurred in 1 patient each. Treatment-related AEs required infusion interruption or dose delay in 11.0% and led to permanent treatment discontinuation in 3 patients (1.6%), including the 2 patients with imAEs that led to death.

Any-grade treatment-related AESIs occurred in 34.6%, most of which were grade 1 or 2 (eTables 2 and 3 in Supplement 2); their cumulative incidence seemed to plateau at around 32 weeks (eFigure 3 in Supplement 2); 4.7% experienced treatment-related grade 3/4 AESIs. Within this population, imAEs occurred in 11.5%; most had grade 1 or 2 events (eTable 3 in Supplement 2). The most common imAEs were hypothyroidism (5.2%), diarrhea (2.1%), and selected hepatic events (2.1%; including increased aspartate aminotransferase level, increased alanine aminotransferase level, increased transaminase levels, and autoimmune hepatitis). Grade 3 or 4 imAEs occurred in 4 patients (2.1%) and were reported in the categories of selected hepatic events (1.0%), and rash and selected renal events (0.5% each). Among patients with any-grade treatment-related AESIs, 7.3% were administered concomitant systemic steroids.

Safety in the Overall Population Receiving Durvalumab, 10 mg/kg Q2W

The types and frequencies of any-grade and grade 3 or 4 treatment-related AEs in all patients who received durvalumab, 10 mg/kg Q2W (ie, patients with any solid tumor; n = 970), were similar to those of patients with UC (Table 3). In addition, time to onset of treatment-related AEs was similar (8.1 weeks; range, 6.1-10.1 weeks) and the cumulative incidences of treatment-related AEs and AESIs also seemed to be highest during the first 8 weeks, plateauing at around 32 weeks (eFigure 3 in Supplement 2). See the eResults in Supplement 2 for further details regarding the safety of durvalumab in this population.

Limitations

Notwithstanding inherent limitations of cross-trial comparisons, the antitumor activity of durvalumab was consistent with those of other, previously studied immunotherapies. In addition, despite different methodologies, most of these studies also showed better outcomes in PD-L1 high patients. Although clinical activity with durvalumab was observed in both PD-L1 high and low or negative patients, ORR was numerically higher in PD-L1 high patients. The combined algorithm therefore supports the effectiveness of the VENTANA PD-L1 (SP263) assay to identify patients most likely to respond to durvalumab, as internally validated using the first 20 enrolled patients to establish the algorithm and the remaining patients in the UC cohort to validate the approach (data not shown); however, external validation is warranted before this approach can be used extensively. In addition, this aligns with the biology of response to immunotherapy and is consistent with external UC data, since PD-L1 expression on either TCs or tumor-associated immune cells can inhibit antitumor immunity. However, while optimal, this approach does not completely exclude patients who may respond. The high negative predictive value of the assay (94.9%) may be especially helpful in educating patients on the likelihood of response to durvalumab; however, it should not be used to exclude patients from therapy, especially because there is no single standard of care in the second-line or greater UC setting, such that some PD-L1 low or negative patients may be better candidates for single-agent durvalumab vs commonly used chemotherapy.