NCT02517398 is an ongoing, phase I, open-label trial of bintrafusp alfa that includes multiple solid tumor expansion cohorts. Planned enrollment for the advanced SCCHN cohort was 30 patients. With 30 subjects treated, the study has approximately 87% power to rule out a ≤15% ORR (null hypothesis) when the true ORR is 35% at the 10% type I error rate (one-sided). Meeting this endpoint would have warranted the clinical development of a phase II or III trial based on a very strong phase I signal. Eligible patients were aged ≥18 years and had histologically confirmed, advanced SCCHN not amenable to curative therapy that progressed/recurred after prior platinum therapy in the recurrent/metastatic setting, or within 6 months after the last dose of platinum received in combination with radiotherapy for the treatment of locally advanced disease. Patients may have received prior cetuximab, but prior treatment with any antibody/drug-targeting T-cell coregulatory proteins such as anti-PD-1, anti-PD-L1 or anticytotoxic T-lymphocyte-associated protein-4 antibody was not allowed. Patients had an Eastern Cooperative Oncology Group performance status of 0 or 1; life expectancy of ≥12 weeks; adequate renal, hepatic and hematological function; measurable disease by RECIST 1.1 and available archival tumor material or fresh biopsies taken within 28 days of the first administration of study drug. Patients were not selected on the basis of PD-L1 expression or other biomarkers.

Integrated analysis of bintrafusp alfa exposure, response and progression-free survival (PFS) supported a flat dose of 1200 mg every 2 weeks,32 and this dose was used in the expansion cohorts. Patients received bintrafusp alfa via intravenous infusion over 1 hour once every 2 weeks until confirmed progressive disease (PD), unacceptable toxicity or trial withdrawal. Treatment beyond PD was permitted if adverse events (AEs) were manageable, performance status was maintained, and no further treatment was indicated.

Premedication with an antihistamine and acetaminophen approximately 30–60 min prior to each dose of bintrafusp alfa was mandatory for the first 2 infusions and was optional afterward. Dose reduction was not permitted; however, interruption or discontinuation of bintrafusp alfa was allowed as described in online supplementary table S2.

Tumor response was assessed by radiographic imaging 6 weeks after initiating treatment, then every 6 weeks for the first year and every 12 weeks thereafter. Response was confirmed by repeated radiographic assessment no sooner than 4 weeks from the first documentation of response. AEs were monitored throughout treatment, for the first 28 days after the last study dose, at 10 weeks post-treatment and every 12 weeks thereafter and were assessed according to NCI-CTCAE V.4.03. Immune-related AEs were identified using a prespecified list of MedDRA preferred terms followed by medical review.

The primary endpoint for this expansion cohort was to assess the confirmed best overall response (BOR) according to RECIST 1.1 as adjudicated by an independent review committee (IRC). Secondary endpoints included BOR per investigator and safety and tolerability. Exploratory endpoints included PFS, duration of response (DOR) and disease control rate (DCR) according to RECIST 1.1 as adjudicated by IRC, and OS. Exploratory biomarker analyses of efficacy were also performed.

Efficacy and safety were analyzed in all patients who received ≥1 dose of bintrafusp alfa. DCR—defined as the proportion of patients with a BOR of complete response (CR), partial response (PR), non-CR/non-PD or stable disease (SD) —was tabulated with an exact 95% Clopper-Pearson CI. PFS, DOR and OS were analyzed using the Kaplan-Meier method. Safety and tolerability were analyzed using descriptive statistics.

Tumor PD-L1 protein expression was measured by immunohistochemical staining of formalin-fixed, paraffin-embedded (FFPE) tumor tissue blocks using a proprietary assay (Dako, Carpinteria, California, USA) and an anti-PD-L1 rabbit monoclonal antibody clone 73-10 under license from Merck KGaA (Darmstadt, Germany). PD-L1 positivity in tumor cells was scored as the proportion of tumor cells showing membranous PD-L1 staining.

HPV status was determined by RNA sequencing (RNASeq) using FFPE tissue samples and standard protocols. Kallisto33 (V.0.43.1) was used to align sequencing results to a transcriptome containing the human Ensembl transcripts (V.91, GRCh38) and viral genomes from the National Center for Biotechnology Information collection. HPV content in each sample was assessed as the fraction of reads mapping to any papillomavirus genome. To determine a cut-off (for fraction of viral reads), the same process was applied to SCCHN samples from SRP063006/GSE72536 and SRP066090/GSE74927. The HPV statuses deposited with these sequences were used to choose a cut-off that provided perfect separation between the 28 positive and 51 negative training samples.

For gene expression quantification, RNASeq data were generated from FFPE tissue samples and standard protocols. Bowtie2 V.2.2.334 was used to align sequencing results to the Ensembl 75 human genome (GRCh37 February 2014). Gene expression was determined using RSEM V.1.2.31 with the Ensembl gene annotations.35 Hypothesis testing was performed by comparing RSEM-computed expected counts. Transcript-per-million (TPM) values were upper-quartile normalized and log-transformed for further analysis. Signature scores were defined as the mean log 2 (fold-change) among all genes in each gene signature (online supplementary table S3).19 36 These were calculated by adding a pseudocount of 0.5 TPM to all genes and samples, determining the log 2 (TPM), then subtracting the median log 2-TPM for each gene across all samples from the log 2-TPM for each gene.

Tumor samples were used to determine immune phenotype based on available immunohistochemistry data (PD-L1 stain and PD-L1 negative control) and H&E staining. An exploratory classification system was used to categorize tumors as having an immune-inflamed (immune cells in direct physical contact with tumor cells), immune-excluded (≥1% of tumor stroma area populated by lymphocytes, immune cells possibly located in immediate vicinity of tumor cells but not efficiently infiltrating tumor cell clusters, and very infrequent physical contact between lymphocytes and tumor cells), or immune desert (<1% of tumor stroma area populated by lymphocytes, no dense immune cell infiltrates and no contact of immune cells with tumor cells) phenotype. This system was developed from methods previously described.19 A pathologist who was masked to the response data scored the scanned slides and determined the corresponding immune phenotype.

Between October 6, 2016 and January 12, 2017, 46 patients were screened. Of these, 32 patients were enrolled and treated (full analysis set) at 20 sites in the Asia-Pacific region, Europe and the USA. At the time of data cut-off on August 24, 2018, patients had been followed for a median of 86.4 weeks (range 2–97 weeks); 4 patients (13%) remained on treatment and the median duration of therapy was 12.1 weeks (range 2–96 weeks). The primary reason for treatment discontinuation was disease progression (n=20 (63%)), AE (n=5 (16%)), consent withdrawal (n=1 (3%)), death unrelated to study treatment (hemorrhage; n=1 (3%)) and investigator decision (n=1 (3%)). The death unrelated to study treatment occurred 2 weeks after the patient received the first and only dose of bintrafusp alfa. Twenty-eight patients (88%) had progressed on platinum-based therapy in the second-line or later-line, recurrent and/or metastatic setting, and 4 patients (13%) had progressed on platinum-based therapy in the locally advanced setting (table 1).

The confirmed ORR by IRC was 13% (95% CI 4% to 29%; 4 PR), and an additional 4 patients had SD, resulting in a DCR of 34% (95% CI 19% to 53%) (table 2 and figure 1A,B). The median DOR was 18.1 months (range 6–20+ months). Two of 4 patients with PR had an ongoing response of 16.5+ and 19.6+ months as of data cut-off (figure 1C). Median PFS was 1.4 months (95% CI 1.3 to 4.0 months) and the 6-month and 12-month PFS rates were 28% (95% CI 14% to 45%) and 21% (95% CI 9% to 37%), respectively; the 18-month PFS rate was stable at 21% (table 2 and figure 2A). Median OS was 9.1 months (95% CI 6.6 months to not estimable (NE)) and the 12-month and 18-month OS rates were 48% (95% CI 30% to 64%) and 44% (95% CI 26% to 61%), respectively (figure 2B).

By investigator assessment, ORR and DCR were 16% and 34%, respectively (table 2). Two patients developed delayed PR after initial PD (time to PR 5.1 and 8.5 months); the patient with time to PR of 8.5 months by investigator assessment did not have target lesions and was therefore assessed as non-CR/non-PD by IRC based on the non-target lesions. The total clinical response rate by investigator assessment was 22%.

Thirty-one patients (97%) experienced a treatment-emergent AE (TEAE) and 24 (75%) experienced a grade ≥3 TEAE. Treatment was discontinued in 8 patients (25%) due to TEAE. Twenty-two patients (69%) experienced treatment-related AEs (TRAEs); of these, 11 (34%) experienced grade 3 TRAEs, which included rash maculopapular (n=2 (6%)) and increased liver enzymes, anemia, diabetic ketoacidosis, colitis, hyperglycemia and hyperthyroidism (n=1 each (3%)) (table 3). No grade 4 or 5 TRAEs were observed. TRAEs led to treatment discontinuation in 2 patients (colitis and SCC of the skin n=1 each (6%)). Keratoacanthomas (n=3 (9%)) and SCC of the skin (n=2 (6%)) were managed with excision and observation. No patient experienced an infusion-related reaction and no infusion rate reductions were necessary. Immune-related AEs were observed in 13 patients (41%), and 4 experienced grade 3 immune-related AEs (colitis and increased alanine and aspartate aminotransferases (n=1 each) and rash maculopapular (n=2)); no grade 4 or 5 immune-related AEs were observed (online supplementary table S4). The occurrence of immune-related colitis led to treatment discontinuation in 1 patient (3%).

Tumor PD-L1 expression status was available in 31 of 32 patients. Of 25 patients with PD-L1 expression on ≥1% of tumor cells (PD-L1-positive), 3 achieved PR and 3 had SD, resulting in an ORR of 12% (95% CI 3% to 31%) and DCR of 24% (95% CI 9% to 45%). One of 6 patients with PD-L1 expression on <1% of tumor cells (PD-L1-negative) achieved PR (figure 3A and online supplementary table S5). Median PFS in patients with PD-L1-positive tumors was 1.4 months (95% CI 1.2 to 4.0 months) and 3.5 months (95% CI 1.2 months to NE) in patients with PD-L1-negative tumors (figure 3B and online supplementary table S5). Median OS in patients with PD-L1-positive tumors was 8.8 months (95% CI 6.3 months to NE) and NE (95% CI 3.2 months to NE) in patients with PD-L1-negative tumors (figure 3C and online supplementary table S5).

Tumor HPV status was evaluable in 31 of 32 patients. Objective responses were observed in 3 of 9 patients with HPV-positive tumors (33% (95% CI 5% to 70%)) and in 1 of 22 patients with HPV-negative tumors (5% (95% CI 0% to 23%)) (figure 3D, online supplementary table S1 and online supplementary figure S1). Median PFS in patients with HPV-positive tumors was 2.7 months (95% CI 1.2 to 19.6 months) and 1.3 months (95% CI 1.2 to 4.0 months) in patients with HPV-negative tumors. Median OS in patients with HPV-positive tumors was 8.0 months (95% CI 2.7 months to NE) and 9.1 months (95% CI 6.3 months to NE) in patients with HPV-negative tumors (online supplementary table S1 and online supplementary figure S1).

A total of 30 patients had both HPV and PD-L1 data available. Among 9 patients with HPV-positive tumors, 7 (78%) were PD-L1-positive, and among 21 patients with HPV-negative tumors, 17 (81%) were PD-L1-positive, suggesting no association between HPV status and PD-L1 expression.

RNAseq data were analyzed to evaluate prespecified genes and gene signatures related to the TGF-β pathway and the immune-related pathways versus response to bintrafusp alfa. An extensive exploratory analysis was also conducted to identify genes or gene signatures associated with response to bintrafusp alfa in a biology/pathway agnostic manner. Several genes and signatures that reflect immune activity had higher levels in HPV-positive tumors compared with HPV-negative tumors, although the differences were not statistically significant after correcting for the size of the set tested (figure 3E).

Evaluable patient tumors were characterized according to immune phenotype as either inflamed (tumor infiltration of both CD4+ and CD8+ T cells and other immune cells), immune-excluded (immune cells are present but do not infiltrate the tumor) or immune-desert (lack of immune cells in and around the tumor). Of the 10 patients with a tumor of the inflamed immune phenotype, 3 had responses, and of 21 patients with a tumor of the immune-excluded phenotype, 1 had a response. A response following pseudoprogression was observed in a patient with a tumor of the immune-excluded phenotype (figure 3F).