Immunotherapy agents have shown great promise for the treatment of patients with head and neck cancer; however, only a small subset of patients reap these benefits. Thus, the question becomes: which patients are more likely to benefit from these therapies, and why?
Robert L. Ferris, MD, PhD
Immunotherapy agents have shown great promise for the treatment of patients with head and neck cancer; however, only a small subset of patients reap these benefits. Thus, the question becomes: which patients are more likely to benefit from these therapies, and why? As such, researchers are investigating potential predictive biomarkers of response to identify the patients who would respond best to these agents.
“Squamous cell carcinoma of the head and neck [SCCHN] is an immunosuppressive disease, with lower absolute lymphocyte counts than those found in healthy subjects, impaired natural killer cell activity, and poor antigen-presenting function. Impairment of tumor-infiltrating T lymphocytes has also been reported in SCCHN and other cancers, with a strong impact on clinical outcome. Therefore, immunomodulatory therapies that overcome immune suppressive signals in patients with SCCHN have therapeutic promise,” Robert L. Ferris, MD, PhD, said in a report published in theJournal of Clinical Oncology.1
Two checkpoint inhibitors have already been approved as immunotherapy agents for patients with head and neck cancer. Pembrolizumab (Keytruda) is for the treatment of patients with platinum-refractory SCCHN, and nivolumab (Opdivo) is for patients with recurrent or metastatic SCCHN with disease progression on or after a platinum-containing regimen. Both are inhibitors of PD-1, which is a T-cell molecule that binds to the ligands PD-L1 or PD-L2.
In the KEYNOTE-012 trial, pembrolizumab was investigated in patients with recurrent and/or metastatic SCCHN, and patients showed durable responses to the treatment. The objective response rate (ORR) was 18% by central review and 20% by investigator review.2
Patients were not selected for the trial based on biomarkers, but were tested for human papillomavirus (HPV) and PD-L1 expression. Twenty-one percent were associated with HPV; among this group, the ORR was 32%. Expression of PD-L1 was noted in 77.9% of all patients.3 PD-L1positive patients (≥1% expression in the tumor cells alone) showed an ORR of 18% (95% CI, 12-26) versus 19% (95% CI, 10-30; P = .461) in PD-L1–negative patients (<1% expression). When measuring PD-L1 expression as a composite score from both tumor cells and inflammatory cells, PD-L1–positive patients had an ORR of 21% (95% CI, 15-28) compared with 6% (95% CI, 1-19) in PD-L1– negative patients (P = .023). A similar trend was noticed with progression-free survival (PFS) and overall survival (OS).4PD-L1 expression is a potential biomarker that has received a great deal of attention, as it can be found in ≥50% of patients with SCCHN.1However, there are con icting reports of its significance. HPV-associated disease also holds interest as a possible biomarker of response. Ferris, co-leader of the Cancer Immunology Program, University of Pittsburgh Cancer Institute, and chief in the Division of Head and Neck Oncologic Surgery, University of Pittsburgh School of Medicine, noted that tumor infiltration by PD-1positive regulatory T cells may be more common among patients with HPV-associated SCCHN than in patients with HPV-negative disease. In a group of patients with SCCHN, 55% had high levels of PD-L1 expression, yet there was no association between the PD-L1 expression and the HPV status in the tumor.5
The utilization of PD-L1 expression as a biomarker is complicated by the factors that go into measuring PD-L1 expression levels.6Trials often use different cutoff points to indicate PD-L1 positivity in a patient, with the base starting at ≥1%, as in the KEYNOTE-012 trial. However, improved survival rates in patients with PD-L1 expression of ≥50% in the tumor cells was noticed in the KEYNOTE-024 study, which investigated pembrolizumab in patients with nonsmall cell lung cancer.7A true verification of the predictive value of PD-L1 expression may require a more standardized approach.
Expression levels are also often measured using different assays or by testing different cells. KEYNOTE-012, and the ongoing KEYNOTE-055 trial, which presented preliminary results during the 2016 ASCO Annual Meeting, de ned PD-L1 positivity as expression of ≥1% on tumor and inflammatory cells, using the Dako assay.8KEYNOTE-055 also showed a difference in ORRs (17% vs 8%, respectively) between patients with positive and negative PD-L1 expression at this cutoff point.
However, a recent report suggested that PD-L1 expression is only a “favorable” prognostic biomarker when measured using tumor-infiltrating immune cells instead of tumor cells in head and neck cancer.9The analysis investigated PD-L1 expression on tumor cells and tumor-in ltrating immune cells, divided into 4 groups based on degree of expression (TC0/1/2/3 and IC0/1/2/3), and the abundance of tumor-infiltrating lymphocytes in 402 patients with SCCHN.
The study showed that there was no significant difference in recurrence-free survival (RFS; P = .206) or OS (P = .317) rates between patients with PD-L1positive or –negative expression when measured on tumor cells alone. On the other hand, higher RFS and OS rates were noted for PD-L1–positive patients when measured on immune cells compared with patients with negative expression. This association continued even after adjusting for factors such as sex, smoking, age, pathologic stage, and HPV infection.
Kim et al suggested that this was because the frequency of higher PD-L1 expression was greater in immune cells than in tumor cells, with little to no overlap. In these patients with the highest PD-L1 expression levels (TC3 or IC3), expression levels of effector T-cell markers, including interferon-gamma and GZMB, were found to be higher in the IC3 group of patients compared with the TC3 group.
PD-L2 expression as a biomarker was also investigated in an analysis by Laura Quan Man Chow, MD, and colleagues of responses to pembrolizumab treatment in patients with head and neck cancer. In patients with positive PD-L2 expression on their tumor and inflammatory cells, the ORR was 23% (95% CI, 15-31) versus 10% (95% CI, 4-20) in PD-L2negative patients (P = .022). The ORR among patients who were positive for both PD-L1 and PD-L2 expression was 23% (95% CI, 16-32).4
“We believe that PD-L2 expression on both tumor and inflammatory cells was predicative of response to pembrolizumab,” Chow, associate professor of medical oncology at the University of Washington School of Medicine, said when presenting the ndings during the 2016 ASCO Annual Meeting.
When looking for further predictive biomarkers to treatment with pembrolizumab, investigators tested a 6-gene interferon-gamma (IFN-γ) signature, consisting of the genes (CXCL9, CXCL10, IDO1, IFNG, HLA-DRA, and STAT1) that had been identified in a cohort of patients with melanoma from the KEYNOTE-001 study, as a composite score.10The biomarker analysis demonstrated that the gene signature was associated with an inflamed tumor phenotype and diverse immunoregulatory activities, including immune tolerance.
There was a significant association with best overall response (P = .0047) and PFS (P = .0009) for the composite score, which demonstrated a positive predictive value of 40% and a negative predictive value of 95%. The authors noted that the composite score could be used to identify patients who would not respond to treatment with pembrolizumab, although further validation was recommended. The study of a potential association between the IFN-γ signature and PD-L1positive expression was also suggested as patients with higher PD-L1 expression levels and IFN-γ–related genes showed greater antitumor activity.
In the phase III CheckMate-141 trial, 361 patients with recurrent SCCHN were treated with second-line therapy of nivolumab or standard therapy. The PD-1 inhibitor showed a median OS of 7.5 months (95% CI, 5.5-9.1) over 5.1 months for patients treated with standard therapy (HR, 0.70; 97.73% CI, 0.51-0.96; P = .01). Estimated 1-year OS rates were 36% with nivolumab versus 16.6% with standard therapy.11
PD-L1 expression was evaluated on tumor cells in a prespecified analysis of the trial, in which 57.3% of the 260 evaluable patients demonstrated PD-L1 expression of ≥1%. PD-L1 expression was grouped into different levels of positivity: ≥1%, ≥5%, and ≥10%. The median OS did not differ significantly between the 3 groups (8.7, 8.8, and 8.7, respectively). Alternatively, patients with <1% expression showed a median OS of 5.7 months while patients with <5% and <10% expression levels showed median OS rates of 7.0 and 7.2 months, respectively (TABLE).
The analysis also investigated tumor HPV status by p16 immuno-histochemical testing in 178 evaluable patients. Regardless of p16 status (≥1% or <1%), the median OS was longer with nivolumab compared with standard therapy. Yet patients with p16-positive tumors showed a median OS of 9.1 months compared with 7.5 months for patients with p16-negative tumors and 4.4 months for p16-positive patients treated with standard therapy. However, p16-negative patients did better with standard therapy than p16-positive patients.
The HPV status was especially important among patients with oropharyngeal cancer, as HPV-positivity is more frequently noted in oropharyngeal cancers.9In the analysis by Kim et al, the authors also noted that PD-L1expressing immune cells were more frequently noted in patients with HPV-positive oropharyngeal cancer than in patients with HPV-negative oropharyngeal cancer.
Combined, p16 status and PD-L1 expression levels showed the greatest effect on response in patients with positive p16 and negative PD-L1 expression on tumor cells (n = 24), for a median OS of 10 months. Patients with positive PD-L1 expression ≥1% and negative p16 status (n = 17) showed the lowest hazard ratio of 0.44 (95% CI, 0.18-1.10).11
“Although we observed preliminary evidence that patients with a tumor PD-L1 expression level of 1% or more or p16-positive tumors (or both) may have a greater magnitude of effect from nivolumab therapy than those whose PD-L1 level was less than 1% or who had p16-negative tumors, the interactions were not significant and were not corrected for multiple comparisons,” Ferris et al wrote in their conclusion.
When analyzing for additional potential predictive biomarkers of response to nivolumab treatment, Ferris et al investigated peripheral blood lymphocytes both at baseline and during treatment and found that levels of CD8-positive T cells were higher in the patients who responded to treatment with nivolumab than in those who did not respond (23% vs 13%, respectively), both at baseline and during treatment (P <.05).12
Patients who responded to nivolumab also showed a decrease in CTLA-4-positive CD8-positive effector T cells while on treatment. However, responders and nonresponders both showed a 2-fold decrease in PD-1-positive CD8-positive and PD-1 CTLA-4-positive CD8-positive effector T cells following treatment with nivolumab (P <.05). PD-1-positive regulatory T cells (Tregs) were found to be at lower levels in responders than in nonresponders both at base- line (19% vs 33%, respectively; P <.01) and during treatment (12% vs 20%; P <.001). Also, levels of Ki67-positive Tregs at baseline were lower in nonresponders than responders (P <.05).
The authors suggested that both PD-1 and CTLA-4 pathways be targeted in patients with SCCHN to overcome suppressive signals in CD8-positive effector T cells and Tregs that express both PD-1 and CTLA-4. The ongoing CheckMate-651 trial is doing just that, investigating nivolumab in combination with ipilimumab (Yervoy), a CTLA-4 inhibitor, against the standard of care in patients with recurrent or metastatic SCCHN (NCT02741570).