Concurrent Definitive Immunoradiotherapy for Patients with Stage III–IV Head and Neck Cancer and Cisplatin Contraindication

A standard treatment for locally advanced head and neck squamous cell carcinoma (HNSCC) is the combination of radiation and cisplatin. However, cisplatin causes numerous toxicities including hearing loss, tinnitus, neuropathy, and nephropathy that can lead to contradictions to its use. Thus, there is a great clinical unmet need for an effective systemic agent for combination with radiation in cisplatin-ineligible patients with potentially curable disease.

Pembrolizumab is a fully humanized IgG mAb against the programmed death receptor 1 (PD-1) that is active against head and neck cancer. At the time of study design and initiation, multiple studies had shown activity in platinum-refractory recurrent/metastatic disease, leading to approval. Recently, on the basis of data reported in KEYNOTE-048 (1), pembrolizumab was approved in first-line metastatic/recurrent disease. However, in contrast to other cancers, such as non–small cell lung cancer, most head and neck cancer (∼90%) does not present with metastatic disease, but rather with local or loco-regional disease. Because of this curability, as well as the large number of quality life years gained when cure is achieved, this population is particularly important.

We hypothesized that pembrolizumab and radiation could provide synergistic benefit in patients with locally advanced HNSCC. Radiation results in type I IFN induction through a stimulator of IFN genes–dependent pathway. Furthermore, it can increase HLA expression, upregulate HLA-associated tumor antigen complexes, and increase both T-cell receptor clonality and diversity. Critically for the combination of a PD-1 agent, radiotherapy can increase PD-1 and PD-L1 expression as well as improve T-cell trafficking to the tumor microenvironment (2). Preclinically, concurrent PD-L1 blockade has been shown to sensitize cancer cells to the effects of ionizing radiation (3) and ionizing radiation has been shown to sensitize tumors to PD-L1 inhibition (4). We hypothesized that additional pembrolizumab subsequent to the last dose of radiotherapy could provide ongoing sensitization to radiation's effects and, based on the preceding rationale, could be particularly effective just after radiotherapy.

We therefore embarked upon a single-arm, phase II study for patients with locally advanced HNSCC for whom chemoradiotherapy would be a standard of care, except with a contraindication to cisplatin. Patients were treated with three cycles of pembrolizumab concurrent with standard of care, definitive radiation (70 Gy), followed by three additional adjuvant cycles of pembrolizumab. In the context of more than 150 studies exploring the addition of immunotherapy to standard full dose radiotherapy (with or without chemotherapy; ref. 5), we herein present data on toxicity, progression-free survival (PFS; primary endpoint), overall survival (OS), correlative studies, quality of life (QoL) measures, and patient-reported outcomes (PRO).

Patient were required to have untreated stage III–IV (nonmetastatic) HNSCC which was histologically or cytologically confirmed as per American Joint Committee on Cancer (AJCC) version 7.0, which was current staging criteria at the time of study initiation. Nasopharyngeal carcinoma was excluded. Patients were required to have a contraindication to cisplatin as defined by: abnormal renal function [glomerular filtration rate (GFR) < institutional lower limit of institutional normal], abnormal hearing (patient or audiology defined), preexisting tinnitus, neuropathy (bilateral paresthesias or loss of deep tendon reflexes in upper and/or lower extremities), diabetes mellitus, oncologist certification that patient would not be considered eligible for high-dose cisplatin when given as standard of care (e.g., due to age or another medical problem, reason was required to be documented), and patient refusal for high-dose cisplatin. Of note, other than for GFR as defined above, thresholds for other indications for cisplatin ineligibility were provider defined consistent with standard-of-care practice, without specific thresholds. Patients were required to have Eastern Cooperative Oncology Group (ECOG) performance status 0–1 and normal organ function. Patients with active autoimmune disease, a requirement for immunomodulating medications, active infection, or noninfectious pneumonitis were excluded. Full details of inclusion criteria may be found in the protocol, provided as Supplementary Data S1.

This study was conducted as a single-arm, multicenter phase II study at the University of North Carolina (Chapel Hill, NC), Fox Chase Cancer Center (Philadelphia, PA), and Johns Hopkins (Baltimore, MD). Patients were treated with 70 Gy of intensity modulated radiotherapy in 35 fractions. To ensure appropriate and uniform treatment across sites, details of this treatment were specified in the protocol and can be seen in detail in Supplementary Data S1. Patients were treated with three cycles of pembrolizumab at a flat dose of 200 mg i.v. concurrent with radiation, at time points 0, 3, and 6 weeks and for three adjuvant cycles following. Toxicity was assessed via Common Terminology Criteria for Adverse Event (CTCAE) v4.0 and PRO-CTCAE (6). Continuous safety monitoring (7) was conducted to ensure that patients received at least 95% of the intended radiation dose. The primary endpoint was PFS. Secondary endpoints included survival, toxicity, and QoL (via FACT-HN). Correlative endpoints were exploratory.

This trial was approved by the Lineberger Comprehensive Cancer Center (Chapel Hill, NC) Protocol Review Committee and the University of North Carolina Institutional Review Board (Chapel Hill, NC) and the relevant regulatory bodies at each collaborating site. It was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent. The study was registered at clinicaltrials.gov (NCT02609503). The study was designed and the article was written by the authors, who vouch for the accuracy and completeness of the data reported and adherence to the study protocol. No one who is not an author contributed to the writing of the article. Merck supplied pembrolizumab and funds for the conduct of the study. Additional funding for correlative analysis was supplied by Lineberger Comprehensive Cancer Center.

We hypothesized a median PFS of at least 16 months and assumed a median null hypothesis of 10 months, uniform accrual, no loss to follow-up, exponentially distributed times, a one-sided alpha of 0.1, 24-month accrual, and follow-up time of 12 months to accrue 29 patients to achieve 80% power. The null hypothesis of 10 months was derived from the Bonner and colleagues' study (8, 9), which compared radiotherapy alone to radiotherapy plus cetuximab. In that study, which did not select patients for cisplatin ineligibility, PFS was improved from 12.4 months to 17.1 months with the addition of cetuximab to radiotherapy. We adjusted the null hypothesis to 10 months to account for the accrual of less fit patients due to the inclusion requirement of cisplatin ineligibility and sought to provide a greater benefit than that seen with the addition of cetuximab to radiotherapy.

Sequential boundaries were used to monitor for unacceptable toxicity, defined as failing to receive at least 95% of the intended dose of radiotherapy (XRT), that is, if the number of patients is equal to or exceeds b_n out of n patients (see Supplementary Table S3). This was a Pocock-type stopping boundary (7) that yields the probability of crossing the boundary at most 0.05 when the rate of failing to receive at least 95% of the intended dose of XRT is equal to the acceptable rate of 10%.

The Kaplan–Meier method was used to estimate OS and PFS. PFS was defined as the time from the start of treatment to the first occurrence of progressive disease (RECIST or clinically defined) or death as a first event and subjects without an event were censored at their last evaluation date. OS was defined as the time from the start of treatment to death from any cause, with censoring of those alive and the last date known to be alive. All patients, regardless of total treatment exposure, were included in analyses of both safety and efficacy. Patients who withdrew from treatment were followed for progression or survival if they gave permission, or censored at the time of withdrawal of permission.

PD-L1 staining was conducted by Qualtek Molecular Laboratories as described previously (5) and required by Merck in supported investigator-initiated trials. For both the modified H-score (MHS) and modified percent score (MPS), the percent of tumor with membrane-specific staining was directly estimated at four intensity levels [negative (<1), low (1+), moderate (2+), and high (3+)] and reported as a percentage quantitatively as any one of the following: 0%, 1%, 2%, 5%, 10%, 15%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 85%, 90%, 93%, 95%, 99%, or 100%. To calculate MPS, the percents of tumor staining at low, moderate, and high levels were summed. To calculate MHS, the percent of membrane staining at low levels was multiplied by a factor of 1, the percent of staining at moderate levels was multiplied by 2, and the percent of staining at high levels was multiplied by 3; the three products were then summed to arrive at a final MHS. The abundance of tumor-infiltrating lymphocytes (TIL) present was also noted as a number between 0–3 with 0 being the absence of TILs and 3 indicating high profusion of TILs.

Peripheral blood was obtained for flow cytometry at baseline, week 20, and week 40.

Flow cytometry gates are provided in Supplementary Figs. S1 and S2. Briefly, CD19⁺ B cells were focused by removing T/natural killer/monocytic cell lineages and divided into mature (CD10⁻) and immature (CD10⁺) subsets. Transitional B cells were defined as IgD⁺CD38⁺ under immature subsets. The mature B-cell subset was subsequently divided into plasmablasts, unswitched B cells, and memory B cells based on IgD and CD38 expression. Marginal zone B cells and naïve B cells were distinguished by CD27 expression in unswitched B cell subset. Memory B cells were subcategorized into four subsets based on CD21 and CD27 expression under IgD⁻CD38⁻ gates, including active memory, resting memory (RM), tissue-like memory (TLM), and intermediate memory B cells. T cells were split into naïve (CD45RA⁺) and memory/effector (CD45RA⁻) subsets. T-regulatory (Treg) cells were defined by transcription factor Foxp3 and CD25⁺CD127⁻ surface phenotypes. Exhaustion markers (TIGIT, PD1, CD39, and CTLA4) were monitored in CD4 T cells, whereas only PD-1 was analyzed in CD8 T cells.

Blood was obtained for cytokine and chemokine analysis at baseline, week 4, week 20, and week 40. Plasma cytokine and chemokine concentrations were determined with the Bio-Rad Bio-Plex Pro Cytokine, Chemokine, and Growth Factor Assays 27-Plex Panel (catalog no. M50-0KCAFOY). Plasma samples cryopreserved at −80°C were thawed and centrifuged at 1,000 × g for 15 minutes at 4°C before transferring the cell and platelet-free supernatants to the 96-well assay plate. The bead-bound immunoassay was performed according to the manufacturer's instruction and data were acquired on the Bio-Plex 200 fluorescent plate reader. Cytokine, chemokine, and growth factor concentrations were acquired and analyzed by the Bio-Plex Manager. Among 27 cytokines, 12 were detected, whereas other 15 cytokines were found out of range of the standard curve for each protein. Mixed-effects analysis plus Tukey multiple comparisons test was performed at alpha = 0.05 by GraphPad analyses.

FACT-HN and PRO-CTCAE were evaluated at baseline, week 10, and week 20. Of note, higher FACT scores indicate superior QoL. The ranges for physical, social, and functional were 0–28; emotion was 0–24; and head/neck was 0–40. PRO-CTCAE symptoms were asked as a combination of severity, frequency, and interference items on a 0–4 scale. A composite score (0–3) for each symptom was created combining information from the three items (10). G-tube placement (yes/no) and duration of use (weeks) was recorded.

Between May 2016 and July 2018, we accrued 29 patients. At the time of data cutoff, median follow-up time for PFS was 21 months (range 10–40 months). Baseline characteristics describe a population typical for platinum-ineligible head and neck cancer (Table 1). Median age was 63, and most patients were male. The most common primary contraindication to cisplatin was otopathy, due to preexisting hearing loss (48.3%) and tinnitus (20.7%). Various anatomic locations were represented. Of 20 patients with oropharynx cancer, 14 were p16 status positive, of whom 8 also had ≤10 pack-year smoking history. The stage matrix (Supplementary Table S1) shows the population to advanced stage, consistent with the inclusion criteria requiring stage III or IV disease in the AJCC7 system. The updated AJCC version 8 staging system incorporates site of origin, tumor–node–metastasis stage, and human papillomavirus (HPV) status to provide an integrated risk of death. Therefore, we also provided staging of study patients based on AJCC8 with 21 of 29 patients (72.4%) having stage III or IV disease. Additional patient details are provided in Supplementary Table S2.

Toxicities were mostly grade 1–2 and consistent with those expected from radiation alone (Table 2). The major exception was grade 3–4 lymphopenia, experienced in 17 patients (58.6%). These were not associated with infections typically associated with lymphopenia. Median absolute lymphocyte count (ALC) on C1D1 was 1.5/L (0.8–3.3), 0.3/L (0.1–2.1) at the end of radiation, 0.6/L (0.2–4.3) at week 20, and 0.6/L (0.2–2.7) at week 40 (Fig. 2A).

We further characterized the nature of the lymphopenia through peripheral blood flow cytometry (Fig. 2B). Memory/effector CD4 T cells, naïve T cells, and B cells declined, while CD8 T cells, both memory/effector and naïve were preserved. Within B cells, significant but modest increases were observed in TLM B cells and transitional B cells, while decline was observed in RM cells (Fig. 3B). Within memory/effector T cells, the greatest declines observed, as expected, were in CD4⁺PD1⁺ and CD8⁺PD1⁺ cells (Fig. 3C); a significant decline in naïve CD8⁺PD1 cells was also observed (Fig. 3D). Although an increase in CD4⁺CD39⁺ cells was observed, we noted an absence of change in CD4⁺TIGIT⁺. Although CD4⁺CD25⁺ cell increased from baseline through week 20, with subsequent decline by week 40, we noted the absence of significant change in CD4⁺CD25⁺CD127⁻.

Absolute lymphopenia did not predict for progression. In comparing changes in cell types in patients with and without progression (Fig. 3A), we noted that patients with progression had higher baseline naïve B cells and lower circulating marginal zone B cells.

No major changes were seen in peripheral blood cytokines (measured by Luminex) over time (Supplementary Fig. S3) and peripheral blood cytokines did not predict for progression (data not shown).

With a median follow-up of 21 months, 1 year PFS and OS rates were 76% (95% confidence interval, 56%–88%) and 86% (67%–95%), respectively. Estimated 2-year PFS was 71% (49%–84%) and estimated 2-year OS was 75% (51%–88%). The primary PFS endpoint exceeded the hypothesized median of 16 months and its median was not reached (Fig. 1). PD-L1 did not predict for progression. Patients with p16-associated oropharynx cancer had a 1-year PFS of 86% (54%–96%) and OS of 93% (59%–99%) while others (non-oropharynx plus p16⁻ oropharynx) had 1-year PFS of 67% (38%–85%) and 1-year OS of 80% (50%–93%). In evaluating p16⁺ oropharynx cancers, patients with ≤10 pack-year smoking history had a 1-year PFS of 88% (39%–98%) and 1-year OS of 100%, while those with a ≥10 pack-year smoking history had a 1-year PFS of 83% (27%–97%) and 1-year OS of 83% (27%–97%). Although accrual was defined by AJCC7, we recategorized subjects by AJCC8 to provide a measure of outcomes by integrated stage-based and biologic risk. As shown in Supplementary Fig. S4, subjects with lower stage had numerically superior PFS and OS.

The patient experience was further characterized via FACT-HN and PRO-CTCAE. Key symptoms of pain, decreased appetite, swallowing difficulty, dry mouth, and fatigue climaxed at 10 weeks and improved from climax at week 20 (Fig. 4). A similar time course was seen in FACT subscales including the head and neck subscore.

Of 24 patients with data on G-tube use, 4 (16.7%) had G-tubes placed with a median duration of 50 days. Minimum radiation dose was 69.75 Gy; all patients received 70 Gy except for one. Only 4 patients received less than the six planned doses of pembrolizumab with 2 patients receiving five doses and 2 patients receiving three doses. Reasons for treatment discontinuation included adverse events, progression on study, death on study, and patient preference. In patients with progressive disease, subsequent therapies included salvage surgery (1), carboplatin/nab-paclitaxel/cetuximab (1), and carboplatin/paclitaxel (1).

In this phase II study of pembrolizumab concurrent with radiation for locally advanced head and neck cancer in cisplatin-ineligible patients, the primary endpoint of a median PFS of at least 16 months was met. To our knowledge, this represents the first published efficacy data on definitive immunoradiotherapy in this population. Treatment was safe with few grade 3–4 events. The most common toxicities, including radiation dermatitis, dry mouth, oral mucositis, and dysgeusia are expected from radiotherapy and were not worsened by pembrolizumab. We did not observe any immune-related adverse events.

The major toxicity observed was lymphopenia. In this study, 59% of patients experienced grade 3–4 lymphopenia. Lymphopenia was not reported with pembrolizumab alone in the KEYNOTE-048 study (1). When radiotherapy was combined with cetuximab or with cisplatin in the RTOG 1016 study (1), the rate of grade 3–4 events was 17% with either regimen, although other studies have shown that the radiation fractionation used in this study can result in severe lymphopenia (11, 12). Safety data on HPV⁺ patients in a phase II trial of pembrolizumab, cisplatin, and radiotherapy reported a 94% rate of grade 3–4 lymphopenia (13). Given that the mechanism of action of pembrolizumab is checkpoint inhibition of T cells, this toxicity is relevant to efficacy, safety, and future directions. We therefore further characterized the lymphopenia through flow cytometry. While B cells and CD4 T cells declined, CD8 T cells were relatively preserved. Peripheral blood flow cytometry subcategorized changes in lymphocyte subsets over the course of treatment. We observed an increase in TLM B cells and transitional B cells, while RM cells declined. In comparing baseline immune phenotypes, we observed more naïve B cells and fewer marginal zone B cells in patients with progression.

The patient perspective on nonhematologic toxicity and on QoL was further characterized with select PRO-CTCAE measures and the FACT-HN subscales. Both measures showed decline at 10 weeks compared with baseline, with incomplete recovery by 20 weeks. These patient-derived outcomes confirm CTCAE results that the acute toxicity of pembrolizumab and radiation is acceptable for the definitive treatment of HNSCC. The absence of long-term data is a weakness of this work and would be of use in future studies.

We acknowledge multiple additional limitations of this study. First, the study is small and nonrandomized. We have therefore avoided comparison of efficacy to landmark randomized studies. However, we do believe that the favorable efficacy data observed in this phase II study are sufficient to justify phase III study. The PRO and CTCAE data provide a more comprehensive characterization of the patient experience than is typically provided in small phase II studies. We caution against over interpretation of this data, but note the consistency between toxicity, PRO-CTCAE, and FACT and believe that this data will be useful in phase III design considerations.

Similarly, the small sample size of the peripheral blood flow cytometry data and cytokine/chemokine data, together with the absence of tissue studies and functional studies, precludes our ability to define the functional B cells' differences observed in peripheral blood. However, given emerging evidence for the role of B cells in the antineoplastic response (14), we consider these findings worthy of further study. While neither PD-L1 nor cytokine/chemokine profiling of peripheral blood predicted recurrence [the former consistent with results from the PACIFIC (15) trial in lung cancer], we again caution against overinterpretation given small patient numbers and encourage their evaluation in larger studies. Third, the use of nonclinical staining conditions and scoring precludes meaningful analyses or interpretation of PD-L1 expression; we note only that PDL1⁺ TILs were observed at high level in most patients. Finally, the optimal number of pembrolizumab cycles was not defined. Currently trials are evaluating 1 year of maintenance immunotherapy and it is possible that longer treatment could have resulted in superior efficacy.

Multiple clinical studies are already underway to further define the role of immunoradiotherapy in HNSCC. Most relevant to this work is GORTEC-2015-01, which randomizes patients ineligible for cisplatin to radiotherapy with either cetuximab or pembrolizumab. Other future work should investigate immune modulations concurrent with radiation, evaluate optimal sequence and timing, and consider the role of immune cells other than T cells.

J. Weiss is a paid consultant for AstraZeneca, EMD Serono, and Genentech, and reports receiving commercial research grants from Merck and AstraZeneca. S. Patel reports receiving commercial research grants from AstraZeneca. C.J. Shen reports receiving speakers bureau honoraria from and is an unpaid consultant/advisory board member for Nanbiotix. D.N. Hayes is a paid consultant for Merck and GeneCentric Therapeutics. R. Mehra is a paid consultant for Bayer and reports receiving commercial research grants from Merck and AstraZeneca. J.R. Bauman is a paid consultant for AstraZeneca and Pfizer. B.S. Chera holds ownership interest (including patents) in Naveris. B.G. Vincent is a paid consultant for GeneCentric Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Weiss, A.M. Deal, J.E. Grilley Olson, A.M. Zanation, D.N. Hayes, B.G. Vincent

Development of methodology: J. Weiss, A.M. Deal, T.G. Hackman, D.N. Hayes, K.P. McKinnon, B.G. Vincent

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Weiss, J.E. Grilley Olson, T.G. Hackman, J.M. Blumberg, A.M. Zanation, C.J. Shen, D.N. Hayes, C. Hilliard, R. Mehra, K.P. McKinnon, H.-H. Wang, M.C. Weissler, J.R. Bauman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Weiss, S. Sheth, A.M. Deal, J.E. Grilley Olson, S. Patel, D.N. Hayes, R. Mehra, K.P. McKinnon, H.-H. Wang, J.R. Bauman, B.S. Chera, B.G. Vincent

Writing, review, and/or revision of the manuscript: J. Weiss, S. Sheth, A.M. Deal, J.E. Grilley Olson, S. Patel, T.G. Hackman, J.M. Blumberg, T.J. Galloway, S. Patel, C.J. Shen, D.N. Hayes, R. Mehra, H.-H. Wang, J.R. Bauman, B.S. Chera, B.G. Vincent

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Weiss, S. Sheth, A.M. Deal, D.N. Hayes, C. Hilliard, K.P. McKinnon, B.G. Vincent

Study supervision: J. Weiss, S. Sheth, D.N. Hayes, J.R. Bauman, B.S. Chera, B.G. Vincent

S. Sheth gratefully acknowledges grant 18-40-12-SHET for the 2018 AACR-AstraZeneca-MedImmune Clinical Immuno-oncology Training Fellowship.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.