Articles

Safety and clinical activity of pembrolizumab for treatment

of recurrent or metastatic squamous cell carcinoma of the
head and neck (KEYNOTE-012): an open-label, multicentre,
phase 1b trial
Tanguy Y Seiwert, Barbara Burtness, Ranee Mehra, Jared Weiss, Raanan Berger, Joseph Paul Eder, Karl Heath, Terrill McClanahan, Jared Lunceford,
Christine Gause, Jonathan D Cheng, Laura Q Chow

Summary
Lancet Oncol 2016; 17: 956–65 Background Patients with recurrent or metastatic squamous cell carcinoma of the head and neck have few treatment
Published Online options. We aimed to assess the safety, tolerability, and antitumour activity of pembrolizumab, a humanised
May 27, 2016 anti-programmed death receptor 1 (PD-1) antibody, in patients with PD-L1-positive recurrent or metastatic squamous
http://dx.doi.org/10.1016/
cell carcinoma of the head and neck.
S1470-2045(16)30066-3
See Comment page 856
Methods This study was an open-label, multicentre, phase 1b trial of patients with recurrent or metastatic squamous cell
Department of Medicine,
University of Chicago, IL, USA
carcinoma of the head and neck. Patients were eligible for enrolment if they were aged 18 years or older, had a confirmed
(T Y Seiwert MD); Department diagnosis of recurrent or metastatic squamous cell carcinoma of the head and neck, and had any level of PD-L1 expression
of Medical Oncology, Fox Chase (ie, at least 1% of tumour cells or stroma that were PD-L1-positive by immunohistochemistry). Patients received
Cancer Center, Philadelphia, PA, pembrolizumab 10 mg/kg intravenously every 2 weeks. Primary outcomes were safety in the per-protocol population and
USA (B Burtness MD, R Mehra);
Lineberger Comprehensive
the proportion of patients with centrally reviewed overall response per Response Evaluation Criteria In Solid Tumors
Cancer Center at the University (RECIST, version 1.1). Overall response was analysed in the full analysis set, which was defined as all patients who had
of North Carolina, Chapel Hill, received at least one dose of pembrolizumab, had measurable disease at baseline, and one post-baseline scan or patients
NC, USA (J Weiss MD); Sheba without a post-baseline scan who discontinued therapy because of disease progression or a drug-related adverse event.
Medical Center, Tel Hashomer,
Israel (R Berger MD);
The study is registered with ClinicalTrials.gov, number NCT01848834 and is ongoing, but no longer enrolling patients.
Department of Medicine, Yale
University School of Medicine Findings Of the 104 patients screened between June 7, 2013, and Oct 3, 2013, 81 (78%) were PD-L1-positive. Of these,
and Yale Cancer Center, New
60 patients with PD-L1-positive squamous cell carcinoma of the head and neck were enrolled and treated: 23 (38%)
Haven, CT, USA (J P Eder MD);
Merck & Co, Kenilworth, NJ, were HPV-positive and 37 (62%) were HPV-negative. Pembrolizumab was well tolerated, with 10 (17%) of 60 patients
USA (K Heath MPH, having grade 3–4 drug-related adverse events, the most common of which were increases in alanine aminotransferase
T McClanahan PhD, and in aspartate aminotransferase, and hyponatraemia, each occurring in two of 60 patients; one patient developed a
J Lunceford PhD, C Gause PhD,
grade 3 drug-related rash. 27 (45%) of 60 patients experienced a serious adverse event. There were no drug-related
J D Cheng MD); and University
of Washington, Seattle, WA, deaths. The proportion of patients with an overall response by central imaging review was 18% (eight of 45 patients;
USA (L Q Chow MD) 95% CI 8–32) in all patients and was 25% (four of 16 patients; 7–52) in HPV-positive patients and 14% (four of
Correspondence to: 29 patients; 4–32) in HPV-negative patients.
Dr Tanguy Seiwert, Section of
Hematology and Oncology,
Interpretation Pembrolizumab was well tolerated and demonstrated clinically meaningful antitumour activity in
University of Chicago Medical
Center, Chicago, IL 60637, USA recurrent or metastatic squamous cell carcinoma of the head and neck, supporting further study of pembrolizumab
[email protected]. as anticancer therapy for advanced head and neck cancers.
uchicago.edu
Funding Merck & Co.

Introduction mutational burden that is probably linked to tobacco use,

Squamous cell carcinoma of the head and neck is the whereas HPV-positive cancers might also have mutations
seventh most common cancer worldwide.1 Patients with attributed to expression of APOBEC cytidine deaminases.4–6
recurrent or metastatic disease have a poor prognosis Genes involved in immunity and inflammation, such as
and few treatment options.2,3 The combination of HLA A or B and nuclear factor-κB pathway genes, are
cetuximab, platinum, and fluorouracil is commonly used frequently mutated in patients with squamous cell
as first-line treatment for recurrent or metastatic head carcinoma of the head and neck.4,5 The misregulation of
and neck squamous cell carcinoma,3 although taxanes these particular genes might explain the frequent
and methotrexate are used in later lines of treatment. observation of an inflamed phenotype in squamous cell
However, a more effective and less toxic treatment is carcinoma of the head and neck with the presence
needed in this palliative setting. of tumour-infiltrating lymphocytes in roughly 40% of
Human papillomavirus (HPV)-negative squamous cell tumours, and provide a rationale for the potential efficacy of
carcinoma tumours of the head and neck carry a high immunomodulatory drugs.6 In the healthy immune system,

956 www.thelancet.com/oncology Vol 17 July 2016

 Articles

Research in context
Evidence before this study PD-L1-positive recurrent or metastatic squamous cell carcinoma
To identify other published studies of the programmed death 1 of the head and neck. Pembrolizumab was well tolerated and
(PD-1) receptor or PD-L1 inhibitors in squamous cell carcinoma showed clinically significant antitumour activity in this heavily
of the head and neck, we searched the PubMed database using pretreated population. Results of this study also show that
the following search terms, not limited by date or language pembrolizumab efficacy was associated with PD-L1 and
restrictions: “PD-1 OR PD-L1 OR MK-3475 OR lambrolizumab OR interferon-γ-related gene expression, supporting further
nivolumab OR BMS-936558 OR MPDL3280A OR BMS-936559 assessment of predictive biomarkers for pembrolizumab in
AND HNSCC.” Although preclinical evidence strongly supports future studies.
the use of PD-1 blockade to enhance antitumour activity in
Implications of all the available evidence
squamous cell carcinoma of the head and neck (eg, Malm and
The high proportions of patients achieving an overall
colleagues, 2015), this search did not identify any other clinical
response to PD-1 blockade and durability of responses and
studies assessing PD-1 or PD-L1 inhibitors in patients with
stable disease observed with pembrolizumab support the
squamous cell carcinoma of the head and neck.
importance of the PD-1 pathway in squamous cell carcinoma
Added value of this study of the head and neck, and warrants further study of
To our knowledge, this is the first report of the efficacy and pembrolizumab as an anticancer treatment for advanced head
safety of a PD-1 or PD-L1 inhibitor in patients with and neck cancers.

the programmed death 1 (PD-1) receptor is expressed on of pembrolizumab in patients with advanced solid
activated T cells and interacts with its ligands, PD-L1 and tumours; here we report on the squamous cell head and
PD-L2, to protect healthy cells from excessive inflammatory neck cohort. Centres enrolling patients in this cohort were
or autoimmune responses.7–10 Tumour-associated regulation located throughout the USA and one was located in Israel
of the PD-1 pathway might lead to escape from immune (appendix p 5). Patients were eligible for enrolment if they See Online for appendix
surveillance. Tumour cells expressing PD-L1 can reduce were aged 18 years or older and had a confirmed diagnosis
T-cell effector activity and terminate immune responses.11,12 of recurrent (not amenable to locally curative options) or
Host tumour-infiltrating T lymphocytes6,13,14 mediate metastatic squamous cell carcinoma of the head and neck
PD-L1 expression via interferon-γ secretion.11 with at least 1% PD-L1 expression as determined by
Pembrolizumab is a high-affinity, humanised, IgG4-κ immunohistochemistry. Additional inclusion criteria
monoclonal PD-1 antibody. In clinical studies, pembro- were measurable disease based on Response Evaluation
lizumab has shown efficacy in patients with various Criteria In Solid Tumors (RECIST, version 1.1), Eastern
advanced solid tumours and is approved for the treatment Cooperative Oncology Group performance status of
of melanoma.15–17 Additionally, PD-L1 expression has been 0 to 1,20 adequate organ function determined by tests done
correlated with a higher treatment response to anti-PD-1 within 10 days of treatment initiation, and provision of
antibodies in many cancer types.17,18 However, patients with tumour tissue for PD-L1 expression analyses, HPV status,
negative PD-L1 staining also benefit from treatment with and biomarker assessment. The number of previous
PD-1 inhibitors, albeit at a lower frequency than those treatments the patients had received was not limited for
patients with PD-L1 expression.17,18 In this study we aim to inclusion, and treatment-naive patients were also allowed.
provide, to the best of our knowledge, the first safety and Patients who received previous treatments specifically
efficacy report for pembrolizumab in patients with recurrent targeting T-cell co-stimulation or checkpoint pathways
or metastatic PD-L1-positive squamous cell carcinoma of were excluded. Previous systemic immunosuppressive
the head and neck. Additionally, we also assessed RNA treatment had to be concluded within 7 days, chemotherapy
expression of six interferon-γ-regulated genes using a within 2 weeks, and anticancer monoclonal antibody
formalin-fixed paraffin-embedded (FFPE) tissue-compatible treatment within 4 weeks from the start of study treatment.
analysis method. A six-gene signature (consisting of Patients with additional progressing malignancies, CNS
CXCL9, CXCL10, IDO1, IFNG, HLA-DRA, and STAT1) was metastases, autoimmune diseases, interstitial lung
identified in a melanoma cohort in the KEYNOTE-001 disease, infections requiring systemic therapy, HIV, or
study19 of pembrolizumab, used in a predefined analysis in hepatitis B or C were excluded.
this cohort, and might serve as a predictive biomarker in the Patients were allocated to HPV-negative and
form of a composite score. HPV-positive subgroups based on investigator HPV
determination. HPV status was assessed at the local
Methods institution using p16 immunohistochemistry (where at
Study design and participants least 70% of cells staining positive for p16 were counted
This was an open-label, multi-cohort, multicentre, as p16 positive) as a surrogate marker; this analysis was
phase 1b trial assessing the safety and antitumour activity not reviewed centrally.

www.thelancet.com/oncology Vol 17 July 2016 957

 Articles

The study was done in accordance with the Declaration study with Nanodrop ND1000 and quality parameters
of Helsinki and the International Conference on built into the Nanostring nSolver software (version 2.5)
Harmonisation Good Clinical Practice guidelines, and according to manufacturer recommendations. Gene
was approved by relevant regulatory and independent expression values of the 680-gene NanoString panel
ethics committees. All patients provided written informed were calculated from the NanoString assay by
consent before study entry. quantile normalisation of log-transformed data.
Ten interferon-γ-related genes were previously tested as a
Procedures prespecified signature in melanoma samples from the
Enrolled patients received pembrolizumab 10 mg/kg KEYNOTE-001 study19 and found to be associated with
intravenously every 2 weeks until documented disease clinical outcome. The ten gene signature was then pared
progression, intolerable adverse events, intercurrent down to six member genes (CXCL9, CXCL10, IDO1,
illness that prevented further treatment, or completion of IFNG, HLA-DRA, and STAT1) on the basis of findings in
24 months of treatment. Treatment could be discontinued patients with melanoma before collection of outcome data
at physician or patient discretion. CT scans or MRI were for patients in the head and neck cohort. The interferon-γ
done at baseline and at 8-week intervals after treatment composite score evaluated here was calculated by
initiation to assess response. Patients with complete averaging the normalised values of these six genes.
response could discontinue study treatment, and clinically
stable patients with progressive disease could remain on Outcomes
treatment until progressive disease confirmation by a The primary endpoints were to assess the safety of
follow-up scan performed at least 4 weeks thereafter. pembrolizumab and the proportion of patients who
A patient could continue on treatment if the follow-up achieved an overall response, defined as the proportion of
scan showed a reduction in tumour burden compared patients who achieved complete response or partial
with the initial scan that showed progressive disease. response as assessed per RECIST version 1.1 by central
Incidence of adverse events was monitored and review. Secondary efficacy endpoints included proportion
graded using the National Cancer Institute Common of patients with an overall response per RECIST version 1.1
Terminology Criteria for Adverse Events (version 4.0). in HPV-positive patients and in patients previously treated
Throughout the study, the pembrolizumab dose could with cetuximab and platinum, proportion of patients with
be modified by delaying administration by a week an overall response per RECIST version 1.1 by investigator
(ie, administered every 3 weeks instead of every 2 weeks); assessment, and duration of response, progression-free
the dose could not be reduced. For patients who survival, and overall survival in the total patient population.
experienced grade 3 or worse drug-related adverse events, A separate analysis was done to assess additional
pembrolizumab treatment was withheld until toxicity predictive biomarkers such as interferon-γ-related gene
resolved to grade 0–1. Patients had to discontinue the expression. Progression-free survival was defined as the
study treatment if drug-related toxicity had not resolved time from allocation to the first documented disease
within 12 weeks of the last pembrolizumab infusion. progression according to RECIST version 1.1 or death due
Adverse events of special immunological interest to any cause, whichever occurred first. Duration of
regardless of causality were recorded by the investigators. response represented time from first RECIST version 1.1
Laboratory safety assessments (ie, haematology, response to disease progression in patients who achieved
chemistry, and urinalysis) were done within 10 days of a partial response or better. Overall survival was defined as
the first study treatment and up to 72 h before each dose. the time from allocation to death.
Blood samples were collected for pharmacokinetic and
antidrug antibody analyses on the first and second Statistical analysis
treatment cycle and every four cycles thereafter. If 22 assessable PD-L1-positive patients with
Immunohistochemistry for PD-L1 was done centrally HPV-negative head and neck cancer were enrolled, the
with the 22C3 anti-PD-L1 antibody (Merck & Co, study would provide approximately 80% power to detect
Kenilworth, NJ, USA) on newly obtained or archival a 25% difference in the proportion of patients with an
tumour samples in a research assay under Clinical overall response under the null hypothesis of the
Laboratory Improvement Amendments (CLIA) con- proportion of patients achieving overall response equal
ditions. PD-L1 expression prevalence was scored on a to 10% with a type I error rate of 2·5% if the true
scale from 0% to 100% and was predefined as positive if proportion of patients achieving an overall response is
detected in at least 1% of tumour or inflammatory cells or 35%. For the HPV-positive cohort, if 12 assessable
stroma. Further details about scoring are provided in the PD-L1-positive patients with HPV-positive head and
appendix (p 2). neck cancer were enrolled, the study would provide
Tumour RNA was extracted from FFPE slides with approximately 73% power to detect a 35% difference in
tumour macrodissected content of at least 60% and the proportion of patients achieving an overall response
assessed on the NanoString nCounter (Seattle, WA, USA). under the null hypothesis of overall response equal to
The quality and amount of RNA were assessed in our 20% with a type I error rate of 5% if the true overall

958 www.thelancet.com/oncology Vol 17 July 2016

 Articles

response is 55%. Per protocol, the actual number of overall response analysis. Central imaging vendor review
HPV-negative PD-L1-positive patients and HPV-positive was done on all patients, and an additional analysis was
PD-L1-positive patients enrolled might have exceeded done on patients with a confirmed response per central
the target to ensure that a minimum number of patients review. Overall survival was assessed using the intention-
were assessable for analysis. to-treat population, and progression-free survival was
All patients who received at least one dose of assessed using the full analysis set population.
pembrolizumab, had measurable disease at baseline, had For the proportion of patients with overall response
at least one post-baseline scan, or who did not have a (investigator and central review) per RECIST version 1.1,
post-baseline scan but discontinued therapy because of the 95% CI and p values were provided using exact
clinical disease progression or a treatment-related adverse binomial distribution. Statistical tests were done at the
event (full analysis set population) were included in the α=0·05 (two-sided) level. Patients without response
efficacy analyses. All patients who received at least one data were regarded as non-responders. HPV-positive
dose of pembrolizumab were included in the safety and HPV-negative patients were assessed separately.
analyses (per-protocol population). All responses as For estimates of progression-free survival, overall survival,
defined per investigator review, regardless of whether they and duration of response, Kaplan-Meier statistics were
were confirmed on a follow-up scan, were included in the applied. SAS (version 9.3) or R (version 3.1.3) was used

104 patients screened

81 PD-L1-positive

20 patients were not enrolled

61 PD-L1-positive enrolled

1 did not receive pembrolizumab

60 received pembrolizumab 10 mg/kg

intravenously every 2 weeks
(safety population)
49 discontinued total
8 adverse events
2 deaths
35 progressive disease
3 withdrawal by patient
1 withdrawal by physician
11 unknown*

23 HPV-positive 37 HPV-negative
18 discontinued 31 discontinued
3 adverse events 5 adverse events
14 progressive disease 2 deaths
1 withdrawal by patient 21 progressive disease
5 unknown* 2 withdrawal by patient
1 withdrawal by physician
6 unknown*

3 excluded 4 excluded 1 excluded because no post-baseline assessment or

2 no baseline measureable disease 2 no baseline measureable disease did not discontinue because of a drug-related
1 no post-baseline assessment or did not 2 no post-baseline assessment or did not adverse event or progressive disease
discontinue because of a drug-related adverse discontinue because of a drug-related adverse
event or progressive disease event or progressive disease

20 included in efficacy analysis† 56 included in efficacy analysis† 36 included in efficacy analysis†

Figure 1: Study profile

*A disposition record did not exist at the time of reporting; these patients might have been on treatment at the data cutoff. †Patients who receive at least one dose of pembrolizumab and had a
baseline and at least one post-baseline tumour measurement.

www.thelancet.com/oncology Vol 17 July 2016 959

 Articles

for all analyses. Additional statistical methodology is Role of the funding source
provided in the appendix (pp 2,3). The study is registered Academic advisers and the funder designed this study.
with ClinicalTrials.gov, number NCT01848834. All data were collected by the investigators and their site
personnel. Statisticians employed by the funder analysed
Total (n=60) the data, which were subsequently interpreted by all
Age (years) 63 (20–83) authors. The corresponding author (TYS) wrote the first
Sex draft of the manuscript, and all authors contributed to the
Female 11 (18%) writing, reviewing, and amending of the manuscript.
Male 49 (82%) A medical writer contracted by the funder provided
ECOG performance status assistance in preparing the manuscript. All authors had
0 18 (30%) full access to the data, vouch for their accuracy, and attest
1 41 (68%) that the study conformed to the protocol. TYS had the
Unknown 1 (2%) final responsibility to submit the publication.
HPV expression
Positive 23 (38%) Results
Negative 37 (62%) Between June 7, 2013, and Oct 3, 2013, 104 patients were
Previous tobacco use 51 (85%) screened, and 81 (78%) screened patients were found to
Sum of measurable target lesions at baseline (mm) 84 (10–336) be PD-L1-positive (figure 1; appendix p 4). Of these
Location of primary tumour PD-L1-positive patients, 61 (75%) were enrolled and
Oropharynx 16 (27%) 60 (74%) received at least one dose of pembrolizumab and
Tongue 14 (23%)
were included in the safety analysis set (one enrolled
Oral cavity 11 (18%)
patient who was randomly assigned in error, should not
Nasal cavity 4 (7%)
have been enrolled in the trial, and did not receive
Nasopharyngeal 3 (5%)
pembrolizumab and was excluded): 23 (38%) of 60 patients
were HPV-positive and 37 (62%) of 60 were HPV-negative
Larynx 2 (3%)
(of whom, two patients with unknown HPV status were
Hypopharynx 1 (2%)
considered to be HPV-negative and are included in the
Other or unknown 9 (15%)
HPV-negative subgroup; figure 1). 56 patients met the
Previous adjuvant or neoadjuvant therapy
prespecified criteria for the efficacy analyses. Of those
Yes 25 (42%)
enrolled, five patients were excluded from efficacy analyses
No 33 (55%)
for the following reasons: no study treatment administered
Unknown 2 (3%)
(one HPV-negative patient), no baseline measurable
Number of recurrences
disease within the data window (two patients; both
0 16 (27%)
HPV-positive), and no post-baseline assessment or did not
1 16 (27%)
discontinue because of a drug-related adverse event or
2 10 (17%) progressive disease (two patients; one HPV-positive and
3 11 (18%) one HPV-negative). For gene expression analyses,
≥4 7 (12%) 43 patients had adequate tissue, with 40 assessable for
Number of previous lines of treatment for recurrent or metastatic disease overall response. 13 of these patients were HPV-positive
0 7 (12%) and 26 were HPV-negative. One patient had unknown
1 9 (15%) HPV status and was considered to be HPV-negative.
2 20 (33%) Median follow-up was 14 months (IQR 4–14). The cutoff
≥3 22 (37%) date for the analyses was Dec 9, 2014.
Unknown 2 (3%) Baseline characteristics of patients are shown in table 1.
Previous treatment 39 patients (65%) had disease progression within
Cetuximab and platinum 38 (63%) 6 months of completion of platinum monotherapy or
Platinum without cetuximab 15 (25%) platinum and cetuximab treatment. HPV-positive patients
Taxanes 43 (71%) had a median of three lines of previous treatments for
Fluorouracil or capecitabine 30 (50%) recurrent or metastatic disease, and HPV-negative
Methotrexate 5 (8%) patients had a median of two. Most HPV-positive patients
Previous surgical procedure 48 (80%) had tumours in the oropharynx, and three patients
Previous radiotherapy 56 (93%) had non-oropharynx HPV-positive disease (one with
unknown primary tumour location and two with primary
Data are n (%) or median (range). Details regarding distant recurrence and
locoregional recurrence were not collected. ECOG=Eastern Cooperative Oncology
tumours in the nasopharynx). Pack-year information was
Group. HPV=human papillomavirus. missing for 19 (37%) of 51 patients who ever used tobacco
because of missing or incomplete number of cigarettes
Table 1: Baseline and demographic characteristics
smoked per day or missing or incomplete number of

960 www.thelancet.com/oncology Vol 17 July 2016

 Articles

years smoked. 56 patients (23 [100%] of 23 HPV-positive of 36). For patients who had had previous radiotherapy,
and 33 [89%] of 37 HPV-negative) reported previous data as to whether responses were observed within or
radiotherapy. At the cutoff date, 49 (82%) patients outside the previously irradiated field were not available
discontinued treatment because of progressive disease for all patients. However, responses generally encom-
(35), adverse events (eight), withdrawal by patient (three), passed all tumour sites.
death (two), or withdrawal by physician (one; figure 1). Overall, 26 (51%) of the 51 patients who had a baseline
The overall proportion of patients with drug-related scan and a post-baseline scan experienced any reduction
adverse events of any grade was 63% (n=38), with the most in tumour burden, per investigator review (figure 2A).
common events being fatigue, pruritus, nausea, decreased 12 patients who were HPV-positive and 14 patients who
appetite, and rash (table 2). Ten (17%) of 60 patients had were HPV-negative had some reduction in tumour size
grade 3 drug-related adverse events, which included (figure 2A). At the cutoff date, 11 patients, including
increased alanine aminotransferase, increased aspartate three with stable disease, remained on treatment
aminotransferase, hyponatraemia, fatigue, rash, atrial (figure 2B). Individual changes in tumour size from
fibrillation, congestive heart failure, diarrhoea, lympho-
penia, musculoskeletal pain, and neck abscess (table 2).
Grade Grade 3 Grade 4
The proportion of HPV-positive patients with a drug-related 1 or 2
adverse event of any grade was 16 (70%) of 23 patients and
Fatigue 12 (20%) 1 (2%) 0
of grade 3 was five (22%). The proportion of HPV-negative
Pruritus 7 (12%) 0 0
patients with a drug-related adverse event of any grade was
Rash 3 (5%) 1 (2%) 0
22 (59%) of 37 patients and of grade 3 was five (14%). Nine
Lymphopenia 0 1 (2%) 0
patients had prespecified adverse events of special
Atrial fibrillation 0 1 (2%) 0
immunological interest regardless of whether they were
Congestive cardiac failure 0 1 (2%) 0
deemed to be related to treatment: four (7%) patients had
hypothyroidism, one patient had grade 1 and three patients Diarrhoea 1 (2%) 1 (2%) 0

had grade 2; two (3%) had adrenal insufficiency, one patient Neck abscess 0 1 (2%) 0
had grade 1 and one patient had grade 2; and one patient Alanine aminotransferase increase 0 2 (3%) 0
each (2%) had grade 1 hyperthyroidism, grade 2 myositis, Aspartate aminotransferase increase 0 2 (3%) 0
and grade 3 rash. 12 (20%) of 60 patients had a treatment Hyponatraemia 0 2 (3%) 0
interruption because of an adverse event, five (8%) patients Musculoskeletal pain 1 (2%) 1 (2%) 0
had pembrolizumab withdrawn because of an adverse Data are n (%) for n=60 patients.
event, three (5%) patients experienced a treatment
interruption and subsequently withdrew because of an Table 2: Treatment-related adverse events
adverse event, and one (2%) patient experienced a treatment
interruption and subsequent reduction in dosing frequency
(every 2 weeks to every 3 weeks) because of a non-treatment- Total HPV-positive HPV-negative
(%; 95% CI*) n (%; 95% CI*) n (%; 95% CI*)
related grade 3 stridor. Neither of the two deaths that
occurred during the study were deemed to be related to Central review (n=45)† (n=16) (n=29)
study treatment (one due to cardiac arrest and one due to Overall response 8 (18%; 8–32) 4 (25%; 7–52) 4 (14%; 4–32)
malignant neoplastic progression). Complete response 1 (2%; <1–12) 1 (6%; <1–30) 0 (0%; 0–12)
The proportion of patients with an overall response Partial response 7 (16%; 7–30) 3 (19%; 4–46) 4 (14%; 4–32)
(confirmed per central imaging review) was 18% (8 of 45; Stable disease 8 (18%; 8–32) 3 (19%; 4–46) 5 (17%; 6–36)
95% CI 8–32; table 3). With an overall response Progressive disease 25 (56%; 40–70) 8 (50%; 25–75) 17 (59%; 39–77)
(unconfirmed per investigator review) of 21% (12 of 56; Data unavailable 3 (7%; 1–18) 1 (6%; <1–30) 2 (7%; <1–23)
12–34), a consistency of effect between proportion of Images not assessable 1 (2%; <1–12) ·· 1 (3%; <1–18)
patients with an overall response assessed by central review Investigator review (n=56)† (n=20) (n=36)
and investigator review was observed (table 3). Several Overall response 12 (21%; 12–34) 5 (25%; 9–49) 7 (19%; 8–36)
patients who were clinically stable or improving continued Complete response 1 (2%; 0–10) 1 (5%; <1–25) 0 (0–10)
the study treatment beyond progressive disease. The patient Partial response 11 (20%; 10–32) 4 (20%; 6–44) 7 (19%; 8–36)
with the complete response had initial resolution of distant Stable disease 15 (27%; 16–40) 7 (35%; 15–59) 8 (22%; 10–39)
metastatic lesions (lung metastases) and subsequently, Progressive disease 25 (45%; 31–59) 7 (35%; 15–59) 18 (50%; 33–67)
resolution of large primary tumour as well. This patient Data unavailable 4 (7%; 2–17) 1 (5%; 0·1–25) 3 (8%; 2–23)
also had an initial tumour flare (ie, increase in tumour
Confirmed responses per RECIST version 1.1 per central review. Confirmed and unconfirmed responses per modified
size) followed by response that became a complete
RECIST per investigator assessment. RECIST=Response Evaluation Criteria In Solid Tumors. *Based on binomial exact
response and was the only case of an atypical response. confidence interval method. †Patients who received at least one dose of pembrolizumab and had baseline and at least
Per investigator review, the proportion of patients one post-baseline tumour measurement.
with an overall response was higher in HPV-positive
Table 3: Antitumour activity of pembrolizumab
(five [25%] of 20) than HPV-negative patients (seven [19%]

www.thelancet.com/oncology Vol 17 July 2016 961

 Articles

A baseline show reduced tumour burden over multiple

100 HPV-positive assessments as well as durability of the responses for up
HPV-negative
80
to 68 weeks (figure 2C). Median time to response was
8 weeks (95% CI 7–17) and the median duration of
Change from baseline in sum of longest diameter

60 response was 53 weeks (13 to not reached). Of the

40 ten patients with a response to pembrolizumab by
investigator assessment, two who had a partial response
of target lesion (%)

20 20% increase discontinued treatment because of an adverse event, and

0 eight patients had responses that were ongoing.
Median progression-free survival was 2 months (95% CI
–20
2–4) in the full analysis set population (n=56), 4 months
–30% decrease
–40 (2–10) in HPV-positive patients, and 2 months (2–4) in
–60 HPV-negative patients (figure 3A). Median overall survival
was 13 months (95% CI 5 to not reached) in the intention-
–80
to-treat population (n=61), not reached for HPV-positive
–100 patients (8 to not reached), and 8 months (4 to not reached)
for HPV-negative patients (figure 3B). Patient survival at
12 months is 51% (figure 3B). The median progression-
B
free survival in the 12 patients with unconfirmed response
by investigator assessment was approximately 14 months
(5 to not reached). The median overall survival in this
responder subgroup was not reached.
PD-L1 expression levels and presence of stromal
staining showed associations with both best overall
response (p=0·010) and progression-free survival
(p=0·020, mean intensity scoring; appendix p 6). The
area under the receiver operating characteristics (ROC)
Patients

curve was 0·65 for immunohistochemistry mean

intensity scoring, and the proportion of patients with a
HPV-positive
HPV-negative response above and below the Youden Index21 threshold
Treatment ongoing were 50% and 14%, respectively.
Complete response
Partial response
All six interferon-γ-related genes had significantly
Progressive disease as best higher mean expression values in responder versus
response non-responder patients (figure 4). Using the composite
Progressive disease after
complete response, partial score, a significant association with best overall response
response, or stable disease (p=0·0047) and progression-free survival (p=0·0009) was
Last pembrolizumab dose
observed using logistic or Cox regression models. Area
0 20 40 60 80 under the ROC curve for the interferon-γ composite
Weeks score was 0·80 (95% CI 0·61–0·95), with a 40% response
C
above the Youden Index threshold (and thus had a 40%
100 HPV-positive
HPV-negative positive predictive value) and a 5% response below it
80 (and thus had a 95% negative predictive value).
60
The proportional hazards assumption was checked for
both the immunohistochemistry and gene expression
40 Cox regressions and no evidence of a violation was found.
Change from baseline (%)

20 20% increase

–20 Figure 2: Efficacy of pembrolizumab based on RECIST assessed by

–30% decrease investigator review
–40 (A) The maximum percentage reduction from baseline in target lesions. Includes
only patients with RECIST measurable lesions at baseline and at least
–60 one complete follow-up scan of target lesions (n=51). (B) Treatment exposure
and response duration. Four patients did not have a valid post-baseline
–80
assessment so were not included in the population. There is one additional
patient here than in part A because there was an additional patient who had
–100
0 10 20 30 40 50 60 70 progressive disease as a best response who did not have all target lesions
Weeks measured post-baseline. (C) Change from baseline in tumour size.
RECIST=Response Evaluation Criteria In Solid Tumors.

962 www.thelancet.com/oncology Vol 17 July 2016

 Articles

Discussion A
Pembrolizumab showed a manageable safety profile 100 Overall population
and promising antitumour activity in patients with HPV-positive
PD-L1-positive recurrent or metastatic squamous cell 90 HPV-negative

carcinoma of the head and neck. Present treatment options

for advanced squamous cell carcinoma of the head and 80

neck are limited. To our knowledge, this study is the first to

70
report the efficacy and safety of an anti-PD-1 antibody in

Progression-free survival (%)

patients with advanced PD-L1-positive recurrent or
60
metastatic squamous cell carcinoma of the head and neck.
To our knowledge, this is also the first study to present 50
clinical results showing the effectiveness of immuno-
therapy for recurrent or metastatic squamous cell 40
carcinoma of the head and neck, paving the way for future
studies of immune-modulating drugs in squamous cell 30
carcinoma of the head and neck.
Pembrolizumab monotherapy showed substantial and 20

clinically significant antitumour activity in patients with

10
heavily pretreated recurrent or metastatic squamous cell
carcinoma of the head and neck, with 18% of patients
0
achieving an overall response by central review, and 21% of 0 2 4 6 8 10 12 14
patients achieving an overall response by investigator Number at risk
Overall population 56 29 20 16 15 12 9 3
assessment. Overall survival was 13 months (95% CI 5 to HPV-positive 20 13 8 7 7 5 4 1
not reached) and duration of response was approximately HPV-negative 36 16 12 9 8 7 5 2
53 weeks (12·2 months). Although cross-trial comparisons B
100
must be made with caution, this level of antitumour
activity and the duration of response compares favourably 90
with single-drug cetuximab (proportion of patients with an
overall response, 13%; duration of response, 4 months).22 80
Despite the fact that 70% of patients in this study had
received two or more lines of treatment, survival was 70
similar to the median overall survival of treatment-naive
Overall survivial (%)

60
recurrent or metastatic patients treated with doublet
chemotherapy plus cetuximab.3 Although the patients 50
included in this study are generally consistent with those
typically enrolled in similar studies for the population 40
under investigation, this trial had a higher proportion of
patients with HPV-associated disease, a group now 30
recognised to have longer survival after recurrence.23
20
Because the incidence of HPV-positive cancer has been
rising steadily in recent decades, we believe that the study 10
population reflects the rapidly changing mix of the disease
in the US coastal areas and represents the dominant 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13
demographics seen at the centres that accrued patients. Months
Because tumour inflammation and PD-L1 expression are Number at risk
Overall population 60 59 54 49 42 38 33 32 31 28 26 25 22 15
present to a higher degree in HPV-positive tumours,13 it HPV-positive 23 23 22 21 20 17 16 16 15 14 13 13 11 7
could be expected that HPV-positive and HPV-negative HPV-negative 37 36 32 28 22 21 17 16 16 14 13 12 11 8
patients might derive different benefit from
Figure 3: Kaplan-Meier estimates of progression-free survival (A) and overall survival (B)
pembrolizumab. In this study, pembrolizumab was
Progression-free survival was analysed using patients who received at least one dose of pembrolizumab, had
efficacious and the proportions of patients with an overall measurable disease at baseline, and at least one post-baseline scan, or discontinued because of progressive disease
response were similar irrespective of HPV status. or drug-related adverse events. Overall survival was assessed using the intention-to-treat population.
However, one limitation of this analysis was the use of
p16 immunohistochemistry as a surrogate for HPV the present study, there were three patients who had
infection. Despite its common use to determine HPV non-oropharynx HPV-positive disease.
status and high positive and negative predictive values in Pseudo-progression or tumour flares and delayed
oropharyngeal tumours, p16 is a poor surrogate for HPV responses are uncommon with PD-1 inhibitors in head
infection in non-oropharyngeal tumours.24 However, in and neck cancers and most solid tumours.25,26 One patient

www.thelancet.com/oncology Vol 17 July 2016 963

 Articles

A STAT1 p=0·007 HLA.DRA p=0·008 IFNG p=0·011

B Responder status
4
Normalised expression level

3 2·5

Normalised expression level

1

0 2·0

IDO1 p=0·039 CXCL9 p=0·014 CXCL10 p=0·014

4
Normalised expression level

2 1·5

0 0
Non− Responder Non− Responder Non− Responder Non−responder Responder
responder responder responder

Responder status Responder status

Figure 4: Dotplot of distribution of six individual interferon-γ-related genes (A) and the total composite score (B) by response status
Responders were defined as complete responders or partial responders (not stable disease) by RECIST by investigator review. RECIST=Response Evaluation Criteria In
Solid Tumors.

had a tumour flare or pseudo-response followed by a However, the small number of events restricts the
response in this study. Additionally, atypical response interpretation of these findings, and their clinical
patterns were visible in some patients; patients with usefulness remains the subject of further study. It should
stable disease tended to have undulating tumour be noted that the prototype assay did not allow for a
burdens, some of which might have been due to technical distinction between predictability based on tumour
or measurement inaccuracy, but might also reflect versus inflammatory cell expression, and the small
tumour and host interaction. number of patients included based on stromal or
Pembrolizumab was safe and well tolerated, with immune cell-only staining prohibited such an analysis.
38 (63%) of patients experiencing treatment-related The six interferon-γ-related genes are associated with
adverse events, most commonly grade 1–2 pruritus, an inflamed tumour phenotype19 and diverse immuno-
fatigue, or rash that were transient. Grade 3–5 treatment- regulatory activities, including immune tolerance.27,28
related adverse events occurred in 17% of patients. The method used to determine expression of the gene
This toxicity profile was generally consistent with safety signature (quantification of RNA from FFPE tissue) is
data reported for other tumour types,15–17 except that no highly robust and reproducible.29 Significant associations
cases of pneumonitis or colitis requiring intervention with best overall response and progression-free survival
occurred in this cohort. No severe toxic effects specific to were noted for the composite score of these genes in
squamous cell carcinoma of the head and neck were noted. squamous cell carcinoma of the head and neck and
Given that survival data with pembrolizumab were on provide the rationale to develop a predictive biomarker
par with those for first-line combination therapy,3 future gene expression signature. The gene expression
studies to assess the efficacy of pembrolizumab as composite score provided a positive predictive value of
first-line therapy are warranted. Standard therapies may 40% and a negative predictive value of 95%, potentially
alter the immune environment of squamous cell permitting identification of those patients who will not
carcinoma of the head and neck, generating conditions benefit from pembrolizumab treatment. Notably, the
favouring response to pembrolizumab, and trials of proportion of patients with an overall response in the
pembrolizumab in combination or in sequence with biomarker-selected population was nearly double that
chemotherapy or radiotherapy are also warranted. seen in the trial overall. Because the genes in the signature
Validation of the predictive biomarkers identified here or are highly correlated in their expression, one would not
in other studies will facilitate selection of subpopulations be able to show only a single gene as being the dominant
of patients with squamous cell carcinoma of the head predictive factor statistically. Although all genes exhibit
and neck who are most likely to benefit from different differences between responders and non-responders, use
treatment combinations. of a single gene would be a less robust approach and
The degree of PD-L1 expression by immuno- would probably be more prone to technical error or
histochemistry was found to be predictive of best overall biological variation. If confirmed in a larger validation
response and improved progression-free survival. cohort using a clinical-grade CLIA assay, this multigene

964 www.thelancet.com/oncology Vol 17 July 2016

 Articles

analysis will have substantial clinical application as the 6 Keck MK, Zuo Z, Khattri A, et al. Integrative analysis of head and
high negative predictive value of this signature allows the neck cancer identifies two biologically distinct HPV and
three non-HPV subtypes. Clin Cancer Res 2015; 21: 870–81.
selection and potential exclusion of patients from 7 Keir ME, Butte MJ, Freeman GJ, Sharpe AH. PD-1 and its ligands in
ineffective therapy. The requirement for PD-L1 positivity tolerance and immunity. Annu Rev Immunol 2008; 26: 677–704.
for study enrolment precludes exploration of a possible 8 Carter LL, Fouser LA, Jussif J, et al. PD-1:PD-L inhibitory pathway
affects both CD4+ and CD8+ T cells and is overcome by IL-2.
interaction effect between PD-L1 and the interferon-γ Eur J Immunol 2002; 32: 634–43.
signature. Future studies are necessary to characterise the 9 Freeman GJ, Long AJ, Iwai Y, et al. Engagement of the PD-1
biology of non-inflamed tumours, to elucidate whether immunoinhibitory receptor by a novel B7 family member leads to
negative regulation of lymphocyte activation. J Exp Med 2000;
tumour inflammation can be induced, and to determine 192: 1027–34.
if there is an interaction effect between PD-L1 and the 10 Latchman Y, Wood CR, Chernova T, et al. PD-L2 is a second ligand
interferon-γ signature. for PD-1 and inhibits T cell activation. Nat Rev Immunol
In summary, pembrolizumab showed clinically 2001; 2: 261–68.
11 Pardoll DM. The blockade of immune checkpoints in cancer
significant activity in patients with heavily pretreated immunotherapy. Nat Rev Cancer 2012; 12: 252–64.
squamous cell carcinoma of the head and neck irrespective 12 McDermott DF, Atkins MB. PD-1 as a potential target in cancer
of HPV status. Greater antitumour activity was recorded in therapy. Cancer Med 2013; 2: 662–73.
patients with squamous cell carcinoma tumours of the 13 Lyford-Pike S, Peng S, Young GD, et al. Evidence for a role of the
PD-1:PD-L1 pathway in immune resistance of HPV-associated head
head and neck that expressed higher levels of PD-L1 and and neck squamous cell carcinoma. Cancer Res 2013; 73: 1733–41.
interferon-γ-related genes. Thus, pembrolizumab might 14 Saloura V, Zou Z, Koeppen H. Correlation of T-cell inflamed
represent a new treatment approach for patients with phenotype with mesenchymal subtype, expression of PD-L1, and
other immune checkpoints in head and neck cancer.
squamous cell carcinoma of the head and neck. Proc Soc Am Clin Oncol 2014; 32 (suppl 5): 6009 (abstr).
Contributors 15 Hamid O, Robert C, Daud A, et al. Safety and tumor responses with
TYS, BB, JDC, and LQC contributed to the study conception, design, or lambrolizumab (anti-PD-1) in melanoma. N Engl J Med 2013;
planning. TYS, BB, RM, RB, JW, JPE, KH, TMC, and LQC gathered data. 369: 134–44.
TYS, BB, JW, TMC, JL, CG, and JDC did or supervised data analysis. 16 Robert C, Ribas A, Wolchok JD, et al. Anti-programmed-death-
TYS, BB, RM, JW, JL, CG, JDC, and LQC interpreted the data. TYS, BB, receptor-1 treatment with pembrolizumab in ipilimumab-refractory
advanced melanoma: a randomised dose-comparison cohort of a
RM, JW, JPE, and LQC provided study materials or patients. TYS, JL,
phase 1 trial. Lancet 2014; 384: 1109–17.
and LQC wrote sections of the initial manuscript. All authors critically
17 Garon EB, Rizvi NA, Hui R, et al. Pembrolizumab for the treatment
reviewed iterations of the manuscript and approved the final draft for
of non-small-cell lung cancer. N Engl J Med 2015; 372: 2018–28.
submission.
18 Topalian SL, Hodi FS, Brahmer JR, et al. Safety, activity, and
Declaration of interests immune correlates of anti-PD-1 antibody in cancer. N Engl J Med
TYS has served as a consultant for Merck Sharp & Dohme, Merck Serono, 2012; 366: 2443–54.
Amgen, and Bristol-Myers Squibb. BB has received grant support and 19 Ribas A, Robert C, Hodi S, et al. Association of response to
personal fees from Merck Sharp & Dohme. RM has received personal fees programmed death receptor 1 (PD-1) blockade with pembrolizumab
from Bristol-Myers Squibb, Bayer, Novartis, and Genentech. JPE has (MK-3475) with an interferon-inflammatory immune gene
received personal fees from Merck Sharp & Dohme. JW has received signature. Proc Soc Am Clin Oncol 2015; 33 (suppl): 3001 (abstr).
grant support from Merck Sharp and Dohme. KH, TMC, JL, CG, and 20 Oken MM, Creech RH, Tormey DC, et al. Toxicity and response
JDC are employees of Merck & Co. LQC has received personal fees for criteria of the Eastern Cooperative Oncology Group.
serving as an advisory board member from Merck Sharp & Dohme. Am J Clin Oncol 1982; 5: 649–55.
The other authors declare no competing interests. 21 Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32–35.
22 Vermorken JB, Trigo J, Hitt R, et al. Open-label, uncontrolled,
Acknowledgments multicenter phase II study to evaluate the efficacy and toxicity of
Funding for this research was provided by Merck & Co, Inc, Kenilworth, cetuximab as a single agent in patients with recurrent and/or
NJ, USA. The authors thank the patients and their families and metastatic squamous cell carcinoma of the head and neck who
caregivers for participating in the study. The authors also thank failed to respond to platinum-based therapy. J Clin Oncol 2007;
Mark Ayers, Michael Nebozhyn, Erin Murphy, and Jennifer Yearley for 25: 2171–77.
their contributions to the NanoString analysis, and QualTek Molecular 23 Argiris A, Li S, Ghebremichael M, et al. Prognostic significance of
Laboratories (Newtown, PA, USA) for development of the PD-L1 human papillomavirus in recurrent or metastatic head and neck
immunohistochemistry assay. Medical writing assistance was provided cancer: an analysis of Eastern Cooperative Oncology Group trials.
by Matthew Grzywacz, PhD, of ApotheCom, Yardley, PA, USA. Ann Oncol 2014; 25: 1410–16.
This assistance was funded by Merck & Co, Inc, Kenilworth, NJ, USA. 24 Seiwert TY. Ties that bind: p16 as a prognostic biomarker and the
need for high-accuracy human papillomavirus testing. J Clin Oncol
References 2014; 32: 3914–16.
1 International Agency for Research on Cancer. GLOBOCAN 25 Burtness B. Moving forward in the management of squamous cell
2012: estimated cancer incidence, mortality, and prevalence carcinoma of the head and neck: promising immuno-oncology
worldwide in 2012. http://globocan.iarc.fr (accessed Jan 13, 2016). approaches. Am J Hematol Oncol 2015; 11: 28–31.
2 Seiwert TY, Salama JK, Vokes EE. The chemoradiation paradigm in 26 Chiou VL, Burotto M. Pseudoprogression and immune-related
head and neck cancer. Nat Clin Pract Oncol 2007; 4: 156–71. response in solid tumors. J Clin Oncol 2015; 33: 3541–43.
3 Vermorken JB, Mesia R, Rivera F, et al. Platinum-based 27 Vacchelli E, Aranda F, Eggermont A, et al. Trial watch: IDO
chemotherapy plus cetuximab in head and neck cancer. inhibitors in cancer therapy. Oncoimmunology 2014; 3: e957994.
N Engl J Med 2008; 359: 1116–27.
28 Spivey TL, Uccellini L, Ascierto ML, et al. Gene expression profiling
4 Seiwert TY, Zuo Z, Keck MK, et al. Integrative and comparative in acute allograft rejection: challenging the immunologic constant
genomic analysis of HPV-positive and HPV-negative head and neck of rejection hypothesis. J Transl Med 2011; 9: 174.
squamous cell carcinomas. Clin Cancer Res 2015; 21: 632–41.
29 Geiss GK, Bumgarner RE, Birditt B, et al. Direct mulitplexed
5 The Cancer Genome Atlas Network. Comprehensive genomic measurement of gene expression with color-coded probe pairs.
characterization of head and neck squamous cell carcinomas. Nat Biotechnol 2008; 26: 317.
Nature 2015; 517: 576–82.

www.thelancet.com/oncology Vol 17 July 2016 965