SUPREME-HN A Retrospective Cohort Study of PD-L1 in Recurrent and Metastatic Squamous Cell Carcinoma of Head and Neck (SCCHN)

Background/Rationale:

Programmed cell death protein 1 (PD-1) is an immune inhibitory receptor that interacts with
two ligands, programmed death ligand 1 (PD-L1) and ligand 2 (PD-L2). PD-1 pathway is a major
immune checkpoint which has been implicated in adaptive immune resistance of squamous cell
carcinoma of the head and neck (SCCHN) tumors, particularly in those associated with human
papillomavirus (HPV) infection. Based on an internal analysis performed by AstraZeneca (AZ),
it is reported that approximately 25% of cases of SCCHN express PD-L1. Tumoral PD-L1
expression status correlates closely with response to anti-PD-1/anti PD-L1 antibodies
Durvalumab (MEDI4736) is an immunoglobulin G1 kappa monoclonal antibody with high affinity
and selectivity for PD-L1 and no binding to PD-L2, which is in development for the treatment
of patients with recurrent or metastatic SCCHN.

This non-interventional study (NIS) aims to generate and provide data on the prognostic value
of PD-L1 status in patients with recurrent/metastatic SCCHN.

Objectives and Hypotheses:

Primary objective:

To determine the prognostic value of PD-L1 status in terms of overall survival (OS) in
patients with recurrent/metastatic SCCHN. Overall survival after diagnosis of
recurrent/metastatic SCCHN will be assessed in PD-L1 positive and PD-L1 negative patients and
in predefined sub-groups (e.g., HPV status, HIV status, smoking history, heavy alcohol use,
anatomical sub-site of primary tumor, and prior exposure to radiation therapy).

Secondary objectives:

Secondary objectives are to perform the following in a) all patients meeting the eligibility
criteria, b) all patients per availability of tumor sample (not available, available at any
setting), and c) all patients with available tumor sample at any time, at the time of first
line therapy, or at the second line therapy per PD-L1 status (positive, negative, unknown/not
done):

1. To describe the relevant demographic and clinical characteristics of patients,
stratified by PD-L1 status

2. To describe first line treatment choices, and where available, subsequent treatment
choices

3. To describe investigator-assessed tumor response for first line and second line of
therapy (if any), including: best response, duration of response where applicable, and
objective response rate (ORR)

4. To describe investigator-assessed progression-free survival (PFS) for first line and
second line of therapy (if any).

Exploratory objectives:

1. To perform additional biomarker research on tumor samples (depending on tissue
availability and volume).

2. If feasible, to assess agreement of PD-L1 status in samples obtained at different time
points (e.g., before and after chemotherapy or radiation therapy).

Methods:

Study design:

This is a retrospective international, multi-center, non-interventional cohort study based on
use of data derived from established medical records and secondary analysis of archival tumor
samples. The study will collect data on patient and tumor characteristics, PD-L1 status,
patterns of treatment, and clinical outcomes, in up to 600 adult patients with
recurrent/metastatic SCCHN.

For patient selection, the date of diagnosis of recurrent/metastatic disease will be used as
the index date. The patient selection period extends from the 1st March 2011 to the 30th June
2015. This allows the inclusion of patients with tumor samples of approximately ≤ 5 years
age, and ensures approximately 10 months follow-up for living patients recruited at last day
of the enrolment window. All patients with a diagnosis of recurrent/metastatic squamous cell
carcinoma (SCC) of oral cavity (tongue, gum, floor of mouth, and other/unspecified part of
the mouth), oropharynx, hypopharynx, or larynx during that period will be considered for
inclusion in the study. Patients will be identified and followed up through their medical
records until death or end of data collection (approximately Q4 2016).

Patients' demographic and clinical characteristics, and medical history will be described.
Clinical outcomes including PFS, best response, duration of response, and ORR will be
described for the first and second lines of therapy(if any), and OS will be collected. A
mandatory archived tumor sample will be used to determine PD-L1 status. If a patient has more
than one suitable tissue sample, the most recent sample will be used as the mandatory tissue
sample. Where available, additional tumor samples obtained at any other time points of the
disease will be also collected (optional).

Statistical analyses will be performed for the whole cohort, per PD-L1 status and for
predefined subgroups.

Data Source(s):

The study will be implemented in a total of approximately 20 sites in the US, Asia and
Europe. Data will be abstracted from patient's hospital medical records until death, or end
of data collection. Archival tumor samples will be retrieved and tested by
immunohistochemistry to assess the PD-L1 status of the tumor. Where there is sufficient
tissue quantity, the expression of other biomarkers will also be evaluated.

Study Population:

Unselected patient population representative of patients who may receive MEDI4736 in the real
life setting.

Exposure(s): The primary variable is the PD-L1 status (positive or negative) in the overall
patient population. Archived tissue samples, including samples from the primary site, lymph
nodes or distant metastatic sites, will be used to determine PD-L1 status. PD-L1 status will
be determined by immunohistochemistry (IHC) using a validated assay.

Outcome(s):

Overall survival will be assessed. Additionally, the following clinical outcomes will be
described for first line and second line treatments for recurrent and/or metastatic disease:
investigator-assessed best tumor response, duration of response where applicable, and
investigator-assessed PFS. ORR (sum of complete and partial responses) will be derived from
investigator-assessed best responses and described for first line and second line therapies
(if any).

Patient characteristics, disease characteristics, and patterns of treatment will be captured
and described.

Sample Size Estimation:

The primary objective of this study is to estimate the prognostic value of PD-L1 status in
terms of OS in patients with recurrent/metastatic SCCHN. The available sample size is not
known a priori and will be driven by the number of patients at selected sites with available
tissue. However, assuming that PD-L1 status distribution is approximately 25% positive and
75% negative status, that the median OS in these patients is 10 months, that the study will
accrue uniformly over 52 months with 10 months follow-up from the last patient entering and
that survival times are exponentially distributed, the illustrations of hazard ratios it
would be possible to detect with 80% power (2-sided alpha 0.05) over various sample sizes are
given in the table below.

HR to detect Number of deaths Number of patients (Total) 0.3 30 40 0.4 51 68 0.5 74 112 0.6
136 196 0.7 278 396 HR: Hazard ratio These illustrations do not take account of the fact that
the binary prognostic factor of interest (PD-L1 status) may be correlated with other
covariates and hence could over-estimate the power of the various sample sizes (Bernardo et
al, 2000). The sample size (number of events) required when no correlation is assumed should
be multiplied by 1/(1-R2) to account for an R2 > 0 (Hsieh and Lavori, 2000). The correlation
is not currently known. Additionally, a proportion of patients will not have PD-L1 status
available. Given these uncertainties, a sample size of up to 600 patients is felt to be
adequate.

Statistical Analysis:

Descriptive analyses on patient characteristics, treatment choices and treatment outcomes
will be conducted.

Time to event data (OS, PFS) including rates of affected patients will be assessed described
using the Kaplan-Meier method. Two-sided 95% CIs will be provided for the main statistical
estimators.

OS and PD-L1 status will be described in subgroups, including but not limited to the
following:

- Per anatomical sub-site of primary tumor

- Per HPV status

- Per HIV status

- Per smoking history

- Per alcohol consumption history

- Per prior exposure to radiation therapy

- Per types of treatment regimens

- By performance status at diagnosis of recurrent/metastatic disease. The prognostic value
of PD-L1 status will be investigated using a Cox proportional hazards model. Additional
covariates to be included in the model will be described in the statistical analysis
plan.

Inclusion criteria:

1. Provision of subject informed consent (or consent from next of kin/legal representative
if applicable) for use of the data and retrieval of tumor sample, according to local
regulations 2. Adult patient (≥ 18 years old) 3. Patient with histologically confirmed
SCCHN of oral cavity (tongue, gum, floor of mouth, other/unspecified part of the mouth),
oropharynx, hypopharynx or larynx 4. Patient with recurrent or metastatic SCCHN diagnosed
between 01 March 2011 and 30 June 2015 5. Mandatory archival tissue sample (most recent)
from the primary site, a lymph node or a distant metastatic site:

- Tissue sample less than 5 years old (compared to date of retrieval) if provided as
complete block (preferred option) or as section cut within 60 days of shipment from
site prior to testing 6. Optional archival tissue samples taken at other time points
of the disease from the primary site, a lymph node or a distant metastatic site (where
available):

- Tissue sample less than 5 years old (compared to date of retrieval) if provided as
complete block (preferred option) or as section cut within 60 days of shipment from
site prior to testing.

Exclusion criterion:

1. Treatment for SCCHN with anti-CTLA-4, anti-PD-1, anti-PD-L1, or anti-PD-L2 antibodies,
or any other antibody with known immunomodulatory effect.