Critical Reviews in Oncology/Hematology 100 (2016) 8898

Contents lists available at ScienceDirect

Critical Reviews in Oncology/Hematology

journal homepage: www.elsevier.com/locate/critrevonc

PD-L1 expression in cancer patients receiving anti PD-1/PD-L1

antibodies: A systematic review and meta-analysis
Sara Gandini a , Daniela Massi b , Mario Mandal c,
a
b
c

Division of Epidemiology and Biostatistics, European Institute of Oncology, Milan, Italy

Division of Pathological Anatomy, Department of Surgery and Translational Medicine, University of Florence, Italy
Unit of Medical Oncology, Department of Oncology and Hematology, Papa Giovanni XXIII Hospital, Bergamo, Italy

Contents
1.
2.

3.

4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Meta-analysis methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.1.
Literature search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Statistical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.2.
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.1.
Description of literature search and data from included studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.2.
Summary estimates of clinical objective response rates (SORR) in PD-L1 positive vs. PD-L1 negative patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Summary estimates for mortality in PD-L1 positive vs. PD-L1 negative melanoma patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.3.
3.4.
Sensitivity analysis assessing the impact of PD-L1 positivity cut-off on estimates of clinical objective responses rates . . . . . . . . . . . . . . . . . . . . . . . 92
Between-study heterogeneity analysis of clinical objective response according to line of treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.5.
3.6.
Sensitivity analysis of summary estimates in patients receiving anti PD-1 therapy (excluding anti PD-L1 trials) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Appendix A.
Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

a r t i c l e

i n f o

Article history:
Received 1 December 2015
Received in revised form 12 January 2016
Accepted 8 February 2016
Keywords:
PD-L1 overexpression
Melanoma
Anti-PD-1/PD-L1 antibodies
Meta-analysis
Clinical trials

a b s t r a c t
Background: Despite the success of immunotherapy directed at inhibiting of programmed death-1 (PD1)/PD-ligand (L)1 signaling, it is not established whether PD-L1 expression correlates with the clinical
response and outcome in different tumors. The present meta-analysis investigates whether the PD-L1
status, detected by immunohistochemistry, is associated with clinical response and mortality in patients
treated with anti-PD-1/PD-L1 therapy.
Methods: A systematic literature search and quantitative analysis were planned, conducted and reported
following CONSORT and QUORUM checklists, up to December 2015, to identify clinical trials with information on cancer outcome by PD-L1 immunohistochemical expression in tumor tissues. We used random
effects models to estimate Summary Objective Response Rates (SORRs) and Summary Odd Ratio (SOR)
for the comparison of PD-L1 positive and negative patients.
Results: We summarized 20 trials carried out in metastatic melanoma (MM), non-small cell lung
cancer (NSCLC), and renal cell carcinoma (RCC) patients receiving anti-PD-1/PD-L1 antibodies
(4230 MM, 1417 NSCLC and 312 RCC patients). Positive PD-L1 MM patients showed a signicant decrease (53%) in the risk of mortality vs. negative cases with no heterogeneity. Furthermore,
SORRs were 45% and 27% in PD-L1 positive and negative patients, respectively, and SOR indicates
a signicant difference in term of responses: 2.14 (95% CI: 1.65, 2.77), with low between-study
heterogeneity (I2 = 35%). Furthermore, results from randomized clinical trials on MM showed that
PD-L1 expression is signicantly associated with greater clinical response rates to anti-PD1 treatments (SOR 1.89; 95%CI: 1.35, 2.64) but not to other treatments (SOR 0.96; 95%CI: 0.5, 1.87).

Corresponding author at: Unit of Medical Oncology, Department of Oncology and Haematology, Papa Giovanni XXIII Cancer Center Hospital, Piazza OMS 1, 24100 Bergamo,
Italy.
E-mail address: mariomandala@tin.it (M. Mandal).
http://dx.doi.org/10.1016/j.critrevonc.2016.02.001
1040-8428/ 2016 Elsevier Ireland Ltd. All rights reserved.

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

89

In non-squamous NSCLC SORRs were 29% and 11% in PD-L1 positive and negative patients, respectively,
and SOR indicates a signicant difference between responses: 3.78 (1.54, 9.24), with no between-study
heterogeneity. Squamous NSCLC and RCC did not show any signicant difference in response according
to the PD-L1 status.
Conclusion: PD-L1 expression is signicantly associated with mortality and clinical response to anti-PD1/PD-L1 antibodies in MM patients and with clinical response in patients with non-squamous NSCLC.
2016 Elsevier Ireland Ltd. All rights reserved.

1. Introduction
A dynamic relationship exists between host and tumor, and the
ability of the tumor to evade immune recognition often determines the clinical course of the disease (Shin and Ribas, 2015).
Multiple mechanisms of immune suppression are known to prevent effective antitumor immunity, including increased secretion
of immunosuppressive cytokines, i.e., interleukin (IL)-10 and tumor
growth factor (TGF)-, reduced expression of major histocompatibility antigens on cell surface, enhanced differentiation of
immune effector cells to a regulatory phenotype, as well as an
inux of myeloid-derived suppressor cells and tumor associated
macrophages (Pardoll, 2012). Receptors expressed in the plasma
membrane of immune cells, by ne tuning cellular activation or
inhibition, are key regulators of infections, autoimmunity and
cancer. While some of these receptors exert positive activities,
others produce detrimental effects. Antibodies, which target regulatory proteins, either inhibiting repressor receptors or stimulating
of activating receptors, represent potential strategies to enhance
immune responses, and ameliorate cancer outcome (Pardoll, 2012).
The use of immunomodulatory monoclonal antibodies that
directly enhance the anti-tumor immune response by T cells, or
block immunologic checkpoints that would otherwise restrain
effective anti-tumor immunity, have boosted signicant enthusiasm in cancer treatment (Postow et al., 2015a). Cancer development
is facilitated by dis-regulation of physiological pathways that,
under normal circumstances, down-regulate immune activation
and maintain tolerance to self. Among these pathways an important
contribution is given by the programmed death-1 (PD-1)/PD-ligand
(L) 1 axis.
PD-1 is a key immune-checkpoint receptor expressed by activated T cells, which mediates immunosuppression. PD-1 functions
primarily in peripheral tissues, where T cells may encounter the
immunosuppressive PD-1 ligands, namely PD-L1 (B7-H1) and PDL2 (B7-DC), which are expressed by tumor cells, stromal cells, or
both (Pardoll, 2012). In agreement with the proposed function of
the PD-1/PD-L1 axis in the induction and maintenance of peripheral tolerance (Pardoll, 2012), surface expression of PD-L1 in some
tumors has been reported to be an independent predictor of adverse
clinical outcome (Sznol and Chen, 2013), although the prognostic
signicance of PD-L1 expression remains controversial (Puzanov
et al., 2015; Taube et al., 2012; Konishi et al., 2004; Thompson
et al., 2004; Massi et al., 2014). Antibodies blocking the PD-1, or
its ligand (PD-L1), were shown in phase I trials to induce a 30% to
50% response in several cancer types (Topalian et al., 2012). Recent
phase II and III studies backed the accelerated approval of anti-PD1 antibodies for metastatic melanoma (MM), non-small cell lung
cancer (NSCLC) and renal cell cancer (RCC) (Robert et al., 2015a,
2015b; Gettinger et al., 2015; Garon et al., 2015).
Biomarker-driven selection of immunotherapy responders and
non-responders would minimize unnecessary exposure of patients
to potentially permanent and life-threatening immune-related toxicities and reduce the nancial burden for health systems due to
these expensive treatments. Although PD-L1 expression appeared
to correlate with response to treatment from exploratory analyses

of early reported trials, whether the level of expression of PD-L1

predicts response rate, progression free (PFS) and overall survival
(OS) in the context of anti-PD-1/PD-L1 therapy is still object of
debate.
At the time of the present manuscript the Food and Drug
Administration (FDA) approved anti-PD-1 therapies for three
tumor histotypes: MM, NSCLC and RCC. All of them are the
focus of the present systematic review and meta-analysis. All current clinical evidence in cutaneous MM, NSCLC and RCC patients
receiving anti-PD-1/PD-L1 therapy has been reviewed and quantitatively summarized. We performed a formal systematic literature
search and meta-analysis of the controlled clinical trials evaluating
anti-PD-1 antibodies with the aim at identifying whether PD-L1
expression is associated with response to treatment and survival
benet. Accordingly, the main endpoint of the study are objective
clinical response and mortality. Our study provides the rst systematic review of the results of all implemented clinical trials and of the
summary estimates of the associations between PD-L1 expression
and cancer disease outcomes.
2. Meta-analysis methods
A systematic literature search and quantitative analysis were
planned, conducted and reported following CONSORT and Quality of Reporting of Meta-analyses (QUORUM) checklists regarding
meta-analysis of clinical trials (Moher et al., 2001; Clarke, 2000).
2.1. Literature search
Published reports were obtained from the following databases
using validated search strategies (Yoshii et al., 2009): PubMed
(http://www.ncbi.nlm.nih.gov/entrez/query.fcgi), Ovid Medline
(Ovid Technologies, Inc., New York, 1950-April 29, 2011), EMBASE
(Elsevier, Amsterdam, the Netherlands, 1980- April 29, 2011), ISI
Web of Knowledge (Thomson Scientic Technical Support, New
York, 1945-May 4, 2011), the Cochrane Library and clinicaltrial.gov
up to December 2015. We also performed manual searches of references cited in the retrieved articles and preceding reviews on the
topic. From personal contacts we retrieved also presentations given
at international conferences [American Society of Clinical Oncology (ASCO), American Association for Cancer Research (AACR),
European Society of Medical Oncology (ESMO), European CanCer
Organization (ECCO), Society for Melanoma Research (SMR), European Association of Dermato Oncology (EADO)]. Ecological studies,
case reports, observational studies, reviews and editorials were not
considered eligible.
The following keywords or corresponding MeSH terms were
used: anti-PD-1, Nivolumab, Pembrolizumab, Atezolizumab, Lambrolizumab, MPDL3280A, anti-PD-L1, antibody, PD-L1 overexpression,
immunohistochemical (IHC), cancer, non-small cell lung cancer, squamous cell, lung cancer, non squamous cell, melanoma, renal cell
carcinoma, advanced, metastatic, therapy, randomized, prospective,
phase I, phase II, phase III, study and trial. The search for keywords
in the title and in the abstract was carried out systematically. A
manual search of references cited in the selected articles and in

90

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

selected reviews or abstracts and presentations from conferences

was carried out. Two of the authors (D.M. and M.M.) independently
selected the articles; abstracts and presentations.
We screened titles and looked at abstracts when the title suggested a study possibly meeting the three main criteria. If the
abstract content was relevant, full copies of the articles were
retrieved and thoroughly read by at least two co-authors.
We selected clinical trials reporting the minimum information
on Objective Response (OR that is complete response or partial
response, as dened in the clinical trials following RECIST criteria), or deaths necessary to perform adequate meta-analysis, with
the following main criteria:
1. Sufcient information to obtain the estimates and 95% condence intervals for the association between PD-L1 expression
with clinical objective response or deaths (odds ratios, relative
risks, hazard ratio or crude data and corresponding standard
errors, variance, condence intervals or P-value of the signicance of the estimates), and/or to estimate Objective Response
Rates (ORRs) by PD-L1 expression, with condence intervals, in
patients treated with anti-PD-1/PD-L1 antibodies.
2. Studies had to be independent and not duplicate results
published in another article/presentation. When several articles/presentation concerned the same study, results from the
publication using the largest sample of subjects were used.
3. Studies designs had to be clinical trials assessing efcacy of antiPD-1/PD-L1 treatment in cancer patients.
In each study we evaluated if immune-related response criteria were used in addition to the standard RECIST criteria (Therasse
et al., 2000).
When possible we retrieved data and estimates for the two
different cut-off points for the PD-L1 expression most frequently
reported (1% and 5%). When trials also reported information also
for an arm evaluating efcacy of other treatments (e.g., Ipilimumab)
we included estimates only for the anti-PD-1/PD-L1 treatment arm.
There were no language, time or geographical restrictions.
Review articles not reporting original data were also excluded but
checked for references. A standardized data-collection protocol was
used for gathering the relevant data from each selected article. For
each trial selected for this meta-analysis, we extracted information
on authors, journal/conference, year of publication, cancer type,
types of treatment, Clinical Trial Registry Numbers, study phase,
line of treatment, type of treatment, previous treatment, treatment
with Ipilimumab in MM patients, number of patients included with
PD-L1 expression evaluation, methods and cut-off for assessment
for PD-L1 positivity, number of clinical responses and deaths by
PD-L1 expression, hazard ratio, and median survival for PD-L1 positivity.
2.2. Statistical analysis
We extracted data from anti-PD1 treatment arm of clinical trials
to calculate response rates and deaths by PD-L1 expression, and we
estimated odd ratios comparing response rates and deaths by PDL1 expression. In order to evaluate if the expression of PD-L1 is a
predictive response marker or a prognostic marker we summarized
randomized trials comparing response rates by PD-L1 expression in
antiPD1 treatments arms with responses in other treatment arms.
Every measure of association (Odd ratio and Hazard ratio) with
corresponding condence intervals was transformed into log relative risks and the corresponding variance was calculated using the
formula proposed by Greenland (1987). When no estimates were
given, crude estimates were calculated from tabular data. We used
Woolfs formula to evaluate the standard error of the log relative
risk.

The Summary Odd Ratio (SOR) was estimated by pooling

the study-specic estimates with the random effects models as
described by van Houwelingen et al. (2002) with summary estimate
obtained from maximum likelihood estimation, with a hierarchical model when two estimates are extracted from a single study.
Condence intervals were computed assuming an underlying tdistribution to be conservative.
We used random effects models, with maximum likelihood
estimates, also to estimate Summary Objective Response Rates
(SORRs) by PD-L1 positive and negative. We used transformation of proportions into quantities suitable for the usual random
effects summaries. The pooled proportion is calculated as the
back-transform of the weighted mean of the transformed proportions, using inverse arcsine variance weights. We used the
FreemanTukey arcsine transform in order to stabilize variances
(Stuart, 1994).
The homogeneity of the effects across studies was assessed
using the large sample test based on the Chi-square statistic. Since
the Chi-square test has limited power, we considered statistically
signicant heterogeneity at the P = 0.10 level of association. A further measure of heterogeneity I2 has been considered in order to
compare between heterogeneities for different numbers of pooled
studies. It can be interpreted as the percentage of total variation across several studies that is attributable to heterogeneity:
larger values of I2 indicate greater heterogeneity (Higgins and
Thompson, 2002). A threshold of I2 below 50% is generally considered an acceptable level of variability. Sub-group analyses and
meta-regressions were carried out to investigate between-study
heterogeneity focusing on line of treatment, type of control, length
of follow-up, PD-L1 cut-off, tumor type, and study phase.
We presented SORs separately for each cancer type and we produced forest plots including both the study specics and the pooled
risk estimates. Publication bias was evaluated graphically with a
funnel plot; and we conducted the Macaskill test, which is more
powerful when less than 20 estimates are included in the analysis
(Macaskill et al., 2001).
All the statistical analyses were performed using SAS software
(SAS Institute Inc., Cary, NC; version 9.2) and R software, version
2.12.2 (http://www.r-project.org).

3. Results
3.1. Description of literature search and data from included
studies
A total of 173 articles, abstracts, or presentations at international
conferences were retrieved and checked for relevance in terms of
intervention, design and reporting of data both on responses and
survival (Fig. 1). We identied 90 articles or abstracts that published
information on studies evaluating the effects of anti-PD-1/PD-L1
reporting data on clinical responses or deaths by PD-L1 status.
Twenty-ve of these articles/abstracts were not included in the
meta-analysis for the following reasons: (i) Nine were reviews; (ii)
three reported information on trials including other cancer types;
(iii) in thirteen trials, the intervention was not an anti-PD-1/PD-L1
treatment. Sixty-ve articles/abstracts were considered for inclusion; among them, 38 were excluded for the following reasons:
29 were not independent from other studies, 6 did not report
information on responses/deaths by PD-L1 expression, 2 included
only PD-L1 positive patients. Furthermore, one study (Gibney et al.,
2015) was excluded because reported only information on relapse
and not the number of deaths by PD-L1 expression. We eventually
included 27 full-text articles or abstracts/presentations (Supplementary Tables 1S-3S) regarding 22 estimates from 20 trials: 11
in MM patients (n = 4230), 8 in NSCLC patients (n = 1417), and 3

Idencaon

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

Records idened through

database searching
n = 88

91

Addional records idened

through other sources
n = 85

n = 25
9 reviews
3 other cancer sites than
CM, NSCLC or RCC
13 other treatments

Eligibility

Records screened
aer duplicates removed
n = 90

Full-text arcles/presentaons/abstracts
assessed for eligibility
n = 65

Eligibility

Screening

Records excluded:

Full-text arcles/presentaons/abstracts
included in the meta-analysis:
n =27
20 trials and 22 cancer esmates:
3 RCC, 8 NSCLC and 11CM

Arcles/abstracts excluded
n = 38
29 not independent
6 no informaon on PDL1
expression or not enough
data to esmate objecve
response rate or deaths by
PD-L1
2 only PDL1+ paents were
included
1 only informaon in
Progression free survival

Fig. 1. Flowchart of studies retrieved and included in the meta-analysis.

in RCC patients (n = 312); for two trials (NCT00730639/CA209-003

and NCT01295827/Keynote 001) we retrieved information for two
cancer types (MM and NSCLC).
Five studies in MM patients were phase I, two phase II and
four phase III trials, three assessed efcacy of anti-PD-1 therapy concomitant to anti-CTLA-4 Ipilimumab. All trials in MM
patients assessed Nivolumab except for three that evaluated Pembrolizumab. Four trials in NSCLC patients were phase I, two phase
II and two phase III trials. Five trials in NSCLC patients assessed
Nivolumab, one Pembrolizumab and two trials considered anti-PDL1 therapies. All trials in RCC assessed Nivolumab; two were phase
I trials and one a phase II trial. For some studies estimates for different cohorts of patients, different treatments or different cut-off
points of PD-L1 positivity were included, as detailed in Supplementary Tables 1S-3S. Table 2bS presents NSCLC estimates found by
squamous and non-squamous histotypes.
IHC assays for PD-L1 evaluation adopted in the studies selected
in the current meta-analysis included Dako/BMS clone 288, Merck
mAb clone 22C3 and Ventana (Genetech/Roche) mAb clone SP142.
Additional information concerning cellular/subcellular localization
and scoring are reported in Supplementary Tables 4S.
For some studies the number of events by PD-L1 were estimated from percentages, histograms or hazard ratio reported by
the authors. In the study by Wolchok et al. (2015), information on
responses by PD-L1 status were obtained from the histograms and
in the study by Robert et al. (2015b) the number of deaths by PD-L1
status were presented for patients treated with anti-PD-1 and Ipilimumab, therefore we estimated the number of events for patients
treated with anti-PD-1 from the hazard ratio comparing anti-PD-1
and Ipilimumab.
ORRs have been dened by RECIST in all evaluated studies.
However in 4 studies (Weber et al., 2013, Wolchok et al., 2013,
Daud et al., 2014, Hodi et al., 2014) the immune-related response

criteria were used to determine whether patients should remain

on treatment in case of a mixed response.
No evidence for publication bias was found for any cancer type,
neither for deaths nor for ORRs.

3.2. Summary estimates of clinical objective response rates

(SORR) in PD-L1 positive vs. PD-L1 negative patients
Table 1 reports, by cancer types, SORs comparing objective
response rates by PD-L1 status, as well as SORRs according to PD-L1
status.
In the context of metastatic melanoma SORRs were 45% and 27%
in PD-L1 positive and negative patients, respectively, and SOR indicates a signicant difference between responses: 2.14 (1.65, 2.77)
(Table 1, Fig. 2a), with low between-study heterogeneity (I2 = 35%).
When we summarized randomized trials (Table 1, Fig. 2b)
comparing treatment with anti PD-1 vs other treatments we conrm that the expression of PD-L1 is signicantly associated with
response rates in anti-PD1 treatments arms, as indicated by the
signicant SOR (1.89; 95%CI: 1.35, 2.64), but not in other treatment
arms (SOR = 0.96; 95%CI: 0.5, 1.87), that presented lower response
rates (16% and 12% in positive and negative PD-L1). Differences in
responses by PD-L1 expression between the two types of treatments are signicant (P = 0.05 from meta-regression model).
In non-squamous NSCLC the SORRs were 29% and 11% in PD-L1
positive and negative patients, respectively, and SOR indicates a signicant difference in terms of responses: 3.78 (1.54, 9.24) (Fig. 3a,
Table 1). While in squamous NSCLC the SORRs were 26% and 15%
in PD-L1 positive and negative patients, respectively, and SOR indicates a non-signicant difference between responses: 1.49 (0.48,
4.64) (Fig. 3b, Table 1). No between-study heterogeneity was found
for NSCLC (I2 = 0% for both histotypes).

92

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

Table 1
Summary estimates according to PD-L1 status for clinical objective response and deaths.
Cancer Type

n. of estimates (n.
of trials)

Subgroups

PD-L1 status

12 (10)

Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative

MM
5 (6)

In anti-PD-1 treatment
arms
In anti-PD-1 other
treatment arms

5 (6)
NSCLC

9 (8)
5 (5)

Squamous

4 (4)

Non-squamous

RCC

4 (3)

MM

6 (4)

Summary
Objective Response
Rate (95% CI)
45% (35, 55)
27% (17, 39)
46% (27, 65)
35% (19, 53)
16% (11, 2)
12% (5, 23)
25% (20, 31)
14% (10, 18)
26% (16, 38)
15% (8, 24)
29% (19, 4)
11% (5, 19)
29% (8, 57)
25% (0, 76)

Deaths

Odd Ratio

I2 %

2.14 (1.65, 2.77)

40

1.89 (1.35, 2.64)

0.96 (0.5, 1.87)

2.12 (1.23, 3.66)

26

1.49 (0.48, 4.64)

3.78 (1.54, 9.24)

1.70 (0.32, 9.02)

33

0.47 (0.30, 0.75)

First Author, PY, Clinical Trial name

Weber, 2013, CA209-006
Wolchok, 2013, CA209-004 (Cuncurrent)
Wolchok, 2013, CA209-004 (Sequence)
Daud, 2014, Keynote 001 (Validation Set)
Daud, 2014, Keynote 001 (Pooled set)
Hodi, 2014, CA209-003
Robert, 2014, CheckMate 066
Postow, 2015, CheckMate 069
Puzanov, 2015, Keynote 002
Weber, 2015, CheckMate 037
Wolchok, 2015, CheckMate 067 (Nivolumab+Ipilimumab)
Wolchok, 2015, CheckMate 067 (Nivolumab)

Summary OR: 2.52 (95%CI: 1.65, 3.87)

I2=40%

0.05

0.20

0.40

0.80

1.50

4.00

8.00

20.00

Fig. 2. Forest plot of ORs of objective clinical response for PD-L1+ (cut-off 5%) vs. PD-L1- in treated metastatic melanoma patients.

In RCC the association did not reach statistical signicance, probably because the number of trials was low (n = 3) (Fig. 4).

3.4. Sensitivity analysis assessing the impact of PD-L1 positivity

cut-off on estimates of clinical objective responses rates

3.3. Summary estimates for mortality in PD-L1 positive vs. PD-L1

negative melanoma patients

The majority of trials reported estimates for objective response

and death, with 5% or 1% PD-L1 cut-off. Few studies presented the
estimates for both the cut-off values and the majority presented
the data for 5%. Thus, we carried out the main analysis choosing
the estimates with 5% and when not available with 1% of cut-off.
As a sensitivity analysis, we calculated summary estimates also
choosing the estimates with 1% when available but the summary
estimates did not show a considerable difference: for MM SOR was
2.14 (95% CI: 1.51, 3.03) with I2 = 15%, while for NSCLC SOR was
2.09 (95% CI: 1.22, 3.59) with I2 = 19%. We did not nd any relevant

We were able to retrieve information to obtain mortality risk

estimates by PD-L1 expression from 4 trials (Table 1) (Robert et al.,
2015a, 2015b; Daud et al., 2014; Hodi et al., 2014). Overall, we summarized 374 deaths from 1274 MM patients, and SOR indicates
a signicant 53% decrease in risk of mortality in PD-L1 positive
vs. negative patients (Table 1, Fig. 5), with no between-study
heterogeneity (I2 = 0%). The majority of estimates were obtained
evaluating PD-L1 expression with 1% cut-off.

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

93

First Author, PY, Clinical Trial name

Rizvi, 2014, CA209-012

Brahmer, 2015, CheckMate 017
Garon, 2015, NCT01295827 (Validation set)
Gettinger, 2015, CA209-003
Rizvi, 2015, CheckMate 063

Summary OR: 1.49 (0.48, 4.64)

I2=0%

0.2

0.4

0.6

0.8

1.5

2.0

4.0

8.0

20.0

First Author, PY, Clinical Trial name

Rizvi, 2014, CA209-012

Garon, 2015, NCT01295827 (Validation set)

Gettinger, 2015, CA209-003

Paz-Ares, 2015, CheckMate 057

Summary OR: 3.78 (1.54, 9.24)

I2=0%

0.2

0.4

0.6

0.8

1.5

2.0

4.0

8.0

20.0

Fig. 3. Forest plot of ORs of clinical response in PD-L1+ (cut-off 5%) vs. PD-L1- treated non-small cell lung carcinoma patients.

variation in the estimates considering different cut-off points for

PD-L1 expression.
In the context of each study we tried to evaluate the variability
of responses by cut-off of PD-L1, but we were able to retrieve only
2 trials in MM patients, 4 in NSCLC patients and 1 in RCC patients

(Table 5S) presenting ORR by PD-L1 status with both 5% and 1%

cut-off points. Trials in MM patients showed that with 5% cutoff the association between PD-L1 status and objective response
was always signicantly positively associated (SOR with 95% CI not
crossing zero) but signicance is lost with 1% cut-off. Similarly, SOR

94

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

First Author, PY, Clinical Trial name

Amin, 2014, CA209-016 (P arm)

Amin, 2014, CA209-016 (S arm)

Choueiri, 2014, CA209-009

Motzer, 2015, CA209-010

Summary OR for objective clinical response:

1.70 (95%CI: 0.32, 9.02)
I2=33%
0.2

0.4 0.6

1.0 1.5

4.0

8.0

20.0

Fig. 4. Forest plot of ORs of clinical response in PD-L1+ (cut-off 5%) vs. PD-L1- treated renal cell cancer patients.

First Author, PY, N. Clinical Trial

Hodi, 2014, NCT00730639

Daud, 2014, NCT01295827 (Validation set)

Daud, 2014, NCT01295827 (Pooled set)
Robert, 2014, NCT01721772
Robert, 2015, NCT01866319 (Pembro at 2 weeks)
Robert, 2015, NCT01866319 (Pembro at 3 weeks)

Summary OR for death: 0.47 (95%CI: 0.30, 0.75)

I2=0%

0.2

0.4

0.6

0.8

1.0

1.5

2.0

Fig. 5. Forest plot of ORs for deaths in PD-L1+ (cut-off 1%) vs. PD-L1- treated metastatic melanoma patients.

for response showed consistently higher responses in PD-L1 positive vs. negative when PD-L1 status was assessed with 5% cut-off
but the results were not consistent with 1% cut-off, both in NSCLC
and RCC patients.
3.5. Between-study heterogeneity analysis of clinical objective
response according to line of treatment
We studied heterogeneity among trials in MM investigating
anti-PD-1 antibodies as rst or second line of treatment as well as

concomitant to the anti-CTLA-4 Ipilimumab (Table 2). The analysis was carried out considering 5% as cut-off when available. We
found that responses in trials with anti-CTLA-4 given concomitantly with anti-PD-1 were higher than with anti-PD-1 alone, even
if not statistically signicant different.
We could evaluate responses by PD-L1 status in patients treated
with anti-PD-1 as rst line in only two trials (Robert et al., 2015a;
Wolchok et al., 2015) and responses were higher than in six trials
that investigated anti-PD-1 therapy in second line treatment, without concomitant anti-CTLA-4 (Ipilimumab), both in PD-L1 negative

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

95

Table 2
Summary estimates according to PD-L1 status for clinical objective response and deaths by possible heterogeneity factors.
Heterogeneity factors

MM

P-value

Anti-PD-1 concomitant
to Ipi
No Ipi concomitant

4 (3)

Objective Response Rate

8 (8)

Objective Response Rate

Anti-PD-1 in rst line

2 (2)

Objective Response Rate

Anti-PD-1 as second
line and no Ipi

6 (6)

Objective Response Rate

Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative

62% (34, 86)

48% (18, 79)
43% (32, 54)
23% (14, 34)
55% (21, 86)
38% (10, 71)
37% (25, 50)
15% (10, 21)

0.21
0.18

0.34
0.07

I2 % Percentage of heterogeneity; Odd Ratio: Odd Ratio for positive objective response and PD-L1 status (positive vs. negative); Ipi = Ipilimumab; MM = Metastatic melanoma;
NSCLC = Non-Small Cell Lung Cancer. RCC = Renal Cell Cancer. P-value of heterogeneity factors in meta-regression random effect model. The total number of trials is 9
because one study (Wolchok et al., 2015; Gibney et al., 2015) presented one estimate for patients with Ipi concomitant and one with no Ipi.

patients (38% vs. 15%; P = 0.07) and in PD-L1 positive patients (55%
vs. 37%; P = 0.34). In NSCLC, all patients received anti-PD-1 as second line treatment and the summary ORR were lower than in MM:
14% (95% CI: 1018) for PD-L1 negative and 25% (95% CI: 2031) for
PD-L1 positive.
3.6. Sensitivity analysis of summary estimates in patients
receiving anti PD-1 therapy (excluding anti PD-L1 trials)
Since, in the present analysis, we included NSCLC patients who
received anti-PD-1 and anti-PD-L1 antibodies, we performed a sensitivity analysis to assess the impact of anti-PD-1 therapy alone
by excluding patients treated with anti-PD-L1 antibodies. The sensitivity analysis showed no differences in summary estimates by
excluding the only two trials evaluating anti-PD-L1 antibodies
(Spira et al., 2015; Herbst et al., 2014). SOR for PD-L1 positive vs.
PD-L1 negative was 2.67 (95% CI: 1.48, 4.82; I2 = 6%). Furthermore,
SORRs were 13% (95% CI: 917) and 27% (95% CI: 2232) in PD-L1
positive and negative, respectively.
4. Discussion
The present meta-analysis reports three main ndings. The rst
is that the expression of PD-L1 in tumor tissues correlates with
the clinical response to antibodies targeting the PD-1/PD-L1 axis in
MM and in non-squamous NSCLC patients. This contention is supported by the difference in the objective response observed in MM
(45% vs. 27%) and in non-squamous NSCLC (29% vs. 11%) according
to the positive vs. negative expression of PD-L1, respectively. Furthermore, regardless of cut-off for estimating the positivity of PD-L1
expression (either 5% or 1%) the correlation with the response rate
is maintained. In addition, the nding that the higher the IHC PDL1 overexpression (5% vs. 1%), the greater the association of the
response rate with PD-L1 status, further strengthens the clinical
importance of the present proposal. Our ndings are consistent
with those recently reported in a pooled analysis of 655 melanoma
patients enrolled in the KEYNOTE-001 study (Daud et al., 2014) and
in previously-treated, advanced NSCLC patients in the KEYNOTE010 study (Herbst et al., 2014) underlining the clinical benet of
anti-PD-1/PD-L1 therapies in PD-L1 positive melanoma patients.
Although supported by smaller numbers of cases, similar ndings are documented in non-squamous NSCLC patients. Squamous
NSCLC did not show any signicant difference in response according to the PD-L1 status. Consistently with these results, in the
recently reported CheckMate-017 the benet of nivolumab over
docetaxel was independent from PD-L1 expression regardless of
the cut-off used. A potential limit of our present meta-analysis is
the inclusion of patients enrolled in phase I studies, where response
rates were not usually a primary endpoint and are subject of great
bias in the tumor assessment. However, the clinical development
of immunomodulating agents has substantially changed the way

early phase I trials are planned and conducted as demonstrated

by the case of Pembrolizumab whose approval was obtained on
the basis of the largest phase I trial ever performed in clinical trial
history.
The second important nding derives from randomized clinical
trials on MM comparing anti PD-1 therapy with other treatments.
For the rst time in the current literature we show that PD-L1 is
a predictive marker of clinical response. Based on these results
expression of PD-L1 is signicantly associated with greater clinical
response rates in anti-PD1 treatment but not in other treatments.
Thus, based on the present analysis, if it cannot be concluded that
PD-L1 expression is a predictive biomarker of outcome, a correlation was found with the response to anti-PD-1 therapy and this
correlation cannot be found in MM patients receiving other treatments.
The third novel nding derives from the observation that PDL1 expression is associated with a signicant better prognosis (risk
reduction in mortality of 53%) in MM patients receiving anti-PD1 antibodies. Notably, this result was obtained in 1274 melanoma
patients under a remarkable absence of study heterogeneity and
unbiased results. Previous data showed that the differences in
response rates between PD-L1 positive and PD-L1 negative do not
translate to differences in survival benet in melanoma patients
(Robert et al., 2015b). In a phase III trial MM patients treated
with Nivolumab, the median OS was not reached for either PDL1 positive or PDL1 negative subgroups and both subgroups had
improved OS as compared to patients who received dacarbazine
chemotherapy (CheckMate-066 (Robert et al., 2015a)). However,
in the update results of the CheckMate-066 (Atkinson et al., 2015),
OS was statistically and clinically meaningful improved in PD-L1
positive patients (HR = 0.56), consistently with the results of the
present meta-analysis.
Possibly because of the limited number of analyzed studies, a
correlation between PD-L1 status and response rate or outcome
in RCC has been not documented. In non-clear cell RCC, patients
with PD-L1 positive tumors were reported to have worse clinical
outcomes, and PD-L1 positivity in tumor cells was associated with
higher tumor stage and grade (Choueiri et al., 2014). In a recent
study on Nivolumab, in patients with advanced RCC, data on PD-L1
were not reported (McDermott et al., 2015). Thus, further studies
are needed to draw denitive conclusions on the predictive and
prognostic role of PD-L1 in RCC. Peculiar tumor and microenvironment features, specic immune tolerance according to the tumor
types and variation in the levels of T-cell exhaustion are some of
the mechanisms that may account for the different response rate
and disease control in patients with different cancers. Accordingly,
more controlled, randomized prospective trials are clearly needed
in RCC.
Understanding the biology of PD-L1 expression may contribute
to better recognize response and resistance mechanisms to anti
PD-1/PD-L1 antibodies. Some patients with PD-L1positive tumors

96

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

fail to respond to therapy because the PD-L1 expression is constitutive, driven by activation of a pathway within the tumor
(e.g., PI3-kinase), rather than induced by an adaptive response to
tumor-specic CD8+ lymphocyte inltration (Chen et al., 2015;
Berry and Taube, 2015; Taube et al., 2014). Blockade may have no
effect because the tumor microenvironment is lacking the effector immune cells that ght cancer. Actually, pre-treatment samples
obtained from responding patients showed higher numbers of CD8, PD-1- and PD-L1-expressing cells at the invasive tumor margin and
inside tumors, with close proximity between PD-1 and PD-L1, and
a more clonal TCR repertoire (Tumeh et al., 2014).
Our meta-analysis shows that a proportion of PD-L1 negative patients also benets from anti-PD-1 therapy in MM and
squamous-NSCLC. Thus, expression of PD-L1 in tumor tissues,
while cannot be considered as a predictive biomarker of eligibility for treatment with anti-PD-1/PD-L1 antibodies, it may
rather represent a correlation marker for non-squamous NSCLC and
melanoma. Nevertheless, further studies also with longer follow-up
are needed to establish rm conclusions. Observed differences in
squamous and non-squamous NSCLC subpopulations may results
from smoking habits, mutational load and key targets (i.e., EGFR,
ALK) and dissimilarities in immune status or immunomodulatory
mechanisms. Activation of EGFR induced PD-L1 expression may
contribute to escape from the anti-tumor immune response and
blockade of the PD-1 pathway using EGFR-TKIs was found to reduce
PD-L1 expression (Akbay et al., 2013; Azuma et al., 2014; DIncecco
et al., 2015).
Reasons explaining unexpected clinical responses in negative
patients may include intra-tumor and intra-patient heterogeneity in PD-L1 expression (Madore et al., 2015; Callea et al., 2015;
McLaughlin et al., 2015) and the dynamic inducible expression of
PD-L1 in the tumor microenvironment (Taube et al., 2012) as well
as PD-L2 expression (Rozali et al., 2012). Sampling errors, technical
and pre-analytical issues may also contribute to the accuracy on
the assessment in formalin-xed and parafn-embedded samples,
and optimal strategies for PD-L1 testing in tumor tissues remain an
unsettled question. Lung specimens are particularly challenging for
PD-L1 evaluation due to a high rate of sampling errors in small lung
biopsies and evaluation of cytological material in which relevance
of immune cells inltration is undetermined.
PD-L1 expression has been investigated also in the context of
immune checkpoint blockade combination strategy. The combined
administration of anti-cytotoxic T-lymphocyte antigen 4 (CTLA4) immunotherapy (Ipilimumab) plus anti PD-1 immunotherapy
(Nivolumab) has shown a higher level of anti-melanoma activity
than monotherapy with either Nivolumab or Ipilimumab, although
this schedule was associated with increased toxicity (Postow et al.,
2015b). Retrospective analysis of a recently reported phase II randomized study has shown that the response rates in wild-type
BRAF and BRAFV600 mutated disease were similar (Postow et al.,
2015b). The objective response rates for the combination were 58%
in patients whose tumors were PD-L1 positive and 55% in those
whose tumors were PD-L1 negative (Postow et al., 2015b). Findings
of a prospective phase III randomized study that the combination of
anti-CTLA-4 (Ipilimumab) with the anti-PD-1 are associated with
objective responses higher with the combination therapy than with
the anti-PD-1 alone support this contention (Larkin et al., 2015).
Efcacy of combination of Ipilimumab and Nivolumab was apparently independent from PD-L1 status. Subgroup analyses of the
CA 209067 trial suggest that the greatest benet in terms of PFS
with the Nivolumab/Ipilimumab combination vs. Nivolumab alone
was observed in PD-L1 negative tumors (Larkin et al., 2015). However, longer follow-up and OS data from this phase III trial will
be required to determine whether the combination should replace
Nivolumab monotherapy as the preferred approach for checkpoint
inhibition immunotherapy as well as the role of PD-L1 expression.

PD-L1 immunohistochemistry has previously suffered from

poor standardization and validation since reliable PD-L1specic
antibodies have been difcult to develop, resulting in wide assay
variability in frozen and parafn embedded tissues. In addition,
different subcellular localization (membranous vs. cell surface vs.
cytoplasmic) and uncertainty on the role of PD-L1 expression in
immune cells (tumor inltrating lymphocytes and macrophages)
within the tumor microenvironment add further complexity to
data interpretation. Different methods and reagents may also contribute to the variability of results in PD-L1 expression in the
different data sets. The clinical trials selected for the current metaanalysis included different analytically validated IHC assays, such
as Dako/BMS clone 288, Merck mAb clone 22C3 and Ventana
(Genetech/Roche) mAb clone SP142. Staining on tumor cells or
immune cells (Spira et al., 2015; Herbst et al., 2014) vs. tumor cells
only, combined score of percentage and intensity of positive cells
(Garon et al., 2015; Paz-Ares et al., 2015) vs. percentage of positive
cells only, and type and timing of biopsy are differences in methodologies that should be taken into consideration. Interestingly, in one
study (Choueiri et al., 2014), tumoral PD-L1 immunostaining was
associated with the additional analysis of CD3 or CD8+ intratumoral
lymphocytes, further supporting the view that PD-L1 expression
should be interpreted in the context of the microenvironment as
well as immune cells phenotype (Puzanov et al., 2015).
PD-L1 status should be evaluated in light of the complexity
resulting from the dynamics inherent to a rapidly evolving immune
response. T cells that co-express PD-1 along with other inhibitory
molecules such as LAG-3 or TIM-3 may be less responsive than
those expressing PD-1 alone, thus underlining the need for the
blockade of multiple targets (Woo et al., 2012; Sakuishi et al.,
2010). The analysis of the pattern of expression of the ligands and
their receptors in T cells, tumor cells, myeloid cells, and other
components in the tumor microenvironment would be key for
development of combination strategies with greater clinical benet. Identication of biomarkers more reliable than PD-L1 such
as mutation rates, immune scores/cytokine proling, CD8+ T cell
ratios, neo-epitopes, gene signatures or RNA expression proles,
both in the stroma and tumor inltrating lymphocytes, would probably represent the next step forward.
In order to improve the number of patients who ultimately
may benet from immune checkpoint blockade, PD-1/PD-L1 antibodies are being combined with other anticancer agents such as
targeted therapy, radiotherapy, and other immunotherapies, such
as the indoleamine 2,3-dioxygenase (IDO) inhibitor (Spranger et al.,
2014; Wainwright et al., 2014) and the oncolytic virus talimogene laherparepvec (T-VEC) (Kohlhapp and Kaufman, 2015) with
encouraging early results. Many of these combination approaches
are based on solid preclinical data. However, nal results from randomized studies to suggest that any specic combination approach
is more efcacious than single-agent PD-1/PD-L1 blockade are not
yet available.
Interestingly, results obtained in preclinical models suggest that
TKIs combined with the immune checkpoint regulation in cancer
may suppress the activities of the hyperactivated oncoproteins and
also induce anti-tumor memory of the immune system, achieving the synergistic effect from two directions (Wargo et al., 2014).
Furthermore, under certain circumstances resistance to targeted
therapies basically associates with suppressed anti-tumor activity
of T cells and results from impaired immune checkpoint control
(Massi et al., 2015). Although promising in terms of efcacy, relevant safety issues are becoming evident in a few ongoing clinical
trials that evaluate a combination of immunotherapy and targeted
therapy (Ribas et al., 2015; Puzanov et al., 2015). Future discoveries of novel targetable checkpoint regulators hold the promise to

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

strengthen the immunologic anti-tumor efcacy and potentially to

improve survival.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.critrevonc.2016.
02.001.
References
Akbay, E.A., Koyama, S., Carretero, J., Altabef, A., Tchaicha, J.H., Christensen, C.L.,
et al., 2013. Activation of the PD-1 pathway contributes to immune escape in
EGFR-driven lung tumors. Cancer Discov. 3, 13551363.
Atkinson V., Ascierto P.A., Long G.V., Brady B., Dutriaux C., Maio M., et al., 2015.
Two-year Survival and Safety Update in Patients With Treatment-nave
Advanced Melanoma Receiving Nivolumab or Dacarbazine in CheckMate 066,
SMR.
Azuma, K., Ota, K., Kawahara, A., Hattori, S., Iwama, E., Harada, T., et al., 2014.
Association of PD-L1 overexpression with activating EGFR mutations in
surgically resected nonsmall-cell lung cancer. Ann. Oncol. 25, 19351940.
Berry, S., Taube, J.M., 2015. Innate vs: adaptive: pD-L1-mediated immune
resistance by melanoma. Oncoimmunology 4, e1029704.
Callea, M., Albiges, L., Gupta, M., Cheng, S.C., Genega, E.M., Fay, A.P., et al., 2015.
Differential expression of PD-L1 between primary and metastatic sites in
clear-cell renal cell carcinoma. Cancer Immunol. Res. 3, 11581164.
Chen, J., Jiang, C.C., Jin, L., Zhang, X.D., 2015. Regulation of PD-L1: a novel role of
pro-survival signalling in cancer. Ann. Oncol.
Choueiri T.K., Fishman M.N. Escudier B., Kim J.J., Kluger H., Stadler H., et al., 2014.
Immunomodulatory activity of nivolumab in previously treated and untreated
metastatic renal cell carcinoma (mRCC): biomarker-based results from a
randomized clinical trial:ASCO.
Clarke, M., 2000. The QUORUM statement. Lancet 355, 756757.
DIncecco, A., Andreozzi, M., Ludovini, V., Rossi, E., Capodanno, A., Landi, L., et al.,
2015. PD-1 and PD-L1 expression in molecularly selected non-small-cell lung
cancer patients. Br. J. Cancer 112, 95102.
Daud, A., Hamid, O., Robert, C., Hodi, F.S., Wolchok, J.D., Hwu, W.J., et al., 2014.
Relationship between programmed death ligand 1 (PD-L1) expression and
clinical outcome in patients (pts) with melanoma (MEL) treated with
pembrolizumab (pembro; MK-3475). Eur. J. Cancer 50, 4849.
Garon, E.B., Rizvi, N.A., Hui, R., Leighl, N., Balmanoukian, A.S., Eder, J.P., et al., 2015.
Pembrolizumab for the treatment of non-small-cell lung cancer. N. Engl. J.
Med. 372, 20182028.
Gettinger, S.N., Horn, L., Gandhi, L., Spigel, D.R., Antonia, S.J., Rizvi, N.A., et al., 2015.
Overall survival and long-term safety of nivolumab (Anti-Programmed Death 1
Antibody, BMS-936558, ONO-4538) in patients with previously treated
advanced non-small-cell lung cancer. J. Clin. Oncol. 33, 20042012.
Gibney, G.T., Kudchadkar, R.R., DeConti, R.C., Thebeau, M.S., Czupryn, M.P., Tetteh,
L., et al., 2015. Safety, correlative markers, and clinical results of adjuvant
nivolumab in combination with vaccine in resected high-risk metastatic
melanoma. Clin. Cancer Res. 21, 712720.
Greenland, S., 1987. Quantitative methods in the review of epidemiologic
literature. Epidemiol. Rev. 9, 130.
Herbst, R.S., Soria, J.C., Kowanetz, M., Fine, G.D., Hamid, O., Gordon, M.S., et al.,
2014. Predictive correlates of response to the anti-PD-L1 antibody MPDL3280A
in cancer patients. Nature 515, 563567.
Higgins, J.P., Thompson, S.G., 2002. Quantifying heterogeneity in a meta-analysis.
Stat. Med. 21, 15391558.
Hodi F.S., Sznol M., Kluger H.M., McDermott D.F., Carvajal R.D., Lawrence D.P., et al.,
2014. Long-term survival of ipilimumab-nave patients (pts) with advanced
melanoma (MEL) treated with nivolumab (anti-PD-1; BMS-936558;
ONO-4538) in a phase I trial:ASCO.
Kohlhapp, F.J., Kaufman, H.L., 2015. Molecular pathways mechanism of action for
talimogene laherparepvec, a new oncolytic virus immunotherapy. Clin. Cancer
Res.
Konishi, J., Yamazaki, K., Azuma, M., Kinoshita, I., Dosaka-Akita, H., Nishimura, M.,
2004. B7-H1 expression on non-small cell lung cancer cells and its relationship
with tumor-inltrating lymphocytes and their PD-1 expression. Clin. Cancer
Res. 10, 50945100.
Larkin, J., Chiarion-Sileni, V., Gonzalez, R., Grob, J.J., Cowey, C.L., Lao, C.D., et al.,
2015. Combined nivolumab and ipilimumab or monotherapy in untreated
melanoma. N. Engl. J. Med. 373, 2334.
Macaskill, P., Walter, S.D., Irwig, L., 2001. A comparison of methods to detect
publication bias in meta-analysis. Stat. Med. 20, 641654.
Madore, J., Vilain, R.E., Menzies, A.M., Kakavand, H., Wilmott, J.S., Hyman, J., et al.,
2015. PD-L1 expression in melanoma shows marked heterogeneity within and
between patients: implications for anti-PD-1/PD-L1 clinical trials. Pigm. Cell
Melanoma Res. 28, 245253.
Massi, D., Brusa, D., Merelli, B., Ciano, M., Audrito, V., Serra, S., et al., 2014. PD-L1
marks a subset of melanomas with a shorter overall survival and distinct
genetic and morphological characteristics. Ann. Oncol. 25, 24332442.
Massi, D., Brusa, D., Merelli, B., Falcone, C., Xue, G., Carobbio, A., et al., 2015. The
status of PD-L1 and tumor-inltrating immune cells predict resistance and

97

poor prognosis in BRAFi-treated melanoma patients harboring mutant

BRAFV600. Ann. Oncol.
McDermott, D.F., Drake, C.G., Sznol, M., Choueiri, T.K., Powderly, J.D., Smith, D.C.,
et al., 2015. Survival, durable response, and long-term safety in patients with
previously treated advanced renal cell carcinoma receiving nivolumab. J. Clin.
Oncol. 33, 20132020.
McLaughlin, J., Han, G., Schalper, K.A., Carvajal-Hausdorf, D., Pelakanou, V.,
Rehman, J., et al., 2015. Quantitative assessment of the heterogeneity of PD-L1
expression in non-small-cell lung cancer. JAMA Oncol., 19.
Moher, D., Schulz, K.F., Altman, D., CONSORT Group (Consolidated Standards of
Reporting Trials), 2001. The CONSORT statement: revised recommendations
for improving the quality of reports of parallel-group randomized trials. JAMA
285, 19871991.
Pardoll, D.M., 2012. The blockade of immune checkpoints in cancer
immunotherapy. Nat. Rev. Cancer 12, 252264.
Paz-Ares L., Horn L., Borghaei H., Spigel D.R., Steins M., Neal E., et al., 2015. Phase
III, randomized trial (CheckMate 057) of nivolumab (NIVO) versus docetaxel
(DOC) in advanced non-squamous (non-SQ) cell non-small cell lung cancer
(NSCLC):ASCO.
Postow, M.A., Callahan, M.K., Wolchok, J.D., 2015a. Immune checkpoint blockade in
cancer therapy. J. Clin. Oncol. 33, 19741982.
Postow, M.A., Chesney, J., Pavlick, A.C., Robert, C., Grossmann, K., McDermott, D.,
et al., 2015b. Nivolumab and ipilimumab versus ipilimumab in untreated
melanoma. N. Engl. J. Med. 372, 20062017.
Ribas, A., Puzanov, I., Dummer, R., Schadendorf, D., Hamid, O., Robert, C., et al.,
2015. Pembrolizumab versus investigator-choice chemotherapy for
ipilimumab-refractory melanoma (KEYNOTE-002): a randomised, controlled,
phase 2 trial. Lancet Oncol.
Robert, C., Long, G.V., Brady, B., Dutriaux, C., Maio, M., Mortier, L., et al., 2015a.
Nivolumab in previously untreated melanoma without BRAF mutation. N. Engl.
J. Med. 372, 320330.
Robert, C., Schachter, J., Long, G.V., Arance, A., Grob, J.J., Mortier, L., et al., 2015b.
Pembrolizumab versus ipilimumab in advanced melanoma. N. Engl. J. Med.
372, 25212532.
Rozali, E.N., Hato, S.V., Robinson, B.W., Lake, R.A., Lesterhuis, W.J., 2012.
Programmed death ligand 2 in cancer-induced immune suppression. Clin. Dev.
Immunol. 2012, 656340.
Sakuishi, K., Apetoh, L., Sullivan, J.M., Blazar, B.R., Kuchroo, V.K., Anderson, A.C.,
2010. Targeting Tim-3 and PD-1 pathways to reverse T cell exhaustion and
restore anti-tumor immunity. J. Exp. Med. 207, 21872194.
Shin, D.S., Ribas, A., 2015. The evolution of checkpoint blockade as a cancer
therapy: whats here, whats next. Curr. Opin. Immunol. 33, 2335.
Spira A.I., Park K., Mazieres J., Vansteenkiste J., Rittmeyer A., Ballinger M., et al.,
2015. Efcacy, safety and predictive biomarker results from a randomized
Phase II study comparing atezolizumab (MPDL3280A) vs docetaxel in 2L/3L
NSCLC (POPLAR):ASCO.
Spranger, S., Koblish, H.K., Horton, B., Scherle, P.A., Newton, R., Gajewski, T.F., 2014.
Mechanism of tumor rejection with doublets of CTLA-4, PD-1/PD-L1, or IDO
blockade involves restored IL-2 production and proliferation of CD8(+) T cells
directly within the tumor microenvironment. J. Immunother. Cancer 2 (3)
(1426-2-3. eCollection 2014).
Stuart, A.O.K., 1994. Kendalls Advanced Theory of Statistics, 6th ed. John Wiley &
Sons, Hoboken, NJ, USA.
Sznol, M., Chen, L., 2013. Antagonist antibodies to PD-1 and B7-H1 (PD-L1) in the
treatment of advanced human cancer-response. Clin. Cancer Res. 19, 5542,
0432, CCR-13-2234, Epub 2013 Sep 18.
Taube, J.M., Anders, R.A., Young, G.D., Xu, H., Sharma, R., McMiller, T.L., et al., 2012.
Colocalization of inammatory response with B7-h1 expression in human
melanocytic lesions supports an adaptive resistance mechanism of immune
escape. Sci. Transl. Med. 4, 127ra37.
Taube, J.M., Klein, A., Brahmer, J.R., Xu, H., Pan, X., Kim, J.H., et al., 2014. Association
of PD-1, PD-1 ligands, and other features of the tumor immune
microenvironment with response to anti-PD-1 therapy. Clin. Cancer Res. 20,
50645074.
Therasse, P., Arbuck, S.G., Eisenhauer, E.A., Wanders, J., Kaplan, R.S., Rubinstein, L.,
et al., 2000. New guidelines to evaluate the response to treatment in solid
tumors European Organization for Research and Treatment of Cancer. National
Cancer Institute of the United States, National Cancer Institute of Canada. J.
Natl. Cancer Inst. 92, 205216.
Thompson, R.H., Gillett, M.D., Cheville, J.C., Lohse, C.M., Dong, H., Webster, W.S.,
et al., 2004. Costimulatory B7-H1 in renal cell carcinoma patients: indicator of
tumor aggressiveness and potential therapeutic target. Proc. Natl. Acad. Sci. U.
S. A. 101, 1717417179.
Topalian, S.L., Hodi, F.S., Brahmer, J.R., Gettinger, S.N., Smith, D.C., McDermott, D.F.,
et al., 2012. Safety, activity, and immune correlates of anti-PD-1 antibody in
cancer. N. Engl. J. Med. 366, 24432454.
Tumeh, P.C., Harview, C.L., Yearley, J.H., Shintaku, I.P., Taylor, E.J., Robert, L., et al.,
2014. PD-1 blockade induces responses by inhibiting adaptive immune
resistance. Nature 515, 568571.
Wainwright, D.A., Chang, A.L., Dey, M., Balyasnikova, I.V., Kim, C.K., Tobias, A., et al.,
2014. Durable therapeutic efcacy utilizing combinatorial blockade against
IDO, CTLA-4, and PD-L1 in mice with brain tumors. Clin. Cancer Res. 20,
52905301.
Wargo, J.A., Cooper, Z.A., Flaherty, K.T., 2014. Universes collide: combining
immunotherapy with targeted therapy for cancer. Cancer Discov. 4,
13771386.

98

S. Gandini et al. / Critical Reviews in Oncology/Hematology 100 (2016) 8898

Weber, J.S., Kudchadkar, R.R., Yu, B., Gallenstein, D., Horak, C.E., Inzunza, H.D., et al.,
2013. Safety, efcacy, and biomarkers of nivolumab with vaccine in
ipilimumab-refractory or -naive melanoma. J. Clin. Oncol. 31, 43114318.
Wolchok, J.D., Kluger, H., Callahan, M.K., Postow, M.A., Rizvi, N.A., Lesokhin, A.M.,
et al., 2013. Nivolumab plus ipilimumab in advanced melanoma. N. Engl. J.
Med. 369, 122133.
Wolchok J.D., Chiarion-Sileni V., Gonzalez R., Rutkowski P., Grob J.-, Cowey C.L.,
et al., 2015. Efcacy and safety results from a Phase III trial of nivolumab slone
or combined with ipilimumab vs. ipilimumab alone in treatment-nave
patients with advanced melanoma (CheckMate 067); Abstract #LBA1:ASCO.
Woo, S.R., Turnis, M.E., Goldberg, M.V., Bankoti, J., Selby, M., Nirschl, C.J., et al.,
2012. Immune inhibitory molecules LAG-3 and PD-1 synergistically regulate
T-cell function to promote tumoral immune escape. Cancer Res. 72, 917927.

Yoshii, A., Plaut, D.A., McGraw, K.A., Anderson, M.J., Wellik, K.E., 2009. Analysis of
the reporting of search strategies in Cochrane systematic reviews. J. Med. Libr.
Assoc. 97, 2129.
van Houwelingen, H.C., Arends, L.R., Stijnen, T., 2002. Advanced methods in
meta-analysis: multivariate approach and meta-regression. Stat. Med. 21,
589624.
Puzanov, I., Dummer, R., Schachter, J., Pavlick, A.C., Gonzalez Rene, Ascierto,
Antonio, Paolo, et al., 2015. Efcacy based on tumor PD-L1 expression in
KEYNOTE-002, a randomized comparison of pembrolizumab (pembro;
MK-3475) versus chemotherapy in patients (pts) with ipilimumab-refractory
(IPI-R) advanced melanoma (MEL). J. Clin. Oncol. 33, 3012, ASCO Annual
Meeting, 2015.