Cancer Treatment with Anti-PD-1/PD-L1 Agents: Is PD-L1 Expression a Biomarker for Patient Selection?

Transcription

1 Drugs (2016) 76: DOI /s x REVIEW ARTICLE Cancer Treatment with Anti-PD-1/PD-L1 Agents: Is PD-L1 Expression a Biomarker for Patient Selection? Lucia Festino 1 Gerardo Botti 1 Paul Lorigan 2 Giuseppe V. Masucci 3 Jason D. Hipp 4 Christine E. Horak 4 Ignacio Melero 5 Paolo A. Ascierto 1 Published online: 26 May 2016 Springer International Publishing Switzerland 2016 Abstract Strategies to help improve the efficacy of the immune system against cancer represent an important innovation, with recent attention having focused on antiprogrammed death (PD)-1/PD-ligand 1 (L1) monoclonal antibodies. Clinical trials have shown objective clinical activity of these agents (e.g., nivolumab, pembrolizumab) in several malignancies, including melanoma, non-smallcell lung cancer, bladder cancer, squamous head and neck cancer, renal cell cancer, ovarian cancer, microsatelliteunstable colorectal cancer, and Hodgkin s lymphoma. Expression of PD-L1 in the tumor microenvironment appears to be crucial for therapeutic activity, and initial trials suggested positive PD-L1 tumor expression was associated with higher response rates. However, subsequent observations have questioned the prospect of using PD-L1 expression as a biomarker for selecting patients for therapy, especially since many patients considered PD-L1-negative experience a benefit from treatment. Importantly, there is not yet a definitive test for determination of PD-L1 and a cut-off reference for PD-L1-positive status has not been established. Immunohistochemistry with different antibodies and different thresholds has been used to define PD- L1 positivity (1 50 %), with no clear superiority of one threshold over another for identifying which patients respond. Moreover, the type of cells on which PD-L1 expression is most relevant is not yet clear, with immune infiltrate cells and tumor cells both being used. In conclusion, while PD-L1 expression is often a predictive factor for treatment response, it must be complemented by other biomarkers or histopathologic features, such as the composition and amount of inflammatory cells in the tumor microenvironment and their functional status. Multi-parameter quantitative or semi-quantitative algorithms may become useful and reliable tools to guide patient selection. & Paolo A. Ascierto paolo.ascierto@gmail.com Lucia Festino luciafestino1984.lf@gmail.com Gerardo Botti gbotti1@alice.it Paul Lorigan paul.lorigan@christie.nhs.uk Giuseppe V. Masucci giuseppe.masucci@ki.se Jason D. Hipp jason.hipp@bms.com Christine E. Horak christine.horak@bms.com Ignacio Melero imelero@unav.es Istituto Nazionale Tumori Fondazione G. Pascale, Via Mariano Semmola, Naples, Italy University of Manchester/The Christie NHS Foundation Trust, Manchester, UK Department of Oncology-Pathology, Karolinska Institutet and Hospital, Stockholm, Sweden Bristol-Myers Squibb, Princeton, NJ, USA CIMA and CUN, University of Navarra, Pamplona, Spain

2 926 L. Festino et al. Key Points The anti-pd-1/pd-l1 monoclonal antibodies are revolutionizing the treatment of cancer. Recent approval by the FDA of pembrolizumab and nivolumab in melanoma and nivolumab in non-small-cell lung cancer represents only the first evidence of such revolution. In the next few years, approval of anti-pd-1 could be extended to other cancers. The use of PD-L1 expression on tumor tissue as a predictive marker for anti PD-1/PD-L1 treatment response is a current focus of debate. However, data from melanoma II trials and II studies in other cancers suggest PD-L1 expression may not be useful as a predictive biomarker. 1 Introduction After a long period of time during which immunotherapy was typically associated with disappointing results in patients with cancer, recent years have seen dramatic improvements in outcomes at least in certain tumor types, brought about by an understanding of the molecular controls of the immune system and the development of drugs that act on immune system cells. The success of sipuleucel- T in the treatment of prostate cancer [1] and ipilimumab in melanoma [2, 3], the positive results achieved with antiprogrammed death (PD)-1/PD-ligand 1 (L1)agents (Table 1) in melanoma [4 8], non-small-cell lung cancer (NSCLC) [7, 9 11], kidney cancer [7], and urothelial bladder cancer [12] as well as the encouraging data from the combination of ipilimumab and nivolumab in melanoma [4, 13, 14] and renal cell cancer (RCC) [15 18] most likely herald the beginning of a bright future for immunotherapy. On one level, these successes are not unexpected given that human cancer is well known to encompass a number of genetic alterations, including non-synonymous variants, resulting in neo-antigens recognizable by T lymphocytes. Although there is abundant evidence for human immune reactivity against solid tumors [19, 20], this response is commonly ineffective in curtailing tumor development and progression. The tumor microenvironment is able to develop multiple resistance mechanisms, including the ability to induce local and systemic immunosuppression, immune tolerance, and local and systemic dysfunction in T and NK cell signaling [21, 22]. Strategies to help improve the efficacy of the immune system s action against cancer represent an important innovation. One aim of current cancer immunotherapy is to overcome specific obstacles to the activity of the immune system through inhibition of so-called immune checkpoints. The archetype for this kind of therapy is ipilimumab, a monoclonal antibody that blocks the interaction of the co-inhibitory receptor cytotoxic T-lymphocyte-associated protein (CTLA)-4 on T cells with its ligands B7-1 and B7-2 (CD80 and CD86), expressed on professional antigen-presenting-cells. CTLA-4 competes with the stimulatory receptorcd28 for binding to these ligands and provides a negative signal in the T cell, leading to inhibition of T cell activation and expansion. Inhibition of CTLA-4 results in effective treatment of metastatic melanoma [2, 3] and has also shown overall survival benefit as postsurgical adjuvant treatment in earlier stages of the disease [23], leading to approval by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA). It is also possible that anti CTLA-4 monoclonal antibodies cause regulatory T lymphocyte depletion in the tumor microenvironment [24, 25]. However, attention has focused on the PD-1/PD-L1/PD- L2 pathway due to accumulating evidence for unprecedented patient benefit. PD-1is a transmembrane glycoprotein of the Ig superfamily that acts as a T cell co-inhibitory receptor [26, 27]. Binding of PD-1 to any of its ligands (PD-L1 or PD-L2) results in inhibition of T-cell activation and function. The main mechanism of action is that ligated PD-1 drags the SHP-2 tyrosine phosphatase linked to its cytoplasmic tail to the immune synapse. As a result of dephosphorylation of several substrates, PD-1 downregulates signaling from CD28 and thetcr-cd3 complex [26]. Dysfunctional T cells chronically activated by antigen, often described as exhausted T cells, abundantly express PD-1 on their plasma membrane [28]. PD-1 expression on T cells is variable and influenced by a number of factors. After cognate antigen detection, PD-1 is rapidly up-regulated on T cells, while its expression decreases after antigen clearance. Modulation of PD-1 expression on lymphocytes is regulated by a number of inflammatory cytokines, such as interferon (IFN)-c and interleukin (IL)-2. PD-L1 (B7-H1) is the primary PD-1 ligand and is broadly expressed on hematopoietic and non-hematopoietic cells. It is constitutively expressed on T and B cells and on macrophages and is also expressed in non-immune parenchymal tissue, such as endothelial and muscle cells. Following stimulation by inflammatory cytokines, such as IFNs, expression of PD-L1is greatly increased [29]. PD-L2 (B7-DC) is the second PD-1 ligand to be discovered and is expressed more selectively on activated macrophages and dendritic cells while it is not found on epithelial or mesenchymal cells. In cancer PD-L2 is ectopically expressed in a fraction of cases but its functional significance has not been studied, at least in part due to the lack of good

3 PD-L1 Expression as a Predictive Biomarker for Anti-PD-1/PD-L1 Treatment 927 Table 1 Anti PD1/PD-L1 agents in clinical development Target Drug name Source Clinical testing phase PD-1 MEDI0680 MedImmune/AstraZeneca Phase I Nivolumab Bristol-Myers Squibb, Ono Pharmaceuticals Approved, unresectable melanoma (Japan, USA, Europe) and squamous NSCLC (USA and Europe) Pembrolizumab Merck Approved, unresectable melanoma (USA and Europe) Pidilizumab Cure Tech Phase I II PD-L1 BMS Bristol-Myers Squibb Phase I MEDI4736 (durvalumab) MedImmune/AstraZeneca Phase I III MPDL3280A (atezolizumab) Genentech/Roche Phase I III MSB C EMD Serono Phase I II NSCLC non-small-cell lung cancer antibodies for immunohistochemistry. Given that both PD- 1 and PD-L1 are up-regulated in tissues during inflammation, and that their interaction determines a functional inhibition of lymphocytes, PD-1/PD-L1 could play a key role in the regulation of the local inflammatory response and thereby be able to protect the tissue against immunemediated damage [30]. This effect seems to be of extraordinary importance in the placenta to curtail maternal alloreactive immune responses that would otherwise severely damage the embryo [31]. In addition to inflammation, hypoxia operating via hypoxia-inducible factor (HIF)-1a has recently been described to top-regulate PD-L1 expression on both tumor and myeloid cells [32].While CTLA-4 acts primarily during the early activation of T cells in lymphoid tissue, PD-1 is mostly active during the T-cell effector phase. Indeed, its major role is to reduce the damage caused by T cells in the periphery of an inflammatory response to infection rather than to limit autoimmunity. The aim of this review is to present the latest data on PD-1/PD-L1 therapies and to discuss the potential utility of PD-L1 expression as predictor of treatment outcomes and, as such, a biomarker for patient selection. 2 Results from Clinical Trials with Anti-PD-1/PD- L1 The approval of ipilimumab in 2011 for the treatment of advanced melanoma was a milestone in clinical cancer research. The most acclaimed success of ipilimumab is its impact on survival with a very modest effect on PFS [2]. Because response rates to ipilimumab were low, the benefit of ipilimumab was underestimated. This was the first treatment to demonstrate a survival benefit in a randomized trial. Patients treated with ipilimumab experienced a median improvement in survival of approximately 3.5 months compared with those patients receiving vaccine therapy. More importantly, a proportion of patients went on to be long-term survivors [2]. This finding was borne out in subsequent ipilimumab trials and a recent meta-analysis showed that treatment results in long-term survival in approximately 20 % of patients with metastatic melanoma [33]. However, clinical results from -II studies and from large II trials with the anti-pd-1 antibodies nivolumab and pembrolizumab [7, 10, 34 36] and anti-pd- L1 agentsmedi4736 and atezolizumab suggest that blockade of this pathway may be an even more powerful tool for the treatment of cancer than CTLA-4 inhibition. Several very encouraging studies have recently reported on the activity of anti-pd-1/pd-l1 agents against several malignancies, which, in addition to melanoma, also include NSCLC [37 39], bladder cancer [12, 40], squamous cell carcinoma of head and neck (SCCHN) [41, 42], RCC [17, 43], microsatellite-unstable (MSI) colorectal cancer [44] and Hodgkin s lymphoma [45]. 3 PD-1 Targeted Agents Efficacy data of anti PD-1/PD-L1 agents in melanoma, NSCLC, and other tumors are summarized in Tables 2, 3, Nivolumab Nivolumab was the first anti-pd-1 antibody evaluated in cancer patients and has been extensively assessed. It is a fully human IgG4 monoclonal antibody with high affinity to PD-1 that blocks binding of PD-1 to both PDL-1 and PDL-2, thereby co-stimulating proliferation of T cells and the production of cytokines [46]. The activity and safety profile of nivolumab monotherapy was evaluated in a large trial; objective responses were observed in patients with metastatic melanoma (objective response rate [ORR] 17 %), NSCLC (ORR 10 %), RCC (ORR 12 %), and

4 928 L. Festino et al. Table 2 Anti PD1/PD-L1agents in metastatic melanoma Naïve/pretreated Study type No. of patients ORR PFS OS Nivolumab Checkmate 066 [10] Pretreated Prospective II % 5.1 m 1 y OS 72.9 % Checkmate 037 [34] Pretreated Prospective II % 4.7 m Nivolumab plus ipilimumab Melanoma [13] Pretreated Prospective b % 1 y OS 94 % 2 y OS 88 % Checkmate 067 [56] 57.6 % 11.5 m Checkmate 064 [57] Naïve? pretreated Prospective I % Pembrolizumab Melanoma [61] Pretreated Prospective % 22 m 1 y OS % Melanoma [62] Pretreated Prospective I % m Melanoma [36] Naïve? pretreated Prospective II % 6 m PFS % 1yOS % Atezolizumab Melanoma I [5] Pretreated Prospective % 24 w PFS 35 % M month, ORR objective response rate, OS overall survival, PFS progression-free survival, w week, y year Table 3 Anti PD1/PD-L1 agents in lung cancer Naïve/ pretreated Study type No. of patients ORR PFS OS Nivolumab NSCLC [50] Naïve Prospective % 15.6 w 98.3 w NSCLC [51] Pretreated Prospective I % Squamous NSCLC Checkmate 017 [52] Pretreated Prospective II % 3.5 m 9.2 m Non-squamous NSCLC Checkmate 057 Pretreated Prospective II % 2.3 m 12.2 m [53] Nivolumab? Ipilimumab NSCLC [58] Naïve Prospective % 16 w Pembrolizumab NSCLC KEYNOTE 001[37] Pretreated Prospective % 3.7 m 12.0 m NSCLCI KEYNOTE 010 [11] SCLC KEYNOTE 028 [61] Pretreated Prospective I/ III m Pretreated Prospective b % Mesothelioma KEYNOTE 028 [66] Pretreated Prospective b % Atezolizumab NSCLC [75] Pretreated Prospective % 24 w PFS 42 % 1yOS 82 % NSCLC [76] Pretreated Prospective I m Durvalumab NSCLC [39] Pretreated Prospective /II % M month, NSCLC non-small-cell lung cancer, ORR objective response rate, OS overall survival, PFS progression-free survival, SCLC small-cell lung cancer, w week, y year ovarian cancer (ORR 6 %) [5, 7]. A recent update of this trial reported an ORR of 41 %, a median progression-free survival (PFS) of 9.7 months, a median overall survival (OS) of 20.3 months, and 1, 2, and 3-year OS rates of 63, 48, and 41 %, respectively, in patients with metastatic melanoma [47].

5 PD-L1 Expression as a Predictive Biomarker for Anti-PD-1/PD-L1 Treatment 929 Table 4 Anti PD1/PD-L1 agents in other tumors Naïve/pretreated Study type No. of patients ORR PFS OS Nivolumab Bladder cancer [12] Pretreated Prospective % 24 w % Ovarian cancer [55] Pretreated Prospective I % Hodgkin s lymphoma [45] Pretreated Prospective % 24 mpfs 86 % RCC Checkmate 025 [54] Pretreated Prospective II % 4.6 m 25 m Pembrolizumab Urothelial cancer KEYNOTE 012 [40] Naïve? pretreated Prospective b 33 SCCHN KEYNOTE 012 [41] Esophageal Cancer KEYNOTE 028 [67] Naïve? pretreated Prospective b % Pretreated Prospective b % Atezolizumab mrcc [79] Pretreated Prospective % mpfs 24 w mrcc (18) Durvalumab SCCHN [81] Pretreated Prospective Phase I % M month, ORR objective response rate, OS overall survival, (m)pfs (median) progression-free survival, RCC renal cell carcinoma, SCCHN squamous cell cancer of the head and neck, w week Used as first-line therapy within the randomized CA trial, nivolumab showed an ORR of 40 versus 13.9 % with dacarbazine. At 1 year, the OS was 72.9 % with nivolumab compared with 42.1 % with dacarbazine while the median PFS was 5.1 months in the nivolumab group versus 2.2 months in the dacarbazine group [10]. Nivolumab has also been assessed in patients with metastatic melanoma who previously progressed on ipilimumab. In the CheckMate 037 trial, 272 patients were randomly allocated to nivolumab and 133 to chemotherapy. Confirmed responses were reported in 32 % of patients treated with nivolumaband in 11 % of patients in the control arm, with a median time to response of 2.1 versus 3.5 months. Responses with nivolumab were fast and durable, with 95 % of responses ongoing at the time of data cut-off. Grade 3 and 4 toxicity with nivolumab occurred in 12 (5 %) of patients versus nine (9 %) in the chemotherapy arm and included increased lipase, increased alanine aminotransferase, anemia, and fatigue [34]. With regard to the BRAF mutation status, the ORR in patients with BRAF wild-type was 34 % compared with 23 % in patients with BRAF mutations [34]. Used in combination with a multipeptide vaccine, the ORR was 25 % with a median duration of response (DOR) not reached at a median follow-up of 8.1 months, with no beneficial or detrimental effect attributable to the peptide vaccine [48]. Nivolumab has also been studied in other tumor types. Survival data obtained from the study of nivolumab in solid tumors were particularly encouraging in patients with NSCLC, with a median OS of 14.9 months in 37 patients receiving nivolumab 3 mg/kg, and 1-, 2-, and 3-year OS rates of 56 % (95 % CI 38 71), 42 % (95 % CI 24 58), and 27 % (95 % CI 12 43), respectively [49]. In light of these positive results, nivolumab is under evaluation in several trials in patients with NSCLC as first- and second-line treatment. Used as first-line therapy in a phase I trial in patients with NSCLC, nivolumab resulted in an ORR of 21 %, median PFS of 15.6 weeks, and median OS of 98.3 weeks [50]. In heavily pretreated NSCLC patients the ORR was 14.5 % (95 % CI ) [51]. In the phase III CheckMate 017 study of 260 platinum-pretreated patients with squamous NSCLC (NCT ), nivolumab demonstrated significantly superior OS versus docetaxel, with a 41 % reduction in the risk of death (hazard ratio [HR]: 0.59 [95 % CI ; p = ]) in a prespecified interim analysis. The median OS was 9.2 months in the nivolumab arm (95 % CI ) and 6 months in the docetaxel arm (95 % CI ) [52]. As a result of these data and unpublished results, the FDA granted approval for nivolumab for second-line treatment of squamous NSCLC in March 2015 ( NewsEvents/Newsroom/PressAnnouncements/ucm htm). In the CheckMate 057 study, nivolumab was compared with docetaxel in patients with advanced non-squamous NSCLC after failure of platinum-based doublet chemotherapy. Nivolumab demonstrated superior OS

6 930 L. Festino et al. (median 12.2 months in the nivolumab group vs. 9.4 months in the docetaxel group) showing 27 % reduction in risk of death (HR 0.73; p = ) and improved ORR (19.2 vs %; p = ). Median DOR was 17.2 months in the nivolumab group versus 5.6 months in the docetaxel group, while median PFS was 2.3 months with nivolumab compared to 4.2 months with the docetaxel group and 1-year PFS was 19 versus 8 months [53]. Based on these data, the FDA has also approved the use of nivolumab in patients with non-squamous NSCLC with progressive disease during or after platinum-based chemotherapy ( PressAnnouncements/ucm htm). Within the trial, improved OS with nivolumab was also been observed in patients with RCC with a 1-year OS of 70 % and 2-year OS of 50 % (median duration of over 22 months, 95 % CI 13.6 NR) [7]. Interesting data from a I trial were recently presented, showing an ORR of %, a median OS between 18.2 and 25.5 months, and 36 months OS ranging from 32 to 40 % according to the different dose levels [15]. In the Check- Mate 025 trial that compared nivolumab with everolimus in pretreated mrcc, there was a significant improvement in terms of ORR (25 vs. 5 %, p \ ) and median OS (25.0 vs months, p = ) with nivolumab, which also had a good safety profile [54]. Among other cancers, in a study of 20 patients with platinum-resistant ovarian cancer, two had a complete response and one patient had a partial response to nivolumab. The partial responder had responses for 5 months, and two complete responders survived without disease progression for 14 and 17 months [55]. On the basis of preclinical studies suggesting that Reed-Sternberg cells can evade immune detection and have increased the expression of PD-L1 and PD-L2 due to an alteration in chromosome 9p24.1, nivolumab was also tested in a study enrolling patients with relapsed or refractory Hodgkin s lymphoma. Of 23 patients, 20 (87 %) had an objective response including 17 % with a complete response and 70 % with a partial response; the remaining three patients had stable disease. Responses were durable with PFS at 24 months of 86 % [45]. Combined anti-pd-1 and anti-ctla-4 therapy is also a subject of increasing attention, with nivolumab being investigated in combination with ipilimumab in a study of 86 patients with metastatic melanoma (NCT ) [14]. In the most recent cohort of 41 patients treated at the maximum tolerated dose (MTD), the ORR was 43 %, with 10 % complete responses and 28 % of patients experiencing C80 % reduction in tumor burden. Responses were durable, and 82 % were ongoing at the time of analysis with median DOR not reached. With concurrent nivolumab 3 mg/kg plus ipilimumab 3 mg/kg every 2 weeks for four cycles of induction followed by maintenance with biweekly nivolumab, the 1-year OS was 94 % and 2-year OS was 88 % [13]. In patients receiving sequential nivolumab and ipilimumab treatment, the ORR was 41 %, which reached 63 % in the cohort with nivolumab at a dose of 1 mg/kg. A total of 34 % of patients experienced tumor burden reduction C80 % at 36 weeks, and the 1-year OS was 70 % [52]. Grade 3 and 4 toxicity were relatively less frequent with sequential therapy compared to the combination, occurring in 64 % of patients of the combination cohorts and only 24 % of patients of the sequential cohorts [52]. In the II CheckMate 067 study, 945 untreated patients with metastatic melanoma were randomized to receive nivolumab monotherapy, nivolumab plus ipilimumab, or ipilimumab monotherapy. Nivolumab alone or combined with ipilimumab resulted in significantly longer PFS than ipilimumab alone (6.9 and 11.5 months vs. 2.9 months). The ORR in the three arms was respectively 43.8, 57.6, and 19 % [56]. In the I CheckMate 064 study that evaluated the efficacy and safety of nivolumab followed by ipilimumab versus ipilimumab followed by nivolumab in advanced melanoma, a consistent improvement in ORR (35.3 vs. 10 % at week 15 and 47.7 vs % at week 25 and progression rate (38.2 vs % at week 15 and 38.2 vs % at week 25) was demonstrated with the sequence of nivolumab followed by ipilimumab versus ipilimumab followed by nivolumab [57]. On the basis of these results the combination has been granted FDA approval for the treatment of BRAF wild type unresectable melanoma ( OnDrugs/ApprovedDrugs/ucm htm). In patients with NSCLC, nivolumab 1 mg/kg in combination with ipilimumab 3 mg/kg attained an ORR of 22 % and PFS of 16 weeks [58]. In patients with EGFRmutant NSCLC, nivolumab has been assessed in combination with erlotinib, achieving an ORR of 19 % (4/20) [59]. Median PFS was 29.4 weeks with a 24-week PFS rate of 51 %. Median OS was not reached, while 1-year OS was 73 %.The combination of nivolumab with platinum-based chemotherapy has also been evaluated, with an ORR of %, 24-week PFS of %, and 1-year OS of % [60]. Combined therapy with nivolumab and ipilimumab has also been studied in patients with RCC. Results of expansion cohorts of the CheckMate 016 study showed an ORR of 38 % (18/47) with nivolumab 3 mg/kg plus ipilimumab 1 mg and 43 % (20/47) with nivolumab 1 mg/ kg plus ipilimumab 3 mg/kg. The median PFS was, respectively, 30.3 and 36 weeks, and the OS was not reached at time of analysis [16]. The efficacy and safety of nivolumab plus pazopanib or sunitinib in the treatment of metastatic RCC was also assessed. Nivolumab plus

7 PD-L1 Expression as a Predictive Biomarker for Anti-PD-1/PD-L1 Treatment 931 sunitinib had an ORR of 52 % (17/33) with a disease control rate (DCR) of 82 % (27/33) and a 24-week PFS of 79 %. The median DOR was 37.1 weeks. Nivolumab plus pazopanib gave equally encouraging results with an ORR of 45 % (9/20), a DCR of 80 % (16/20), a median DOR of 30.1 weeks, and a 24-week PFS of 55 % [18]. 3.2 Pembrolizumab Pembrolizumab, another anti-pd-1 antibody, was the first anti-pd1 therapy to ever receive approval by the FDA. It is a humanized monoclonal antibody, an IgG4/kappa isotype, which blocks the interaction between PD-1 and its ligands PD-L1 and PD-L2. In a dose-escalation study pembrolizumab monotherapy at doses of 1, 3, and 10 mg/ kg were well tolerated with no dose-limiting toxicities [35]. In the expansion cohort of this trial, which included 411 patients with advanced melanoma (190 ipilimumab-naïve and 221 ipilimumab-pretreated), pembrolizumab attained an ORR of 34 %. The ORRs were 40 % in ipilimumabnaïve patients and 28 % in ipilimumab-pretreated patients. Pembrolizumab showed durable responses with 88 % ongoing at the time of the analysis and a manageable safety profile across dose and schedules both in ipilimumab-naïve and in ipilimumab-pretreated patients. The median PFS was 5.5 months, while the median OS was not yet reached at a follow-up of 24 months; the latest data available are a 1-year OS rate of 69 % and 18-month OS rate of 62 % [6]. In the expansion phase of this trial,173 patients with ipilimumab-refractory advanced melanoma were randomized to receive pembrolizumab 2 mg/kg or 10 mg/kg. The ORR was 26 % at both doses and median PFS was 22 weeks for the pembrolizumab 2 mg/kg group and 14 weeks for the pembrolizumab 10 mg/kg group. One-year OS rates were 58 % for the 2 mg/kg group and 63 % for the 10 mg/kg group [61]. In a multinational, randomized I trial, both pembrolizumab 2 mg/kg and 10 mg/kg dose regimens substantially improved PFS compared with investigator choice chemotherapy in 540 patients with advanced melanoma. Despite the median PFS being similar in the three arms, the percentage of patients surviving beyond 6 and 9 months of treatment was higher with pembrolizumab compared with chemotherapy. Six-month PFS was 34 % with pembrolizumab 2 mg/kg, 38 % with pembrolizumab 10 mg/kg, and 16 % with chemotherapy, while 9-month PFS was 24, 29, and 8 %, respectively. Mean PFS was also significantly increased with pembrolizumab versus chemotherapy, reaching 5.4 months with pembrolizumab 2 mg/kg, 5.8 months with pembrolizumab 10 mg/kg, and 3.6 months with chemotherapy. The ORRs in these three groups were 21, 25, and 4 %, respectively [62]. Pembrolizumab was compared to ipilimumab in a large II trial in which 834 patients with advanced melanoma were randomized to receive pembrolizumab at a dose of 10 mg/kg every 2 weeks or every 3 weeks or four doses of ipilimumab 3 mg/kg every 3 weeks. The 6-month PFS was 47.3 % for pembrolizumab every 2 weeks, 46.4 % for pembrolizumab every 3 weeks, and 26.5 % for ipilimumab, and the 1-year OS rate was 74.1, 68.4, and 58.2 %, respectively. ORR was also improved with pembrolizumab administered every 2 weeks (33.7 %) and every 3 weeks (32.9 %) compared with ipilimumab (11.9 %) [36]. In the expansion cohort of the KEYNOTE 001 study, 495 patients with NSCLC received pembrolizumab at a dose of either 2 mg/kg or 10 mg/kg every 3 weeks or 10 mg/kg every 2 weeks; ORR was 19.4 %, median DOR was 12.5 months, median PFS was 3.7 months, and median OS was 12.0 months [38]. Pembrolizumab was also used in combination with ipilimumab in patients within the D cohort of the KEYNOTE 021 trial, which enrolled 17 pretreated NSCLC patients and showed a very high DCR, with an ORR of 55 %, with one complete response, seven partial responses, and a stable disease rate of 45 % [63]. Within the same trial, pembrolizumab was combined with chemotherapy in 44 patients with untreated NSCLC, and achieved an ORR of between 30 % and 58 % [64]. In a randomized II study for second-line NSCLC treatment, OS was significantly longer versus docetaxel for pembrolizumab 2 mg/kg (HR 0.71, 95 % CI ; p = ) and pembrolizumab 10 mg/kg (HR 0.61, 95 % CI ; p \ ). Median PFS was 3.9 months with pembrolizumab 2 mg/kg, 4.0 months with pembrolizumab 10 mg/kg, and 4.0 months with docetaxel, with no significant difference versus docetaxel for pembrolizumab 2 mg/kg (HR 0.88, 95 % CI ; p = 0.07) or pembrolizumab 10 mg/kg (HR 0.79, 95 % CI ; p = 0.004). Among patients with at least 50 % of tumor cells expressing PD-L1, OS was significantly longer than with docetaxel for pembrolizumab 2 mg/kg (median 14.9 vs. 8.2 months; HR 0.54, 95 % CI ; p = ) and pembrolizumab 10 mg/kg (17.3 vs. 8.2 months; HR 0.50, 95 % CI ; p \ ) [11]. This study led to FDA approval ( Drugs/InformationOnDrugs/ApprovedDrugs/ucm htm) and, for the first time in immunotherapy, with a companion diagnostic for IHC manufactured by Dako to determine patients with over 50 % of malignant cells expressing PD-L1. In patients with extensive stage SCLC PD-L1 positive after failure or unable to receive standard therapy, pembrolizumab showed an ORR of 35 % (95 % CI 15 59). Responses were durable, with six of seven responses ongoing at time of data cut-off [65]. In the mesothelioma cohort of the KEYNOTE 028 trial, pembrolizumab obtained an ORR of 28 % with a DCR of 76 % in PD-L1

8 932 L. Festino et al. selected pretreated patients [66]. In addition, a small trial that assessed pembrolizumab in 33 patients with urothelial cancer as part of the b KEYNOTE-012 trial reported an ORR of 25 % (95 % CI 11 45), a median PFS of 2 months, and a median OS of 12.7 months, with 1-year OS of 52.9 % [40]. Pembrolizumab has also been evaluated in patients with SCCHN. In the SCCHN cohort of the KEYNOTE 012 trial, 132 patients were treated with pembrolizumab 200 mg every 3 weeks; the ORR was 24.8 % (95 % CI ), with one complete response. The median DOR was not reached at the time of data cut-off [41]. In patients with PD-L1-positive esophageal cancer treated within the KEYNOTE 028 b trial, pembrolizumab showed an ORR of 30.4 % (7/37) with a median DOR of 40 weeks [67]. Promising results with pembrolizumab have also been reported in Asiatic patients with PD-L1-positive gastric cancer, with an ORR of 22.2 months, a median PFS of 1.9 months (95 % CI ), and a median OS of 11.4 months (95 % CI 5.7 NR) [68]. In the ovarian cohort of the b KEYNOTE 028 study, the ORR obtained was 11.5 %, with one complete response and a median DOR not reached at time of data cut-off [69]. In 27 PD-L1- positive triple-negative breast cancer patients, pembrolizumab achieved an ORR of 18.5 %; responses were durable, with median DOR not reached at the time of data cut-off [70]. In a study of patients with heavily pretreated Hodgkin s lymphoma, the ORR obtained was 66 % (19/ 29), with 21 % complete responses. The clinical benefit rate was 86 % (25/29) [71]. 3.3 Pidilizumab Pidilizumab (CT-011) is a humanized anti PD-1 IgG1k in development in both hematology and melanoma. In a phase II trial of 66 patients with diffuse large B-cell lymphoma (DLBCL), the 16-month PFS rate was 72 % overall and 70 % among 24 high-risk patients after salvage chemotherapy [72]. Another I study evaluated the safety and efficacy of pidilizumab in patients with metastatic melanoma and reported an ORR using immune-related response criteria (irrc) of 5.9 % for the entire cohort and 10 % for the cohort receiving 1.5 mg/kg previously treated with ipilimumab. Patients with prior ipilimumab therapy also had immune-related stable disease (irsd) of 53.7 % (vs %) and median PFS of 2.8 (vs. 1.9) months. One-year OS rate was 64.5 % with no significant differences between different subgroups, a rate comparable to that seen with other anti-pd-1 s [73]. The specificity of pidilizumab for PD-1 has been questioned and investigations are exploring this issue. 4 Anti PD-L1 targeted therapies 4.1 BMS BMS is a high-affinity, fully human, PD-L1- specific, IgG4 (with the S228P mutation) that inhibits the binding of PD-L1 to both PD-1 and CD80. A doseescalation study in 207 patients with different solid tumors reported ORRs of 17 % (9/52) in patients with metastatic melanoma, 12 % (2/17) in RCC, 10 % (5/49) in NSCLC, and 6 % (1/17) in ovarian cancer. Eight of the 16 responding patients with at least 1 year of follow-up had a response lasting at least 1 year [9]. In spite of these findings, no trials are listed for this agent in clinicaltrials.gov. 4.2 Atezolizumab Another PD-L1 agent is atezolizumab, a human monoclonal antibody that contains a mutated Fc domain designed to avoid Fc-receptor binding and therefore any PD-L1-targeted antibody-dependent cellular cytotoxicity (ADCC). Atezolizumab was assessed in a still ongoing large study in patients with locally advanced or metastatic solid tumors, and preliminary data have shown broad activity across several solid tumors, including NSCLC, RCC, melanoma, CRC, and gastric cancer, with an ORR of 21 % (25/122) and 24-week PFS of 44 % [74]. Data on 35 patients with metastatic melanoma who initiated treatment at doses of 1 20 mg/kg showed an ORR of 26 % (9/35) with some patients experiencing tumor shrinkage within days of initial treatment. The 24-week PFS was 35 % [5]. In 88 patients with NSCLC with squamous or non-squamous histology, an ORR of 21 % was observed and included several patients experiencing rapid tumor shrinkage. The median DOR was 67 weeks, 24-week PFS was 42 %, and 1-year OS was 82 % [75]. Based on these results, I and III trials are ongoing to evaluate the efficacy of atezolizumab in patients with locally advanced or metastatic NSCLC. In the I POPLAR study, atezolizumab was compared with docetaxel in pretreated patients with locally advanced or metastatic NSCLC. The median OS was 12.6 months (9.7, 16.4) in the experimental arm versus 9.7 months (8.6, 12.0) in the docetaxel arm, with an HR of 0.73 (0.53, 0.99) [76]. In patients with metastatic platinum-pretreated urothelial bladder cancer, the ORR observed was between 11 and 43 %, depending on the level of PD-L1 [11]. The 24-week OS rate was between 71 and 85 %, depending on PD-L1 expression level [77]. In a I trial of 316 patients with locally advanced or metastatic cancer of the bladder, renal pelvis, ureter, or urethra who received atezolizumab

9 PD-L1 Expression as a Predictive Biomarker for Anti-PD-1/PD-L1 Treatment mg every 3 weeks, ORR was 15 % (95 % CI 11 20; p = ), and responses were durable, with median DOR not reached in any subgroup [78]. In the metastatic RCC expansion cohort of a a trial with atezolizumab, the ORR was 15 % with 76 weeks median DOR and 24 weeks median PFS [79]. Promising results have also been reported with the use of atezolizumab in combination with bevacizumab in solid tumors. Objective responses were observed in several types of tumors, particularly in patients with treatment-naïve RCC (ORR 30 %) and metastatic CRC treated with atezolizumab plus bevacizumab and FOLFOX as first-line treatment (ORR 47 %) [80]. 4.3 MEDI4736 (Durvalumab) Durvalumab (MEDI4736) is a human IgG1j antibody anti- PD-L1 with an engineered Fc domain. In an ongoing phase I multicenter, open-label study, preliminary clinical activity was observed with acceptable safety across a range of cancers including SCCHN, pancreatic, gastric, NSCLC, and melanoma. The ORR in all tumors was 9 %, rising to 13 % in NSCLC and 14 % in SCCHN, while DCR it was 30 % in all tumors and 32 % in both NSCLC and SCCHN. Responses were reported to occur early, within 6 weeks of treatment [81]. In the study expansion phase, durvalumab was administered at 10 mg/kg every 2 weeks to 198 patients with NSCLC and achieved an ORR of 14 % (21 % in squamous and ten in non-squamous). Responses were durable with 76 % of responses ongoing at the time of data cut-off [39]. In the SCCHN expansion cohort, the ORR was 12 % with responses ranging from 4 to 42 weeks and median DOR not reached [82]. 5 PD-L1 Evaluation in Tumor Samples: Technical Challenges Expression of PD-L1 at the protein level in the tumor microenvironment is definitively involved in therapeutic activity and has been thoroughly explored as a potential biomarker. However, a major difficulty is that it is very challenging to generate monoclonal antibodies to detect the molecule in formalin-fixed and paraffin-embedded tissue samples. Most antibodies have been generated but all need cumbersome and potentially unreliable epitope retrieval procedures. Multiple PD-L1 immunohistochemistry (IHC) assays have been described and are summarized in Table 5. Figure 1 shows a biopsy of melanoma stained for PD-L1 by IHC using a Dako assay with rabbit monoclonal antibody clone A crucial point is to determine which types of cells should be scored for surface PD-L1 expression. Most published reports have only assessed PD-L1 expression on tumor cells. However, PD-L1 can be expressed on the cells of the immune infiltrate [7, 10, 66 68] and more emphasis has recently been placed on these cell types [83]. A correlation with the ORR has been reported with both approaches, so the question of which is preferable remains unanswered, and it is likely that quantitative assessments Table 5 Different immunohistochemistry (IHC) assays for PD-L1 in several studies Antibody clone Cut-off % of scored cells to determine positivity Type of cells scored Nivolumab Topalian et al [7] 5H1 C5 % Tumor cells Weber et al. [34] mab(28-8) C5 % Tumor cells Grosso et al. [89] mab(28-8) C5 % Tumor cells Hodi et al. [47] mab(28-8) C1 %, C5 % Tumor cells Robert et al. [10] mab(28-8) C5 % Tumor cells McDermott et al. [79] mab(28-8) C5 % Tumor cells Sznol et al. [13] mab(28-8) C5 % Tumor cells Hammers et al. [16] mab(28-8) C5 % Tumor cells Pembrolizumab Kefford et al. [90] 22C3 [1 % Tumor cells Garon et al. [37] 22C3 1 % Tumor cells Herbst et al. [11] 22C % Tumor cells Atezolizumab Herbst et al. [74] SP142 [5 %,[10 % Tumor infiltrate immune cells Powles et al. [12] SP142 [5 %,[10 % Tumor infiltrate immune cells McDermott et al. [17] SP142 [5 %,[10 % Tumor infiltrate immune cells

10 934 L. Festino et al. Fig. 1 PD-L1 IHC staining of melanoma. A biopsy of melanoma stained for PD-L1 by immunohistochemistry using a Dako assay with rabbit monoclonal antibody clone The two images represent two fields of view from the same tissue section demonstrating membrane specific staining of the tumor cells ranging in intensity from 1? to 3? (920 magnification) on both the tumor and immune infiltrate will be required. In the study of atezolizumab in solid tumors, a statistically significant association was observed between response to atezolizumab and tumor-infiltrating immune cell PD-L1 expression (p = in NSCLC; p = in all tumors), while the association with tumor cell PD-L1 expression did not reach statistical significance (p = in NSCLC; p = in all tumors) [84]. The best cut-off percentage of scored cells to determine PD-L1 positivity also remains an unresolved question. Since the first published data, the threshold most often chosen is [5 % expression. In several studies that have evaluated response, this cut-off point seems to be justified. For example, in the CA trial of patients with RCC treated with combined nivolumab plus ipilimumab, no difference in ORR was detected between patients who were either PD-L1 positive or negative using the threshold of 1 % (50 % in PD-L1-positive vs. 55 % in PD-L1-negative patients). However, a difference was observed when raising the threshold to 5 % (56 vs. 25 %) [16]. Similarly, in the study of nivolumab monotherapy in metastatic melanoma, a threshold of 5 % identified a greater number of responders (ORR 44 % in PD-L1-positive vs. 13 % in PD-L1- negative patients), whereas with a cut-off of 1 % the difference in ORR was smaller (35 vs. 13 %) [47]. Nevertheless, in some studies the use of the 1 % threshold has shown a significant difference between PD-L1-positive and -negative patients. In the study of nivolumab plus multipeptide vaccine in metastatic melanoma, the group with [1 % PD-L1 positivity obtained a meaningful advantage in terms of ORR compared with the \1 % PD-L1 group [48]. Recent data with the 1 % PD-L1 cut-off also showed significant differences in ORR in advanced melanoma patients treated with pembrolizumab. Moreover, the cut-off of 10 % that was also used in this study was not any better in identifying response [74]. Recently in the KEYNOTE 010 study in which all patients were PD-L1 [1 %, all had a response to pembrolizumab; using a cut-off of [50 % expression identified a subgroup of patients with higher ORR and OS than the general population [11]. Thus, when

11 PD-L1 Expression as a Predictive Biomarker for Anti-PD-1/PD-L1 Treatment 935 interpreting these data, it is important to consider that different cut-offs may be required for different types of tumors as well as for combination therapy versus monotherapy. The size of specimen that is to be analyzed is essential for the proper evaluation of PD-L1 expression. Larger tissue specimens are preferred since PD-L1 expression follows a geographical pattern and accumulates in lymphocyte-rich areas [83]. However, no relationship has yet been demonstrated between the size of biopsy (larger excisional specimen, core needle biopsy, or fine needle aspirate) and PD-L1 expression nor between time since sample collection and IHC staining [83]. From a topographical point of view, PD-L1-positive tumor cells are often localized adjacent to the tumor-infiltrating lymphocytes (TILs), often forming a line of demarcation between the tumor tissue and the immune infiltrate. An increased concentration of cytokines, such as INF-c, has also been highlighted along this border [84]. This distribution is not casual, because the expression of PD-L1 on tumor cells is induced by the presence of cytokines produced by the cells of the immune system. Indeed, PD-L1-induced expression is considered a prominent mechanism of immune evasion. This topographical spatial pattern could affect the interpretation of a small biopsy to evaluate PD-L1 expression. PD-L1 is not a fixed marker, but a dynamic immunological marker, and for this reason it is highly dynamic and subjected to change over time in relation to the state of the host immune anti-tumor response. In most cases, except rare occasions where it can be attributed to a form of intrinsic resistance of the tumor, the expression of PD-L1 is related to stimulation by cytokines produced by TILs. The up-regulation of PD-L1 is an indicator of immunosuppression but also should be interpreted as evidence of a tumor response to the endogenous inflammatory response [85]. The balance between these two forces determines the outcome of the patient. In this context, immunotherapy should be interpreted as an enhancement of an ongoing immune response already activated in the individual against cancer and not as a strategy to elicit it de novo. The expression of PD-L1 should be evaluated at a time point as close as possible to the beginning of treatment with anti- PD1/PD-L1 blockers, in order to be able to reliably assess the status of the in-progress immune response against the tumor at time of therapy onset. Evidence is that the expression of PD-L1 in the primary tumor does not correlate with prognosis in patients with melanoma, although there was a correlation between PD-L1 expression in tumor samples obtained from metastasis and outcome, which seems to confirm this hypothesis [83]. Recent data from 175 melanoma specimens from 57 patients showed that PD-L1 expression was often discordant between primary tumors and metastases, and even between metastases in individuals, hence diminishing PD- L1 s potential utility as a predictive biomarker in melanoma management. A significant association was found between melanoma-specific survival and PD-L1 expression in metastatic disease but, as mentioned, not for the primary melanoma sample [86]. Recently it was shown that tumor regression after therapeutic PD-1 blockade requires the presence of pre-existing CD8? T cells at the invasive tumor margin; T cells are negatively regulated by PD-1/PD-L1-mediated adaptive immune resistance and could be activated by inhibitions of this pathway. Tumeh et al. [87] showed that the presence of CD8? T cells at the invasive tumor margin could serve as a predictive biomarker of response to treatment with anti- PD-1/PD-L1 inhibitors. A quantitative predictive image analysis algorithm taking into account T-cell infiltrates and PD-L1 expression at the tumor margins and tumor cores was developed that accurately predicted a validation cohort of cases from a different center. Furthermore, in cases in which serial intra-treatment biopsies were available, it was possible to observe the dynamic changes of PD-L1 and T-cell infiltration in those patients responding to treatment. 6 PD-L1 Expression as a Potential Predictive Biomarker An important difference between the anti-pd-1/pd-l1 agents and ipilimumab is the possibility that there is a potential biomarker to predict patient response to treatment in the case of anti-pd-1/pd-l1. The improvement in OS seen with ipilimumab is not reflected in changes in surrogate endpoints such as ORR and PFS. Subgroup analysis has shown that ipilimumab is effective in different groups of patients (e.g., with elevated lactate dehydrogenase [LDH], asymptomatic brain metastases, different disease stages, etc.) and so it is not possible to select patients on the basis of response, PFS, LDH, or other clinical characteristics. While CTLA-4 expression cannot be considered a marker to predict the response to treatment with ipilimumab, with anti-pd-1/pd-l1 therapy, the expression of PD-L1 in the tumoral cells might be a useful predictor of patient response. A relationship between expression of PD- L1 on tumor cells and response to treatment was first apparent in the study with nivolumab in solid tumors (CA ).Of 46 patients with melanoma, NSCLC, and RCC, 25 were PD-L1 positive and 21 were negative. Tumor surface PD-L1 expression was evaluated by IHC using the murine antihuman PD-L1 monoclonal antibody 5H1 and PD-L1-positive status was defined as C5 % cell membrane staining of any intensity. A treatment response was seen in nine PD-L1 positive patients (36 %) but none of the PD-L1-negative patients. It is important to

12 936 L. Festino et al. note that multiple tumor slides were tested for each patient and the patient was considered PD-L1 positive if just one biopsy sample was positive [88]. Using data from the same study, the association between tumor PD-L1 expression and clinical activity in patients with advanced solid tumors treated with nivolumab was also assessed using a sensitive and specific PD-L1 rabbit monoclonal antibody (clone 28-8) co-developed by Dako and Bristol-Myers Squibb and distinct from the conventional murine 5H1 antibody. PD-L1 positivity was again defined by a 5 % expression threshold. In the subgroup of 38 patients with melanoma, patients who were PD-L1 positive (n = 16) compared with patients who were PD-L1 negative (n = 18) had an ORR of 41 versus 14 %, PFS of 9.1 versus 2.0 months, and OS of 21.1 versus 12.5 months [89]. Recently updated data from a study of 107 metastatic melanoma patients treated with nivolumab in this same trial confirmed the association between ORR and expression of PD-L1. Using a cut-off of 1 % PD-L1 expression, median OS was 25 months for PD-L1 positive compared with 12 months for PD-L1-negative patients. Raising the cut-off to 5 % expression meant median OS was not reached in the PD-L1-positive group and was 13 months in the PD-L1-negative group [47]. Other studies of nivolumab have also shown an association between PD-L1 status and response to therapy. In the study of nivolumab combined with multipeptide vaccine in advanced melanoma, PD-L1 expression was assessed for 44 biopsies with 12 patients defined as positive using a 5 % cut-off and 23 using a 1 % cut-off. ORR was 67 % (8/12) in the positive group versus 19 % (6/32) in the negative group using the C5 % PD-L1 cut-off, and 39 % (9/23) versus 23 % (5/21) using the 1 % cut-off [48]. Similarly, in ipilimumab-pretreated patients, nivolumab had an ORR of 44 % in PD-L1-positive patients and 20 % in PD-L1-negative patients (PD-L1 positive defined as C5 % tumor cell membrane staining measured using the clone 28-8 antibody). As observed in this and other studies, the absence of PD-L1 expression does not necessarily seem to result in a lack of response in all patients [35]. This is also shown by the CA trial in which nivolumab was associated with a better ORR than dacarbazine irrespective of patients PD-L1 status. In PD-L1-positive patients, the ORR was 52.7 % with nivolumab versus 10.8 % with dacarbazine. However, in patients with negative (or indeterminate) PD-L1 status, the ORR was also 33.1 % with nivolumab compared with 15.7 % with dacarbazine although the difference was not as marked [9]. Most studies of PD-L1 expression and response to nivolumab have been in melanoma but studies in other tumor types have also shown a similar association. In the I study of nivolumab in metastatic RCC, median OS across doses was 29.9 months in the PD-L1-positive subgroup and 18.2 months in the PD-L1-negative subgroup [15]. In patients with NSCLC, use of first-line nivolumab therapy resulted in an ORR of 31 % in PD-L1-positive patients (C5 % expression assessed by clone 28-8 antibody) while in PD-L1-negative patients it was 10 %. Median PFS also appeared to be associated with PD-L1 expression, being 15.4 weeks in PD-L1-positive patients versus 10.4 weeks in PD-L1-negative patients. The median OS in PD-L1-negative patients was 98 weeks, while it was not reached in PD-L1-positive patients [50]. In NSCLCpretreated patients enrolled in the study, no association was found between ORR and PD-L1 expression, with the ORR 15 % in PD-L1 positive- and 14 % in PD- L1-negative patients. The median OS was 10.5 months in PD-L1-negative compared with 7.8 months in PD-L1- positive patients [49]. In platinum-pretreated squamous NSCLC patients, PD-L1 expression was not predictive of any efficacy endpoint. The OS and PFS rates favored nivolumab in all patient subgroups and a similar rate of objective responses was observed in PD-L1-positive and PD-L1-negative patients, and were consistently higher in the nivolumab group. ORR with nivolumab in the PD-L1- positive and PD-L1-negative groups was 20 versus 13 % using a 1 % cut-off and 24 versus 14 % using a 5 % cut-off [51]. Conversely, in the study of nivolumab versus docetaxel in non-squamous NSCLC, PD-L1 expression was associated with a benefit from nivolumab. In PD-L1-positive patients, nivolumab showed improved efficacy across all endpoints at predefined 1, 5, and 10 % cut-off points. PD-L1 expression was predictive of benefit with nivolumab, starting at the lowest expression level (1 %). Median OS nearly doubled with nivolumab compared with docetaxel across the PD-L1 expression continuum. No difference in OS was seen when PD-L1 was not expressed in the tumor. The ORR nearly tripled in PD-L1-positive patients. Median DOR was longer with nivolumab versus docetaxel in both PD-L1-positive and -negative patients [53]. Use of PD-L1 as a potential biomarker has also been assessed in patients with melanoma treated with combined ipilimumab and nivolumab. In this cohort, the identification of a biomarker seems to be even more necessary considering the high impact of severe treatment-related toxicity. In the CheckMate 004 trial, patients treated with concurrent ipilimumab and nivolumab had ORRs of 57 % (8/14) and 35 % (9/22) in PD-L1-positive and -negative groups, respectively. In patients who received sequential treatment, the ORRs were 63 % (5/8) in PD-L1-positive and 20 % (3/15) in PD-L1-negative patients [13]. The ORR obtained in the PD-L1-negative subgroup in this study looks particularly clinically meaningful. The pivotal CheckMate 067 trial compared the combination of nivolumab plus ipilimumab with nivolumab and ipilimumab monotherapies with patients stratified according