J-C. Soria

A direct comparison of CellSearch and ISET for circulating tumour-cell detection in patients with metastatic carcinomas

Background:Circulating tumour cells (CTCs) can provide information on patient prognosis and treatment efficacy. However, there is no universal method to detect CTC currently available. Here, we compared the performance of two CTC detection systems based on the expression of the EpCAM antigen (CellSearch assay) or on cell size (ISET assay).Methods:Circulating tumour cells were enumerated in 60 patients with metastatic carcinomas of breast, prostate and lung origins using CellSearch according to the manufacturers protocol and ISET by studying cytomorphology and immunolabelling with anti-cytokeratin or lineage-specific antibodies.Results:Concordant results were obtained in 55% (11 out of 20) of the patients with breast cancer, in 60% (12 out of 20) of the patients with prostate cancer and in only 20% (4 out of 20) of lung cancer patients.Conclusion:Our results highlight important discrepancies between the numbers of CTC enumerated by both techniques. These differences depend mostly on the tumour type. These results suggest that technologies limiting CTC capture to EpCAM-positive cells, may present important limitations, especially in patients with metastatic lung carcinoma.

809OIMMUNE CORRELATES AND LONG TERM FOLLOW UP OF A PHASE IA STUDY OF MPDL3280A, AN ENGINEERED PD-L1 ANTIBODY, IN PATIENTS WITH METASTATIC RENAL CELL CARCINOMA (MRCC)

ABSTRACT Aim: PD-L1, which can mediate cancer immune evasion, is broadly expressed in RCC. As RCC can respond to immune-based therapy, blocking PD-L1 represents a strategy to restore tumor-specific T-cell immunity. MPDL3280A, a human mAb containing an engineered Fc-domain, targets PD-L1 to prevent binding to its receptors PD-1 and B7.1 on activated T cells. Interim results showed that MPDL3280A had a manageable safety profile and clinical activity in advanced NSCLC, melanoma and mRCC pts. Here, we report long term follow up and correlative data for the RCC cohort. Methods: mRCC pts were enrolled in a Ph I expansion study. All pts who received MPDL3280A IV q3w at doses of 3-20 mg/kg were evaluable for safety. Pts were treated for ≤ 1 y. Response was assessed by RECIST v1.1. PD-L1 IHC was centrally assessed in tumor biopsies. Results: As of Jan 1, 2014, among 69 mRCC pts evaluable for safety, median age was 61 y and all were ECOG PS 0-1 (48% PS 1). 94% of pts had prior nephrectomy and 87% received prior systemic therapy, including cytokines (39%), VEGF-inhibition (64%) and mTOR inhibitors (26%). Pts received MPDL3280A for a median duration of 7.7 m (0.7-24.3 m). 80% of pts had a treatment-related AE; however, the frequency of G3 related events was 16%, including anemia, dehydration, fatigue and hypophosphatemia (3% each). No related G4 AEs or deaths occurred. Clinical activity was evaluated in 58 pts with clear cell histology dosed prior to Jul 1, 2013; 51 pts (88%) had an evaluable baseline PD-L1 IHC status. The ORR was 14% (8/58 PRs, 95% CI: 6, 25) and the median duration of response was 54 wks (2.7+ to 68.1+ wks). The 24-wk PFS rate was 53% (95% CI: 40, 66). An association was seen between PD-L1 intensity and response to MPDL3280A. Finally, antitumor activity that included immune-related responses was observed in non-clear cell histology pts (n = 10). Updated data including biomarkers will be presented. Conclusions: MPDL3280A was well tolerated, with no treatment-related deaths. Evidence of increased activity was observed in pts with elevated PD-L1. Durable responses and prolonged SD were observed in mRCC pts, warranting further study. Disclosure: D.F. McDermott: has participated in advisory boards for Genentech, BMS and Merck; M. Sznol: has consulted for Immune Design, Merus, Lion Biotechnologies, Kyowa-Kirin, Aztra-Zeneca-Medimmune, BMS, Amgen, Medimmune, Genentech, Symphogen, Nektar, and Anaeropharma; J. Soria: has received honoraria from Genentech and Roche; M.S. Gordon: has served in an advisory/consultant role and has received research funding from Roche/Genentech; O. Hamid: has been a speaker and consultant for Genentech, and research funding has been provided by Genentech; M. Fasso, Y. Wang, J. Bruey and G. Fine: is employed by Genentech. All other authors have declared no conflicts of interest.

1222OA RANDOMIZED, OPEN-LABEL, PHASE III TRIAL OF AFATINIB (A) VS ERLOTINIB (E) AS SECOND-LINE TREATMENT OF PATIENTS (PTS) WITH ADVANCED SQUAMOUS CELL CARCINOMA (SCC) OF THE LUNG FOLLOWING FIRST-LINE PLATINUM-BASED CHEMOTHERAPY: LUX-LUNG 8 (LL8)

ABSTRACT Aim: A is an irreversible ErbB family blocker that has shown promising clinical activity in pts with SCC of the head/neck and lung. Here, we report results of LL8, a phase III trial that prospectively compared A and E in pts with SCC of the lung following failure of first-line chemotherapy. Methods: Eligible pts with stage IIIB/IV SCC were randomized 1:1 to receive A (40 mg/day; 50 mg/day from Cycle 2 for pts meeting AE criteria) or E (150 mg/day) until disease progression. All pts had progressed after ≥4 cycles of platinum-based doublet chemotherapy and had not received prior EGFR TKIs. Pts were stratified based on race (eastern Asian vs other). The trial was powered for PFS and OS. The primary endpoint was PFS (RECIST 1.1) as assessed by independent radiological review (IRR). Secondary endpoints included overall survival (OS; planned at 632 events), objective response rate (ORR), disease control rate (DCR) and safety. Overall 795 pts were recruited between March 2012 and January 2014. Planned primary analysis was based on 414 PFS events by IRR in the first 669 patients randomized (A: 335, E: 334) while recruitment was ongoing. Results: Baseline characteristics were well balanced between arms: median age 65y; male 85%; eastern Asian 22%; never smoker 5%. Median PFS was significantly higher for A than E, both by IRR (2.4 vs 1.9 months; HR [95% CI]: 0.82 [0.68–1.00]; p = 0.043) and investigator review (2.7 vs 1.9 months; HR [95% CI]: 0.78 [0.65–0.93]; p = 0.005). Furthermore, ORR (4.8% vs 3.0%; p = 0.233) and DCR (45.7% vs 36.8%; p = 0.020) were higher with A vs E. Overall AE profile was comparable (pts with ≥G3 AEs: 50.2 and 49.1% for A and E) with higher incidence of drug-related ≥G3 diarrhoea (9.7 vs 2.4%) and G3 stomatitis (3.3 vs 0.0%) with A and higher incidence of G3 rash/acne with E (5.5 vs 9.0%). Conclusions: LL8 is the largest prospective trial to compare EGFR TKIs in pts with relapsed/refractory SCC. PFS and DCR were significantly better for A than E. AEs were comparable and consistent with the mechanistic profile of EGFR inhibition. Disclosure: G. Goss: Advisory board for Boehringer Ingelheim prior to initiation of study; E. Felip: Advisory Boards for: Boehringer Ingelheim, Novartis, Roche and Bristol Myers Squibb; S. Lu: Advisory Boards for: AstraZeneca and Boehringer Ingelheim; K.N. Syrigos: Advisory boards for Roche, Novartis and Leo; A. Morabito: Advisory Boards for: Lilly and Italfarmaco; A. Ardizzoni: Advisory Boards for: Boehringer Ingelheim, MSD, BMS, Lilly, GSK, Daiichi, Pierre Fabre, Pfizer; S. Gadgeel: Advisory Boards for: BI, Novartis and Genentech; J. Love: Employee of Boehringer Ingelheim; V. Chand: Employee of Boehringer Inhelheim J. Soria: Advisory Boards for: Boehringer Inhelheim. All other authors have declared no conflicts of interest.

Outcome of patients with sarcoma and other mesenchymal tumours participating in phase I trials: a subset analysis of a European Phase I database

BACKGROUND Although sarcomas account for only 1% of all solid tumours, patients with sarcomas comprise a larger proportion of patients entering phase I trials, due to the limited number of registered or active drugs for these diseases. To help in patient selection, we evaluated the utility of the predictive Royal Marsden Score which had been derived in carcinoma patients. In addition, we analysed efficacy and toxicity regarding the sarcoma population enrolled in phase I trials. PATIENTS AND METHODS We used data from a European Database comprising 2182 patients treated in phase I trials in 14 European institutions between 2005 and 2007. RESULTS One hundred and seventy-eight patients diagnosed with advanced sarcoma or other mesenchymal tumours were identified and accounted for 217 phase I trial participations during the study period. Histological type, class of drug, number of metastatic sites, high serum lactate dehydrogenase activity (LDH), low albumin and high white blood cell count were independent prognostic factors. Poor performance status (PS), liver metastases and high leucocyte count were associated with increased risk of early death. The class of drug used was the strongest predictor of progression-free survival (PFS) duration, inhibitors of angiogenesis and histone deacetylase giving the best results. Poor PS, high serum LDH and low lymphocyte count correlated with shorter PFS. In this heterogeneous population, PFS with investigational agents appeared comparable with that previously published for patients receiving standard treatments beyond first line. CONCLUSION Prognostic factors in sarcoma patients do not differ from a broader phase I population. Efficacy measures suggest that some patients with sarcoma derive benefit from therapy in this setting which could therefore be considered for patients with no remaining standard therapeutic option.

1322PBIOMARKERS ASSOCIATED WITH CLINICAL ACTIVITY OF PD-L1 BLOCKADE IN NON-SMALL CELL LUNG CANCER (NSCLC) PATIENTS (PTS) IN A PHASE I STUDY OF MPDL3280A

ABSTRACT Aim: Cancers can evade immune surveillance by upregulating PD-L1. MPDL3280A, a human anti-PD-L1 mAb with an engineered Fc domain, blocks PD-L1 binding to its receptors PD-1 and B7.1 on activated T cells. Predictive factors of response to PD-L1 blockade are not well characterized. Here, we examine biomarkers that might predict MPDL3280A response. Methods: NSCLC pts received MPDL3280A 1-20 mg/kg IV q3w. Pts were treated for ≤ 1 y. Responses (ORR, including unconfirmed) were assessed by RECIST v1.1. FFPE tumor samples were analyzed by IHC and Genentech immunochip measuring 90 immune-related genes to characterize the tumor immune microenvironment at baseline. Tumor-infiltrating immune cells (ICs) or tumor cells (TCs) were scored as IHC 0, 1, 2 or 3 if Results: As of Apr 30, 2013, 53 NSCLC pts dosed by Oct 1, 2012, were evaluable for efficacy. In this trial ≈ 30% of NSCLC pts were PD-L1 IHC 2 or 3. ORRs were associated with PD-L1 expression in ICs (p = .02): IHC 3: ORR 83% (5/6), IHC 2: 14% (1/7), IHC 1: 15% (2/13), IHC 0: 20% (4/20). PD-L1 expression in TCs did not significantly correlate with response (p = .92). The 24-wk PFS rate was 45%. Expression of checkpoint markers was a weaker predictor of MPDL3280A response and did not appear to confer MPDL3280A resistance in PD-L1 IHC2/3 pts (Table). Additionally, elevated IFNγ expression, and IFNγ-inducible genes (e.g., IDO1 and CXCL9), did not appear to associate with MPDL3280A response. Response Rates in Biomarker-Defined Subpopulationsa,b Subpopulation n ORR, % Subpopulation n ORR, % PD-L1 IHC 2 or 3 10 50 CTLA4 + 18 39 PD-L1 IHC 2 or 3; CTLA4 + 6 67 PD-L2 + 20 30 PD-L1 IHC 2 or 3; PD-L2 + 8 63 LAG3 + 18 28 PD-L1 IHC 2 or 3; LAG3 + 6 67 IDO1 + 19 26 PD-L1 IHC 2 or 3; IDO1 + 8 63 B7-H3 + 20 25 PD-L1 IHC 2 or 3; B7-H3 + 5 80 TIM3 + 20 25 PD-L1 IHC 2 or 3; TIM3 + 6 67 B7-H4 + 18 22 PD-L1 IHC 2 or 3; B7-H4 + 5 40 a For pts with both PD-L1 IHC and immunochip data only (N = 37). b PD-L1 IHC scores based on ICs. Conclusions: These data suggest PD-L1 expression has a stronger correlation with ORR compared to other immune checkpoints. PD-L1 appears to be a leading predictive biomarker for response to MPDL3280A. These findings help provide a better understanding of biomarkers and MPDL3280A clinical activity. Disclosure: J-C. Soria has received honoraria from Roche/Genentech; M. Gordon has served as an advisor/consultant to and has received research funding from Roche/Genentech; R.S. Heist has received clinical research funding to her institution from Roche, GSK, EMD Serono, Exelixis and Debiopharm; L. Horn: Research funding from Astellas. Served as an adviser for Bristol Myers Squibb, Clovis, Helix bio (compensated) and PUMA, Xcovery (uncompensated). Steering Committee: Bayer (Uncompensated). Received honoraria from Boehringer Ingelheim; D.R. Spigel has served in an uncompensated role as advisor to Genentech; M. Kowanetz, A. Mokatrin, Y. Xiao and A. Sandler: is an employee of Genentech, Inc.; E. Felip has served as an advisor to BI, Novartis, Roche, BMS and Eli Lilly. All other authors have declared no conflicts of interest.

3D waterfall plots: a better graphical representation of tumor response in oncology

Compared to a waterfall plot, swimmer plots additionally provides us with information both on the quality of response (date of best response) and quantity (duration of response in time). Nevertheless, the information is given in terms of categorical variables, such as complete or partial response, and does not represent the magnitude of the response and how it numerically changes over time. Furthermore, only those patients who respond are represented in this type of graphic, not reporting data from patients that have progressive or stable disease. In contrast, waterfall plots do provide individual patient information as they show a more accurate estimation of the response (adding relative quantification of tumor size change compared to baseline measures) (3). However, waterfall plots show “static” information since only the best response presented is reported. In order to overcome this caveat, spider plot graphics were designed and they show the individual response of each patient over time. So spider plots, not only allows us to follow the change in tumor dimension according to RECIST criteria, but also to observe the evolution and response duration and indirectly, to determine the length of the best response.

22INOVERVIEW OF TRIALS TESTING TUMOUR MOLECULAR PROFILING

ABSTRACT The feasibility to incorporate molecular screening approaches based on highthroughput technologies in cancer patients is now clearly established and is tested across prospective trials around the world. Such trials can be divided in 2 main groups: 1) First generation trials which are molecular triage trials without randomization and frequently without a build-in approach to gain access to the ad hoc molecular targeted therapies. 2) Second generation trials that encompass within the trial the access to targeted therapies, a randomization component and usually test the relevance of a new algorithm rather than the efficacy of a given compound and matched abnormality SAFIR01 is one of main prospective first-generation precision medicine trial ever conducted in breast cancer and demonstrated not only the feasibility of such an approach but also the high interest of patients and the potential associated benefit for patients with specific molecular alterations of FGFR/EGFR or PI3K pathway (Andre F et al. The Lancet Oncology 2014). The table herebelow provides a non-exhaustive list of some of the ongoing trials. Trial or program name and origin Genomics analysis technics Cancer types Clinical trial reference SHIVA Curie Institute, Paris, France IonTorrent/PGM (Amplicon sequencing) + CytoscanHD Solid tumors NCT01771458 MOSCATO Gustave Roussy, Villejuif, France IonTorrent/PGM* (Amplicons sequencing) CGHa (WG) Solid tumors eligible for Phase I trials NCT01566019 SAFIR02 Gustave Roussy, Villejuif, France IonTorrent/PGM (Amplicons sequencing) + CGHa (WG) Breast metastatic Cancer or lung cancer NCT02117167 IMPACT MD Anderson Cancer Center, Houston, US Targeted – HostSpot (PCR based and FISH) Solid tumors NCT00851032 IMPACT Princess Margaret Cancer Centre, Toronto, Canada Targeted - Host Spot (Sequanom) – MiSeq (sequence) Selected Solid Tumors NCT01505400 WINTHER WIN Consortium, Inernational Fondation On - CGHa - Whole Transcriptome Solid tumors NCT01856296 Battle-2 MD Anderson Cancer Center, Houston, US Targeted - Host spot (PCR-based and FISH) NSCLC, PD on chemotherapy NCT01248247 IMAGE Study Johns Hopkins, Baltimore, US NGS (Foundation One) Metastatic Breast Cancer NCT01939847 Disclosure: The author has declared no conflicts of interest.

1331TiPPHASE I STUDY OF TREMELIMUMAB (TREM) IN COMBINATION WITH GEFITINIB (GEF) IN EPIDERMAL GROWTH FACTOR RECEPTOR MUTANT (EGFR-MUT) NON-SMALL CELL LUNG CANCER (NSCLC)

ABSTRACT Background: Treatment options for patients (pts) with EGFR-mut NSCLC who have failed standard chemotherapy and first generation tyrosine kinase inhibitors (TKIs) are limited. Treatment with EGFR TKIs influences the release of chemokines leading to increased T cell recruitment, and tumor cell death induced by TKIs leads to increased antigen release and immune priming. Trem is a human immunoglobulin G2 (IgG2) kappa monoclonal antibody (mAb) that is directed against the CTLA-4 (cytotoxic T lymphocyte-associated antigen 4). This trial is aimed at elucidating whether or not adding anti-CTLA-4 therapy to existing Gef would complement and enhance its effects. Trial design: This phase I multicenter, single-arm study is designed to evaluate Gef (oral 250mg once-daily) plus Trem (starting dose of 3mg/kg IV every 4 weeks for 6 cycles and beyond every 12 weeks) in pts with EGFR-mut advanced NSCLC who have failed EGFR TKI. A rolling 6-cycle design and a dose limiting toxicity period of 42 days will be implemented. Three escalating doses of Trem are planned (3, 6 and 10mg/kg). Six pts will be enrolled for each dose level (up to 24 evaluable pts will be enrolled). Pts may have received chemotherapy between the EGFR TKI and inclusion in this study but must present a systemic objective progression. To prevent a tumor flare when starting Gef + Trem therapy, pts not previously treated with Gef will receive Gef treatment for 2 weeks before the addition of Trem. Pts with ECOG performance status ≥2, history of chronic inflammatory or autoimmune disease, untreated brain metastases or requiring systemic steroids are not eligible. The primary objective is to determine the safety and tolerability of the combination and to establish a recommended phase 2 dose. Secondary/exploratory objectives include evaluating response endpoints (overall response rate, progression free survival according to RECIST 1.1), determining immunogenicity and identifying immune biomarkers of this combination (peripheral blood mononuclear cell, sera and sequential tumor biopsies), and describing the pharmacokinetics for each agent in combination. The study is recruiting in 3 sites in France since February 2014. ClinicalTrials.gov identifier NCT02040064. Disclosure: D. Planchard: Advisory board with Astra-Zeneca to disclose; F. Barlesi: Consultant or advisory relationship with AstraZeneca to disclose; N. Byrne: Nathalie Byrne is an employee of AstraZeneca and owns stock/stock options in AstraZeneca; Besse: research grants from AstraZeneca; A. Nash: Anthony Nash is an employee of AstraZeneca and owns stock/stock options in AstraZeneca; A. Di Pietro: Alessandra DiPietro is an employee of MedImmune and owns stock/stock options in AstraZeneca; J. Soria: JC Soria have honoraria from AstraZeneca to disclose. All other authors have declared no conflicts of interest.

1629OPROGNOSTIC FACTORS OF TUMOUR CELLULARITY IN IMAGE-GUIDED BIOPSIES: RESULTS FROM A PROSPECTIVE MOLECULAR TRIAGE TRIAL (MOSCATO)

ABSTRACT Aim: MOSCATO (molecular-screening for cancer treatment optimization) is a prospective molecular triage trial based on on-purpose tumour biopsies to perform molecular portraits. We aimed at identifying factors that predict a high cellularity in image-guided tumour biopsy. Methods: The percentage of tumour cells was evaluated on each biopsy sample. The primary endpoint was the maximum percentage of tumour cells across the different samples. All CT-scan images were centrally reviewed. Non-parametric tests were used to compare the distribution of the maximum percentage of tumour cells, between subsets of patients (pts). Paired samples (central vs. peripheral) were compared when multiple samples were obtained. Results: Among the 460 enrolled pts from November 2011 to March 2014, 335 pts (73%) had an image-guided needle biopsy of the primary tumour (n=39, 12%) or a metastatic lesion (n=296, 88%). Biopsies were performed on liver (n=127, 38%), lung (n=71, 21%), lymph nodes (n=71, 21%), bone (n=11, 3%), and other tumour sites (n=55, 16%). Sixteen pts (5.7%) experienced a complication: pneumothorax in 8 pts treated medically, and haemorrhage in 8 pts requiring embolisation in 3 cases. The median percentage of tumour cells was 50% (Q1-Q3, 30-70%). Age, gender, on-going chemotherapy, target origin (primitive vs. metastasis), number of lesions in the biopsied organ, target lesion size, tumour growth rate, presence of necrosis and size of the needle were not statistically related to a high cellularity (P Conclusions: The majority of tumour biopsies provided molecular screening efficiency with a low rate of complications. Imaging modality used to guide to the biopsy, multiple samples, peripheral and central, and the chosen organ were the prognostic factors of cellularity. Disclosure: All authors have declared no conflicts of interest.

1300PBUPARLISIB (BKM120) IN PATIENTS WITH PI3K PATHWAY-ACTIVATED, METASTATIC NON-SMALL CELL LUNG CANCER (NSCLC): RESULTS FROM THE BASALT-1 STUDY

ABSTRACT Aim: In vitro, NSCLC cell lines with PIK3CA mutations have shown increased sensitivity to the oral pan-PI3K inhibitor buparlisib (BKM120). BASALT-1 is an open-label, two-stage, Phase II study to assess the safety and efficacy of buparlisib in patients (pts) with pre-treated metastatic NSCLC (squamous [sq] and non-squamous [non-sq] histology) and activated PI3K pathway (NCT01297491). Here we report results from the Stage 1 futility analysis (FA) for each histology. Methods: Pts (≥18 yrs) with previously treated, metastatic NSCLC (sq or non-sq histology) with activated PI3K pathway (defined as PIK3CA mutation and/or PTEN mutation and/or PTEN-negative staining [ Results: Thirty pts had been treated in each histologic group at the time of FA (>1200 pts screened). In the sq group, 21 pts were male (median age: 65.5 yrs); in the non-sq group, 19 pts were male (median age: 62.0 yrs). Median treatment exposure was 6.9 (sq group) and 7.2 (non-sq group) wks. Median PFS was 2.79 (95% CI: 1.38, 4.11) and 2.69 (95% CI: 1.41, 2.89) months for the sq and non-sq groups, respectively. PFS rate at 12 wks was 23.3% (95% CI: 9.9, 42.3) for the sq group and 20.0% (95% CI: 7.7, 38.6) for the non-sq group. Best overall responses (RECIST 1.1): partial response (1 pt [3.3%] in each group), stable disease (14 pts [46.7%] in each group), progressive disease (7 pts [23.3%] sq / 9 pts [30.0%] non-sq), unknown (8 pts [26.7%] sq / 6 pts [20.0%] non-sq). Most common (≥30%) AEs possibly related to study drug: hyperglycemia (36.7%), pruritus (33.3%), diarrhea (30%), and nausea (30%) for sq, and asthenia (30%) for non-sq. Conclusions: In this PI3K pathway-activated group of NSCLC pts, the futility criterion in Stage 1 was met in both histology groups. Thus, enrollment of Stage 2 was not initiated. Results of this study are in line with previous observations where single agents targeting the PI3K pathway have shown marginal activity. Studies of buparlisib in combination with docetaxel in pts with sq NSCLC are ongoing. Disclosure: J. Soria: - Research funds from Novartis (Instititution) - Compensated advisory board Novartis F. Grossi: Novartis advisory board; E. Felip: Advisory Board: BI, Novartis, Roche, BMS, Lilly; M. Thomas: AD-Board honoraria: Novartis, Roche, Lilly, BMS Speaker honoraria: BMS, Lilly; P. Roussou: - Novartis employee - Shares in Novartis; G. Atalla-Vidam: Novartis employee; P. Aimone: Novartis employee. All other authors have declared no conflicts of interest.