Search

Immunotherapies offer the promise of greatly improved outcomes, but often in only a minority of patients. It is of great importance to be able to identify which patients are likely to benefit from new immunotherapy advances. Developing tests to identify patients likely to benefit from immunotherapeutics presents special issues due to the lack of clear endpoints able to demonstrate which patients benefit from therapy. Traditional categorical endpoints, such as response, may not identify all patients who benefit and may be poor surrogates for long term progression-free survival and overall survival. Using methods adapted from the field of deep learning, we have developed an approach to multivariate test development for benefit from therapy able to work with the most clinically meaningful endpoints, by simultaneously refining the definition of relative benefit groups, the subset of most relevant genetic or proteomic information, and the diagnostic test itself. During the iterative process of test creation, the system learns what subset of the available molecular data is useful for the particular problem being addressed. We will demonstrate how this method has been used to develop a test predictive of benefit from an immunotherapy in adjuvant pancreatic cancer.

Summary Background An established multivariate serum protein test can be used to classify patients according to whether they are likely to have a good or poor outcome after treatment with EGFR tyrosine-kinase inhibitors. We assessed the predictive power of this test in the comparison of erlotinib and chemotherapy in patients with non-small-cell lung cancer. Methods From Feb 26, 2008, to April 11, 2012, patients (aged ≥18 years) with histologically or cytologically confirmed, second-line, stage IIIB or IV non-small-cell lung cancer were enrolled in 14 centres in Italy. Patients were stratified according to a minimisation algorithm by Eastern Cooperative Oncology Group performance status, smoking history, centre, and masked pretreatment serum protein test classification, and randomly assigned centrally in a 1:1 ratio to receive erlotinib (150 mg/day, orally) or chemotherapy (pemetrexed 500 mg/m 2 , intravenously, every 21 days, or docetaxel 75 mg/m 2 , intravenously, every 21 days). The proteomic test classification was masked for patients and investigators who gave treatments, and treatment allocation was masked for investigators who generated the proteomic classification. The primary endpoint was overall survival and the primary hypothesis was the existence of a significant interaction between the serum protein test classification and treatment. Analyses were done on the per-protocol population. This trial is registered with ClinicalTrials.gov, number NCT00989690. Findings 142 patients were randomly assigned to chemotherapy and 143 to erlotinib, and 129 (91%) and 134 (94%), respectively, were included in the per-protocol analysis. 88 (68%) patients in the chemotherapy group and 96 (72%) in the erlotinib group had a proteomic test classification of good. Median overall survival was 9·0 months (95% CI 6·8–10·9) in the chemotherapy group and 7·7 months (5·9–10·4) in the erlotinib group. We noted a significant interaction between treatment and proteomic classification (p interaction =0·017 when adjusted for stratification factors; p interaction =0·031 when unadjusted for stratification factors). Patients with a proteomic test classification of poor had worse survival on erlotinib than on chemotherapy (hazard ratio 1·72 [95% CI 1·08–2·74], p=0·022). There was no significant difference in overall survival between treatments for patients with a proteomic test classification of good (adjusted HR 1·06 [0·77–1·46], p=0·714). In the group of patients who received chemotherapy, the most common grade 3 or 4 toxic effect was neutropenia (19 [15%] vs one [ vs 22 [16%]) was the most frequent in the erlotinib group. Interpretation Our findings indicate that serum protein test status is predictive of differential benefit in overall survival for erlotinib versus chemotherapy in the second-line setting. Patients classified as likely to have a poor outcome have better outcomes on chemotherapy than on erlotinib. Funding Italian Ministry of Health, Italian Association of Cancer Research, and Biodesix.

A method was developed that allows the evaluation of complex biological processes from mass spectrometry of human serum samples. Applying gene set enrichment analysis ideas to matched protein expression data from a panel of 1129 proteins and deep MALDI® mass spectral data from a set of 49 human serum samples with lung cancer (n=45) or non-cancer (n=4), subsets of mass spectral features associated with selected biological functions were identified. Biological functions included acute response, complement system, and wound healing and the protein members of each function were assigned using the intersection of gene ontologies and the protein panel. Using mass spectral data collected from an independent NSCLC sample cohort (n=85) from patients treated with targeted therapy, principal component analysis was used to derive scoring functions (combinations of the feature subsets) for each biological class. These scoring functions were validated in an additional NSCLC sample set (n=123) treated with chemotherapy. The scoring functions can be applied to mass spectra obtained from any human serum sample to generate scores associated with individual biological processes. We have developed scoring functions for several biological processes. Acute response score was associated with outcome in Cox proportional hazard analysis in several independent patient cohorts across multiple therapies and indications, including lung cancer, as mentioned, and ovarian cancer (n=165) treated with platinum doublets following surgery. Choice of a single cutoff allowed stratification of patients into groups with significantly better or worse outcome. In a cohort of nivolumab treated NSCLC patients (n=67) with available longitudinal samples, outcome was also found to depend on changes in scores during therapy. Interestingly, while the distributions of acute phase scores were quite similar across multiple tumor types, scores for other biological functions, such as wound healing, varied considerably. This may reflect differences in relative importance of individual biological functions between tumor types. A scoring system based on biological functional categories has a wide range of uses that could be tested and applied in a clinical setting. Currently, tests that measure a single biomarker for monitoring a particular disease have questionable utility and limit the usefulness to the clinician and patient treatment decision making. A tool that measures the activity of a complex biological process could be useful in deciding when an intervention is needed (screening), or for monitoring the effects during therapy. Citation Format: Carlos Oliveira, Julia Grigorieva, Krista D. Meyer, Joanna Roder, Heinrich Roder. Development of scores reflective of biological processes underlying human disease states from mass spectrometry of serum [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 210. doi:10.1158/1538-7445.AM2017-210

Introduction Previously presented results from NExUS, a Phase 3 study of sorafenib in combination with G + C (GC) vs placebo + GC in first line NSCLC patients, showed no improvement in overall survival (OS) and a small statistically significant improvement in progression free survival (PFS) for sorafenib + GC. VeriStrat® (V), a proteomic test using Matrix Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF MS), was used in the present study to classify baseline plasma samples from NExUS patients into V Good (VG) and V Poor (VP) categories based on a pre-specified 8-peak mass spectral signature. The overall objective was to determine if V status was predictive of sorafenib + GC clinical activity. Methods Patients with Stage IIIB or IV NSCLC, ECOG PS = 0 or 1, were randomized 1:1 to receive G (1250 mg/m2 on Days 1 and 8) and C (75 mg/m2 on Day 1) for up to six 21-day cycles, in combination with sorafenib 400 mg bid or placebo. V status was determined for 403 of 774 non-squamous patients. Analyses were performed blinded to clinical outcomes. PFS was compared between treatment arms within VP and VG groups. Hazard ratios (HR) and Kaplan-Meier curves were evaluated and multivariate analyses performed. Results Consistent with previous reports, approximately 30% of subjects had VP status. PFS and HRs in treatment arms in patients with and without assigned V status were similar. VG status was associated with a longer PFS in placebo + GC patients (HR = 0.51, log-rank p Conclusions VG status is associated with a better prognosis in first line NSCLC patients treated with placebo + GC. V is also predictive, with VP patients receiving sorafenib + GC showing improved PFS compared to placebo + GC. Disclosure L. Paz-Ares: Luis Paz-Ares has received honoraria from Bayer HealthCare Pharmaceuticals, Lilly, Roche and Pfizer. T.Q.G. Eisen: Tim Eisen has received consulting fees and honoraria from Bayer, Pfizer, Roche, GSK, Aveo, owns stock in Astrazeneca, and has received research funding from Astrazeneca, GSK, Pfizer, Bayer. W. Eberhardt: Wilfried E. E. Eberhardt has received honoraria from Bayer, Roche, Astrazeneca, Pfizer, Boehringer Ingelheim, Sanofiaventis, Novartis, BMS, GSK, Imclone, Eli Lilly, and Pierre Fabre. M. Thomas: Michael Thomas has received consulting fees from Bayer Healthcare. C. Lathia: Chetan Lathia is an employee of Bayer HealthCare Pharmaceuticals. H. Roder: Heinrich Roder is a founder and stockholder of Biodesix, Inc. All other authors have declared no conflicts of interest.