1573PD - Patient-derived xenografts accurately capture clinical responses to treatment

Date 29 September 2014
Event ESMO 2014
Session Trials and tribulations in oncology: Future approaches
Topics Basic Science
Translational Research
Presenter Justin Stebbing
Citation Annals of Oncology (2014) 25 (suppl_4): iv546-iv563. 10.1093/annonc/mdu358
Authors J. Stebbing1, S. Zacharoulis2, A. Gaya3, W. McGuire4, W. Harris5, R.G. Maki6, M. Hidalgo7, A. Davies8, D. Vasquez-Dunddel8, D. Ciznadija8, A. Katz8, D. Sidransky9, K. Paz8
  • 1Surgery And Cancer, Imperial College London-Charing Cross Hospital, SW7 2AZ - London/GB
  • 2Pediatric Oncology, Royal Marsden Hospital, London/GB
  • 3Clinical Oncology, Guy's and St Thomas' Hospital, London/GB
  • 4Medical Oncology, Franklin Square Hospital Center, Falls Church/US
  • 5Clinical Research, Fred Hutchinson Cancer Research Center, Seattle/US
  • 6Department Of Medicine And Pediatrics, Mount Sinai School of Medicine, 10029-6574 - New York/US
  • 7Ciocc, Hospital Madrid Norte San Chinarro Centro Integral Oncologico Clara Campal, ES-28029 - Madrid/ES
  • 8Personalized Oncology Services, Champions Oncology, 07601 - Hackensack/US
  • 9Otolaryngology-head And Neck Cancer Research, Johns Hopkins University, Baltimore/US

 

Abstract

Aim

The inability to distinguish patient populations that are most likely to respond to different treatments is a continuing challenge for medical oncologists. Technologies that can be broadly applied to all patients, which enable multiple therapeutic regimens to be evaluated simultaneously to identify the ones most likely to be clinically beneficial, are needed. Drug sensitivity screening in patient-derived xenografts (PDXs) is a viable solution. In this study we examined the capacity of PDXs to replicate patient outcomes across a variety of solid tumors and treatments and report performance metrics highlighting the clinical utility of this tool.

Methods

Tumor tissue from 495 cancer patients was engrafted into immunodeficient mice to generate PDX models. Of these, 65 were screened against the treatments received by the corresponding patient. Patient clinical responses and model drug responses were assessed and correlated using RECIST criteria, with parameters, including sensitivity, specificity, and predictive values, calculated to determine the capacity of PDX responses to capture patient responses.

Results

Positive (complete and partial response as well as stable disease) and negative (progressive disease) patient outcomes to treatment were accurately replicated by PDXs, regardless of tumor type or treatment. Based on 96 correlations in 65 patients, we calculated a sensitivity of 99% (72/73 correlations) and specificity of 70% (16/23 correlations) for our PDX drug screens, as well as predictive values of 91% (72/79 correlations) and 94% (16/17 correlations). Despite patient exposure to first-line regimens, PDX models generated from the first resection retained the ability to reproduce outcomes to treatments used for recurrent disease.

Conclusions

PDXs accurately reflect clinical responses and have strong potential to correctly guide an oncologist to treatments most likely to yield a favorable patient response, whilst avoiding those that would not. Integration of PDX technology into routine cancer care will re-shape clinical decision-making paradigms in oncology, generating better patient outcomes and eventually help cancer be managed as a chronic disease.

Disclosure

M. Hidalgo: I am on the Scientific Advisory Board of Champions Oncology; A. Davies: I am an employee of Champions Oncology and I own stock in the company; D. Vasquez-Dunddel: I am an employee of Champions Oncology and I own stock in the company; D. Ciznadija: I am an employee of Champions Oncology and own shares in the company; A. Katz: I am an employee of Champions Oncology and I own stock in the company; D. Sidransky: I am on the Board of Directors for Champions Oncology and I own stock in the company; K. Paz: I am an employee of Champions Oncology and I own stock in the company. All other authors have declared no conflicts of interest.