Reliable predictive assays will provide

information to help inform clinical

treatment decisions.

Past Flights. Photograph courtesy of Craig Damlo. www.soapboxrocket.com.

Special Technologies for Ex Vivo Analysis of Jenny M. Kreahling, PhD, and Soner Altiok, MD, PhD

Background: Predictive assays for cancer treatment are not new technology, but they have failed to meet the criteria necessary for standardized use in clinical decision-making. Methods: The authors summarize the use of predictive assays and the challenges and values associated with these assays in the clinical setting. Results: Predictive assays commercially available in the clinical setting are not standardized, have significant obstacles to overcome, and cannot be relied upon by health care professionals due to the limited value these assays provide to the decision-making process for the treatment of patients. Conclusions: A method that more closely recapitulates the human tumor microenvironment and accurately predicts response with high reproducibility would be beneficial to patient outcomes and quality of life.

Introduction events endured by patients. Cancer is the second leading cause of death in the The treatment of cancer currently relies on clas- United States, despite the many agents sical chemotherapy regimens based on histology and and hundreds of combinations available for treat- targeted agents for patients who carry specific genom- ment.1 However, health care professionals generally ic alterations.1 In the era of genomics, we are learning make treatment decisions based on standard protocols that cancer is an individual disease and, with a push developed in clinical trials to successfully match a toward personalized medicine, treatment is shifting to patient to a treatment regimen.2 This generalized ap- a more tailored approach. However, genomic testing proach can lead to a response, but oftentimes patients still fails to capture the factors that ultimately deter- fail to show improvement. A predictive assay to help mine how tumor cells will behave inside the body. guide treatment decisions has the potential to lead to One of these factors is intratumoral heterogeneity. improved outcomes and fewer unnecessary adverse Large-scale sequencing of solid has revealed that different cells within a tumor can have distinct profiles, including in gene expression, proliferation, From the Departments of Experimental Therapeutics (JMK) and Anatomic Pathology (SA), H. Lee Moffitt Cancer Center & Research and metastatic potential, among others. These differ- Institute, Tampa, Florida. ences can contribute to treatment failures and have Submitted December 15, 2014; accepted April 1, 2015. consequences related to personalized medicine.3 The Address correspondence to Soner Altiok, MD, PhD, Department of impact on predictive assays lies in the sampling of Anatomic Pathology, Moffitt Cancer Center, 12902 Magnolia Drive, tumors that commonly rely on biopsy specimens or Tampa, FL 33612. E-mail: [email protected] small resections that do not fully represent the com- No significant relationships exist between the authors and the companies/organizations whose products or services may be plexity of the tumor, thus resulting in the failure of referenced in this article. drugs selected through personalized screening based

226 Cancer Control April 2015, Vol. 22, No. 2 on a small population of biologically identical cells.4-6 several fold below physiological relevance to several The information gained from studying tumor hetero- fold above physiological relevance. Some cell viability geneity will be invaluable in optimizing predictive assays exist to examine chemosensitivity or chemo- assays, such as serial biopsies, to gain a broader pic- resistance; however, few are commercially available ture of individual tumor response to drug therapy.4 and periodically used in the clinic (Table).13 Functional predictive assays providing informa- The differential staining cytotoxicity assay in- tion on personalized responses to drugs are need- volves mechanical disaggregation of cells from sur- ed to help guide treatment decisions and improve gical or biopsy specimens.14 Three-dimensional fresh outcomes. Chemosensitivity assays offer potential in tumor cell clusters are cultured in anchorage-indepen- predicting treatment response; however, after more dent conditions and treated with the drugs of interest than 20 years, controversy still exists on the predictive at 3 dose levels; the middle dose is that which could value of such assays.7 be achieved in therapy: 10-fold lower than the phys- iologically relevant dose and 10-fold higher from 4 to Assay-Guided Therapy 6 days. At the conclusion of the culture period, fast Chemosensitivity assays are described as any green dye is added to the microwells, the contents are laboratory analysis that tests whether tumor growth then sedimented onto permanent Cytospin (Thermo has been inhibited by a known chemotherapy drug or Fisher Scientific, Waltham, Massachusetts) centrifuge a panel of drugs, including both classical and targeted slides, and then they are counterstained with hema- agents and their combinations. All chemosensitivity toxylin and eosin (H & E). The dead cells take up assays share the same basic steps, including isolating the fast green dye, and H & E allows the tumor cells tumor cells, incubating cells with drugs, assessing to be differentiated from normal cells. Rates of drug cell growth or survival, and interpreting the results. sensitivity are measured by the ratio of live cells in These assays are often referred to as chemore- the treated samples to the number of live cells in the sistance assays because ineffective drugs (resistant to untreated controls.8,13,14 This assay is capable of mea- treatment) are identified with significantly more accu- suring both apoptotic- and nonapoptotic-mediated cell racy more frequently than effective drugs (sensitive to death in a population of cells; in addition, the assay treatment).8 This leads to questions regarding distin- can be applied to solid and hematological neoplasms, guishing drug-resistance assays from chemosensitivity does not require a pure population, and can be used assays and the reliability of these types of assays.2,9 with a wide variety of drugs.15,16 For any type of available chemopredictive assay, The ex vivo analysis of programmed cell death cancer tissue samples are obtained from biopsy speci- assay also measures apoptotic and nonapoptotic cell mens or during . These samples are then disso- death markers in tumor samples exposed to che- ciated, grown in a laboratory, and exposed to a range motherapeutic agents. Tumor specimens obtained of concentrations of chemotherapeutic drugs, usually through biopsy specimens or surgical resections are chosen by the ordering physician and regularly take at least 3 weeks. The intent is that cancer cells treated Table. — Commercially Available Assays in the laboratory will reflect the same response in the patient; however, these results are often ambiguous Assay Method at best because the cancer cells show unexpected be- Differential Drug sensitivity measured by the ratio of live cells havior, such as resistance to the drugs typically used staining in the treated samples to number of live cells in cytotoxicity untreated controls to treat that particular cancer and sensitivity to drugs 10,11 Ex vivo analysis End point of interest is cell death as assessed by not generally used. These and other obstacles have of programmed number of cells differentially stained due to slowed the use of chemopredictive assays in routine cell death changes in the integrity of the cellular membrane 7 clinical testing. 3H-thymidine Measures uptake of tritiated thymidine by DNA of viable cells Cell Viability Assays Histoculture drug Drug sensitivity evaluated by quantification of cell Tissue culture–based chemopredictive assays have resistance growth in the 3-dimensional collagen matrix. An inverse relationship exists between drug sensitivi- been used since chemotherapy was first used in can- ty of the tumor and cell growth cer treatment. Growth inhibition or cell death has ATP Measures ATP to quantify the viable cells in a been used in previous iterations of assays of sensi- bioluminescence culture using a luminometer. Decrease in ATP tivity to conventional chemotherapeutic agents.12 All indicates drug sensitivity, whereas no loss of ATP assays use characteristics of cell physiology to distin- suggests drug resistance guish between viable and nonviable cells to quantify Chemosensitivity Utilizing ATP quantification with a luminometer, cells are grown in a monolayer rather than a cell death following exposure to a particular drug of 3-dimensional matrix interest. Drug doses used in the assays are variable, but all assays require drug exposures ranging from ATP = adenosine triphosphate.

April 2015, Vol. 22, No. 2 Cancer Control 227 disaggregated using DNAse and type 4 collagenase to 2 mm in size placed on a collagen matrix so the frag- yield tumor clusters of 50 to 100 cell spheroids. These ments can grow 3-dimensionally and maintain their microaggregates are thought to more closely approx- signaling pathways. After 24 hours, explants are incu- imate the human tumor microenvironment because bated with a particular drug for 3 days. Subsequently, they are not proliferated. Aggregates are then treated they are fixed in formalin and embedded in paraf- with dilutions of drugs and incubated for 3 days. After fin. Radioactivity is quantified in slide sections using drug exposure is completed, a mixture of nigrosin B autoradiography.13 Extreme drug resistance assays and fast green dye with glutaraldehyde-fixed avian have been shown to have a high negative predictive erythrocytes is added to the cellular suspensions.17 value for identifying ineffective drugs. In this way, The samples are then agitated and Cytospin-centri- by process of elimination, results of chemoresistance fuged and then counterstained with H & E. The end dictate the positive selection of potentially effective point of interest for this assay is cell death, which is drugs. However, avoiding potentially unnecessary tox- assessed by observing the number of cells differen- icity is not a straightforward outcome and, although tially stained due to changes in the integrity of the the eliminated drug will not be used, health care pro- cellular membrane.13,17,18 According to the results of a fessionals may be faced with making a choice to sub- single phase 2 clinical trial, functional profiling using stitute it with another drug. This convoluted outcome ex vivo analysis of programmed cell death doubles the suggests that these assays may still be inadequate response rate and improves time-to-progression and for routine oncology care until further randomized survival rates in patients with advanced lung cancer.17 controlled testing is performed.13,18 Several methods measure the incorporation of Drug sensitivity assays are evaluated by quantify- radioactive precursors by macromolecules in viable ing cell growth in the 3-dimensional collagen matrix cells. Incorporating tritiated thymine measures the evaluated by an inverse relationship between the drug uptake of tritiated thymidine by the DNA of viable sensitivity of the tumor and cell growth. However, cells.19 Using proteases and DNAse to disaggregate the concentrations of drug and incubation times are not tissue, samples are seeded into single-cell suspension standardized and vary depending on drug combi- cultures on soft agar and treated with a particular drug nation and tumor type. The adenosine triphosphate for 4 days and, after 3 days, tritiated thymidine is add- (ATP) bioluminescence assay relies on measuring the ed. After 24 hours of additional incubation, cells are level of ATP after single cells or small aggregates are lysed, and radioactivity is quantified and compared cultured and then exposed to particular drugs. Fol- with a blank control. Only viable and proliferating lowing incubation with the drug, the cells are lysed, cells will take up the radioactive thymidine.29 There- cytoplasmic components are solubilized, and then fore, an inverse relationship exists between the uptake luciferin and firefly luciferase are added to the cell of radioactivity and the sensitivity rate of the cells and lysis product, catalyzing the conversion of ATP to the agent of interest.13 adenosine diphosphate and monophosphate and light The extreme drug resistance assay is method- is proportionally emitted to metabolic activity and is ologically similar to the thymidine incorporation as- quantified with a luminometer.13 The number of cells say.20 By incorporating metabolic tritiated thymidine can be calculated from the measurement of light. A to measure cell viability, small tissue samples, rather decrease in ATP indicates drug sensitivity, whereas no than single cell suspensions, are incubated with a loss of ATP suggests that the tumor is resistant to the particular drug for 5 days at doses ranging from 5-fold agent of interest. ChemoFX (Precision Therapeutics, below to 80-fold above concentrations that would Pittsburgh, Pennsylvania) uses this type of technology; reflect physiological relevance. Tritiated thymidine is however, cells are grown in a monolayer rather than subsequently added to the culture, and the uptake rate a 3-dimensional matrix.18 is quantified after various incubation times. Only live The rationale for chemosensitivity assays is stron- (resistant) cells will incorporate tritiated thymidine, gest where a variety of therapeutic options exists but and the level of incorporation is directly related to the no clear selection criteria have been chosen for any rate of chemoresistance, rather than responsiveness particular drug regimen in an individual patient.13 (chemosensitivity). Extreme resistance is identified Some studies have reported a correlation between when the incorporation of thymidine is inhibited in in vitro prediction or response and clinical response, the presence of the particular drug by less than 1 stan- and, although these studies may have internal validi- dard deviation of the median cell inhibition measured ty, they cannot determine whether patients given as- for several hundred reference tumor samples. In this say-guided therapy or empirical therapy have different test, a positive resistance to a specific drug would outcomes.18,22-26 To determine whether assay-guided prevent the selection of an “ineffective” drug.13,18,21 treatment results in overall different outcomes than Another resistance assay is the histoculture empirical treatment, it is important to take into ac- drug-resistance assay that tests tissue fragments of 1 to count response rates, survival rates, the presence of

228 Cancer Control April 2015, Vol. 22, No. 2 adverse events, and patient quality of life. Chemosen- data had not yet been published.2,31 Previously, reports sitivity and chemoresistance assays are used in some in soft-tissue sarcoma xenograft models showed low centers for decisions related to future chemotherapy responses to conventional agents and were unable in situations with multiple equivalent treatment op- to find a strong correlation between selected genetic tions available and when “the current level of evi- markers and chemotherapeutic effect.32 Furthermore, dence is not sufficient to supplant current standard xenograft models of non–small-cell lung cancer were of care chemotherapy.”27 A need exists to further val- treated with 3 hemotherapeutic combinations.33 A 90% idate these assays with direct evidence gathered from engraftment rate was achieved, and approximately prospective trials comparing individuals empirically one-half of the tumors were tested against all 3 che- treated with those given assay-directed therapy. In motherapeutic regimens.33 The xenografts histologi- this way, response rates, survival rates, the incidence cally resembled the primary tumor. Nearly one-third of adverse events, and patient quality of life can all of the tumors were not sensitive to any of the tested be taken into consideration.13,18,21,28-30 combinations.33 In addition, a series of 11 patients received one of the tested combinations (vinorelbine Concerns and cisplatin), and 7 of the patients had recurrences, Concern exists as to whether disaggregated tumor 6 of whom had corresponding nonresponsive xeno- cells in culture represent the 3-dimensional structure grafts.33 One xenograft was sensitive to the combi- of an intact tumor, including the tumor microenviron- nation and also had a recurrence. Results from this ment, and their effect on drug treatment.2 Further- assay took 6 to 8 weeks.33 The authors concluded that more, cells exposed to a particular drug in a culture a correlation was present between the assay results do not take into account the pharmacokinetics of and clinical data in the recurrence group and that drug metabolism and delivery in a human being. The xenograft-based testing may help identify new active reliability and consistency of chemopredictive test- combinations in non–small-cell lung cancer.33 ing are also concerning. Many studies have produced In another study, 29 advance sarcoma xenografts conflicting results, some of which have suggested a were collected and 76% were engrafted.31 Of these 22, small benefit in assay-directed treatments, whereas a total of 16 were correlated with patient outcomes, others have demonstrated no difference at all.12 Spe- 13 of which had a positive correlation, including cific problems with previous assays include the low 6 tumors prospectively tested and 7 tumors tested percentage of successfully completed assays that yield after the study volunteer had already received chemo- clinically useful results, with rare interpretations that therapy.31 According to the authors, all 7 prospectively differed from standard of care, technical complexity determined regimens, including 2 novel, nonstandard that made their application beyond a single labora- approaches, resulted in disease control or response tory/institution unlikely, and the long time needed in the xenograft system and were matched with clin- to study completion, thus delaying therapy. Reviews ical benefit to the 5 individuals who received these of the medical literature, government studies, and agents.31 Although this study had a small data set, its health care insurance evaluations concluded that the results are suggestive of a positive predictive value, evidence is insufficient to support the use of che- and further investigation might identify tumor- and mopredictive assays in routine clinical practice and patient-specific treatments.2,31 that such assays are not recommended, nor covered, by most insurance companies.18,21,28-30 Still, support Conclusions continues for a predictive assay in the context of a The pursuit of reliable predictive assays is an im- clinical trial. Proposals for a successful assay include portant task, and the information these assays could standardizing materials, testing a range of concen- provide would help inform clinical treatment deci- trations to provide a dose-response curve, applying sions. However, after many years, these assays are simple techniques with the possibility of automation, still experimental and cannot be recommended for and measuring cell survival using clear interpretation routine clinical use.7 Many still rely on traditional techniques. approaches of dissociating tumors and establishing cell lines maintained in vitro in serum-based growth Patient-Derived Xenograft Models media. Cultured cells are not completely representa- An advance over assay systems is the tive of the parent tumor and such differences are of patient-derived mouse xenograft model that allows concern when predicting drug response as well as to for the recapitulation of 3-dimensional tumor archi- the basic study of cancer. In particular, culture selec- tecture and incorporates aspects of tumor stroma.12 tion in cell lines may disturb the in vitro relationship Two major limitations of animal studies are their cost between the cancer stem cell and its progeny, and it and intensity of labor required. Until recently, a com- also removes tumor–stromal interactions essential to parison of xenograft responses with clinical outcome the 3-dimensional biology of solid tumors .12

April 2015, Vol. 22, No. 2 Cancer Control 229 The field is lacking studies published with ran- 2014;30C:1-6. 5. Gerlinger M, Swanton C. How Darwinian models inform therapeu- domized, prospective designs to evaluate the clinical tic failure initiated by clonal heterogeneity in cancer medicine. Br J Cancer. utility of chemoresistance and chemosensitivity as- 2010;103(8):1139-1143. 6. Gillies RJ, Verduzco D, Gatenby RA. Evolutionary dynamics of car- says. The data are insufficient to determine whether cinogenesis and why targeted therapy does not work. Nat Rev Cancer. use of a particular test to select chemotherapy regi- 2012;12(7):487-493. 7. Jenks S. Chemosensitivity assays: still eyeing the clinic. J Natl Cancer mens for individual patients will improve outcomes. Inst. 2012;104(23):1775-1777. Limitations exist and include sampling bias due to 8. Weisenthal LM, Kern DH. Prediction of drug resistance in cancer che- motherapy: the Kern and DiSC assays. Oncology (Williston Park). 1991;5(9):93- heterogeneity of tumors and insufficient biospecimen 104, 111-114, 117, 118. processing resulting in nonevaluable data; therefore, 9. Fruehauf JP, Alberts DS. In vitro drug resistance versus chemosensi- tivity: two sides of different coins. J Clin Oncol. 2005;23(15):3641-3643. all chemosensitivity assays are currently considered 10. Health Net. National medical policy. Chemotherapy sensitivity and re- to be investigational.13,18 sistance assays. Policy No. NMP210. Published March 2005. Updated March 2015. https://www.healthnet.com/static/general/unprotected/pdfs/national/policies The foremost priority in clinical oncology is pro- /ChemotherapySensitivityandResistanceAssays.pdf. Accessed April 17, 2015. viding personalized treatment approaches for more 11. Health Net. National medical policy. ChemoFX assay for prediction of response to chemotherapy. Policy No. NMP459. Published June 2009. Up- effective and better-tolerated therapies. One major dated May 2014. https://www.healthnet.com/static/general/unprotected/pdfs step toward offering personalized therapy in cancer /national/policies/ChemoFXAssayforPredictionofResponsetoChemotherapy. pdf. Accessed April 17, 2015. is the identification of actionable mutations so that 12. Reed D, Altiok S. Metastatic soft tissue sarcoma chemotherapy: an health care professionals can select the most appro- opportunity for personalized medicine. Cancer Control. 2011;18(3):188-195. 13. The Regence Group; Blue Cross/Blue Shield of Oregon, Utah, Ida- priate treatment for their patients. New technologies, ho, and Select Counties of Washington. Medical policy manual: in vitro che- such as next-generation sequencing (NGS), offer a moresistance and chemosensitivity assays. Policy No. 6. Published January 1996. Updated March 2014. http://blue.regence.com/trgmedpol/lab/lab06.pdf. comprehensive analysis of specific mutations present Accessed April 17, 2015. in the cancer of each individual patient and are like- 14. Mason JM, Drummond MF, Bosanquet AG, et al. The DiSC assay. A cost-effective guide to treatment for chronic lymphocytic leukemia? Int J ly to yield information regarding tumor biology and Technol Assess Health Care. 1999;15(1):173-184. response to various treatment modalities. However, 15. Solé RV, Valverde S, Rodriguez-Caso C, et al. Can a minimal rep- licating construct be identified as the embodiment of cancer? Bioessays. mutations identified through NGS analysis may not 2014;36(5):503-512. be actionable driver mutations, in that they do not 16. Weisenthal LM. Differential staining cytotoxicity assay: a review. Meth- ods Mol Biol. 2011;731:259-283. lend themselves to specific and effective therapies. 17. Nagourney RA, Blitzer JB, Shuman RL, et al. Functional profiling to Rather, we propose that combining ex vivo drug select chemotherapy in untreated, advanced or metastatic non-small cell lung cancer. Anticancer Res. 2012;32(10):4453-4460. testing with NGS analysis may help distinguish ran- 18. Aetna. Clinical policy bulletin: tumor chemosensitivity assays. No. 0245. dom passenger mutations from true drivers. Fresh Published May 26, 1998. Updated May 12, 2014. http://www.aetna.com/cpb /medical/data/200_299/0245.html Accessed April 1, 2015. tumor samples can be directly treated ex vivo with 19. Yung WK. In vitro chemosensitivity testing and its clinical application a particular drug or drug combination for the tar- in human gliomas. Neurosurg Rev. 1989;12(3):197-203. 20. Kern DH, Weisenthal LM. Highly specific prediction of antineoplastic geted blockade of signaling pathways aberrantly ac- drug resistance with an in vitro assay using suprapharmacologic drug expo- tivated through specific mutations in each cancer. sures. J Natl Cancer Inst. 1990;82(7):582-588. 21. BlueCross BlueShield of North Carolina. Corporate medical policy: in The drug-mediated inhibition of proximal and distal vitro chemoresistance and chemosensitivity assays. Published March 1997. signaling pathways accompanied by decreased cell Updated April 2014. https://wwwps.bcbsnc.com/assets/services/public/pdfs /medicalpolicy/in_vitro_chemoresistance_and_chemosensitivity_assays.pdf. viability would indicate a therapeutic opportuni- Accessed April 9, 2015. ty with a specific drug for each individual patient. 22. Gazdar AF, Steinberg SM, Russell EK, et al. Correlation of in vitro drug sensitivity testing results with response to chemotherapy and survival A similar approach could also be used as an enrich- in extensive stage small cell lung cancer: a prospective clinical trial. J Natl ment strategy in designing clinical trials in which Cancer Inst. 1990;82(2):117-124. 23. Agiostratidou G, Sgouros I, Galani E, et al. Correlation of in vitro cy- the tumor samples of study volunteers are tested ex totoxicity nd clinical response to chemotherapy in ovarian and breast cancer vivo for their responses to drugs used in the clinical patients. Anticancer Res. 2001;21(1A):455-459. 24. Konecny G, Crohns C, Pegram M, et al. Correlation of drug response trials and then volunteers are assigned to clinical trials with the ATP tumor chemosensitivity assay in primary FIGO stage III ovarian based on their ex vivo response to treatments. It is cancer. Gynecol Oncol. 2000;77(2):258-263. 25. Salom E, Penalver M, Homesley H, et al. Correlation of pretreatment plausible that ex vivo functional assays may serve as drug induced apoptosis in ovarian cancer cells with patient survival and clinical the basis for inclusion or exclusion for clinical trials response. J Transl Med. 2012;10:162. 26. Strickland SA, Raptis A, Hallquist A, et al. Correlation of the microculture-ki- that can incorporate targeted agents in addition to netic drug-induced apoptosis assay with patient outcomes in initial treatment of cytotoxic therapy.12 adult acute myelocytic leukemia. Leuk Lymphoma. 2013;54(3):528-534. 27. Bhagwandin S, Naffouje S, Salti G. Utility of chemoresponse assay in patients undergoing cytoreductive surgery plus hyperthermic intraperitoneal References chemotherapy. Ann Surg Oncol. 2015. [Epub ahead of print]. 28. Burstein HJ, Mangu PB, Somerfield MR, et al. American Society of 1. American Cancer Society. Cancer Facts & Figures 2015. Atlanta, GA: Clinical Oncology clinical practice guideline update on the use of chemotherapy American Cancer Society; 2015. http://www.cancer.org/acs/groups/content/@ sensitivity and resistance assays. J Clin Oncol. 2011;29(24):3328-3330. editorial/documents/document/acspc-044552.pdf. Accessed April 9, 2015. 29. Schrag D, Garewal HS, Burstein HJ, et al. American Society of Clinical 2. Wagner AJ. Patient-derived sarcoma xenografts for individual-patient se- Oncology technology assessment: chemotherapy sensitivity and resistance lection of chemotherapy-ready for prime time? Cancer. 2014;120(13):1917-1919. assays. J Clin Oncol. 2004;22(17):3631-3638. 3. Gerlinger M, Rowan AJ, Horswell S, et al. Intratumor heterogeneity 30. Laboratory and Diagnostic Services Panel. Human tumor assay and branched evolution revealed by multiregion sequencing. N Engl J Med. systems. Presented at the Medicare Evidence Development & Coverage 2012;366(10):883-892. Advisory Committee Meeting, November 15–16, 1999; Baltimore, MD. 4. Barber LJ, Davies MN, Gerlinger M. Dissecting cancer evolution at the http://www.cms.gov/medicare-coverage-database/details/medcac-meeting macro-heterogeneity and micro-heterogeneity scale. Curr Opin Genet Dev. -details.aspx?MEDCACId=14&year=1999&bc=AAAIAAAAAAAAAA%3d%3d&.

230 Cancer Control April 2015, Vol. 22, No. 2 Accessed April 9, 2015. 31. Stebbing J, Paz K, Schwartz GK, et al. Patient-derived xenografts for individualized care in advanced sarcoma. Cancer. 2014;120(13):2006-2015. 32. Ryan CW, Dolan ME, Brockstein BB, et al. A phase II trial of O6-ben- zylguanine and carmustine in patients with advanced soft tissue sarcoma. Cancer Chemother Pharmacol. 2006;58(5):634-639. 33. Dong X, Guan J, English JC, et al. Patient-derived first generation xeno- grafts of non-small cell lung cancers: promising tools for predicting drug responses for personalized chemotherapy. Clin Cancer Res. 2010;16(5):1442-1451.

April 2015, Vol. 22, No. 2 Cancer Control 231