Precision Medicine Lessons from meta-analyses of 70,253 patients Razelle Kurzrock, MD Senior Deputy Director, Clinical Science Director, Center for Personalized Cancer Therapy and Clinical Trials Office Chief, Division of Hematology/Oncology UCSD Moores Cancer Center
Meta-Analyses Conducted 1) Trials leading to FDA approval from trastuzumab (1998) until June 2013 38,104 patients; 112 trials 2) Phase II studies published between 2010 through 2012 32,149 patients; 570 trials Maria Schwaederle, PharmD
Background Treatment selection based on biomarkers reflecting the underlying cancer biology or specific features have brought remarkable advances in oncology. However, the evidence supporting the benefit of a personalized or biomarker-based approach to cancer research and treatment is still a matter of debate. GOAL: compare efficacy outcomes (RR, PFS and OS) and treatment-related mortality between agents developed under a biomarker-based rationale (personalized therapy) versus those that did not.
META ANALYSIS STUDY #1 PHASE III TRIALS
Impact of a Biomarker-Based Strategy on Oncology Drug Development: A Meta-analysis of Clinical Trials Leading to FDA Approval Denis L. Fontes Jardim, MD 1,2, Maria Schwaederle, PharmD 3, Caimiao Wei, PhD 4, J. Jack Lee, PhD 4, David S. Hong, MD 5, Alexander M. Eggermont, PhD 6,7, Richard L. Schilsky, MD, FACP 7,8, John Mendelsohn, MD 7,9, Vladimir Lazar, PhD 6,7, Razelle Kurzrock, MD 3,7 1 Department of Clinical Medicine, Hemocentro da Unicamp, University of Campinas, Sao Paulo, Brazil 2 Department of Clinical Oncology, Hospital Sirio Libanes, Sao Paulo, Brazil 3 Center for Personalized Cancer Therapy and Division of Hematology and Oncology, University of California, San Diego, CA, USA 4 Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA 5 Department of Investigational Cancer Therapeutics (Phase I Clinical Trials Program), The University of Texas MD Anderson Cancer Center, Houston, TX, USA 6 Department of Functional Genomics, Institut Gustave Roussy, University Paris-Sud, Villejuif, France 7 Worldwide Innovative Network For Personalized Cancer Therapy 8 American Society of Clinical Oncology, Alexandria, VA, USA 9 The University of Texas MD Anderson Cancer Center, Houston, USA.
Trials leading to FDA approval Search: Newly approved agents from September 1998 (trastuzumab) until June 2013 PubMed or ASCO meetings abstracts. Excluded: Pediatric cancer, supportive care, loco-regional treatment, hormonal therapies, vaccines Endpoints: Response rate (RR), progression-free survival (PFS), overall survival (OS), and toxicity-related deaths.
Biomarker-based approach (personalized strategy) definition Treatment met one of the following criteria: Cognate biomarker used to select patients for treatment No biomarker used, but at least 50% of patients are known to harbor the cognate biomarker.
Search Results TOTAL 112 trials 57 randomized 55 non-randomized trials Enrolled a total of 38,104 patients
Characteristic (%) Median number of patients per experimental arm (range) Trial design Randomized Non-randomized Class agent Cytotoxic Targeted Tumor type Solid Hematologic Route Intravenous Oral Treatment Single agent Combination Population Treatment-naïve Previous treatment Control arm (for randomized trials) Active treatment Placebo/BSC Cross-over allowed (for randomized trials) Yes No Personalized trials; n=44, (%) Non-personalized trials; n=68, (%) P value 152 (7-553) 204.5 (33-862) 0.29 18 (41) 26 (59) 0 44 (100) 20 (45) 24 (55) 14 (32) 30 (68) 39 (89) 5 (11) 13 (30) 31 (70) 13/18 (72) 5/18 (28) 12/18 (67) 6/18 (33) 39 (57) 29 (43) 24 (35) 44 (65) 47 (69) 21 (31) 44 (65) 24 (35) 48 (71) 20 (29) 18 (26) 50 (74) 28/39 (72) 11/39 (28) 11/39 (28) 28/39 (72) 0.12 <.001 0.02 0.001 0.04 0.83 1.00 0.009
Benefit of personalized therapy in randomized registration trials (RR, PFS, and OS) Statistical analysis: meta-analysis of relative response rate ratio (RRR) and hazards ratios (HRs) for PFS and OS for personalized trials versus not (random effect model) RRR: higher likelihood of response with a personalized compared to non-personalized strategy (RRR=3.82 [95%CI: 2.51-5.82] vs. 2.08 [95%CI: 1.76-2.47], (P=0.03 in meta-regression). HR for PFS: 0.41 (95%CI: 0.33-0.51) for personalized compared to 0.59 (95%CI: 0.53-0.65) for non-personalized studies, (P<.001 in meta-regression). HR for OS: 0.71 (95%CI: 0.61-0.83) for personalized compared to 0.81 (95%CI: 0.77-0.85) for non-personalized studies (P=0.07 in meta-regression)
Benefits of personalized therapy in all trials (N=112) (RR, PFS, and OS) Stat analysis: random effect meta-analysis for RR; pooled analysis for PFS, and OS for personalized trials versus not (weighted multiple linear regression models) RR: 48% for personalized strategy [95%CI 42-55%] vs. 23% [95%CI 20-27%], P<.001 (also P<.001 after adjustement). PFS: 8.3 months for personalized strategy vs. 5.5 months, P<.001 (P=0.002 after adjustement). OS: 19.3 months for personalized strategy compared to 13.5 months, P=0.01 (P=0.04 after adjustement). Treatment-related mortality was 1.58 percent for personalized versus 1.44 percent for non-personalized trials, which was not statistically different (P=0.74).
Summary of results Multivariable analysis: N = 38,104 Randomized trials meta regression Characteristic P-value RRR meta-regression Personalized therapy strategy 0.03 Ctrl arm placebo vs. active drug <0.001 Cross-over allowed <0.001 Progression Free Survival Personalized therapy strategy <0.001 Ctrl arm placebo vs. active drug <0.001 Hematologic tumor vs solid 0.004 Cross-over allowed <0.001 Overall Survival Personalized therapy strategy 0.07 Hematologic tumor vs solid 0.006 All trials meta-regression (RR) and weighted pooled multilinear regression (PFS/OS) Characteristic Response rate P-value Personalized therapy strategy <0.001 Chemotherapy-naïve patients <0.001 Hematologic tumor vs solid <0.001 Progression Free Survival Personalized therapy strategy 0.002 Overall Survival Personalized therapy strategy 0.041
META ANALYSIS STUDY #2 PHASE II TRIALS
Impact of Precision Medicine in Diverse Cancers: a Meta-Analysis of 32,149 Patients in Phase II Clinical Trials Journal of Clinical Oncology, in press Maria Schwaederle, PharmD 1, Melissa Zhao, BS 1, J. Jack Lee, PhD 2, Alexander M. Eggermont, MD, PhD 3,4, Richard L. Schilsky, MD 4,5, John Mendelsohn, MD 4,6, Vladimir Lazar, MD, PhD 3,4, Razelle Kurzrock, MD 1,4 1 Center for Personalized Cancer Therapy and Division of Hematology and Oncology, University of California, La Jolla, U.S. 2 Department of Biostatistics, University of Texas MD Anderson Cancer Center, Houston, TX, USA. 3 Department of Functional Genomics, Institut Gustave Roussy, University Paris-Sud, Villejuif, France. 4 Worldwide Innovative Network for Personalized Cancer Therapy. 5 American Society of Clinical Oncology, Alexandria, VA, USA. 6 The University of Texas MD Anderson Cancer Center, Houston, USA.
Meta-Analysis of 32,149 Patients in Phase II Clinical Trials A PubMed search was conducted (2010-2012). Only single agent s arms were included in the analysis. Exclusion criteria: pediatric cancers, supportive care, loco-regional treatments, hormonal therapies, and cellular or vaccine therapy. 570 Phase II studies were included, comprising 32,149 patients (641 single-agent arms). Maria Schwaederle, PharmD
Baseline characteristics Personalized or not PERSONALIZED Strategy NON-PERSONALIZED Strategy N of arms N of arms N of P-Value* Characteristics N of patients (%) (%) patients Total studies 112 (100%) 8078 529 (100%) 24071 Study Design a 0.330 Randomized 15 (14%) 1300 92 (18%) 5538 Non-Randomized 95 (86%) 6734 418 (82%) 17829 Chemotherapy status 0.151 Chemo naïve 28 (25%) 1792 99 (19%) 4944 Prior chemotherapy 84 (75%) 6286 430 (81%) 19127 Number of patients per arm b 0.467 35 54 (48%) 1307 277 (52%) 6249 > 35 58 (52%) 6771 252 (48%) 17822 Agent Class <0.001 Cytotoxic 1 (1%) 18 211 (40%) 9647 Targeted 111 (99%) 8060 318 (60%) 14424 Administration route <0.001 Oral 94 (84%) 7216 258 (49%) 11059 Injection 18 (16%) 862 271 (51%) 13012 FDA or EMA approval 0.149 No 16 (14%) 934 110 (21%) 4837 Yes 96 (86%) 7144 419 (79%) 19234 Tumor type 0.120 Solid 88 (79%) 4168 449 (85%) 20505 Hematologic 24 (21%) 3910 80 (15%) 3566 Number of treating centers c 0.079 Single center 32 (29%) 1259 110 (21%) 3299 Multiple centers 79 (71%) 6798 415 (79%) 20609
Response rate (%) Months Months 40 35 30 25 Response Rate (%, CI 95%) 8 7 6 5 Median PFS (Months, CI 95%) 20 18 16 14 12 Median OS (Months, CI 95%) 20 15 10 5 0 Personalized Not personalized 4 3 2 1 0 Personalized Not personalized 10 8 6 4 2 0 Personalized Not personalized Pooled analysis Meta-analysis
Forest Forest plots plots Response Response Rate Rate (RR): (RR): 31% 31% vs. vs. 10.5% 10.5% P-Values univariables P-Values Meta-reg, z=13 0.0002 0.032 0.258 0.159 0.017, z=4.6, z=10.7, z=7.5 0.0002, z=3.7 0.006, z=2.7 0.058, z=1.9 N/A N/A 0.024, z=2.3 Maria Schwaederle, PharmD
Progression-free survival (PFS): 5.9 vs. 2.7 months Personalized Non-Personalized Chemo naïve Prior therapy Solid Hematologic Cytotoxic Targeted Randomized Non-randomized 10 journal IF > 10 journal IF 35 patients/arm > 35 patients/arm Oral Injection Not approved FDA/EMA approved Single center Multiple centers 2 Months 4 Months 6 Months Maria Schwaederle, PharmD 8 Months P-Values univariables 0.012 0.114 0.353 P-Values Meta-reg, z=11.1, z=5.3, z=5.6, z=4.9 N/A, z=4.4 0.041, z=2.0 0.0001, z=3.9 0.050, z=2.0 N/A
Overall survival (OS): 13.7 vs. 8.9 months Personalized Non-Personalized Chemo naïve Prior therapy Solid Hematologic Cytotoxic Targeted 5 Months 10 Months 15 Months 20 Months P-Values univariables 0.0003 0.575 0.184 0.458 P-Values Meta-reg 0.0001, z=3.8 N/A N/A N/A Randomized Non-randomized 10 journal IF > 10 journal IF 35 patients/arm > 35 patients/arm Oral Injection Not approved FDA/EMA approved Single center Multiple centers Maria Schwaederle, PharmD 0.040 0.129 0.003 0.059 0.549 0.040, z=2.1 N/A 0.005, z=2.8 N/A N/A 0.187, z=1.3
Summary of results A personalized strategy was independently associated with higher RRs, longer median PFS and OS (all P 0.0001), as well as fewer toxic deaths (P<0.001; 1.5% vs. 2.3%) Both a personalized-direct (alteration was the direct target of the drug tested) and personalized-indirect correlated with better outcomes (all P<0.001). Personalized arms using a genomic alteration (vs. protein overexpression) as biomarker had higher RR, prolonged PFS and OS (all P<0.05). Targeted arms using a personalized strategy had statistically improved outcomes compared to targeted arms that lacked a personalized approach (All P 0.0001).
CONCLUSIONS Non-personalized targeted arms led to poorer outcomes than cytotoxics arms (All P, except P=0.048 for OS meta-analysis). Worst outcome Best outcome ARMS type Non-personalized targeted RR (%) POOLED Analysis PFS (Mos) OS (Mos) RR (%) Meta-analysis PFS (Mos) OS (Mos) 4 2.6 8.7 7.5 2.5 8.3 Cytotoxic 12 3.3 9.4 16.1 3.3 9.3 Personalized targeted 30 6.9 15.9 31.3 6.1 13.7
THANK YOU for your time and interest Questions?? rkurzrock@ucsd.edu teoam2011@gmail.com