![Page 1: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/1.jpg)
Better planning through design: Modeling and forecasting enrollment and event arrivals
to optimize trial execution Yannis Jemiai, Ph.D.
2/22/2012 ©2012 Cytel Inc. 1
![Page 2: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/2.jpg)
• Introduction • Benefits of modeling and forecasting recruitment
and event accruals • A simulation study • Case study: Phase 3 in oncology • Concluding remarks
2/22/2012 ©2012 Cytel Inc.
Outline
2
![Page 3: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/3.jpg)
• According to a recent report in Cutting Edge Information, clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines in Phase III - all because of recruitment delays. (Beasely, "Recruiting". 2008)
• Six percent of clinical trials are completed on time, and 72% of trials run over schedule by more than one month. (www.fda.gov)
• Eighty percent of total trials are delayed at least one month because of unfulfilled enrollment. (Lamberti, "State of Clinical Trials Industry", 292)
2/22/2012 ©2012 Cytel Inc.
Some facts and figures
3
![Page 4: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/4.jpg)
• Out of all of the research sites in the United States, less than 1/3 contain 70% of the valuable subjects. Therefore 70% of the research sites under-perform, and somewhere between 15%-20% never enroll a single patient. (Pierre, "Recruitment and Retention". 2006)
• Fifty percent of clinical research sites enroll one or no patients in their studies. (Pierre, "Recruitment and Retention". 2006)
2/22/2012 ©2012 Cytel Inc.
Some facts and figures
4
![Page 5: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/5.jpg)
Benefits of modeling Meeting patient recruitment challenges in clinical trials
2/22/2012 ©2012 Cytel Inc. 5
![Page 6: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/6.jpg)
• Significant resources and strategic planning are contingent upon the timing of interim and final data analyses
• Modeling recruitment and event accrual based on current accumulated data allows early and accurate predictions of interim analysis times and study termination
• Statistical modeling also provides confidence levels for predictions
2/22/2012 ©2012 Cytel Inc.
Why model patient recruitment?
6
![Page 7: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/7.jpg)
• Randomness is natural and accounts for many observed phenomena in recruitment (Anisimov, 2009)
• Randomness in recruitment coupled with randomization can substantially affect the statistical validity of study results; and have major implications on the very costly drug supply chain process
2/22/2012 ©2012 Cytel Inc.
Impact of predictions?
7
![Page 8: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/8.jpg)
• Opening new centers to meet recruitment deadlines
• Optimizing (avoiding drug shortage or overage) drug supply chain to meet predicted demand based on recruitment, retention, and randomization forecasts
• Cost savings: – Can centers be closed? – Can drug supply be redirected to other sites? – Can site monitoring be better planned for CRAs to spend time
where most needed?
2/22/2012 ©2012 Cytel Inc.
Taking corrective actions
8
![Page 9: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/9.jpg)
Simulation Study
2/22/2012 ©2012 Cytel Inc. 9
![Page 10: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/10.jpg)
• Number of subjects to be enrolled = 640
• Target recruitment period = 200 days
• 21 sites in US, estimated approximately equally spaced Site Initiation Visit (SIV) times between day 1 and day 50, av. time = 25 days
• Average enrollment rate per site to achieve target given estimated SIV times is about 3.5 patients/site/month (0.17 patients/site/day)
• Clinical Operations estimated Standard Deviation of enrollment rate between sites to be about 50% of the average (0.085 patients/day)
How the simulation model works
2/22/2012 10 ©2012 Cytel Inc.
![Page 11: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/11.jpg)
• Actually recruitment took 255 days • Site opening took longer than anticipated: last site SIV
was 143 days and average SIV time was about 60% longer than planned
11
What really happened
2/22/2012 ©2012 Cytel Inc.
![Page 12: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/12.jpg)
12
Estimated SIV times
2/22/2012 ©2012 Cytel Inc.
![Page 13: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/13.jpg)
13
100 simulations at study design time
2/22/2012 ©2012 Cytel Inc.
![Page 14: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/14.jpg)
14
Study design time recruitment forecast
2/22/2012 ©2012 Cytel Inc.
![Page 15: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/15.jpg)
15
Study design time forecast and actual
2/22/2012 ©2012 Cytel Inc.
![Page 16: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/16.jpg)
16
Forecast after 50 days
2/22/2012 ©2012 Cytel Inc.
![Page 17: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/17.jpg)
17
Forecast after 50 days with actual
2/22/2012 ©2012 Cytel Inc.
![Page 18: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/18.jpg)
18
Forecasts after 0 and 50 days
2/22/2012 ©2012 Cytel Inc.
![Page 19: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/19.jpg)
19
Forecasts after 50 and 100 days
2/22/2012 ©2012 Cytel Inc.
![Page 20: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/20.jpg)
20
Forecasts after 100 and 150 days
2/22/2012 ©2012 Cytel Inc.
![Page 21: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/21.jpg)
21
Forecast after 200 days with actual
2/22/2012 ©2012 Cytel Inc.
![Page 22: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/22.jpg)
22
Forecast after 230 days with actual
2/22/2012 ©2012 Cytel Inc.
![Page 23: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/23.jpg)
23
Drilling down: Site 7 high recruiting (69 subj)
2/22/2012 ©2012 Cytel Inc.
![Page 24: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/24.jpg)
24
Drilling down: Site 8 low recruiting (8 subj)
2/22/2012 ©2012 Cytel Inc.
![Page 25: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/25.jpg)
Case Study Phase 3 trial in patients with relapsed multiple myeloma
2/22/2012 ©2012 Cytel Inc. 25
![Page 26: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/26.jpg)
2/22/2012 ©2012 Cytel Inc.
Probability of achieving milestones
26
![Page 27: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/27.jpg)
2/22/2012 ©2012 Cytel Inc.
500 subject enrollment prediction
27
![Page 28: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/28.jpg)
2/22/2012 ©2012 Cytel Inc.
Event prediction to reach 187‐375 targets
28
![Page 29: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/29.jpg)
2/22/2012 ©2012 Cytel Inc.
Enrollment prediction to reach 750 subj
29
![Page 30: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/30.jpg)
2/22/2012 ©2012 Cytel Inc.
Event prediction to reach 562 target
30
![Page 31: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/31.jpg)
2/22/2012 ©2012 Cytel Inc.
Milestones probability distribution
31
![Page 32: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/32.jpg)
2/22/2012 ©2012 Cytel Inc.
Event milestones probability distribution
32
![Page 33: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/33.jpg)
2/22/2012 ©2012 Cytel Inc.
Monitoring site performance
33
![Page 34: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/34.jpg)
2/22/2012 ©2012 Cytel Inc.
Monitoring site performance
34
![Page 35: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/35.jpg)
2/22/2012 ©2012 Cytel Inc.
Monitoring country performance
35
![Page 36: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/36.jpg)
Concluding remarks
2/22/2012 ©2012 Cytel Inc. 36
![Page 37: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/37.jpg)
• Adaptive designs rely on being able to make decisions early enough (before end of recruitment) to change important aspects of the study design, for example: – sample size – early stopping / dropping treatment arms – randomization / treatment allocation ratios – patient inclusion/exclusion criteria
• Usually require rapidly observed endpoints and slow recruitment rates
• If recruitment needs to be slowed, then the trade-off between the benefits of adapting and those of rapid enrollment need to be investigated and assessed
2/22/2012 ©2012 Cytel Inc.
Adaptive considerations
37
![Page 38: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/38.jpg)
• Modeling, simulating and forecasting enrollment and the arrival of events can help a sponsor or CRO get a grip on the uncertainty inherent in trial timelines
• This quantitative exercise combines with the “art of patient recruitment and retention” to run more efficient and successful studies
• Helps track issues such as poor-performing countries or sites, and guide decisions such as when to open new sites, or how and when to resupply sites with drug
• Cost savings can be substantial if impact on NPV, drug supply chain, and site monitoring is taken into account and acted upon
2/22/2012 ©2012 Cytel Inc.
Take home messages
38
![Page 39: Better planning through design: Modeling and forecasting ... · clinical trials last 42% longer than expected in Phase I, 31% longer in Phase II, and 30% beyond planned deadlines](https://reader035.vdocuments.site/reader035/viewer/2022081406/5f0fc7e37e708231d445d957/html5/thumbnails/39.jpg)
2/22/2012 ©2012 Cytel Inc. 39
47ofthetop50biopharmaceu3calfirmsuseCytelso9wareandservicestodesign,simulate,analyzetheirclinicaltrials.
The Pharm Exec 50—the world’s top 50 pharmaceutical companies according to PHARMACEUTICAL EXECUTIVE
Making science & technology work for you
Learn more at www.cytel.com