Handout for talk entitled Who Needs Bayesian Phase I Trials? in Session 378 by Rick Chappell, University of Wisconsin-Madison (chappell@stat.wisc.edu) Table 4: Design Types Used in the Simulation Design Type Design Traditional 3+3 Design Cohort # DLTs Size or or 10 Traditional 5+5 Design Up-and-down, cohort size adaptive to two sub-DLTs; increment size adaptive then # SubDLTs Action 0/3 1/3 ≤ 1/6 >1/6 >1/3 any any any any any Escalate to next cohort Expand cohort to pts at same dose Escalate to next cohort Stop escalation – MTD defined Stop escalation – MTD defined 0/5 1/5 ≤ 1/10 >1/10 >1/5 any any any any any Escalate to next cohort Expand cohort to 10 pts at same dose Escalate to next cohort Stop escalation – MTD defined Stop escalation – MTD defined 0/2 0, 1/2 Escalate to next cohort of subjects 0/2 2/2 Expand cohort, subjects at same dose; all future cohort sizes = 1/2 any Expand cohort, subjects at same dose; all future cohort sizes = 2/2 any De-escalate; all future cohort sizes = and increments = 6.25 0/5 ≤ Escalate to next cohort 0/5 >2 Escalate to next cohort; all future increments = 6.25 1/5 ≤ Expand cohort, 10 subjects at same dose >2 Expand cohort, 10 subjects at same dose; all future increments = 6.25 any De-escalate; all future increments = 6.25 >1/5 Sample Size Phase Dose Random One below the stopping dose Random ~20-24 Highest dose with ≤ 20% DLTs Highest dose with ≤ 20% DLTs Table 5: Doses Used in High-Toxicity (HT) Scenario Simulations Dose # start 1.5 2.5 goal 3.5 4.5 5.5 6.5 7.5 Dose (units omitted) 6.25 12.5 18.75 25 31.25 37.5 43.75 50 56.25 62.5 68.75 75 81.25 87.5 93.75 100 Chance of DLT (%) Chance of Sub-DLT (%) 12 16 20 24 30 36 44 52 60 70 80 100 100 12 18 24 30 36 45 54 66 78 90 100 100 100 100 Table 6: Simulation Results for High-Toxicity (HT) Scenario Performance Phase dose identified within 6.25 Phase dose identified within 12.5 Mean number of DLTs Mean % DLTs Mean % exposed to sub-therapeutic doses1 Mean % exposed to toxic doses2 Mean (range)3 sample size Mean (range)3 duration, weeks4 Design 31% 78% 2.8 20% 45% 17% 14.9 (3, 27) 39.7 (8, 72) Design 32% 80% 2.9 15% 55% 7% 19.9 (5, 35) 31.9 (8, 56) Design 57% 85% 4.3 19% 37% 27% 22.2 (20, 24) 53.6 (40, 64) Sub-therapeutic is defined to be at least 12.5 below the true recommended dose (here, 25 or less) A toxic dose is defined to be at least 12.5 above the true recommended dose (here, 50 or more) A modified definition of range (minimum, 99 percentile), rather than the typical minimum/maximum, is used here in order to lessen the influence of extremely rare trial outcomes For statistics on timing, accrual of all members of a cohort was assumed to be simultaneous and follow-up is weeks Thus, if excess toxicities are observed in the first 3/5 subjects, the simulation does not halt treatment for the last 2/5 if these subjects have not been accrued, as might happen in the actual trial Table 7: Doses used in Low-Toxicity (LT) Scenario Simulations Dose # start 1.5 2.5 3.5 4.5 goal 5.5 6.5 7.5 8.5 Dose (units omitted) 6.25 12.5 18.75 25 31.25 37.5 43.75 50 56.25 62.5 68.75 75 81.25 87.5 93.75 100 106.25 112.5 Chance of DLT (%) Chance of Sub-DLT (%) 10 12 20 28 36 50 65 80 90 100 100 1.5 4.5 12 15 18 30 42 54 75 100 100 100 100 100 Table 8: Simulation Results for Low-Toxicity (LT) Scenario Performance Phase dose identified within 6.25 Phase dose identified within 12.5 Mean number of DLTs Mean % DLTs Mean % exposed to sub-therapeutic doses1 Mean % exposed to toxic doses2 Mean (range)3 sample size Mean (range)3 duration, weeks4 Design 37% 85% 3.0 15% 61% 11% 20.5 (3, 30) 54.5 (8, 80) Design 36% 78% 3.0 10% 68% 5% 29.8 (5, 45) 47.7 (8, 72) Design 59% 89% 3.3 15% 57% 18% 21.9 (20, 24) 60.9 (40, 72) Sub-therapeutic is defined to be at least 12.5 below the true recommended dose (here, 50 or less) A toxic dose is defined to be at least 12.5 above the true recommended dose (here, 50 or more) A modified definition of range (minimum, 99 percentile), rather than the typical minimum/maximum, is used here in order to lessen the influence of extremely rare trial outcomes For statistics on timing, accrual of all members of a cohort was assumed to be simultaneous and follow-up is weeks Thus, if excess toxicities are observed in the first 3/5 subjects, the simulation does not halt treatment for the last 2/5 if these subjects have not been accrued, as might happen in the actual trial ... A modified definition of range (minimum, 99 percentile), rather than the typical minimum/maximum, is used here in order to lessen the influence of extremely rare trial outcomes For statistics... A modified definition of range (minimum, 99 percentile), rather than the typical minimum/maximum, is used here in order to lessen the influence of extremely rare trial outcomes For statistics... 66 78 90 100 100 100 100 Table 6: Simulation Results for High-Toxicity (HT) Scenario Performance Phase dose identified within 6.25 Phase dose identified within 12.5 Mean number of DLTs Mean %