Phase 2b dose selection, leveraging comparator data through multidisciplinary modeling & simulation Thomas Kerbusch Thomas Kerbusch* (1), Russ Wada (2), Anthe Zandvliet (1), Kuenhi Tsai (6), Jing Su (6), Joanna Zhuoying Peng (4), Yaming Hang (6), Christine Xu (3), Richard Shames (5), Ann Horowitz (3), Diane Neiman (4), Mani Lakshminarayanan (6), Usha Barai (3), Ferdous Gheyas (3), Paul Peloso (4), Devan Mehrotra (6), Nancy Zhang (2), Hanbin Li (2), Jaap Mandema (2), Gary Herman (4), Sandy Allerheiligen (6) (1) Merck Research Laboratories, Oss, The Netherlands; (2) Quantitative Solutions, Menlo Park, CA, USA; (3) Merck Research Laboratories, Kenilworth, NJ, USA; (4) Merck Research Laboratories, Rahway, NJ, USA; (5) Merck Research Laboratories, Palo Alto, CA, USA; 6 Merck Research Laboratories, Upper Gwynedd, PA, USA.
Multi-disciplinary collaboration CRO: clinical Quantitative M&S PK-PD Solutions combine quantitative thinking capture physiology & pharmacol. share workload agree on assumptions FIT FOR Clin Pharm BARDS PURPOSE Clin Res define end-points and decision criteria explore opportunities to optimize trial design
Difficult to establish a dose-response relationship based on Phase 1b data • Cohorts Phase 1b study: – ● 1x mg/kg (n=4) IV – ▲ 5x mg/kg (n=4) IV – ♦ 30x mg/kg (n=8) IV – ▼ 100x mg/kg (n=8) IV – ￭ placebo ratio 3:1 • Strong PoC response: All active treatments resulted in maximal effect at W28 • Need to select Phase 2b dose-range using Phase 1b data • Problem: – Limited Phase 1b data, with n=3-6 patients per dose group. – No robust PKPD model could be established. IV doses at 0, 8, and 12 weeks
Dose selection for Phase 2b study using limited Phase 1b data • Initial proposal for Phase 2b dose selection: based on max injectable SC dose – 25 mg SC at W0, W4, W16 (n=35) – 100 mg SC at W0, W4, W16 (n=70) – 200 mg SC at W0, W4, W16 (n=70) – 200 mg SC every 4 weeks (n=35) – Placebo SC every 4 weeks (n=35) • Decision required: Phase 2b dose-selection – Will the initially proposed dose range (25-200 mg) allow for: • estimation of dose-response? • determination of lowest maximum effective dose? • Actions: Conduct a comparator analysis – Model-based dose selection • Solution to in-house data limitations: borrow strength from published comparator data – Best-in-Class strategy • Explore comparator landscape to understand requirements for Phase 3 doses – need for maximum learning in Phase 2b
Critical assumptions comparator analysis • The maximum efficacy for Merck’s compound is similar to other compounds with similar mechanism-of-action (MoA) • The time-course of the onset of response is similar across compounds • The efficacy of Phase 1b dose regimen (wk0, wk8, wk12) of Merck’s compound is similar to the efficacy of the Phase 2b dose regimen (wk0, wk4, wk16). • Phase 1b and Phase 2 patient populations are assumed the same. • Mean study-arm level data were combined for over 10,000 patients. Compound MoA # Trials # Study arms # pts incl. plac Adalimumab (Humira) Type 1 4 9 1658 Etanercept (Enbrel) Type 1 9 20 2868 Infliximab (Remicade) Type 1 6 15 1695 Ustekinumab (Stelara) Type 2 5 13 2868 Briakinumab (ABT-874) Type 2 2 6 1585 Merck compound 1 5 24
Confidence in estimating efficacy response can be enhanced by co-modeling through correlation Data are plotted across all arms and time points in the database. Symbol size is proportional to the square root of the arm size. The fitted line is the cumulative normal distribution with mean of log (disease score / baseline), standard deviation of 0.66, and cutoff at -1.2.
Dose-response models of in-house compounds and competitors (week 16 , 95% CI) in-house compound SCH900222 adalimumab briakinumab • All compounds were 100 100 100 estimated with 80 80 80 different potencies 60 60 60 40 40 40 • Onset of efficacy % Reduction/Responder 20 20 20 faster for mean % 0 0 0 change than for 0 2 4 6 8 10 0 20 40 60 80 0 200 400 600 800 responder fraction etanercept infliximab ustekinumab 100 100 100 80 80 80 • Limited Ph 1b data 60 60 60 (n=24) resulted in 40 40 40 large uncertainty Disease score 20 20 20 continuous endpoint dichotomous Mean PASI 0 0 0 PASI75 0 10 20 30 40 50 0 2 4 6 8 10 0 50 100 150 Adjusted Dose
Dose-response model of in-house compound: increased response over time (80%CI) Dose-response • Near maximum effect is 4 week 8 week predicted to be achieved ≥50 80 80 mg. 60 60 % response, difference from placebo 40 40 • Doses of 50-200 mg are 20 20 predicted to have little 0 0 separation in time to reach 1 5 10 50 100 1 5 10 50 100 maximum effect. Therefore, 12 week 16 week 200 mg is not predicted to 80 80 have a faster maximum 60 60 effect. 40 40 • 5 mg and 25 mg will allow for 20 20 doses near ED50 (~8.4 mg) 0 0 1 5 10 50 100 1 5 10 50 100 and are predicted to allow for SCH900222 D ( ) Dose in-house compound (mg) establishing dose-response
Positioning of in-house compound in the competitive landscape: a sneak preview • Doses > 50 mg (0, 4 16w) predicted to be superior (positive difference in plot) to etanercept, adalimumab and ustekinumab. • Similar potency and onset of action → no major competitive advantage over ustekinumab adalimumab 40 mg q2w briakinumab 200 mg wk 0,4, 100 mg wk 8 etanercept 50 mg biw 20 20 20 % response, difference from comparator 0 0 0 -40 -40 -40 -80 -80 -80 1 5 10 50 1 5 10 50 1 5 10 50 infliximab 5 mg/kg wk 0,2,6, q8w ustekinumab 45 mg wk 0,4, q12w 20 20 • After Phase 2b a much 0 0 more accurate positioning -40 -40 within the competitor landscape can be -80 -80 determined 1 5 10 50 1 5 10 50 Dose in-house compound (mg)
Clinical trial simulations • Why – Including doses around ED50 in the Phase 2b trial will allow for identification of the lowest dose reaching maximum effect (“cusp of Emax”): best dose for Phase 3. • How – Limitations and uncertainty in the available data are a fact. – Clinical trial simulations should incorporate these and still allow for a robust dose selection decision for Phase 2b. – Dose-range should bracket (predicted) maximum response and ED50. – Doses for Phase 2b were evaluated for being “near” placebo, “near” maximum effect or in between (near ED50) by simulating 200,000 Phase 2b trials.
Making the decision 100% 100% 100% • 100 and 200 mg at plateau (difference at plateau (difference at plateau (difference from maximal from maximal from maximal were predicted to be response <25%) response <25%) response <25%) 80% 80% 80% at the plateau of the dose-response between placebo and between placebo and between placebo and Probability (%) plateau plateau plateau relationship. 60% 60% 60% • Monthly 200 mg similar to placebo similar to placebo similar to placebo (max. feasible (difference from (difference from (difference from 40% 40% 40% exposure) is not placebo response placebo response placebo response <25%) <25%) <25%) informative: drop arm 20% 20% 20% • Establishing dose- original design (empirical response requires a dose selection) dose level between 0% 0% 0% placebo and plateau: 5 mg 5 mg 5 mg 25 mg 25 mg 25 mg 100 mg 100 mg 100 mg 200 mg 200 mg 200 mg 200 mg 200 mg 200 mg new design reasonable (n=35) (n=35) (n=35) (n=35) (n=35) (n=35) (n=70) (n=70) (n=70) (n=70) (n=70) (n=70) (n=35) (n=35) (n=35) (model-based probability of 5 mg ----------- 0, 4, 16W ----------- ----------- 0, 4, 16W ----------- ----------- 0, 4, 16W ----------- Q4W Q4W Q4W dose selection) not being near plateau or placebo add 5 mg (yellow bar). omit 200 mg Q4W
25 mg W0, W4, W16 (n=35) 5 mg W0, W4, W16 (n=35) 100 mg W0, W4, W16 (n=70) 25 mg W0, W4, W16 (n=70) Conclusions 200 mg W0, W4, W16 (n=70) 100 mg W0, W4, W16 (n=70) 200 mg every 4 weeks (n=35) 200 mg W0, W4, W16 (n=70) Placebo every 4 weeks (n=35) Placebo W0, W4, W16 (n=35) Value addition of M&S Comparator Data Analysis • Probability of success of establishing dose-response: – was low for the original design: • 200 mg arm with monthly injections (max. feasible dose- intensity) is not informative, because 200 mg W0, W4, W16 already predicted above Emax – is high for the new design: • A 40-fold dose-range 5-200mg brackets the predicted ED50 (8.4 mg) and maximum effect (≥50 mg). • Model-based dose-response in phase 2b will allow for optimal dose- selection for phase 3. • Re-evaluating Competitive Landscape will allow for optimal Best-in- Class strategy
BACKUP
Comparative efficacy model • The model has a maximum effect that gradually increases over time to a steady-state value. • There is a dose-response relationship at each point in time. • Key assumptions: • The maximum efficacy for in-house compound is similar to competitors with same MoA • The time-course of the onset of response is similar across compounds 100 100 mean disease score / baseline (%) Emax Emax 80 80 response rate (%) PASI75 Responders (%) High Dose PASI Reduction (%) 60 60 High Dose Low Dose 40 40 Low Dose 20 20 Placebo Placebo 0 0 0 5 10 15 0 5 10 15 Time (week) Time (week) 14
Recommend
More recommend
