Posters: 1 Minute Previews - PowerPoint Presentation

1. Posters: 1 Minute Previews

2. Careers andAcademicLiaison Committee Inspiring school students to study statistics at our event, “Maths meets Medicine”An enthusiastic group of volunteers from a range of companies from across the industryInvestigating the need for MSc funding and demand for MSc students in the industry and academiaHelping students find out about careers in the pharmaceutical industry and how to apply at our PSI MSc Careers Fair(53)

3. 3Bootstrap Survival Analysis in a Life-threatening and Ultra-rare DiseaseClare Elkins and Richard Kay In life-threatening and ultra-rare diseases it is often considered unethical to have a contemporary, concurrently placebo control groupStudy Design and EndpointTwo multicenter, open-label studies to assess the safety and efficacy of a treatment for a life-threatening and ultra-rare disease in 37 infants and children (≤ 5 years of age)Study ALXN-NH was a retrospective, non-interventional epidemiologic study of the natural history of 48 patients with the same life-threatening and ultra-rare disease The primary endpoint was overall survival time defined as number of days from birth to deathFor this analysis, historical survival data was used to perform bootstrap survival analysis that matched historical control patients to treated patients by key risk factors to address potential selection biasesThis method allowed the opportunity to evaluate survival data from treated patients against historical control data, utilizing a conservative and statistically valid analysis controlling for risk factors and patients selection bias 1

4. A Practical Application of Multiple Imputation (MI) in a Double Blinded Randomized Phase III Study in Spinal Muscular AtrophyResults from simulation and studyAdvantages in using MIIncrease in powerChallenges Timing of interim analysisDetermination of alpha at final analysis2

5. Use of a composite criterion taking into account rescue medication intake in an orphan disease with a seasonal patternMaëva Dupuis-Deniaud(1), Bruno Scherrer(2), Dahlia Ismail(3), Mourad Amrane(3)Ophtalmologic, rare and debilitating condition : VKC (Vernal Keratocunjunctivits)Double-masked randomized Phase III (Placebo=Vehicle) controlled of 4 months study : VEKTIS STUDYMethodological requirements, discussed with EMA during the Scientific Advice :The main endpoint needs to show a continous effect during the allergic season, not only after 4 monthsA rescue therapy is needed for ethic reason but need to be taken into accountA corticosteroïds-sparing effect study is not possible (no AMM in this indication)As a orphan disease, the number of patients needs to be minimized, and several countries around the world have to be includedThe allergic specificity of the disease has to be consideredLack of previous data (one phase II, with a reduced number of patients)Pediatric population specificity to be addressed (QOL, VAS…)A composite endpoint, including the main measurement (Corneal Fluorescein staining), the rescue therapy intake and ulceration (as penalties) was built, methodologically discussed and endorsed as the main endpoint for the pivotal phase III study.4

6. Monte Carlo SimulationApplication of Monte Carlo Simulation to Relative Survival and BeyondGenerating comparable control populationNo control Arm?Life TablesNational StatisticsKaplan Meier estimatorEstimation of risk relative to background populationempirical solution!Log-Rank testCox RegressionNon-parametric survival modelsEderer I, IIHakulinenBy Mansour Sharabiani PhD5

7. Bayesian methods for leveraging existing clinical data in paediatric trialsNicky Best, Dawn Edwards, Wen WuRecent regulatory interest in partial extrapolation of adult efficacy to paediatrics to reduce data collection requirements in childrenAdolescent study design N = 130 per armSuccess criteria: Posterior Pr(Δ>0) > q% q  (80%, 90%, 95%, 97.5%)Total prior weight on adult information: w w  (0, 0.1, ..., 0.9, 1)Simulation study to investigate operating characteristics for different choices of q and wCase study describing plans to use partial extrapolation of adult efficacy in a phase III trial of an experimental v control regimen in adolescents with a respiratory diseasew/2Vague priorRescaled adult prior Adult prior3 component robust mixture prior for adolescent treatment effectw/21-w6

8. BayesianApplying Bayesian stopping rules to a pilot study with a rare diseaseRare diseaseLimited sample sizeStopping rulesPrior distributionPosterior probabilitiesBeta distributionsResponse RateFutilityFrequentistSimon Two Stage designBy Komel Khabra7

9. Bayesian Prior’s Essential Role in Innovative Trial Designs for Rare Diseases Drug DevelopmentHan S., Palmer, JP, Sun P., Early Clinical Development, Pfizer Inc. Cambridge MA, USAResults: Simulations suggest, assuming the correlations in NH longitudinal data , enrollment of subjects with BL 6MWD <300m or >400m can be stopped once 30-45 subjects enrolledPre-specify N candidate study populations (N ≤ 4) by distinct Baseline functional score intervals and the projected progression trajectories|where “mid” stands for the population with BL 6MWD in [300, 400m] interval, and the prior probability is approximately within [60%, 75%]Decision criterion: drop sub-optimal study populations once | 8

10. Bayesian Belief Networks and success of phase 3 trialsWilmar Igl, Laura Barker, Jonathan Bartlett, Ozgur Ozkan, Bart Willigers(AstraZeneca)Factors to adjust prior PTS value include:Basic Factors, eg biological modelDesign Factors, eg control groupDesign Risk Factors, eg drop-outsDrug Dev Quality Factors, eg phase 2 trialIntegration of PTS procedure using a web appCreate projectCollect beliefs of subject matter expertsValidate across multiple trials after final analysisPSI, London, 2017-05-14/17, Poster presentationEstimation of PTS of phase 3 trials important, but often not done right due to lack of dataUse subjective beliefs of team members and other experts since relevant data is not availableBayesian Belief Networks can help integrate objective data and subjective beliefs9

11.              Choose and treatment optimally:  Choose optimally:  Quantify the additional value brought by using: (i) Combination Testing (ii) Group Sequential Designs     Overall ‘Gain’10

12. 11

13. Stéphanie CADOURA unique model by applying a Dirichlet distribution to the weights of the criteriaCriterion parameters are random variablesDoes not require the elicitation of preferences to weigh the criteriaAllows more flexibility by making the variance of the weights varyAll parameters have a natural interpretation: treatment effects, decision-makers’ preferences and their strength of confidenceDirichlet SMAASince 2010, quantitative methodologies have been developed to assess the benefit-risk balance of drugs. In particular:An improved model in drug benefit-risk assessmentStephanie Cadour1, Gaelle Saint-Hilary2, Veronique Robert1, Mauro Gasparini21 Department of Biostatistics, Institut de Recherches Internationales (IRIS), Suresnes ,France, 2 Dipartimento di Scienze Matematiche (DISMA) Giuseppe Luigi Lagrange, Politecnico di Torino, ItalyStochastic Multicriteria Acceptability Analysis (SMAA) Criterion parameters are random variables Does not require the elicitation of preferences to weigh the criteriax Increased complexity making the interpretation more difficultx High degree of uncertainty in the resultsProbabilistic Multi Criteria Decision Analysis (pMCDA) Criterion parameters are random variablesx Weights are fixed values and need to be defined by decision-makers12

14. Software from Cytel and AstraZeneca for early clinical study designThe Target Value (TV) for a new medicine is defined as the desirable level of clinical activity, which would encourage physicians to switch their patients to the new treatment.The Lower Reference Value (LRV) is defined as the lowest level of efficacy that would be considered clinically meaningful. This could be a zero treatment effect in some cases -- in an active controlled trial, for example, where the key differentiator is expected to be safety.Example input:Example output:13

15. Re-engineering Early Drug Development – GSK’s ApproachEarly phase drug development should de-risk late stage attrition.Hypothesis testing and p-values do not address questions of interest:“Given what we know now, what is the probability that the study will continue to be successful?”“What is the smallest dose leading to an effect of size X?”“How should historical data impact study design in a transparent way?”GSK developed a transformational initiative, Re-Engineering Phase 2 (RP2), with the aim of embedding more quantitative and efficient study designs in early phase development. Maria J. Costa, Jon Robertson and Graeme Archer, GSK R&D, UKPSI 2017RP2 PillarsFutility and Predictive InferenceDose-Response ModellingHistorical DataPeer Review14

16. 15

17. crmPack’s dual endpoint approach to jointly model safety and PD: Operating characteristics for a single-ascending dose-escalation study with a continuous biomarker responseCharles Warne, Roche Products UKWhile oncology First-In-Human SAD studies traditionally escalate towards a Maximum Tolerated Dose, this may not be appropriate for other indications if the Optimal Biological Dose is expected to be lower.Dose escalation designs incorporating both safety and PD response may therefore be safer and more efficient at gaining information at the optimal dose.This poster evaluates the dual endpoint CRM model in the R package crmPack for dose escalation decision making in a FIH healthy volunteer study where the goal is to: maximise the probability of observing a biomarker value of 400-450, andlimit to <5% the probability of observing a Dose Limiting Event rate >20%.Simulations were used to investigate the operating characteristics of the dual endpoint CRM model under a broad range of DLE and biomarker scenarios. Results showed that the dual endpoint CRM approach successfully controlled the risk of observing a DLE rate >20% for all scenarios explored, while maximizing the information gained around doses that resulted in the target biomarker level of 400-450.This leads to more relevant information to inform later studies, and smaller trials when safety and biomarker levels are favourable16

18. 18

19. The Design of a 3-part Phase II Dose-finding Trial: A Case Study using MCP-modRebecca FinchMCP-mod19

20. A more efficient approach for renal/hepatic impairment clinical pharmacology studiesTraditionally~6-8 subjects in each of 4 renal/hepatic impairment groups (N=24–32)Strict matching on demographic characteristics (e.g. age, gender, BMI)Group-level pairwise comparisons of pharmacokinetic parametersAllows greater flexibility in subject matchingReduces required study sizeEliminates need to enrol subjects with normal renal/hepatic functionKatie Patel, Poster #20New & Improved!Direct modelling of renal/hepatic function and drug clearance20