Demonstrated with: Head of Statistics - PowerPoint Presentation

1. Demonstrated with:

2. Head of StatisticsnQuery Lead ResearcherFDA Guest SpeakerGuest LecturerWebinar HostHOSTED BY: Ronan Fitzpatrick

3. Webinar OverviewAdaptive Designs and Sample Size Re-estimation (SSR)Blinded Sample Size Re-estimation GSD, Conditional Power & Unblinded SSRDiscussion and Conclusions

4. Worked Examples OverviewTwo Means Blinded SSRTwo Proportions Blinded SSRTwo Means Group SequentialConditional PowerUnblinded SSR ExampleWORKED EXAMPLES

5. In 2017, 90% of organizations with clinical trials approvedby the FDA used nQuery for sample size and power calculation

6. Part ISSD & Adaptive Design

7. ContextSSD finds the appropriate sample size for your study Common metrics are statistical power, interval width or cost SSD seeks to balance ethical and practical issuesCrucial to arrive at valid conclusions, Type M/S errorsHigh cost of failed clinical trials drug developmentCurrently, only 10-20% of trial pathways end in successAdaptive designs presented as part of cost reductionWhere does SSD fit into the adaptive design framework?

8. Adaptive Trials OverviewAdaptive Trials are any trial where a change or decision is made to a trial while still on-goingEncompasses a wide variety of potential adaptionsE.g. Early stopping, SSR, enrichment, seamless, dose-findingAdaptive trials seek to give control to trialist to improve trial based on all available informationAdaptive trials can thus decrease costs and lead to better and more efficient inferences

9. Adaptive Trials Pros & ConsAdvantagesEarlier DecisionsReduced Potential CostHigher Potential SuccessGreater ControlBetter Seamless DesignsDisadvantagesMore ComplexLogistical Issues (IDMC)Modified Test StatisticsGreater ExpertiseRegulatory Approval?

10. Adaptive Trials Regulatory BackgroundDraft FDA CBER/CDER Guidance published in 2010“Well-understood” and “Less well-understood” DesignsEMA published similar reflection paper (2007)Increase in interest in encouraging adaptive designUS: Innovative Cures Act, EU: Adaptive Pathways New FDA Guidance expected later this yearWill likely see proliferation of new designs soonSee recent guidance on “expansion cohort” design

11. Sample Size Re-estimation (SSR)Will focus here on specific adaptive design of SSRAdaptive Trial focused on higher sample size if neededObvious adaption target due to intrinsic SSD uncertaintyNote that more suited to knowable/short follow-upCould also adaptively lower N but not encouragedTwo Primary Types: Blinded SSR & Unblinded SSRDiffer on whether decision made on blinded data or notBoth target different aspects of initial SSD uncertainty

12. Blinded Sample Size Re-estimationPart II

13. Blinded Sample Size Re-estimationBSSR uses interim blinded nuisance parameter estimateUse of blinded data reduces logistical/regulatory issuesConsidered a “well understood” type of adaptive designMultiple methods but focus on internal pilot approachUpdate N based on parameter estimate from internal pilot Use same methods as fixed term trial incl. pilot dataSmall error inflation but negligible for most casesOther methods control error but use p-value combination

14. Blinded SSR nQuery Summary (Autumn 2018)Blinded SSR MeansSSR Criteria: VarianceThree σ2 Estimate MethodsTwo Sample InequalityTwo Sample NITwo Sample EquivBlinded SSR PropsSSR Criteria: Overall Success RateAssumes effect size trueTwo Sample InequalityTwo Sample NI

15. Two Sample Mean Blinded SSR ExampleSource: NEJM (2006)ParameterValueSignificance Level (2-Sided)0.05Mean Difference (%)-9Standard Deviation (%)16Dropout Rate15%Target Power90%Nuisance Parameter?Standard Deviation“We estimated that we would need to enrol 160 patients, given an expected mean (±SD) annual decline in the FVC of 9±16 percent of the predicted value and a dropout rate of 15 percent, to achieve a two-sided alpha level of 0.05 and a statistical power of 90 percent.”Example 1

16. Two Sample Proportion BSSR Example“This number was based on results from a previous trial … where the success rate of biliary cannulation was 89.1 % for TPPP and 63.6 % for DGT [8]. The estimated probabilities of success were 60 % for the DGT group and 90 % for the TPPP group. Assuming a group difference of 30%, 32 patients per arm would provide a power of over 80 %, enough to detect a difference between the DGT group and the TPPP group, using a two-sided chi-squared test at a 5 % level of significance.” Source: Endoscopy (2018)ParameterValueSignificance Level (2-Sided)0.05Control Rate 0.6Intervention Rate0.9n per Group32Target Power (%)90%Nuisance ParameterOverall Success RateExample 2

17. GSD, Conditional Power & Unblinded SSRPart III

18. Group Sequential Designs (GSD)GSD facilitates interim analysesInterim analyses occur while trial on-going Accrued data analysed at pre-specified timesE.g. After 1/2 subjects measuredCan stop for benefit or futility If neither, continue til end/next lookMust account for multiple analysesUse “spending” of α and/or β errorsGSD ChangesFutility Only DesignsAdditional OutputsNew Two Sample TTEOne Sample Mean GSDOne Sample Prop GSD

19. Error Spending (Lan & DeMets)Two Criteria for early stoppingEfficacy (α-spending)Futility (β-spending)Multiple Error Spending Functions O’Brien Fleming, Pocock etc.Both α and β spending work similarlyCan be very liberal or conservativeAt each interim analysis, spending a proportion of the total errorMakes analysis at endpoint more conservative 

20. Group Sequential Example“A sample size of 242 subjects (121 per treatment group) provides at least 80% power to detect a relative difference of 53% between botulinum toxin A and standardized anticholinergic therapy, assuming a treatment difference of -0.80 and a common SD of 2.1 (effect size = 0.381), and a two-sided type I error rate of 5%. Sample size has been adjusted to allow for a 10% loss to follow-up over the 6-months of treatment as well as one interim analysis to stop early for benefit.”ParameterValueSignificance Level (2-sided)0.05OnabotulinumtoxinA Mean-2.3Anticholinergic Mean-1.5Standard Deviation (Both)2.1Power80%# Interim Analyses1α Spending FunctionO’Brien-FlemingExpected Dropout10%Source: NEJM (2012)Example 3

21. Conditional Power (CP)CP gives prob. of rejecting null given interim test statisticCalculation still depends on what “true” difference set toOften used as ad-hoc criteria for futility testing in GSDMore flexible than β-spending but less error guaranteeFocus here on CP as measure of “promising” results“Promising” meaning less than target but close to target powerNote existence of related Bayesian Predictive PowerEssentially conditional power averaged over prior for effect

22. Conditional Power & Unblinded SSRMost common criteria proposed for unblinded SSR is CPSSR suggested when interim results “promising” (Chen et al)Gives third option vs GSD: continue, stop early, increase N“Promising” user-defined but based on unblinded effect sizePower for optimistic effect but increase N for lower relevant effects?2 methods here: Chen, DeMets & Lan; Cui, Hung & Wang1st uses GSD statistics but only penultimate look & high CP2nd uses weighted statistic but allowed at any look and CPInitial nQuery Adapt release will be two means & proportions

23. Conditional Power & Unblinded SSR ExampleAssume same design as GSD Example (Example 3) futility variant (n = 115 per group)Assume interim difference = 0.65, interim common SD = 2.1and interim N of 58 per group with nominal alpha of 0.0245 for final look.What will required sample size be for SSR if CP between 50% and 80% where maximum n = 230?ParameterValueNominal Final Look Sig. Level0.0245Treatment Interim Mean-2.3Control Interim Mean-1.5Interim SD (Both)2.1Initial N per Group115Interim N per Group1Lower CP Bound50%Upper CP Bound80%

24.

25. Part 4Discussion & Conclusions

26. Discussion and ConclusionsAdaptive Trials expected to become more commonReduction of costs, greater regulatory interest etc.SSR will be one common type of adaptive trialBlinded SSR already widely accepted, unblinded growingBlinded SSR targets initial variance under-estimatesInternal pilot allows easy re-estimate w/o new test statisticUnblinded SSR targets effect size and extends GSDConditional power allows ability to save promising study

27. nQuery Spring 2018 UpdateInitial release focused on Survival & Bayesian tables.April release adds 72 new tables in following areas:New Bayes tables in April updateNew tables in April updateEpidemiology Non-inferiority/EquivalenceCorrelation/ROCBayesianSample Size5220

28. nQuery Autumn 2018 UpdateAutumn 2018 release adds nQuery Adapt module, 32 new tables & undo/redo New Core TablesProportions + Crossover AssuranceConditional PowerGST + SSR20nQuery Bayes Tables12nQuery Adapt Tables15

29. Q&AAny Questions?For further details, contact at: info@statsols.comThanks for listening!

30. ReferencesFriede, T., & Kieser, M. (2006). Sample size recalculation in internal pilot study designs: a review. Biometrical Journal: Journal of Mathematical Methods in Biosciences, 48(4), 537-555..Tashkin, D. P., Elashoff, R., Clements, P. J., Goldin, J., Roth, M. D., Furst, D. E., ... & Seibold, J. R. (2006). Cyclophosphamide versus placebo in scleroderma lung disease. New England Journal of Medicine, 354(25), 2655-2666.Sugiyama, H., Tsuyuguchi, T., Sakai, Y., Mikata, R., Yasui, S., Watanabe, Y., ... & Nishikawa, T. (2018). Transpancreatic precut papillotomy versus double-guidewire technique in difficult biliary cannulation: prospective randomized study. Endoscopy, 50(1), 33-39.Jennison, C., & Turnbull, B. W. (1999). Group sequential methods with applications to clinical trials. CRC Press.Visco, A. G., et al (2012). Anticholinergic therapy vs. onabotulinumtoxina for urgency urinary incontinence. New England Journal of Medicine, 367(19), 1803-1813.Chen, Y. J., DeMets, D. L., & Gordon Lan, K. K. (2004). Increasing the sample size when the unblinded interim result is promising. Statistics in medicine, 23(7), 1023-1038.Cui, L., Hung, H. J., & Wang, S. J. (1999). Modification of sample size in group sequential clinical trials. Biometrics, 55(3), 853-857.