TUTORIAL on ICH E9 and Other Statistical Regulatory Guidance Session 1: PSI Conference, May 2011 Kerry Gordon, Quintiles 1
E9, and how to locate it 2
ICH E9 Statistical Principles for Clinical Trials (Issued 1998) http://www.ema.europa.eu/docs/en_gb/document_library/scientific_guideline/2009/0 9/WC500002928.pdf, 37-page document Confirmatory / Exploratory Trials Study population Primary and secondary endpoints Blinding Randomisation Trial designs Multicentre trials Superiority / Equivalence / Non-inferiority Sample size Sequential designs / interim analyses Pre-specification of analyses Full analysis set (and the intention-to-treat principle) / Per-protocol analysis set Analysis considerations (including estimation, CIs and hypothesis testing) Multiplicity Subgroup analyses Evaluation of safety and tolerability Reporting 3
ICH E9 Statistical Principles for Clinical Trials (Issued 1998) http://www.ema.europa.eu/docs/en_gb/document_library/scientific_guideline/2009/0 9/WC500002928.pdf, 37-page document Confirmatory / Exploratory Trials Study population Primary and secondary endpoints Blinding Randomisation Trial designs Multicentre trials Superiority / Equivalence / Non-inferiority Sample size Sequential designs / interim analyses Pre-specification of analyses Full analysis set (and the intention-to-treat principle) / Per-protocol analysis set Analysis considerations (including estimation, CIs and hypothesis testing) Multiplicity Subgroup analyses Evaluation of safety and tolerability Reporting 4
E9 Introduction Intent for sponsors assists preparation of application summaries assessing evidence of efficacy and safety from later phase trials Scope guidance mainly for confirmatory trials some principles may apply for earlier trials Trial Statistician education/experience to implement these principles ensures that trial protocol covers all relevant statistical issues 5
E9 Introduction Many principles concern bias from trial design from conduct or analysis important to evaluate the robustness of results and conclusions And are mainly Frequentist guidance mainly concerns frequentist methods Bayesian approaches may be considered when reasons are clear and conclusions are sufficiently robust 6
E9 Overall Clinical Development Number and Type of Trials Confirmatory trials are necessary to provide firm evidence of efficacy or safety should address a limited number of questions should address each claim clearly and definitively adequately controlled trial primary objective and hypothesis stated in advance estimate of effect size is as important as hypothesis testing Exploratory trials should have clear and precise objectives choice of hypothesis could be data dependent changes in design could be made in response to accumulating data cannot be the basis of formal proof of efficacy 7
E9 Overall Clinical Development Primary and Secondary Variables Primary variable directly related to the primary objective relevant to the standards of the indication sought reliable and validated measure of clinically relevant treatment benefit used in estimating the sample size needs careful definition pre-specified in the protocol along with rationale Secondary variables either supportive to primary objective or measurements of effects related to secondary objectives 8
E9 Overall Clinical Development Design Techniques to Avoid Bias Blinding double blind is the standard treatments must look, feel and taste the same double-dummy devices can be used no-one must know the treatment assignments if some sponsor staff must know, SOPs should guard against inappropriate dissemination of treatment codes Breaking the blind for an individual only needed if the knowledge of the treatment code will alter further rescue treatment Unblinding a study any intentional or unintentional unblinding must be reported 9
E9 Overall Clinical Development Design Techniques to Avoid Bias Randomisation minimises selection and assignment bias arising from predictability of assignments randomisation schedule should be reproducible and secure advantages to randomising in blocks advisable to stratify by centre using > 2 or 3 stratification factors is rarely necessary stratification factors should be taken account of in analysis emergency access to treatment code must be provided dynamic allocation is possible, but must not be deterministic 10
E9 Trial Design Design configuration Multicentre trials Types of comparison Group sequential designs Sample size Data capture 11
E9 Trial Conduct Interim Analysis and Early Stopping Interim analyses Planned: specified in the protocol Unplanned: protocol amendment before interim analysis Reasons stop for efficacy, futility, or safety Protocol schedule of analyses, stopping guidelines and properties protection of type I error Process completely confidential, possibly with an independent DMC Unplanned interim analyses should be avoided 12
E9 Data Analysis Pre-specification of analysis Analysis sets Missing values and outliers Data transformation Estimation, CIs and testing Adjusting significance and confidence levels Subgroups, interactions and covariates Data integrity and software validity 13
E9 Data Analysis Pre-specification of analysis Analysis sets Missing values and outliers Data transformation Estimation, CIs and testing Adjusting significance and confidence levels Subgroups, interactions and covariates Data integrity and software validity 14
E9 Data Analysis Prespecification of Analysis Principle features described in the protocol proposed confirmatory analysis of primary variable how anticipated analysis problems will be handled Statistical analysis plan written after finalising protocol detailed procedures for analysis of primary, secondary and other data Blinded review SAP reviewed and possibly updated in blinded review should be finalised before breaking the blind (and prove it) only results envisaged in the protocol (+ amendments) can be regarded as confirmatory. 15
E9 Data Analysis Analysis Sets Account for all patients that began trial procedures collect baseline demography and disease status minimise protocol violations, withdrawals and missing data Full Analysis Set (FAS) ITT principle implies primary analysis should be based on all randomised patients FAS as complete as possible and as close as possible to the ITT ideal preserves initial randomisation may be conservative 16
E9 Data Analysis Analysis Sets Per Protocol Set (PPS) subset of the FAS who are more compliant with the protocol exclusions should be documented before breaking the blind maximises ability to show efficacy in the analysis analysis may be severely biased PPS includes patients with a pre-specified minimum exposure to treatment available measurements of the primary variable no major protocol violations 17
E9 Data Analysis Adjusting Significance and Confidence Levels Multiplicity may arise from multiple primary endpoints comparisons time points Avoiding or reducing multiplicity is preferred specifying a primary endpoint specifying a primary comparison use of summary measures over time Adjustment should always be considered or an explanation of why it is not necessary 18
Summary of ICH E9 This is a set of principles for sponsors for regulatory reviewers Many of the principles concern bias Some sections required further clarification Points to Consider Guidance Notes for Guidance Reflections papers More detailed guidance exists for specific diseases 19
Structure of E10 Document http://www.ema.europa.eu/docs/en_gb/document_library/scientific_gu ideline/2009/09/wc500002925.pdf, 30-page document Section Introduction Purpose of Control Group Types of Control Purpose of Trials Assay Sensitivity Detailed Considerations of Types of Control Choosing the Control Group 20
Structure of E10 Document http://www.ema.europa.eu/docs/en_gb/document_library/scientific_gu ideline/2009/09/wc500002925.pdf, 30-page document Section Introduction Purpose of Control Group Types of Control Purpose of Trials Assay Sensitivity Detailed Considerations of Types of Control Choosing the Control Group 21
E10 Introduction The correct choice of the control group is critical It affects inferences that can be made ethical acceptability the degree to which bias can be minimized type of subjects recruited and the recruitment rate kind of endpoints that can be studied credibility of the results acceptability by regulatory authorities 22
E10 Purpose of the Control Group Control groups allows the discrimination of patient outcomes caused by the test treatment from outcomes caused by other factors natural progression of disease observer or patient expectations other treatment A concurrent control group is one chosen from the same population as the test treatment group over the same period of time Ensure that test and control groups are similar at the start of the study and are treated similarly randomization blinding 23
E10 Type of Control Placebo concurrent control No-treatment concurrent control Dose-response concurrent control Active concurrent control External control (including historical control) Multiple control groups 24
E10 Assay Sensitivity The ability to distinguish an effective treatment from a less effective (or ineffective) treatment Non-inferiority or equivalence studies historical evidence appropriate trial conduct i.e. aim to avoid: poor compliance use of concomitant medication that interferes with response a population that tends to improve spontaneously poor disease diagnosis criteria assessment bias Superiority studies with active comparator ineffective treatment or an ineffective trial? three arm trial: include both placebo and an active comparator group 25
E10 Choosing the Control Group Usefulness of Specific Concurrent Control Types in Various Situations Type of Control Active Dose Active Trial Non- Active Response Placebo Placebo Active + Placebo Objective Placebo Inferiority Superiority (DR) + Active + DR + DR + DR Measure Absolute Y N N N Y Y N Y Effect Size Show Existence Y P Y Y Y Y Y Y of Effect Show Dose Response N N N Y N Y Y Y Relationship Compare Therapies N P Y N Y N P Y Y= Yes, N=No, P=Possible - depending on historical evidence of sensitivity to drug effects. 26
E10 Choosing the Control Group Choosing the Concurrent Control for Demonstrating Efficacy Is there a proven effective treatment? Yes Is the proven effective treatment life-saving or known to prevent irreversible morbidity? No Is there historical evidence of sensitivity to drug effects for an appropriately designed and conducted trial? No Yes No Yes Options Placebo control Dose response control Active control (superiority) No-treatment control Any combination of these Options Active control (superiority) Active control (non-inferiority) 1 Placebo control Dose response control Options Placebo control Dose response control Active control (superiority) No treatment control Active and placebo control Options Placebo control Dose response control Active control (superiority) Active control (non-inferiority) Active and placebo control 27
Workshop 28
Workshop format Divide into teams Each team to select a captain Each team to devise their Team Name Team game This will involve testing your knowledge of the statistical principles described in Award ceremony Prize for winning team 29