Presentation on theme: "The Statistical Consensus of Non-Inferiority Clinical Trail for NDA in CHINA XIA J.L., PhD. Prof of Biostatistics of FMMU On behalf of CCTS Working Group."— Presentation transcript:
The Statistical Consensus of Non-Inferiority Clinical Trail for NDA in CHINA XIA J.L., PhD. Prof of Biostatistics of FMMU On behalf of CCTS Working Group --
Outline Introduction of CCTS Introduction of the Consensus BACKGROUND GENERAL CONSIDERATION OF NON-INFERIORITY STUDIES CHOOSING THE NON-INFERIORITY MARGIN ANALYZING THE RESULTS OF AN NI TRIAL 4th DIA China Annual Meeting: Collaboration and Innovation in China May 20-23, 2012 | Shanghai, China
Introduction of CCTS China Clinical Trial Statistics working group CCTS is a branch of Biostatistical Theory and Method Committee of China. Most of the members of CCTS are External/Internal Statistical Reviewer in CDE of SFDA. CCTS is a open academic group.
Introduction of CCTS CCTS Academically serves statistical design and analysis for clinical trial for NDA in China Bridge ICH related Guidelines into China Reach Some Statistical Consensus for drug evaluation Put impact on Clinical trial in China
Background NI Trial was introduced to China almost 10yrs ago. Expression of the result of statistical test was totally changed. Absence of evidence is not the evidence of Absence. The Concept of NI was metaphysically adopted and abused in most of trial design.
Background The Consensus of NI trial was reached last OCT. in Tangshan,Nanking Published on The Journal of Chinese Health Statistics. Main references are FDAs Guidance for Industry Non-Inferiority Clinical Trials,ICH E9,E10 and some related topics of EMEAsGUIDELINE ON THE CHOICE OF THE NON- INFERIORITY MARGIN.
NI in FDA 43 of 175 NDAs for new molecular entities that were submitted to FDA from 2002 to 2009 on the basis of evidence from Non- inferiority trials –18 of 43 were approved by FDA – 9 of 43 were poorly designed –16 of 43 trials were failure
Highlight Clarify the Concept of NI trial NI trials may demonstrate that the new drug has some effect that is not too much smaller than the active control. NI, in Chinese, literally is not bad than the active control. Algebraically NI means but.
Highlight Declare the applicability and the limitation of NI trial Principle of Choosing Active control, determining Margin, calculating Sample size and statistical inference etc. NI trial objectives: one is evaluating the effect of new drug (Superior than putative placebo) and another is that the effect is not too much smaller than the active control.
Applicability of NI trial not ethical to use placebo or naive control Only one Primary endpoint historical evidence of sensitivity to drug effects (HESDE) The consistence effectiveness should be assumed (Consistency Assumption AC study with good quality (GQS )
Limitation of NI Trial Assay sensitivity not determined Most of symptom relief medicine not suitable for NI trial (such as some TCM) Smaller effectiveness lead to unpractical large sample size Lack of well designed historical trial of active control Effectiveness of Active Con. Changed, such as some antibiotics
NI Margin M 1 = the entire effect of the active control assumed to be present in the NI study, which was estimated by C-P considering variability M 2 = the largest clinically acceptable difference, M 2 =(1-f) M 1, 0.5f 0.8 Pre-specified < M 1 for superior than Putative Placebo Pre-specified < M 2 for Non-inferiority than active control
Positive In order to stat the margin be positive, endpoints were classified into HB endpoints and LB endpoints HB=the higher the better such as cure rate ) LB=the lower the better such as motality rate or morbidity rate )
Positive Active Control?! For HB, C-P or ln(C/P) is positive < M 1 : < M 2 : For LB, P-C or ln(P/C) is positive < M 1 : < M 2 :
Hypotheses Tab.1 Hypotheses of NI Trial ScenariosDifference (proportion, mean) Ratio RR,HR,OR HB (Higher, better) LB (lower, better) TEST LEVEL
mean Difference When calculating Sample size, commonly, C=T was assumed. In some fewer cases, T may be assumed a little better than C (according to HESDE), but nominal β may be inflated. Sample size
Cause variability puts heavy weight to sample size estimation. One should be conservatively estimate variability at initial and can do sample size reestimation under blind at some pre-specified time. This do no harm to type I error inflation. Sample size
Unblind sample size reestimation, if needed, should be done under the supervision of DMC and must be carefully pre-planed in the protocol, especially, the αspending function should be depicted in detail. Sample size
Generally, Sample size of Ni trial is larger than that of superiority trial. Philosophically, it is reasonable and is the trade off between ethics and science. Sample size of mean difference test
Missing value imputation Missing value is inevitable in clinical trial. The imputation method should be carefully applied. LOCF do not take for granted. If do so, Ni trial will be beneficiate for poor quality trial. MMRM model may be a candidate imputation method. IIT principle is not conservative for NI trial
Inferences Interval estimation of C-T for HB or T-C for LB was recommended, generally 95%CI If the upper limit <,trial objective can be concluded 95%-95% principle, though conservative, was proposed in the consensus
M1M1 NI HESDE NI HESDE M1M1 M2M2 1 95%CI of C/P 95%CI of C/T 1.12 1.25 M1M1 M2M2 1 95%CI of P/C 95%CI of T/C 1.12 1.25 NI HESDE NI HESDE M1M1 M2M2 M1M1 0 95%CI of C-P 95%CI of C-T 0.11 0.22 M2M2 0 95%CI of P-C 95%CI of T-C 0.11 0.22 Fig.1 Statistical inference principle of 95%-95% method LB HB
Switching between NI and Superiority If the upper limit for respective 95%CI be less than 0 in NI trial, Superiority can be claimed without adjustment of type one error If P>0.05 in superiority trial, even if the upper limit for respective 95%CI be less than (post hoc), NI should not be claimed.
epilogue Two arms NI trial should be considered as the last choice. Do pay attention to biocreep. P=T n <…T 2 <T 1 <C Three arms design would be the best choice, especially a good way for R&D of TCM