FDA Industry Workshop Statistics in the FDA Industry

  • Slides: 32
Download presentation
FDA Industry Workshop Statistics in the FDA & Industry The Future David L De.

FDA Industry Workshop Statistics in the FDA & Industry The Future David L De. Mets, Ph. D Department of Biostatistics & Medical Informatics University of Wisconsin School of Medicine & Public Health 1

Topics • • • Training/Certification Needs Academic/Industry Collaborations Attack on Clinical Trials & Statistics

Topics • • • Training/Certification Needs Academic/Industry Collaborations Attack on Clinical Trials & Statistics CT Costs & Data Management Statistical Methodology Issues 2

Globalization of Clinical Trials • Rate of discovery increasing • Translational into practice is

Globalization of Clinical Trials • Rate of discovery increasing • Translational into practice is not fully realized – Screening – Prevention – Treatment • Declining Recruitment in US • More trials becoming multinational 3

NIH Roadmap: Discipline of Clinical Research Clinical Trialist Clinician Common Core Statistician Knowledge Clinical

NIH Roadmap: Discipline of Clinical Research Clinical Trialist Clinician Common Core Statistician Knowledge Clinical Pharm Behavioral Scientist 4

Clinical Research Training: a multidisciplinary workforce • In USA, number of clinical researchers is

Clinical Research Training: a multidisciplinary workforce • In USA, number of clinical researchers is not increasing • Previous training “on the job”, sort of “trial and error” approach • Rigorous training programs in USA are just starting – NIH Roadmap Initiative • Many disciplines now involved in clinical research without formal training in this science • Threat of the “silver tsunami” – 40% of Clinical Researchers in USA over age 50 • World wide training challenges 5

Training Pyramid in Patient-Oriented Research Ph. D MS Degree Certificate Degree Workshops 6

Training Pyramid in Patient-Oriented Research Ph. D MS Degree Certificate Degree Workshops 6

Biostatistician Crises • Increasing demand for statistician/biostatisticians in academia, industry & government • Supply

Biostatistician Crises • Increasing demand for statistician/biostatisticians in academia, industry & government • Supply of MS and especially Ph. D trained biostatisticians relatively constant over past two decades • Domestic students in biostatistics in very short supply • Crises not fully appreciated 7

Academic – Industry CT Partnerships • • Industry CT funding levels similar to NIH

Academic – Industry CT Partnerships • • Industry CT funding levels similar to NIH Need to continue developing relationships Can be a win-win for all Phases I, II & III Four key elements – – Independent Steering Committee Independent Statistical Center Independent Data Monitoring Committee Freedom to publish • Journals beginning to require investigator independence 8

A Clinical Trial Model Steering Committee Sponsor Regulatory Agencies Independent Data Monitoring Committee (IDMC)

A Clinical Trial Model Steering Committee Sponsor Regulatory Agencies Independent Data Monitoring Committee (IDMC) Statistical Analysis Center (SAC) Data Management Center (DMC) Clinical Centers Patients Central Units (Labs, …) Institutional Review Board 9

Challenge: Attack on Clinical Trials & Statistics • Pending Congressional Legislation • Wall Street

Challenge: Attack on Clinical Trials & Statistics • Pending Congressional Legislation • Wall Street & WSJ • Some Patient Advocacy Groups 10

Senate Bill 1956 • A proposed amendment to Federal Food, Drug & Cosmetic Act

Senate Bill 1956 • A proposed amendment to Federal Food, Drug & Cosmetic Act • Known as the ACCESS Ammendment • A three tiered approval system • More responsive to “the needs of seriously ill patients” 11

Proposed Three Tier Approval • Tier I – Based on Phase I information –

Proposed Three Tier Approval • Tier I – Based on Phase I information – Based on clinical, not statistical analysis – May require post approval studies • Tier II – Based on surrogates or biomarkers • Tier III – Traditional requirements 12

Some Issues in Proposed Legislation • Challenge of placebo controlled studies • De-emphasize statistical

Some Issues in Proposed Legislation • Challenge of placebo controlled studies • De-emphasize statistical analysis-no disapprovals solely on the basis of statistical analysis or 95% CIs • Evidence may be based on uncontrolled studies such as case histories, observational studies, mechanism of actions, computer models… • Outcome data may be a surrogate or biological marker 13

CT Statistical Methodology Issues • Surrogate Outcomes • Composite Outcomes • Non-inferiority Designs •

CT Statistical Methodology Issues • Surrogate Outcomes • Composite Outcomes • Non-inferiority Designs • Adaptive Designs • Gene Transfer Designs • Safety Monitoring 14

Surrogate Response Variables • Used as a substitute for Clinical Endpoint • May lead

Surrogate Response Variables • Used as a substitute for Clinical Endpoint • May lead to smaller or shorter studies • Requirements (Prentice, 1989) T = True clinical endpoint S = Surrogate Z = Treatment • Sufficient Conditions 1. 2. S is informative about T (predictive) S fully captures effect of Z on T • Concern: – Correlation is not Causation – Pathways often more complex – Other side effects not seen 15

Failures of Potential Surrogates • Nocturnal Oxygen Therapy Trial (NOTT) – 24 vs 12

Failures of Potential Surrogates • Nocturnal Oxygen Therapy Trial (NOTT) – 24 vs 12 hour oxygen in COPD patients – Pulmonary Function tests (NS) – Survival (p<0. 001) • CAST – Patients with cardiac arrhythmias – Arrhythmias suppressed – Terminated with increased mortality • Ref (Fleming & De. Mets, Annals Intern Med, 1996) 16

Failures of Potential Surrogates • Inotropic Drugs in Heart Failure – Improved heart function

Failures of Potential Surrogates • Inotropic Drugs in Heart Failure – Improved heart function but increased mortality – PROMISE, PROFILE, VEST, …. • Lipid lowering but no survival benefit – Women’s Health Initiative & HRT – Increased risk of clotting (PE, DVTs) • Ref (Fleming & De. Mets, Annals Intern Med, 1996) 17

Composite Endpoint Rationale • Defined as having occurred if any one of several components

Composite Endpoint Rationale • Defined as having occurred if any one of several components is observed – e. g. death, MI, stroke, change in severity, …. . • May reduce Sample Size by increasing event rates – Assumes each component sensitive to intervention – Otherwise, power can be lost • May avoid competing risk problem – Death is a competing risk to all other morbid events, probably not independent 18

Problems with Composite Outcomes • Interpretability if individual components go in different directions –

Problems with Composite Outcomes • Interpretability if individual components go in different directions – e. g. WHI global index– • Death: similar • Fractures: positive • DVTs, PEs: negative • Relevance of a mixed set of components – Trials are adding softer outcomes • Could have a loss of power if some components not responsive • Failure to ascertain components 19

Non-Inferiority Designs • Design to compare a new intervention with an accepted/proven standard –

Non-Inferiority Designs • Design to compare a new intervention with an accepted/proven standard – “As good as” with respect to a primary – Has some other advantage (cost, less toxic, less invasive, …. . ) • Must define a degree of non-inferiority or indifference, δ – Choice is somewhat arbitrary – Absolute or relative scale 20

Difference in Events Test – Standard Drug (Antman et al) 21

Difference in Events Test – Standard Drug (Antman et al) 21

Non-Inferiority Methodology a) Comparison: New Treatment vs. Standard: RRa Upper CI must be less

Non-Inferiority Methodology a) Comparison: New Treatment vs. Standard: RRa Upper CI must be less than δ b) Estimate of standard vs. placebo: RRb Based on literature c) Imputed effect of New Trt vs. placebo (RRc) RRc = RRa x RRb 22

Challenges for Non-Inferiority Designs • Current paradigm makes all noninferiority trials vulnerable • Relevance

Challenges for Non-Inferiority Designs • Current paradigm makes all noninferiority trials vulnerable • Relevance of standard vs placebo historical estimate • Fraction of standard benefit to be retained • Choice of δ for current trial 23

Adaptive Designs • Many Adaptive Designs in Use – Baseline Driven (based on risk

Adaptive Designs • Many Adaptive Designs in Use – Baseline Driven (based on risk profile) – Total Event Driven Designs – Group Sequential Designs • Benefit or Harm • Futility – Drop the Losing Arm • Statistical & Logistical issues worked out for these • Not a Frequentist vs Bayesian Issue 24

Adaptive Designs • Adjusting design during trial – Sample size – Primary outcome •

Adaptive Designs • Adjusting design during trial – Sample size – Primary outcome • • • Current interest very high A need exists to be adaptive or flexible Some statistical methods developed Still many statistical debates Many remaining issues related to logistics & potential for introducing bias 25

Monitoring of Clinical Trials • Shalala – Death of gene transfer patient – NEJM

Monitoring of Clinical Trials • Shalala – Death of gene transfer patient – NEJM (2000) – Press Release (2000) • IRBs often not provided sufficient information to evaluate clinical trials fully • NIH will require monitoring plans for Phase I, II and III trials - guidelines • FDA issued guidelines for Data & Safety Monitoring Boards and IRBs (2001, 2005) • Post Cox II issues – Rapid access vs long term safety 26

IRB Safety Monitoring Problem • IRBs review trial design and ethics • IRBs responsible

IRB Safety Monitoring Problem • IRBs review trial design and ethics • IRBs responsible for patient safety • Drowning in SAE reports, not useful • Inadequate infrastructure to be able to provide adequate safety monitoring • For some multicenter trials, an alternative process exists (i. e. DMC) • For single center trials, patient “safety” monitoring provided is now inadequate 27

Safety & Observational Data • Long term RCT follow-up for low rate SAEs not

Safety & Observational Data • Long term RCT follow-up for low rate SAEs not common • Have turned to observational data as a supplement • Serious limitations to argue causality due to confounding and bias • Statistical analysis can take us only so far • Need to understand better what can be learned 28

Reducing Trial Costs • DCRI Workshop: Hypothetical Trial Example – 60 -70% of cost

Reducing Trial Costs • DCRI Workshop: Hypothetical Trial Example – 60 -70% of cost site related, half due to site monitoring – Could reduce costs 40% by reducing CRFs & monitoring site visits • DCRI CT example: Ongoing site monitoring improved regulatory compliance but little on trial data results & conclusions • Breast Cancer Fraud Case – Academic network; Intense audit did not alter the results (<1% error), NEJM 1995 29

Need for Change in Site Monitoring • • Current system is “out of control”

Need for Change in Site Monitoring • • Current system is “out of control” Educate/train clinical sites & investigators Focus data collected & limit the extraneous Set priorities on monitoring key variables: – eligibility – primary and secondary outcomes, – serious adverse events (SAE) • Sample audit the rest • Use more statistical QC methods • Standardize CRFs and data management 30

Challenge: Gene Transfer Trials • NIH Re-Combinant Advisory Committee (RAC) • RAC reviews new

Challenge: Gene Transfer Trials • NIH Re-Combinant Advisory Committee (RAC) • RAC reviews new gene transfer trials • Mostly very early phase studies • Designs often not appropriate – No objectives clearly stated – Borrowed from other settings that are not relevant • Design guidelines need further development 31

Summary • With current discovery rate, future appears very promising • Significant challenges exist

Summary • With current discovery rate, future appears very promising • Significant challenges exist • Most are solvable but will require collaboration from academia, regulators & sponsors • Failure is not an option – we need evidence based medicine • Every challenge is an opportunity 32