Comments on “Adaptation and Heterogeneity” by Armin Koch Paul Gallo, Willi Maurer PhRMA Adaptive Design KOL Lecture Series November 14, 2008.

Slides:



Advertisements
Similar presentations
Traps and pitfalls in medical statistics Arvid Sjölander.
Advertisements

The Scientific Method.
Testing Hypotheses About Proportions Chapter 20. Hypotheses Hypotheses are working models that we adopt temporarily. Our starting hypothesis is called.
1. Review- What is Science Explain- What kinds of understandings does science contribute about the natural world Form an Opinion- Do you think that scientists.
Critical Thinking: Chapter 10
What is MPC? Hypothesis testing.
Chapter 11 Contingency Table Analysis. Nonparametric Systems Another method of examining the relationship between independent (X) and dependant (Y) variables.
New or existing slides are easily formatted using built-in layouts that can be applied via the Home tab EMA DRAFT GUIDELINE ON SUBGROUPS DISCUSSION April.
Estimation and Reporting of Heterogeneity of Treatment Effects in Observational Comparative Effectiveness Research Prepared for: Agency for Healthcare.
Elements of a clinical trial research protocol
Elementary hypothesis testing
Evaluating Hypotheses Chapter 9. Descriptive vs. Inferential Statistics n Descriptive l quantitative descriptions of characteristics.
Cal State Northridge  320 Ainsworth Sampling Distributions and Hypothesis Testing.
Elementary hypothesis testing Purpose of hypothesis testing Type of hypotheses Type of errors Critical regions Significant levels Hypothesis vs intervals.
1 Practicals, Methodology & Statistics II Laura McAvinue School of Psychology Trinity College Dublin.
Sabine Mendes Lima Moura Issues in Research Methodology PUC – November 2014.
Adaptive Designs for Clinical Trials
T HE NATURE OF QUALITATIVE RESEARCH Gordana Velickovska Guest Professor Centre for Social Sciences.
Copyright © 2005 Brooks/Cole, a division of Thomson Learning, Inc Chapter 11 Introduction to Hypothesis Testing.
Inferential Statistics
Three Common Misinterpretations of Significance Tests and p-values 1. The p-value indicates the probability that the results are due to sampling error.
The problem of sampling error in psychological research We previously noted that sampling error is problematic in psychological research because differences.
Writing the Research Paper
Study Designs By Az and Omar.
Making all research results publically available: the cry of systematic reviewers.
Are the results valid? Was the validity of the included studies appraised?
POSC 202A: Lecture 1 Introductions Syllabus R Homework #1: Get R installed on your laptop; read chapters 1-2 in Daalgard, 1 in Zuur, See syllabus for Moore.
Inference in practice BPS chapter 16 © 2006 W.H. Freeman and Company.
How to Write a Literature Review
Epidemiology The Basics Only… Adapted with permission from a class presentation developed by Dr. Charles Lynch – University of Iowa, Iowa City.
Introduction to Science: The Scientific Method
Testing Theories: Three Reasons Why Data Might not Match the Theory Psych 437.
What’s in the news right now related to science???? Flesh eating bacteria.
Chapter 8 Introduction to Hypothesis Testing
User Study Evaluation Human-Computer Interaction.
Introduction to inference Use and abuse of tests; power and decision IPS chapters 6.3 and 6.4 © 2006 W.H. Freeman and Company.
Hypothesis Testing Introduction to Statistics Chapter 8 Mar 2-4, 2010 Classes #13-14.
Institute of Professional Studies School of Research and Graduate Studies Introduction to Business and Management Research Lecture One (1)
Hypothesis Testing Hypothesis Testing Topic 11. Hypothesis Testing Another way of looking at statistical inference in which we want to ask a question.
Chapter 20 Testing hypotheses about proportions
URBDP 591 I Lecture 3: Research Process Objectives What are the major steps in the research process? What is an operational definition of variables? What.
No criminal on the run The concept of test of significance FETP India.
Copyright © 2008 Pearson Education, Inc. Publishing as Pearson Addison-Wesley Chapter 20 Testing Hypotheses About Proportions.
What is a non-inferiority trial, and what particular challenges do such trials present? Andrew Nunn MRC Clinical Trials Unit 20th February 2012.
Study Designs for Clinical and Epidemiological Research Carla J. Alvarado, MS, CIC University of Wisconsin-Madison (608)
Issues concerning the interpretation of statistical significance tests.
META-ANALYSIS, RESEARCH SYNTHESES AND SYSTEMATIC REVIEWS © LOUIS COHEN, LAWRENCE MANION & KEITH MORRISON.
DESIGNING AN EXPERIMENT.  Scientific Inquiry – the process of gathering evidence about the natural world and giving explanations based on evidence. DESIGNING.
Health and Disease in Populations 2002 Sources of variation (1) Paul Burton! Jane Hutton.
Considerations for Topical Microbicide Phase 3 Trial Designs, an Investigator’s Perspective Andrew Nunn Medical Research Council Clinical Trials Unit London,
APPLIED DATA ANALYSIS IN CRIMINAL JUSTICE CJ 525 MONMOUTH UNIVERSITY Juan P. Rodriguez.
Hypothesis Testing Introduction to Statistics Chapter 8 Feb 24-26, 2009 Classes #12-13.
A significance test or hypothesis test is a procedure for comparing our data with a hypothesis whose truth we want to assess. The hypothesis is usually.
European Patients’ Academy on Therapeutic Innovation Ethical and practical challenges of organising clinical trials in small populations.
Copyright © 2007 Pearson Education, Inc. Publishing as Pearson Addison-Wesley Slide
Uncertainty and confidence Although the sample mean,, is a unique number for any particular sample, if you pick a different sample you will probably get.
Adaptive trial designs in HIV vaccine clinical trials Morenike Ukpong Obafemi Awolowo University Ile-Ife, Nigeria.
Statistics 20 Testing Hypothesis and Proportions.
Critical Appraisal Course for Emergency Medicine Trainees Module 2 Statistics.
June 29, 2017 Suzanne Hendrix, PhD Pentara Corp
Introduction to inference Use and abuse of tests; power and decision
Testing Hypotheses About Proportions
Supplementary Table 1. PRISMA checklist
Crucial Statistical Caveats for Percutaneous Valve Trials
Managed Access to NIHR-funded Research Data
Significance and t testing
Testing Hypotheses About Proportions
Section 2.2: Sampling.
Paul Gallo 2019 Joint Statistical Meetings July 31, 2019
STAT 515 Statistical Methods I Chapter 1 Data Types and Data Collection Brian Habing Department of Statistics University of South Carolina Redistribution.
Presentation transcript:

Comments on “Adaptation and Heterogeneity” by Armin Koch Paul Gallo, Willi Maurer PhRMA Adaptive Design KOL Lecture Series November 14, 2008

2 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Bottom line  We agree with many of Armin’s bottom-line statements.  Differences in opinion or perspective largely boil down to questions of degree. oe.g., how heavy is the burden of proof ?

3 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Motivation  Not simply a matter of obtaining lower standards for sponsor advantage.  Statistical issues: subsets / subgroups always present challenges, and there can be a tendency in many contexts to over-react to signals.  No party wants to falsely invalidate or substantially undervalue meaningful results.  How do we go about this most rationally?

4 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Scope  Our scope: changes which do not strongly structurally change the nature of the data. oe.g., SS re-estimation, dropping an arm (as in seamless II / III), changing randomization allocation, etc.  More substantial structural changes, e.g., involving the nature of endpoints, would likely raise much more challenging issues with regard to any possibility of combining information to test a common hypothesis.

5 PhRMA Adaptive KOL Lecture Series / November 14, 2008 “One trial or two?” A big question:  Where is within-trial heterogeneity important to investigate / understand? oAll trials? (but we don’t, and there don’t seem to be standards) oGroup sequential trials? (information leakage more easily envisioned, and evidence more accessible, but seems not routinely explored) oAdaptive trials?

6 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Meta-analyses (slides 9-10)  Is there an implication that meta-analyses are in some sense less important than confirmatory trials?  Meta-analyses can affect medical practice.  Not all are the same, but we would start from a view that both can be important within their contexts, and the question is one of most accurately interpreting the results.

7 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Meta-analyses (continued)  We may know that trials within a meta- analysis differ in many identifiable aspects: oinvestigators following different protocols and procedures, monitoring standards, calendar times, information from other trials out in the open, “therapeutic drift”, etc., etc.

8 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Meta-analyses (continued)  Adaptive trials without major structural changes may have none of these problems.  When interpreting whether or not non- definitive but potentially important signals of interaction might be real, shouldn’t there be an important role for plausibility / rationale ?

9 PhRMA Adaptive KOL Lecture Series / November 14, 2008 “Natural reasons” (slide 11)  Because other factors can lead to some heterogeneity, “can we ignore the problem?”  No, but this highlights a difficulty and points toward caution in interpretation.  The concern: results might potentially be invalidated by a signal (real or not) unrelated to the adaptation, which would not have raised a concern had it arisen in a conventional trial.

10 PhRMA Adaptive KOL Lecture Series / November 14, 2008 “Natural reasons”  Underlying minor heterogeneity can increase the chance of a formal test flagging a signal.  e.g., a 0.15 level interaction test may have a higher effective false positive (interaction) rate, because its homogeneity null hypothesis is already not true.

11 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Relationship to effect size (slide 12)  This seems not unlike how some important subgroup issues might be viewed.  A situation with a signal of heterogeneity, but with each subgroup showing benefit, would often be judged differently from one in which there was a subgroup that did not appear to benefit. oSome relationship to quantitative / qualitative issues.

12 PhRMA Adaptive KOL Lecture Series / November 14, 2008 “Information leakage” (slide 14)  We “can’t prove that (good procedures) have been followed”  True, but we wouldn’t say that this alone would be sufficient; it’s just one part of the case.  Evidence of confidentiality / compliance, numerical strength of the signal, scientific plausibility or explanation, etc., are all part of the story.

13 PhRMA Adaptive KOL Lecture Series / November 14, 2008 “Price to be paid” (slide 16)  Generally there are statistical prices to paid, but these are usually easily handled.  “Further investigation” of signals because of the added complexity in an adaptive design, as a price, seems fine also.  But the goal should be to most accurately draw conclusions from the data at hand; to falsely discount meaningful results at some non-trivial frequency may not be a good price to pay.

14 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Practical implementation issues  Don’t lose sight of the challenges inherent in implementing an investigation.  Defining stages is not always as clear as it might sound initially. oe.g., the DMC dataset is not the same as the point of implementation.

15 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Practical implementation (continued)  What’s the relevant unit for definition of stages? opatients (when they are randomized) or events (when they occur)?  Too much unstructured exploration with regard to these types of questions may increase the chance of false signals.  Time-to-event analyses may present special challenges because of confounding with non-proportional hazards.

16 PhRMA Adaptive KOL Lecture Series / November 14, 2008 Pre-specification  Not only with regard to methodology...  What am I concerned about in advance, i.e., what’s the mechanism by which a change might be induced, and is it consistent with what I later see in the data?  Shouldn’t a signal be taken more seriously if consistent with sound pre-stated concerns?  Of course we do have to leave room for exploration of unanticipated signals.