Effect Size – Can the Effect Be Too Small Robert J. Temple, M.D. Advisory Committee Mtg April 25, 2006.

Slides:



Advertisements
Similar presentations
Equivalence Testing Dig it!.
Advertisements

Post Research Benefits Mandika Wijeyaratne MS, MD, FRCS Dept. of Surgery, Colombo.
Informed Choice. Overview Brief introduction to cases (ours, yours) Elements of informed choice Capacity Informed choice for research Return to cases.
Evidence Based Advertising “Don’t accept your dog’s admiration as conclusive evidence that you are wonderful” -Ann Landers.
Medication Guides Nancy M. Ostrove, Ph.D. Division of Drug Marketing, Advertising, and Communications.
Statistical Issues in Research Planning and Evaluation
Optimal Drug Development Programs and Efficient Licensing and Reimbursement Regimens Neil Hawkins Karl Claxton CENTRE FOR HEALTH ECONOMICS.
EPIDEMIOLOGY AND BIOSTATISTICS DEPT Esimating Population Value with Hypothesis Testing.
The ICH E5 Question and Answer Document Status and Content Robert T. O’Neill, Ph.D. Director, Office of Biostatistics, CDER, FDA Presented at the 4th Kitasato-Harvard.
STANDARD Anything serving as a type or pattern to which other like things must conform (Stedman’s Medical Dictionary) STANDARD OF CARE The level at which.
Lecture 2: Thu, Jan 16 Hypothesis Testing – Introduction (Ch 11)
Section 7-2 Hypothesis Testing for the Mean (n  30)
Determining Statistical Significance
Sample Size Determination Ziad Taib March 7, 2014.
1 The Chemoprevention of Sporadic Colorectal Cancer Issues Surrounding a Benefit/Risk Analysis in Clinical Trials Mark Avigan MD CM Medical Officer Division.
Copyright © 2005 Brooks/Cole, a division of Thomson Learning, Inc Chapter 9 Introduction to Hypothesis Testing.
Section 9.1 Introduction to Statistical Tests 9.1 / 1 Hypothesis testing is used to make decisions concerning the value of a parameter.
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University ECON 4550 Econometrics Memorial University of Newfoundland.
Inference in practice BPS chapter 16 © 2006 W.H. Freeman and Company.
Significance Tests: THE BASICS Could it happen by chance alone?
LECTURE 19 THURSDAY, 14 April STA 291 Spring
Investigational Drugs in the hospital. + What is Investigational Drug? Investigational or experimental drugs are new drugs that have not yet been approved.
CLAIMS STRUCTURE FOR SLE Jeffrey Siegel, M.D. Arthritis Advisory Committee September 29, 2003.
Introduction to inference Use and abuse of tests; power and decision IPS chapters 6.3 and 6.4 © 2006 W.H. Freeman and Company.
Challenges of Non-Inferiority Trial Designs R. Sridhara, Ph.D.
Hypothesis Testing Introduction to Statistics Chapter 8 Mar 2-4, 2010 Classes #13-14.
Step 3 of the Data Analysis Plan Confirm what the data reveal: Inferential statistics All this information is in Chapters 11 & 12 of text.
Maximum Likelihood Estimator of Proportion Let {s 1,s 2,…,s n } be a set of independent outcomes from a Bernoulli experiment with unknown probability.
Joint Meeting of Anti-Infective Drugs & Drug Safety and Risk Management Advisory Committees December 14-15, 2006 Ketek  (telithromycin) Regulatory History.
Humanitarian Use Devices September 23, 2011 Theodore Stevens, MS, RAC Office of Cellular, Tissue and Gene Therapies Center for Biologics Evaluation and.
Decision-Theoretic Views on Switching Between Superiority and Non-Inferiority Testing. Peter Westfall Director, Center for Advanced Analytics and Business.
What is a non-inferiority trial, and what particular challenges do such trials present? Andrew Nunn MRC Clinical Trials Unit 20th February 2012.
Lecture 16 Section 8.1 Objectives: Testing Statistical Hypotheses − Stating hypotheses statements − Type I and II errors − Conducting a hypothesis test.
Chapter 8 Delving Into The Use of Inference 8.1 Estimating with Confidence 8.2 Use and Abuse of Tests.
Lecture 18 Dustin Lueker.  A way of statistically testing a hypothesis by comparing the data to values predicted by the hypothesis ◦ Data that fall far.
Lecture 17 Dustin Lueker.  A way of statistically testing a hypothesis by comparing the data to values predicted by the hypothesis ◦ Data that fall far.
Hypothesis Testing An understanding of the method of hypothesis testing is essential for understanding how both the natural and social sciences advance.
1 Study Design Issues and Considerations in HUS Trials Yan Wang, Ph.D. Statistical Reviewer Division of Biometrics IV OB/OTS/CDER/FDA April 12, 2007.
Essential Questions How do we use simulations and hypothesis testing to compare treatments from a randomized experiment?
AP Statistics Unit 5 Addie Lunn, Taylor Lyon, Caroline Resetar.
Goals… Define what is meant by a Type I error. Define what is meant by a Type II error. Define what is meant by the power of a test. Identify the relationship.
STA Lecture 221 !! DRAFT !! STA 291 Lecture 22 Chapter 11 Testing Hypothesis – Concepts of Hypothesis Testing.
Hypothesis Testing Introduction to Statistics Chapter 8 Feb 24-26, 2009 Classes #12-13.
+ The Practice of Statistics, 4 th edition – For AP* STARNES, YATES, MOORE Unit 5: Hypothesis Testing.
Placebos 1.Ethical Issues 2.Problems with active control non- inferiority studies 3.Design modifications that may make placebo controls more acceptable.
Section 9.1 First Day The idea of a significance test What is a p-value?
Hypothesis Tests for 1-Proportion Presentation 9.
© 2010 Pearson Prentice Hall. All rights reserved Chapter Hypothesis Tests Regarding a Parameter 10.
Remaining Challenges in Assessing Non-Inferiority Steven Snapinn DIA Statistics Community Virtual Journal Club December 16, 2014 Based on Paper with Qi.
The Fine Art of Knowing How Wrong You Might Be. To Err Is Human Humans make an infinitude of mistakes. I figure safety in numbers makes it a little more.
Copyright © 2010, 2007, 2004 Pearson Education, Inc. Chapter 21 More About Tests and Intervals.
Statistics in Drug Regulation: The Next 10 Years Thomas Permutt Director, Division of Biometrics II Center for Drug Evaluation and Research The views expressed.
+ Homework 9.1:1-8, 21 & 22 Reading Guide 9.2 Section 9.1 Significance Tests: The Basics.
Sample Size Considerations
Logic of Hypothesis Testing
Biostatistics Case Studies 2007
Introduction to inference Use and abuse of tests; power and decision
Unit 5: Hypothesis Testing
Reasonable Assurance of Safety and Effectiveness: An FDA Division of Cardiovascular Devices Perspective Bram Zuckerman, MD, FACC Director, FDA Division.
Donald E. Cutlip, MD Beth Israel Deaconess Medical Center
CHAPTER 9 Testing a Claim
Medical Device Regulatory Essentials: An FDA Division of Cardiovascular Devices Perspective Bram Zuckerman, MD, FACC Director, FDA Division of Cardiovascular.
Critical Reading of Clinical Study Results
Significance Tests: The Basics
Significance Tests: The Basics
Statistical significance using p-value
CHAPTER 9 Testing a Claim
CHAPTER 9 Testing a Claim
Interpreting Epidemiologic Results.
Chapter 9 Hypothesis Testing.
Presentation transcript:

Effect Size – Can the Effect Be Too Small Robert J. Temple, M.D. Advisory Committee Mtg April 25, 2006

2 Effect Size –Can the Effect be Too Small Legal Standards FD & C Act Legislative history Court cases Reinvention statement PRO document Data presentation: mean vs distribution

3 Legal Standard Section 505(d) says that an application can be refused if there is a lack of substantial evidence that it will have the effect it is represented in labeling to have. This implies that a truthful description of any effect would be a basis for approval. But there are two bases for thinking that is not quite so: 1. Safety requirement 2. Warner-Lambert v Heckler

4 Why Might Effect Size Matter? 1. Safety FD & C Act says application can be rejected if tests of drug show it is unsafe or fail to show that it is safe. Since all drugs have adverse effects, we often say that safe can only mean that benefits outweigh risks. This implies that effect size could matter, certainly for a toxic drug. Could it also mean that an effect size too puny could be outweighed by the unknown risk any drug has? Rebuttal: 1938 law didn’t even ask for evidence of benefit so maybe “safe” did not mean B>R, whatever we now think, but only meant nothing too bad was seen.

5 Why Might Effect Size Matter? 2. Warner-Lambert v Heckler (1986) Basically said that not any showing of a statistically significant effect satisfies the Act. The effect shown must be clinically meaningful and not “therapeutically trivial,” rejecting the argument that any effect claimed, if supported statistically, was sufficient and that the size of the effect is irrelevant.

6 On the Other Hand Legislative history is clear in saying there is no relative effectiveness requirement. A new drug need not be better than, or even as good as, available therapy. “The committee believes that this provision strikes a balance between the need for governmental control to assume that new drugs are not placed on the market until they have passed the relevant tests and the need to assure that governmental control does not become so rigid that the flow of new drugs to the market, and the incentive to undergo the expense involved in preparing them for the market, become stifled.” That does not say, however, that any effect, no matter how small, is sufficient.

7 Reinvention Statement FR August 1, 1995 Apparently to reassure the drug development community that FDA was not imposing new comparative standards, FDA announced that “FDA weighs a product’s demonstrated effectiveness against its risks and considers other factors, such as the seriousness and outcome of the disease being treated and adequacy of existing treatments.” We do not require new drugs to be more effective than existing therapies nor do we “necessarily” require comparison with other products. Except that, “for products to treat life-threatening diseases, diseases with irreversible morbidity, and contagious diseases that pose serious health risks to others, it is essential for public health protection that a new therapy be as effective as existing approved therapies.” [Didn’t mean that, of course, as only superiority would show that; really referred to “non- inferiority”]

8 All in All 1. We might say an effect is clinically meaningless but more likely because of the nature of the effect (increased bile flow, suppression of gut fungus), not its size. In general, however, for a non-toxic drug and for a not serious disease we generally have not demanded an effect of a particular size, or required comparisons with other treatment. 2. Clearly could conclude that a small effect is outweighed by toxicity (rejected at least two Alzheimer’s drugs, one for severe N+V, the other for proximal weakness; many, many other examples). In these cases, the benefit did not outweigh risk. 3. Would consider available Rx. For toxic drug could 1) demand superiority 2) demand effect in non-responders (Clozapine, bepridil). In these cases would need comparative data.

9 All in All (cont) 4. Demand comparative data when disease is serious and there is existing Rx. Indeed, comparative (NI) is all you can do. In those cases, we regularly do insist on preserving some fraction of effect (usually 50%), indicating that loss of too much of a valuable effect could not be acceptable.

10 The PRO Document A new effect size standard? Mainly because of a concern that PRO methods are “too sensitive,” “it is important to consider whether the [detected] changes are meaningful.” It therefore calls for specifying a “minimum important difference (MD) as “a benchmark for interpreting mean differences.” The most explicit statement I know of that a statistically significant effect on a valid measure might not be accepted as evidence of effectiveness. Do we really mean it? Are we going to begin to say “not good enough.” And why only for PROs? Growing sample sizes raise similar issues.

11 One Last Point 1. Mean vs Distribution We tend to look at mean effects, but individuals will have a range of effects, some larger, some smaller. Might we be missing meaningful effects in a fraction of the patients by focusing on the mean? Should we more often show both mean and distribution. The latter will show, for a drug with effect on mean, that there are also more people with an effect of any given size. 2. What is effect size, anyway? Point estimate is not effect size. If we’re serious about MID, we’d need to revise the null hypothesis from Ho: T ≤ O to Ho T ≤ MID Ha: T > O to Ha T > MID Where the 97/2% (one sided) lower bound would be the test.

12 A Few Points 3. Do we really want to specify a minimum mean effect (or minimum difference on some dichotomous measure)? Of course, the biggest question is how we would support any particular value?