8/13/2019 Biomedical Statistics Johns Hopkins

1/28

Copyright 2008, The Johns Hopkins University and Mary Foulkes. All rights reserved. Use of these materials

permitted only in accordance with license rights granted. Materials provided AS IS; no representations orwarranties provided. User assumes all responsibility for use, and all liability related thereto, and must independently

review all materials for accuracy and efficacy. May contain materials owned by others. User is responsible for

obtaining permissions for use from third parties as needed.

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike License. Your use of this

material constitutes acceptance of that license and the conditions of use of materials on this site.

http://creativecommons.org/licenses/by-nc-sa/2.5/http://creativecommons.org/licenses/by-nc-sa/2.5/

8/13/2019 Biomedical Statistics Johns Hopkins

2/28

Study Designs, Objectives, and Hypotheses

Mary Foulkes, PhDJohns Hopkins University

8/13/2019 Biomedical Statistics Johns Hopkins

3/28

Section A

Variations in Study Designs

8/13/2019 Biomedical Statistics Johns Hopkins

4/28

4

Many Designs Available

Parallel Cross-over

Factorial

Cluster

Others

8/13/2019 Biomedical Statistics Johns Hopkins

5/28

5

Categories of Trials

Phases Control groups

Chronic vs. short-term treatment

Single vs. multiple centers

Exploratory vs. confirmatory

8/13/2019 Biomedical Statistics Johns Hopkins

6/286

Study Designs

Focus on trials intended to provide primary evidence of safety andefficacy (pivotal trials)

Regulations permit substantial flexibility (adequate and well-

controlled trials)

8/13/2019 Biomedical Statistics Johns Hopkins

7/287

Selecting a Study Design

What are the objectives? What are the expectations?

Major advance Modest advance

Reduction in side effects What are the practicalities?

Available population

Relevant population Potential impact on medical practice

Ability to blind/mask

8/13/2019 Biomedical Statistics Johns Hopkins

8/28

8/13/2019 Biomedical Statistics Johns Hopkins

9/289

Parallel Design

Classical clinical trial approach Two study groups

Randomized assignment

Randomization

Treatment A

Evaluation of

Outcomes

Treatment B

Adapted from Tinmouth A, Hebert P. Interventional trials: an overview of design alternatives. Transfusion. 2007;47:565-67.

8/13/2019 Biomedical Statistics Johns Hopkins

10/2810

Factorial Designs

Evaluates multiple factors simultaneously 2 X 2 most practical, but little used

Sometimes a combination cannot be given (incomplete factorial)

Randomization

Treatment A

Evaluation of Outcomes

Treatment B Treatment A+B None

Adapted from Tinmouth A, Hebert P. Interventional trials: an overview of design alternatives. Transfusion. 2007;47:565-67.

8/13/2019 Biomedical Statistics Johns Hopkins

11/2811Source: Lancet. (1992). 339: 753-70.

ISIS-3 Design

Fibrinolytic agent

SK tPA APSAC

Anti-thrombotic

agent

Aspirin

Aspirin and

heparin

8/13/2019 Biomedical Statistics Johns Hopkins

12/28

12

Factorial Design

Evaluates two interventions simultaneously Four possible treatment combinations

Efficient approach in some circumstances

Potentially more informative approach

Increases proportion getting active treatment

Major concern: interaction of interventions

8/13/2019 Biomedical Statistics Johns Hopkins

13/28

13

Problems with Factorial Design

Patients must be willing and able to take any of the treatmentcombinations

Optimal dose modification strategy for toxicity may be hard to

determine

May require burdensome administration scheme if blinded Interaction complicates interpretation of treatment effects

8/13/2019 Biomedical Statistics Johns Hopkins

14/28

14

Bottom Line on Interaction

You cant rely on detecting modest interactions if studies arepowered for main effects

Interaction is important to study if agents are likely to be used

together

8/13/2019 Biomedical Statistics Johns Hopkins

15/28

15

Crossover Designs

Randomization

Treatment A Treatment B

Treatment A Treatment B

Evaluation of Outcomes

Evaluation of Outcomes

Adapted from Tinmouth A, Hebert P. Interventional trials: an overview of design alternatives. Transfusion. 2007;47:565-67.

8/13/2019 Biomedical Statistics Johns Hopkins

16/28

16

Advantages of Cross-Over Designs

Address question of major interest Will this patient do better on drug A or drug B? Removes patient effect thereby reducing variability and

increasing precision of estimation

Opportunity to receive both treatments (or be assured of receivingactive treatment at some point) is attractive to patients

Under assumption of no carryover effect, design provides more

information than simple parallel design

8/13/2019 Biomedical Statistics Johns Hopkins

17/28

17

Disadvantages of Cross-Over Designs

Assumption of no carryover effects is difficult to test May be difficult to determine appropriate length of washout period

so as to avoid carryover effects

There may be period effects in addition to carryover effects

Progression of disease Dropouts

8/13/2019 Biomedical Statistics Johns Hopkins

18/28

18

Cluster Design

Groups or clusters randomly assigned, not individuals Examples: villages, classrooms, platoons

Randomization

Treatment A

Evaluation of

Outcomes

Treatment B

Adapted from Tinmouth A, Hebert P. Interventional trials: an overview of design alternatives. Transfusion. 2007;47:565-67.

8/13/2019 Biomedical Statistics Johns Hopkins

19/28

19

Study Designs

Treatment allocation method Blinding of assigned treatment

Choice of control group

ICH E10

8/13/2019 Biomedical Statistics Johns Hopkins

20/28

20

In the Next Lecture Section Well Look at . . .

Objectives and hypotheses How to meet objectives Hierarchy of strength of evidence

Phases of trials

8/13/2019 Biomedical Statistics Johns Hopkins

21/28

Section B

Objectives and Hypotheses

8/13/2019 Biomedical Statistics Johns Hopkins

22/28

22

In the Beginning

Begin with a clear statement of the major scientific questions posed

by the study, usually conveyed in quantitative terms

8/13/2019 Biomedical Statistics Johns Hopkins

23/28

23

Objectives by Phase

Phase I

Determine optimal or tolerable dose Describe adverse event or PK profile Establish feasibility of treatment approach

Phase II

Estimation of activity

Comparison of doses or schedules

Estimation of factors for Phase III Phase III

Demonstrate superiority or non-inferiority

Estimate rates of adverse events Phase IV Address remaining outstanding issues

8/13/2019 Biomedical Statistics Johns Hopkins

24/28

24

Examples

To select the optimal dose that satisfies specific criteria

To demonstrate that the two year mortality rate on treatment A is

less than on treatment B

8/13/2019 Biomedical Statistics Johns Hopkins

25/28

25

The Objective Is to . . .

Classify

Order

Estimate differences

Estimate rates

8/13/2019 Biomedical Statistics Johns Hopkins

26/28

26

Cohesive Driving Force

All other properties of the trial, the study population, the primary

endpoint, the sample size, the primary analysis, flow from the study

objective

Photo by Dirk Gently via Flickr.com. Creative Commons BY-NC-ND.

8/13/2019 Biomedical Statistics Johns Hopkins

27/28

27

Study Hypotheses

The study objective corresponds to the primary hypothesis of the

study, e.g., the null hypothesis, H0

The two year mortality rate on treatment A equals the two yearmortality rate on treatment B

8/13/2019 Biomedical Statistics Johns Hopkins

28/28