Navigation
Fall 2008 training series
Fall 2008 training series includes webinars
on the analysis of stratified data in clinical trials, bioequivalence
problems and samplesize analysis.
Webinar 1: Stratified analyses: Tips for
improving power
Presented by Dr. Devan V. Mehrotra (Merck
Research Laboratories) on October 21, 2008 (noon2:00 Eastern
time).
The MantelHaenszel test and the van Elteren
test, both implemented in SAS PROC FREQ, are widely used for stratified
analyses of binary and ranked data, respectively. Both methods
have good power properties, but only under certain restrictive
assumptions; when the assumptions are violated, there can be a
notable loss in power. In this tutorial, we will describe some
alternatives to these popular methods, including the "minimum
risk" weighting strategy for stratified binary data and an "adaptive"
testing strategy for stratified rankbased analyses. We will use
simulations to provide guidance on which methods might be more
appropriate under given conditions. Numerical examples will be
used throughout for illustration, and to reinforce the key points.
Handouts
Questions and answers
Responses
to the questions submitted during the October 21 webinar.
Webinar 2: Bioequivalence
Presented by Drs. Scott Patterson (Wyeth)
and Byron Jones (Pfizer) on December 10, 2008 (10:00noon
Eastern time).
This introductory course will focus on the
design and analysis of bioequivalence studies for orally administered
drug products. It provides a detailed overview of the most wellestablished
method of demonstrating bioequivalence. The following topic will
be covered: 1. Drug development, clinical pharmacology and statistics.
2. History and international bioequivalence regulations. 3. 2x2
crossover designs and average Bioequivalence with examples. The
course offers a wellbalanced mix of theory and applications,
including regulatory considerations. Examples from real trials
are used throughout the discussion to illustrate the statistical
approaches discussed in the course.
Text book
Patterson S, Jones B. (2006). Bioequivalence
and Statistics in Clinical Pharmacology (Chapters 12). Chapman
and Hall, CRC Press, London.
Software implementation
Software implementation of the described
statistical methods may be found on the book's
web site.
Handouts
Questions and answers
Responses
to the questions submitted during the December 10 webinar.
Webinar 3: Classical samplesize analysis
for hypothesis testing (Part I)
Presented by Prof. Ralph O'Brien (Case Western)
on January 29, 2009 (noon2:00 Eastern time).
This session will cover the key concepts,
all by discussing a straightforward, realistic, twoarm clinical
trial being planned to test a fictional lactatelowering drug
(QCA) that promises to reduce mortality in children with severe
malaria. We will first discuss that if QCA has no effect, the
distribution of the pvalue is uniform between 0.00 to 1.00, regardless
of how large the total sample size (N) is, a concept misunderstood
by many who think that pvalues shift towards 1.00 when the null
hypothesis is true. The graphs let us see that under the null
hypothesis, Prob[p <= alpha] = alpha, the Type I error rate. Then
I show how the pvalue shifts more towards 0.00 when QCA is conjectured
to be more effective and how this leftward shift increases when
N increases and when N is allocated in a nearly balanced manner
between the two arms. This lets us see graphically the Type II
error rate, Prob[p > alpha] = beta, and the power = Prob[p <=
alpha] = 1  beta. Time will be devoted to discussing how one
formulates the scenarios for a typical samplesize analysis and
how one presents the results of this exercise in a statistical
considerations section in a research proposal. We will cover some
of the messy issues that arise, such as: Are samplesize analyses
relevant for "pilot" studies? How do we deal with the builtin
tradeoff between alpha and beta. In other words, what are a prudent
a and an acceptable power? How does using a onesided versus twosided
hypotheses affect these matters? What is "statistical gaming"
in samplesize analysis?
Text book
O'Brien R, Castelloe J. (2007). Samplesize
analysis for traditional hypothesis testing: Concepts and issues.
Pharmaceutical
Statistics Using SAS: A Practical Guide. Dmitrienko A, ChuangStein
C, D'Agostino R. (editors). SAS Press: Cary, NC.
Webinar 4: Classical samplesize analysis
for hypothesis testing (Part II)
Presented by Prof. Ralph O'Brien (Case Western)
on February 12, 2009 (noon2:00 Eastern time).
This session will quickly review the essentials
of the first session and then continue with the malaria example
to explore more vital questions that classical samplesize analyses
fails to address. That is, if the planned study yields a significant
pvalue, what is the chance this is a Type I error? Likewise,
if the study turns out nonsignificant, what is the chance this
is a Type II error? By using judgments about the probability that
the null hypothesis is false, we apply Bayes Theorem (taught with
simple calculations in a table, no formulas) to assess these "crucial"
Type I and II error rates, and we show (using a simple a Excel
program) that they can differ greatly from their classical counterparts.
Importantly, both crucial error rates are reduced by increasing
the statistical power. Studies with small N that propose to test
speculative hypotheses are prone to large crucial error rates.
The final phase of the session deals with an actual early trial
of a highly novel treatment for atherosclerosis in which a 0.02
pvalue was deemed to be "the first convincing demonstration"
of efficacy. What the investigators failed to understand, however,
is that their crucial Type I error rate may have been well over
85%. We will end by going though a mock study planning exercise
to design the followup study.
Text book
O'Brien R, Castelloe J. (2007). Samplesize
analysis for traditional hypothesis testing: Concepts and issues.
Pharmaceutical
Statistics Using SAS: A Practical Guide. Dmitrienko A, ChuangStein
C, D'Agostino R. (editors). SAS Press: Cary, NC.
Registration
Registration fee is $44 (Member of the Biopharmaceutical
Section), $59 (ASA Member) and $74 (Nonmember). To register for
individual webinars, visit the
Biopharmaceutical Section's web page.
