Transcript The Education of Clinical Trial Statisticians: Do we need

The Future of the Education of Clinical
Trial Statisticians: What Changes are
Needed and How do We Implement Them?
Karl E. Peace, Ph.D.
GCCDCS, SRS & Professor of Biostatistics
The Jiann-Ping Hsu College of Public Health
徐建萍公共卫生学院
PO BOX 8015
Georgia Southern University
Statesboro, Georgia 30460
912-478-7905 phone, 912-478-5811 fax
[email protected], [email protected]
Graybill Conference VII
June 12, 2008
Abstract:
Recommendations the speaker has
made over the last 25 years regarding
the relative importance of design,
analysis and interpretation of clinical
trials and The education of Clinical Trial
Statisticians are discussed.
7/17/2015
J-P Hsu College of Public Health
2
Question 1: How should the training of clinical trial
statisticians be modified to increase emphasis on design,
monitoring, consulting, and interpretation?
CTM course: emphasis is on critical thinking and understanding
concepts. More emphasis is placed on the design and conduct of
clinical trials more so than their analyses. Several case studies
are presented and then critiqued reflecting sequentially:
- Developing a clear understanding of the objective,
- Identification of required data or endpoints,
- Translation of the objective into statistical questions (testing
or estimation: including symbolism),
- Identification of the experimental design,
- Determining sample size requirements and bases for their
computation,
- Identification of statistical methods,
- Detailing how the trial was conducted (including monitoring
and data management),
- Reviewing and interpreting the results.
7/17/2015
J-P Hsu College of Public Health
3
Question 1 (continued):
CTM course: In addition,
- Each case study is reviewed to identify and classify methods
or procedures: minimize bias, IC, design, diagnostic, treatment,
safety of patients, data management, statistical, etc. and further
what methods apply pre, during and post study.
- The responsibilities of the statistician Pre, during and post
study are identified and explained.
CTM course, Other Features:
- Students learn how to write a Statistical Analysis Section for a
clinical trial protocol.
- Students have to write a protocol as part of final exam.
7/17/2015
J-P Hsu College of Public Health
4
Question 1 (continued):
Statistical Analysis Section
.1 Study Objectives as Statistical Questions
.2 Endpoints
.3 Statistical Methods
.4 Statistical Monitoring Procedures
.5 Statistical Design Considerations
.6 Subset Analyses
References
7/17/2015
J-P Hsu College of Public Health
5
Question 1 (continued):
CTM Final Exam Question: Following are data sets from a clinical trial in essential
hypertension: Dummy Data 1 and Dummy Data 2. The first column represents
observation number, the second represents treatment group assignment. The
remaining columns, in order, represent supine diastolic blood pressure (SDBP,
mmHG) at the baseline (Visit 1, at which assignment to treatment occurs) visit,
and follow-up visits 2, 3, 4 and 5. The data sets are identical, except that data
set 2 contains some missing values (the '.' ).
A. Write a simple, reasonable, protocol for a clinical trial that could have
produced these data. Make sure you include inclusion & exclusion criteria,
reflecting age at least 18 years, either sex, any race or ethnicity, and any
weight, and that the protocol is multi-center (with two centers).
B. After you have done this, you will need to add additional data columns to
include age, sex, race or ethnicity, and weight, and supply data values for
these. Now add a column for center to reflect that the protocol is multi-center
(2 centers): the first 10 patients in each treatment group come from center #1,
and the last 10 patients in each treatment group come from Center #2. Also add
a column for patient number (starting with 1101 for the first patient data row
for center 1 and treatment A, then 1102, etc. until the last patient receiving
treatment A in center 1, and starting with 1201 for the first patient data for
center 2 receiving treatment A, etc.).
7/17/2015
J-P Hsu College of Public Health
6
CTM Final Exam Question (continued):
C. Then write a SAS program to analyze each data set (ignoring center).
1. Provide estimates of mean treatment differences, Standard errors of the
differences, p-values for the differences, and 95% confidence intervals for the
differences by visit, and averaged across visits.
2. Interpret these results.
3. Turn in your SAS program and Output.
D. Then write a SAS program to analyze each data set (including center).
1. Provide estimates of mean treatment differences, Standard errors of the
differences, p-values for the differences, and 95% confidence intervals for the
differences by visit, and averaged across visits.
2. Interpret these results.
3. Test whether the data are poolable across centers.
4. Turn in your SAS program and Output.
E. Please provide assumptions underlying your analysis methodology, including
for Dummy Data 2, how you handled missing data.
7/17/2015
J-P Hsu College of Public Health
7
Question 2: How should the education of clinical trial
statisticians include development of communication skills
(oral and written)?
Approach that has worked:
- I help each class select a research project;
- with my help, they do the research and
write a manuscript style report, and develop
PPTs for presentation at three research
symposia (PKP, COGS, JPHCOPH).
- Every member contributes.
- Provides experience in: planning, research
methodology, consulting, group interaction,
and written and oral communication.
7/17/2015
J-P Hsu College of Public Health
8
Question 3: The power of computing and resampling –
based methods has in some cases replaced use of large
sample statistical theory. How can we identify an
appropriate balance of applied and methodological
statistical skills?
Unsure as to the proper balance:
- Some mix is needed (including
Bayesian
- go to resampling-based methods any
time no validated method exists and/or
when methodological assumptions are
violated?
7/17/2015
J-P Hsu College of Public Health
9
Question 4: How should the education of clinical trial
statisticians include more medical and biological
training?
Not sure, but a mix is needed.
- Georgia Southern University has no Medical
School and JPHCOPH has no formal affiliation
with one.
- JPHCOPH is young and we are exploring
collaborations with proximal medical schools.
We do have collaborations with SWGCC,
SEGACA, GCC, GACORE, and some health
departments and local hospitals for student
practicum experience.
- In the classes I teach (CTM and SIDRD), I
provide key information about the substantive
aspects of the application (plus draw from
students who have MD)
7/17/2015
J-P Hsu College of Public Health
10
Concluding Remarks:
“Analysis and Interpretation are important, but I believe the
greater contribution by the statistician is at the design stage.
Even helping to define the question the experiment is to
answer is an important contribution.”
“…The curriculum should de-emphasize methods and ‘how to’
and place greater emphasis on ‘why.’ The non-statistical
community will also derive greater benefit from courses which
stress sound experimental design concepts and principles and
the importance of scientific rigor than from all the methods
courses. There should also be courses on understanding the
structure of data, and better ways to descriptively profile and
display data.”
[Peace, KE: “Some Thoughts on the Biopharmaceutical Section
and Statistics” ASA Sesquicentennial Meeting, 1989]
7/17/2015
J-P Hsu College of Public Health
11
Concluding Remarks:
“We are confronted with many opportunities for consultation
in areas where we are ill prepared substantively. In the
future we must find the time to gain sufficient substantive
knowledge about the application so that we interact with the
client in a meaningful way – and keep the client coming back.
Now I know that many statisticians have acquired needed
substantive knowledge. I say it in this way to signal that
greater emphasis in the future should be placed on gaining
such knowledge prior to the consultation – even formally in
statistical degree programs.”
[Peace, KE: “Some Thoughts on the Biopharmaceutical
Section and Statistics” ASA Sesquicentennial Meeting, 1989]
7/17/2015
J-P Hsu College of Public Health
12
The Jiann-Ping Hsu College of Public Health
徐建萍公共卫生学院
Thank You.
7/17/2015
J-P Hsu College of Public Health
13
The Jiann-Ping Hsu College of Public Health
徐建萍公共卫生学院
Summary of Biostatistical Training/ Mentoring Activities in the PI
Biostatistical Aspects of Clinical Development Programs
Focus was asking and answering questions
Writing the protocol
No tolerance for the PARC method of research
Biostatistical Aspects of Basic Research Programs
WHY? To improve communication primarily
Eventually reciprocated from Clinical Development and some of the
substantive areas.
Result: substantive area scientists learned more about statistics,
experimental design, the importance of framing the question, etc.,
and the statisticians learned more about the substantive areas – at
least their vocabulary improved; so that communication and
consultation improved
As Chair of the PMA Training Committee of the Biostatistics
Subsection made both programs available to the PMA and taught
them at several Pharmaceutical companies; Later they were cotaught by statisticians from across the industry in DC and perhaps
under PERI.
7/17/2015
J-P Hsu College of Public Health
14
The Jiann-Ping Hsu College of Public Health
徐建萍公共卫生学院
Biostatistical Aspects of Clinical Development Programs
Quality clinical development programs require careful attention to planning,
execution, summarization and reporting. Even planning a single clinical trial
requires attention to these issues (which reflect generally an exercise of
writing a quality protocol):
Asking and answering Questions
Defining the Question (including endpoints)
Variability of Data
Subject / Patient definition and selection
Treatment (or intervention) definition and assignment
Standard Clinical Protocol Template
Answering the question
Multi-center trials
Safety / Informed Consent
Management Structure and Procedures (including monitoring)
7/17/2015
J-P Hsu College of Public Health
15