Web Site Entry For Medical School Educators In
Medical Statistics
|
Your feedback and collaboration are welcome:
medicalstatistic@aol.com
In addition, new cases from the medical literature
of statistical misadventures and comments suggesting improvement are being
collected for 2008. Some of those cases will be
subsequently posted on this website. (Submit
here)
_________________________________
A
few topics have download files in addition to the other information available:
1.
Lessons from an unusual saga of possible fraud in the literature
The Columbia
'Miracle' Study:
"print
format"
download
(pdf)
2.
The hazards of subgroup analysis. Are Geminis really different? "print
format" download
(pdf)
3.
Example of bias apparent in a meta-analysis in 30 seconds:
"print
format" download
(pdf)
4.
A Type II Statistical Error Incorrect conclusions resulting
from insufficient numbers. Death
of the oat bran fad
download
(pdf)
5.
Miscounted and Misinterpreted. How not to conduct a
study.
The
NEJM cath study
"print
format" download
(pdf)
6. The ALLHAT hypertension trial-- a detailed critique
"print
format" download
(pdf)
7
Celebrex-- an unreliable
cross trial comparison
"print
format" download
(pdf)
8.
A Tale of Two Trials- A superbly conducted trial contrasted to a
suboptimally interpreted trial:
"print
format" download
(pdf)
9
Limitations of Meta-analysis: overview:
"print
format"
download
(pdf) .
|
The task of educating medical students in the critical appraisal skills required
for properly evaluating clinical studies is a large one.
This web site has resource
material that may be useful in helping medical students recognize poorly
formulated statistics or inappropriately drawn conclusions in medical studies.
A number of medical studies with significant problems
in either clinical trial design, medical statistics, or
properly formulated conclusions are examined in detail. There
will be differing viewpoints from those on this website
regarding these case studies, but it is hoped those
differences will provide useful starting points for
discussion.
This web site besides being a
collection of suboptimally performed or interpreted medical studies, also reviews
some of the strengths
and weaknesses of meta-analyses and large randomized clinical trials.
The information on this web site which has
been created by the author can be
used for the instruction of medical students and residents without restriction as long as the source is documented.
It
is hoped that this site will prove a useful resource.
The
following is a summary of the content of this web site in regards to usefulness for educators of medical students:
Example 1: This is probably not particularly useful for medical
students. This has to do with a randomized clinical trial with an inherently
biased protocol. It does provide an example where a suboptimal treatment
protocol used in a clinical trial is a crucial factor in regards to the
implications of that trial for clinical practice.
A
trial led by Harvard Medical School faculty with a protocol so biased
(unintentionally) that the trial was
inherently was an invalid comparison.
Example 2:
This would be a beneficial case study for
medical students. An incredibly poorly conceived statistical analysis in a
major medical journal. A statistical analysis in a major medical
journal JACC (Journal of American College of Cardiology) which was so bad,
that it may have led to a subsequent improvement in the review process of JACC.
Example
3: This study is an example of a
trial's conclusions being the opposite of what
the trial data suggests. This
study would be useful for discussion. A potential for a type 2 statistical error is also illustrated as
well. An
example of a study in the New England Journal of Medicine where the conclusions
run exactly the opposite of what the data suggests.
Example 4:
It is probably not worth the time
of the students to fully understand
the details. Though there is some potential interest in understanding how a
poorly done review of literature can make the conclusions of a trial completely
incorrect. Understanding this
example may be fairly tedious.
How
an incorrectly done review of the literature in a single article in the medical
literature could potentially adversely affect patient care.
Example 5:
This ALLHAT trial critiqued would be of interest
only to those with special interest in
clinical trials that affect current hypertension guidelines.
It is unclear whether this example would be conducive to medical student
education. The
ALLHAT trial- how conclusions that extend beyond data can provoke controversy
(ALLHAT
Trial Critique download- pdf format)
Example 6:
This is a classic type II
statistical error with conclusions that do not recognize the limitations of the
trial protocol nor the weakness resulting from the small numbers of subjects
involved. Death of the oat bran fad. Death
of the oat bran fad. (Murdered by a poorly conceived study.)
Example 7:
The topic of alcohol and pregnancy
may be of potential interest. The study
provides a starting point for discussion of 1) the
strengths and weaknesses of descriptive rather than randomized studies, 2)
limitations of epidemiologic derived information that is not yet confirmed by
randomized trials, 3) and consideration of whether the results of a multiple
regression analysis can always be considered the last word on a subject. Pregnancy
and alcohol, what do some of the
studies really suggest?
Example
8: This consists of comments by Dr. Peter Sleight in an
insightful article regarding the hazards of inappropriate subgroup analysis. The Hazards of Subgroup Analysis
(Are
Gemini's really different?)
Example 9: The unfortunate saga of possible fraud in a medical study
with a medical journal and a medical school failing to respond appropriately.
The Columbia University 'Miracle' Study-
Example
10: These topics may be useful as starting points for discussion
with medical students.
The Very Large Randomized
Clinical Trials (Strengths and Limitations)
Some strengths
and limitations of large randomized clinical trials are considered.
A Tale of Two Large Trials
A
very well done large randomized trial is compared to a poorly interpreted large
clinical trial. Both these trials have and will affect clinical practice
guidelines.
Limitations
of Meta-Analyses
Often
a meta-analysis is popularly thought to be the final word concerning a medical topic.
Some of the potential limitations of meta-analyses are discussed
Network meta-analyses,
additional potential for error
A
type of meta-analysis which is more subject to error
Example 11 How medical science
advances when there are problems in the scientific literature. (How
these particular issues might look "10 years
later") This discussion of the self
correcting aspects of the medical literature may be of interest.
Celebrex This may be useful as a
starting point for discussion with medical students regarding a multitude of
topics. One issue, is that of the reliability of cross trial comparison (particularly given that
drug representatives often will try to make invalid cross-trial comparisons). This is an example of comparing the placebo group from one
trial to a treatment group in another trial, which is always problematic.
The study was written by superb investigators, but this one aspect of the statistical
analysis is quite questionable. (On the other hand, this 2001 article in
JAMA brought beneficial widespread attention to the potential of a COX-2
inhibitor being associated with increased cardiovascular risk.)
____________________________________________
The task of educators of medical students in the field
of interpreting clinical trials is a crucial one. Following
graduation, the students as physicians will be confronted with a barrage of
clinical trials. Skills helping to
discern optimally performed and interpreted clinical trials from those that are
inherently weak will be useful to the current student throughout his or her
career in medicine.
.
home
| 