This question highlights the difference between evidence-based medicine (as
it has been defined and practiced in systematic reviews) and technology
assessment. EBM meta-analyses and systematic reviews have confined
themselves almost exclusively to RCTs. Thus, the topics covered by EBM are
limited to questions addressed by RCTs. Technology assessment (TA) does not
have that luxury. We must present decision makers with the current state of
knowledge, regardless of the source; although, it is essential to critically
analyze the reliability of the data.
I recently gave a talk on meta-analysis of uncontrolled studies at the
annual meeting of the International Society for Technology Assessment in
Health Care that was here in Philadelphia in June. Our approach has been to
use an evidence hierarchy only to guide our literature searches and
inclusion criteria, not to assign points by which to weight evidence. Thus,
if there are a number of double-blind RCTs, we do a meta-analysis of those.
Lesser designs (unblinded RCT, other controlled studies, uncontrolled
studies) will then only be looked at for any additional evidence they may
provide, such as on special patient groups, prognostic factors, etc. But if
there are no dbRCTs, we use whatever there is on the next level down the
evidence hierarchy.
In addition to the Ioannidis article cited by Sontheimer, there are other
interesting articles on randomized versus nonrandomized studies. One is
"Randomized, Controlled Trials, Observational Studies, and the Hierarchy of
Research Designs"
Concato J, Sha N and Horwitz RI, N Engl J Med, 2000, 342:1887-92. This
study found little difference in effect sizes in 55 RCTs and 44 controlled
studies of five different medical topics.
On the other hand, another study, "Assignment Methods in Experimentation:
When Do Nonrandomized Experiments Approximate Answers From Randomized
Experiments?" Heinsman DT and Shadish WR, Psych Meth, 1995, 1:154-69, found
substantial differences in effect sizes in 51 RCTs vs. 47 controlled trials
of four topics in education research. These two contrasting findings show
that the problem is topic specific. Furthermore, these latter authors went
on to do multiple regression analysis of various study design and reporting
variables in the studies. That is, they correlated the study variables to
the effect size. What they found was that randomization was seventh in the
top ten ranking of study variables affecting the effect size. Knowing these
correlation coiefficeints, they were then able to adjust the study results
for these variables. After adjustment there was little or no difference in
the effect sizes of the studies.
Sometimes there are not any controlled trials, only uncontrolled case
series. Then it is necessary to go to the literature and synthesize a
historical control. This is also good practice for assessing the validity
of active controls in RCTs without a no-treatment group. This procedure is
problematic and has been examined in the study "Randomized versus Historical
Controls for Clinical Trials" Sacks H, Chalmers TC, Smith H Jr; Am J Med,
1982, 72:233-40. These authors found that using historical controls
frequently exagerates the effect size. While treatment group results were
similar regardless of the comparison design, historical controls usually
fared worse than parallel controls, thus accounting for the exageration in
effect size. Because of this potential exageration, small or modest effect
sizes found with historical controls are not very reliable; however, we have
seen some situations where the effect size with historical controls was so
large and striking that the findings could not be ignored, and in fact were
strong evidence that there was no equipoise, and an RCT might be unethical.
This raises a point that has always puzzled me. RCTs are only considered
ethical if there is equipoise. But what can the evidence be for equipoise?
EBM only recognizes RCTs as valid evidence. As far as I know, EBM is silent
on what the evidence must be for equipoise. Any thoughts anyone?
David L. Doggett, Ph.D.
Senior Medical Research Analyst
Health Technology Assessment and Information Services
ECRI, a non-profit health services research organization
5200 Butler Pike
Plymouth Meeting, Pennsylvania 19462, U.S.A.
Phone: (610) 825-6000 x5509
FAX: (610) 834-1275
http://www.ecri.org
e-mail: [log in to unmask]
-----Original Message-----
From: Sontheimer, Daniel MD [mailto:[log in to unmask]]
Sent: Friday, August 17, 2001 8:30 AM
To: [log in to unmask]
Subject:
I thought someone might have started kicking this one around, particularly
with all the recent discussion on evidence grading. From JAMA, 8/15/2001:
Ioannidis, J et al. "Comparison of Evidence of Treatment Effects in
Randomized and Nonrandomized Studies"
the authors state:
"Although we perused several hundreds of meta-anlyses, the vast majority
regarded the randomized design as a prerequisite for eligibility and most of
them did not even cite the nonrandomized studies. This is unfair for
epidemiological research that may offer some complementary insights to those
provided by randomized trials. We propose that future systematic reviews
and meta analyses should pay more attention to the available randomized
data. It would be wrong to reduce the efforts to promote randomized trials
so as to obtain easy answers from nonrandomized designs. However,
nonrandomized evidence may also be useful and may be helpful in the
interpretation of randomized results."
I can see their point, but have a little trouble with using the term unfair.
Limiting to randomized data provides a uniform framework for building a
systematic review or meta-analysis. There is something to be said for
keeping it simple. Perhaps, citing of nonrandomized trials that are
discrepant would be helpful. Any other thoughts?
Dan Sontheimer
Assoc. Director Spartanburg Family Medicine Residency
Spartanburg, SC USA
DISCLAIMER: The information in this message is confidential and may be
legally privileged. It is intended solely for the addressee. Access to
this message by anyone else is unauthorized. If you are not the intended
recipient, any disclosure, copying, or distribution of the message, or any
action or omission taken by you in reliance on it, is prohibited and may be
unlawful. Please immediately contact the sender if you have received this
message in error. Thank you.
|