There is a wonderful book titled: Super Crunchers by Ian Ayres which
addresses the beauty and strength of randomization of very large data
sets. Also addresses why randomization can be as strong as RCTs.
Donald M. Grubb MSN ARNP
Lakeside Occupational Medicine
Courtesy Staff
University of South Florida School of Nursing
Tampa, FL
-----Original Message-----
From: Evidence based health (EBH)
[mailto:[log in to unmask]] On Behalf Of Klim
McPherson
Sent: Tuesday, April 01, 2008 12:55 PM
To: [log in to unmask]
Subject: Re: Randomized clinical trials versus observational studies: a
shift in paradigm?
I think so long as people (philosophers or whoever) continue to confuse
the point of randomisation with matching this argument will role and
role - and for no purpose. The two are different; matching is clearly
useful but a chore while randomisation combined with proper analyses is
pure magic and moreover is not especially a chore.
Klim McPherson
Visiting Professor of Public Health Epidemiology
Nuffield Dept of Obstetrics & Gynaecology
Oxford University
Womens Centre, Level3
John Radcliffe Hospital
Headington
Oxford OX3 9DU
Tel: **** 01865 740885****
[log in to unmask]
>>> Jaro Wex <[log in to unmask]> 01/04/08 17:20 >>>
Apologies for cross-posting.
Dear Group Members,
Randomized clinical trials (RCTs) are considered the gold standard in
establishing efficacy of therapeutic interventions. Often considerable
financial interest hinges on the "statistical significance" of results
or
lack thereof. At the same time, more "observational" patient level data
is
becoming available for health outcomes research and it is expected that
such data will increasingly be considered to inform recommendations and
decision making in health care, public or private. Economic data can be
derived from both RCTs and observational studies.
The claim that RCTs always stand higher in the hierarchy of evidence
relative to observational studies has been axiomatic and remains
undisputed, particularly in the medical world. Many philosophers and
statisticians, however, do not seem to agree. While there is consensus
that good studies, RCTs or otherwise, are those with less bias, there
are
numerous examples of unbalanced and biased RCTs, as well as very well
designed cohort studies. According to some authors, notably John Worrall
(LSE), randomization itself offers little advantage, and we can never be
sure that the groups are balanced for all known and unknown factors. The
argument that randomization specifically takes care of the unknown
confounders in by no means convincing. While blinding is valuable, thst
can not always be said of randomization. Bayesian statisticians seem to
have been able to deal with issues of correlation versus causality in
clinical trials without resorting to randomization.
That said, there is no reason to claim that the actually conducted RCTs
and observational studies are of comparable weight as evidence. Much
routine data are of poor quality with missing values and are plagued by
numerous biases and coding errors. RCTs on the other hand, are highly
regulated, thoroughly monitored, standardized and scrutinized. But there
is no reason to believe that all RCTs are inherently superior to, say,
prospective studies on consecutive cohorts encompassing nearly all
treated
patients. While retrospective analysis could create potential
statistical
issues, a priori protocols for observational studies would be a
solution.
I am currently working on a Patient Registry & RCT Data Survey as part
of
an ISPOR Working Group. The goal is to assess the degree to which health
plan administrators, policy makers and regulators depend on registry
data
for their coverage and reimbursement decisions. Here is the link below
is
to a short survey which will inform our further research:
http://tinyurl.com/2fdrkz. If you are interested in results, you are
welcome to participate.
And my questions for discussion are: Will the regulators and decision
makers start looking more closely at both RCTs and observational studies
to identify sources of best evidence in both categories? Would RCTs
always
be considered superior to observational studies, regardless of their
respective quality? Are we experiencing a slow shift in paradigm? What
are
the barriers to that shift?
Kind regards,
Ja
ro
__________________
Jaro Wex BA, MBBS, PhD (econ)
DirecT: +44(0)2030 511 424
F: +44(0)2030 511 435
M:+44(0)7716 299 835
E: [log in to unmask]
W:www.pharmarchitecture.com
|