The points I have read are well-based and thought out, and I have enjoyed
reading the discussion.
The example of penicillin is correct and relevant. However, such dramatic
effects are rare. More often, one reads of researchers who draw unwarranted
conclusions from case series. I believe that the logic of analysis of time series
designs, particularly multiple interrupted time series, is relevant.
To establish likely causation from time series data, one must establish
pre-treatment stability on the outcome of interest, and post-intervention
stability too. Trends should be established, and auto-correlation possibilities
should also be examined. Too often, pre-treatment stability of the
health outcome or functional measure is assumed rather than measured.
The treatment should usually be one in which a short term effect is most plausible
(as opposed to a subtle change in recovery curve), as changes in recovery
curves are hard to detect unambiguously. The effects of other coincident events
(sometimes called history effects) that might effect the outcome would need to be
excluded: the best way to do this is to test whether the outcome
increment or effect occurs reliably after the treatment is applied repeatedly
to different individuals or groups.
I learned of these designs studying Cook and Campbell (1979) and in
econometric studies in graduate school, but much work has been done since then.
Best regards,
* Mark J.
Mark V. Johnston, Ph.D.
Director of Outcomes Research, KMRREC,
Professor, Physical Medicine & Rehabilitation, UMDNJ/NJMS,
1199 Pleasant Valley Way,
West Orange, NJ 07052
Phone: 973 243 6810
Fax: 973 243 6963 R.A.: 973 243 6876
>>> Rod Jackson <[log in to unmask]> 11/11/2005 8:51 PM >>>
Dear Kukuh - you mention the GATE tool as a possible evaluation tool
for case series. While the GATE intervention tool (downloadable from
www.epiq.co.nz) is designed for appraising intervention studies that
include both an exposure (i.e.intervention) and comparison group, you
can think of a case series as an intervention study without an
explicit comparison group. With this in mind the GATE tool or any
other appraisal tool for intervention studies can be used to evaluate
a case series.
It is worth noting that if the occurrence of the outcome of interest
is well known in people who don't receive the intervention described
in the case series (i.e the equivalent of the comparison group in a
standard controlled trial), then a case series can provide reasonable
evidence of the effectiveness of an intervention. One of the most
famous case series is the first group of patients (I think 5-10
people with pneumonia) who received penicillin. Prior to penicillin,
mortality was extremely high in patients with similar symptoms but it
fell dramatically in those receiving penicillin - no RCT was required.
regards Rod Jackson
>Dear Anita,
>
>Just recently, I came across a document entitled: Grading the evidence:
>case series scroing system. I believe this is part of the the Gate Tool.
>In this particular,it was employed by the NZ Guidelines Group in
>conducting systematic review on interventional pain management. In the
>context of EBM, I think the list is reasonable. Here is the link:
>http://www.acc.co.nz/wcm001/idcplg?IdcService=SS_GET_PAGE&nodeId=8012&ss
>SourceNodeId=7975
>I hope this helps.
>Thank you.
>
>Kukuh Noertjojo, MD MHSc MSc
>Evidence Based Practice Group,
>Clinical Services, Worker and Employer Services
>WorkSafe BC
>6951 Westminster Highway
>Richmond, BC V7C 1C6
>CANADA.
>
>Phone: 604 231 8417
>Fax: 604 279 7698
>
>
>-----Original Message-----
>From: Evidence based health (EBH)
>[mailto:[log in to unmask]] On Behalf Of Anita von
>Krogh
>Sent: Friday, November 11, 2005 3:48 AM
>To: [log in to unmask]
>Subject: Checklists for case reports or case series?
>
>Dear list-members
>
>I work at the Norwegian Poisons Information Center where we respond to
>telephone calls from the public and health personnel regarding
>poisonings,
>and I am a new member of this discussion list.
>
>We are presently revising the routines for quality grading of the
>documentation on which we base our advice to the public and health
>personnel. Our dilemma is that most of the available literature on
>toxicology in human beings are case reports, summaries of case reports
>or
>case series. The checklists I have been able to locate don't cover these
>
>types of articles.
>
>We have attempted to make a checklist for case reports/case series, but
>would like to hear if such a thing already exists. I wonder if any
>member
>of this list know if such a checklist exists, or of work to make one, or
>
>have opinions regarding it.
>
>I look forward to reading your response!
>
>Kind regards
>Anita von Krogh
>National Poisons Information Center
>Directorate for Health and Social Affairs
>Norway.
--
Rod Jackson
Professor of Epidemiology
Head of Section of Epidemiology and Biostatistics
School of Population Health, Tamaki Campus
Faculty of Medical & Health Sciences, University of Auckland
100 Morrin Rd (cnr Merton & Morrin Roads), Glen Innes, Auckland
Private Bag 92019
Auckland, New Zealand
Phone: +64 (0)9-3737599 ext 86343
Fax: +64 (0)9-3737494
e-mail: [log in to unmask]
EPIQ website: http://www.epiq.co.nz
email usually checked within 1 week of receipt
|