Hi Neil
The consensus amongst my colleagues is that it depends on the study
question:
If you are interested in testing whether the factors are 'predictors' of
AMD, then only pre diagnosis observations should be considered. Post
diagnosis observations would be confounding.
However, if you are only trying to show a 'link' between the factors and
AMD, post diagnosis observations are acceptable.
Murray
On 21/11/06, Neil Shephard <nshephard at gmail.com> wrote:
Hi,
Apologies for the slightly off-topic post but I'm seeking advise on
how to analyse a set of data I've encountered.
This is not my choice of study design, its something I'm being asked
to help a student with (and personally I think its a bit of a
data-dredging exercise, and have explained this to them, but there).
A group of cases with age-related macular degeneration (AMD) have been
collected from a clinic over a number of years, and because medical
records have been digitized for the past 20-30 years details on their
hospital admissions and a lot of other aspects of their health are
available from both before and the onset of AMD up to present date.
A group of controls from the general population without AMD have been
age and sex-matched to each of the cases, and data on hospital
admissions and so forth have also been extracted for the same
variables, from the first record available up to present date
The 'hypothesis' is that there are 'links' between AMD and
atherosclerotic disease, cardiovascular risk factors, systemic
inflamatory disease and neurodegenerative conditions.
The supervisor of this student has suggested that _all_ data should be
used, i.e. the occurence of some atherosclerotic disease in a case or
control _after_ the onset of AMD in the case should be included. This
to me does not make any sense at all. A case-control study has a
temporal relationship of the exposure occuring before disease does.
The aim is to assess whether certain diseases/risk factors
(athersclerosis, classic CVD risk factors etc.) increase the risk of
AMD. Analysing such events after the onset of disease turns that
question around, and mixing the two just completely messes things up
in my view.
I've read through relevant sections of Breslow & Day's "Statistical
Methods in Cancer Research - Vol 1", and Dawson & Trapp's "Basic &
Clinical Biostatistics" as well as sections on case-control and
similar study designs in the Wiley Encyclopedia of Biostatistics but
can't find anything that deals with analysing data in such a manner.
Is it valid to take events before and after the event of interest
(onset of AMD)? Personally I don't think so, but if anyone has any
thoughts or insights into this I'd greatly appreciate them.
Apologies again for the off-topic post, and thanks for your time,
Neil
--
"An intellectual is a person who has discovered something more
interesting than sex." - Aldous Huxley
Email - nshephard at gmail.com / neilshep at cyllene.uwa.edu.au
Website - http://slack.ser.man.ac.uk/
Photos - http://www.flickr.com/photos/slackline/