Re: analyte specificity and sample size
- From: David Winsemius <doe_snot@xxxxxxxxxxx>
- Date: Sat, 15 Jul 2006 14:07:00 -0500
andie_baker@xxxxxxxxxxx wrote in news:1152752094.258076.167400
@h48g2000cwc.googlegroups.com:
The table below shows the incidence of Crohn's Disease and Celiac
Sprue serological markers in disease state samples.
For Celiac sprue the markers and their incidence are as follows:
AGA IgA (80-90%),
tTG IgA (97%),
AGA IgG (75-85%),
tTG IgG (56.6%),
For Crohn's Disease:
ASCA IgA (36.5%),
ASCA IgG (45.5%),
How many diagnosed samples are required in order to calculate the
disease state sensitivity and specificity for each of the 6 analytes
(above) toward their respective diseases?
You have given us are the proportions of positive results among persons
with the clinical disease (the sensitivities). What you now need to
provide is the proportion of positive results in persons _without_ the
diseases is question. (This is 1 minus specificity. Specificity is the
proportion of correct results when testing persons known not to have a
disease.)
The table above has the percent incidence of the serological markers
for the two disease states, Celiac Sprue and Crohn's Disease. I assume
the gating factor will be the analyte of the lowest incidence for each
of the diseases. So, this would be tTG IgG for Celiac and ASCA IgA
for Crohn's. Am I correct?
If the rest of the readers are like me, then you will need to clarify
what you mean by "the gating factor". Perhaps you mean limiting factor?
If so, then the limiting factor in a binomial power analysis may be the
lowest expected count in a table.
I would like to learn how to calculate the required number of diseaseYou may have thrown in a ringer buy bringing up another measure,
state positive samples, and negatives for two levels of uncertainty in
the concordance values: ± 5% and ± 10%.
"concordance values". Did you mean what are generally called p-values or
did you mean concordance of two tests for the same outcome?
[If the] prevalence among the general population of Crohn's Disease in
North America is 0.026 to 0.198%, while that of Celiac Disease is 0.5
to 1%[, then how] many normal samples are required?
Still not exactly clear what you want. Required for the estimation of
what? If you want to put confidence intervals around sensitivity,
specificity, positive predictive value, and negative predictive values,
then this can only be determined by giving us a reasonable estimate of
the proportion of (falsely) positive results in a normal population. (For
the predictive values you will also need the prevalence of the diseases
in the general population.) If you cannot tell us those numbesr or give
reasonable estimates, then you will be forced to do some sort of
sensitivity analyisis that considers a wide range of possible values. You
will also need to clarify whether your goal is to classify persons with
abdominal complaints into (Crohn's vs. celiac) or (Crohn's vs not
Crohn's). I suspect from your presentation that it is the second question
you want answered.
I see this, but you give no clue what part of the article you did not
I am grateful for any help you can offer. Dr. Martin Holt, of
Medstats, has graciously referred me to the article: Determination of
Sample Size by Nyi Nyi Naing, which provides a good starting point.
(Please see his response on Medstats, July 8th to my initial posting of
this question.)
understand or why it does not answer your questions. You are still
basically asking the same question as you posed in Medstats. It is not we
who need a starting point; it is you. Maybe some more reading would help:
http://www.cche.net/usersguides/diagnosis.asp
http://bmj.bmjjournals.com/cgi/content/full/308/6943/1552
http://bmj.bmjjournals.com/cgi/reprint/309/6948/188
--
David
.
- References:
- analyte specificity and sample size
- From: andie_baker
- analyte specificity and sample size
- Prev by Date: Request for information regarding companies which are hiring M.Sc(Statistics) freshers from central universities all over india
- Next by Date: Re: Who first used the T-score?
- Previous by thread: analyte specificity and sample size
- Next by thread: SAS Vacancies in the UK
- Index(es):
Relevant Pages
|
