really interesting problem.
prevalence of carcinoma of lung in females aged 50-60 by RHA from vamp data per
10,000
Northern 1.084
Yorkshire 6.066
Trent 0.718
East Anglia 2.546
NW Thames 3.525
NE Thames 1.003
SE Thames 1.158
SW Thames 4.486
Wessex 0
Oxford 4.946
South Western 3.801
West Midlands 4.133
Mersey 1.307
North Western 4.869
Wales 3.196
Scotland 3.309
Northern Ireland 6.903
ok so her baseline risk is 4 per 10,000
ie 0.04%
say as she smokes that quadruples her risk
0.16%
i would guess the sensitivity of a chest x-ray is not great ie 70-80% and but
probably quite specific either ie 90-95%
this gives a range of +LR of 7 - 16
so if she walks in through the door and you do the x-ray before she talks the
results would be
pre test prob 0.16% post test prob 1 - 2.5%
of course if she coughs for 3 months & lost weight & had haemoptysis her
pre-test might be 50% and her post test 87-94%
i guess that's why we take histories - the best diagnostic test invented is the
ear !!
martin
-----Original Message-----
From: owen dempsey [SMTP:[log in to unmask]]
Sent: 01 October 1998 22:18
To: evidence-based-health
Subject: cxr/predictive value
2 question here really: background below (the example is a hypothetical
one)
1) can anybody point me in the direction of a source for a)the
prevalence of lung cancer in a 55yr old caucasian female, lived in the
UK all her life, who has smoked 10-20 cigs/day for 35 yrs and whose
younger sister (heavier smoker) has just had lung cancer diagnosed; and
b) the sensitivity and specificity of a plain cxr for diagnosing lung
cancer in such an individual (no known co-existing lung diseases)?
2) i'd be very interested to know what peoples' best guesses are for
these values, if other people want to play this game then they could
e-mail me privately and if i get more than 10 responses say i'll put the
results up on the list. if any radiologists reply i'd be interested if
they could identify themselves as such (do they over or under estimate
the accuracy of the cxr i wonder)
background:
on the gp-uk list we've been discussing the relative merits of
performing a cxr at the request of the individual in a) above, and there
were of course pros and antis. we calculated predictive values, false
positive rates etc using a range of values for the variables mentioned
in a) and b) and the results were extremely interesting and i think have
surprised many. it has been suggested that we attempt to formalise the
results a bit by using some values that have a better basis in evidence
if possible.
I have searched Bandolier, the effective health bulletin and performed
some searches using Pubmed: suffice to say that i am aware of the trials
showing a failure of screening smokers to reduce lung cancer mortality;
and seen a snippet in a review by Eddy D that was difficult to interpret
(quote: Mayo Lung Project: Chest roentgenograms and sputum cytology lead
to false-positive test results in smokers of approximately 5% and 0.5%,
respectively. Because of the lack of evidence of benefit and because of
its potential harms and costs, screening for lung cancer is not
recommended). However i am drawing a blank on the questions above.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|