Measurement of confidence interval of forest plot for diagnostic Odd Ratio

Question

[izsyed16]

Hi all, I am using Metadisc for forest plot of Diagnostic odd ratio(DOR), My Input values are TP, TN, FP, FN of individual studies. Forest Plot gives me Odd ratio for each study and pooled value of DOR from all studies along with confidence Interval. My question is that how confidence interval is calculated given only TP,TN,FP, FN values? Thanks

Answer 1

[Kevin Blighe]

Hello friend,

I have previously worked on clinical test statistics such as TP, TN, FP, FN, precision, accuracy, cost functions, etc. It is an area that can sometimes give a big headache...

The odds ratio (OR) is just calculated as (TP / FN) / (FP / TN) = (TP * TN) / (FP * FN)

In order to derive the 95% confidence intervals from an OR, we usually get the standard error of the natural log of the OR, and also make the assumption that the data is normally distributed (it may not be).

The series of commands (in R) would be:

TP=500
TN=20
FP=100
FN=50

OR <- (TP / FN) / (FP / TN)
OR
[1] 2

stndErr <- sqrt( (1/TP) + (1/TN) + (1/FP) + (1/FN) )
stndErr
[1] 0.2863564

upper95CI <- exp(log(OR) + (1.96 * stndErr))
upper95CI
[1] 3.505755

lower95CI <- exp(log(OR) - (1.96 * stndErr))
lower95CI
[1] 1.140981