Is this right?

[pag794]

I was browsing the University of Chicago celiac page, and came across this post.  The math doesn't seem right to me.  If the ttg test is 97-98% specific, doesn't that mean that out of every 100 people who have elevated ttg, 97-98% do have celiac, and 3-4% do not have celiac?  The way they explain it, the specificity would only be 25%, which is definitely not right.  

[cyclinglady]

Alas, my training is in business/marketing!  But I found this link that could help!

Open Original Shared Link

Oh the wonders of YouTube:

 

 

[Sugarcube]

Thats correct, ttg has a relatively high accuracy compared to other celiac blood tests. However it should not be solely relied  upon in isolation. As many other conitions can elevate the ttg.  Its also possible to have a false negative.

 

 

 

17 hours ago, pag794 said:

I was browsing the University of Chicago celiac page, and came across this post.  The math doesn't seem right to me.  If the ttg test is 97-98% specific, doesn't that mean that out of every 100 people who have elevated ttg, 97-98% do have celiac, and 3-4% do not have celiac?  The way they explain it, the specificity would only be 25%, which is definitely not right.  

 

[RMJ]

I used to work with these calculations but it has been a few years!  They are technically correct about the 1 in 4 IF (a big if) you are testing one hundred people in the general population, i.e., people that are NOT going to the doctor because of symptoms but are just chosen randomly off the street.  The 1 in 4 is the positive predictive value and depends VERY much on the population being tested.  If you test 100 people who are suspected to have celiac, the positive predictive value (proportion of positives that are true positives, i.e., that really do have celiac) will be much, much higher.

This is why screening tests in asymptomatic people for disorders with a low prevalence are not always a great idea.  Even scientists often have a poor understanding of these concepts.