According to what I know for accuracy of diagnostic tests meta-analysis study, we need TN, TP, FP, FN and by having these items, sensitivity and specificity can be obtained.
What should be done if these values are not stated in studies and only sensitivity and specificity of test are expressed?
I check and Metadisk software necessarily wants this ( TN, TP, FP and FN ) information.
Do you know any other software that can do meta-analysis studies with this sensitivity and specificity informations?
Cockran review has the following that use TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative. It may give you examples of how to do:
Article Galactomannan detection for invasive aspergillosis in immuno...
Article Rapid, point-of-care antigen and molecular-based tests for d...
As the sensitivity and specificity are composite measure utilising the T,P TN, FP and FN rates you might be able to back calculate these variables using other information in the article. if you can nit access thee data then you might consider contacting the corresponding authors of the relevant papers and asking them directly for the data
Dear Shin;
Thank you for your answer.
Unfortunately, I have some studies that all of them did not report TN, TP, FN, FP values and only expressed sensitivity and specificity.
And I do not know how these can be used in meta-analysis!!!
Shin Murakami
Dear Marsha
Thank you for your answer.
Unfortunately, All my studies are like this (did not report TN, TP, FN, FP values and only expressed sensitivity and specificity) and it is not possible to make contact and get information from all of them.
Is this the only possible solution?
Thanks.
Marsha Y Morgan
I have some ideas, made one DTA review.
First, above mentioned back calculation is useful but you have already prevalence rate.
It is possible in Calculator of RevMan.
Example; when you have prevalence 10% and sensitivity 90% of 100 subjects. 10 subjects have a disease out of 100. Only 9 subjects can be detected in test (TP). Remaining 1 subject is FN. So on.
Second, this calculator is the another option.
https://www.google.com/url?sa=t&source=web&rct=j&url=http://araw.mede.uic.edu/cgi-bin/testcalc.pl&ved=2ahUKEwjfk_7didLzAhW3gP0HHf6mC7cQFnoECBEQAQ&usg=AOvVaw2vlftJsNX-DluoOcvqbjku
Dear Nyan Min Aung;
Thank you for your answer.
Unfortunately, these studies have not reported prevalence rate.
In addition, I haven't information want in this site. I have only the sensitivity and specificity value of biomarkers that I want to study.
I thought that TN, TP, FN, FP values could not be obtained from sensitivity and specificity values. But I have seen studies in my field that claim to have extracted such information with this condition. In other words, in the articles, there are only sensitivity and specificity values. But in the meta-analysis study, it is stated that the 2 × 2 tables of TN, TP, FN, FP have been extracted.
Nyan Min Aung
In case of biomarker the cut-off point which defined disease condition maybe reported.
From that point you may find the variables.
Best of luck.
Dear all;
I realized that I can't calculate the TP, TN, FP, FN only with sensitivity and specificity values and I need more information.
I found data that was like prevalence rate (not exactly but similar meaning) and with using it, I was finally able to calculate these quantities.
I appreciate all of you.
Nyan Min Aung Marsha Y Morgan Shin Murakami
Dear Khalid Saeed Khan;
Thank you for your answer.
Khalid Saeed Khan
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis