You can use the combination of (imperfect) test outcomes in a composite reference standard (see for example Alonzo & Pepe, Stat Med, 1999). Another option is latent class analysis (see for example Hadgu et al., Epidemiology, 2005).

Article Evaluation of Nucleic Acid Amplification Tests in the Absenc...

In this case, the best you will obtain is the agreement between methods. You can also assume one of then as a gold standard based on theoretical grounds. Then you will be able to see how close the others are in relation to this one. The gold standard can be either the most used/accepted method.

You can use the combination of (imperfect) test outcomes in a composite reference standard (see for example Alonzo & Pepe, Stat Med, 1999). Another option is latent class analysis (see for example Hadgu et al., Epidemiology, 2005).

Article Evaluation of Nucleic Acid Amplification Tests in the Absenc...

You can help to get a better idea about how good any of your tests are by having a good positive control if you know what concentration or level this is at then you can calibrate all the tests more successfully to get a better idea about which gives the best result. High values from one particular test mean nothing if they are not specific ideally you should have a range of negative and positive controls perform a standard curve of the most relevant positive control and relate this back to what you would expect with an ideal assay. The assay that performs closest to expected results should be the best test, clearly this will depend on what you know and have to start with, good luck.

Thanks a lot all for the answers.

I have another practical challenge. Say, I have two tests to compare. I decide to use a golden standard that is the combination of true positive of the two tests. What will happen? I just diferrentiate the performance of the two tests right? But what do the sensitivity and specificity mean here for the practical use of the test? I can only say that one test is better than the other, right? Or can I withdraw other conclusions?

Thanks again.

You should use one of them as your standard. I recommend to use the best according literature, you should take care also the characteristic of your population. there are some supplements in Nature about this topic.

I agree with Sandro Sperandei. The gold standard is the most used and accepted one. lalitha akbilan

A possible and different option is to adopt a fuzzy classification to your reference standard. I'm working with fuzzy classification applied to leprosy diagnostic based on PCR analysis and the results are good so far.