1
$\begingroup$

When you say the Venereal Disease Research Laboratory test (VDRL) has a specificity of x% and sensitivity of y%, what are you comparing to? What is considered as the truth?

Because, Wikipedia says, even a gold standard needn't be perfect. It is just the best test available(?). It too has its specificity and sensitivity (<100%)

So, my question is, when we say that the VDRL has a 75% sensitivity, what are we comparing it to? Is it the clinical diagnosis of syphilis according it the case definition? Wouldn't it be much less accurate?

$\endgroup$
  • 1
    $\begingroup$ This is usually meant in a clinical setting: So I think this is about detecting 75% of the infections. $\endgroup$ – Chris Oct 18 '16 at 15:07
  • $\begingroup$ @Chris essentially. See my answer for a bit more (I tried to keep it short). Specificity is also related to how specific the reagents are for their intended targets, vs. binding to related proteins/hormones/chemicals/whatever the test is measuring. Some tests are yes/no readouts, but I focused more on quantitative tests, which tell you how many virus particles per ml, for example. $\endgroup$ – MattDMo Oct 18 '16 at 15:25
2
$\begingroup$

Specificity and sensitivity are terms used in the pharmaceutical and in vitro diagnostics (IVD) industries that have specific meanings relating to how the particular assay in question was validated. Generally the values are being compared to spiked standards in matrix using recombinant protein or purified virus or whatever the antigen is (depending on how the test operates) where the analyte concentrations are known. With a certain number of titrations and assay repeats, along with a bunch of statistical calculations, you can then compute your expected and actual recovery, as well as values like the LLOQ (lower limit of quantitation) which gives you your sensitivity (we can reliably detect 1e4 viral particles per ml, for example).

Specificity is calculated using related antigens, such as proteins from other types of bacteria that are of varying degrees of relation to syphilis, or perhaps of different serotypes that aren't claimed to be detected by the test, or whatever. Again, you spike known amounts of these materials into your testing matrix (blood, sputum, urine, semen, whatever) and see what your test returns. So, say I have a virus with 2 serotypes (A and B) that I claim recognizes A specifically. I spike in perhaps equal amounts (for example, 100 units) of A and B, then run the A-specific quantitation assay. The readout might then say the sample only had 75 units of A, meaning 75% specificity of the reagents.

For more reading, search for "clinical assay validation" or something similar. There are a ton of papers out there giving guidance for particular types of assays, as well as regulatory agencies' interpretations of the relevant laws covering the tests. There are also many books, reviews, etc. if you're just looking for an overview. It's a huge field, and I've just given a very basic description to avoid writing pages and pages.

$\endgroup$
  • $\begingroup$ Thanks! How would these then account for the biological variation in the population? The same amount of bacterial load in one person may test positive but not in another. Do these terms not talk of that? Because, in syphilis there is a special term called biologically false positives (BFP) for VDRL, are they just a subset of normal FPs or excluded from it? $\endgroup$ – Polisetty Oct 18 '16 at 15:34
  • $\begingroup$ @Polisetty I don't know the specifics of this test, other than what is on Wikipedia, so I can't answer directly. I do know that the VDRL does not test the levels of the spirochete directly, rather it looks for levels of anti-cardiolipin antibody that increase after syphilis infection, and apparently have been shown to have a good correlation with viral load in the absence of confounding conditions like lupus. There are other direct tests that identify the bacterium specifically - they are used for diagnosis. I'm not sure how the BFPs (such as lupus cases) are differentiated from other FPs. $\endgroup$ – MattDMo Oct 18 '16 at 15:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.