Re: Source for 99+% accuracy claim 24 November 2004
Previous Rapid Response Next Rapid Response Top
Nicholas Bennett,
Infectious Disease Postdoc/Clinician
Department of Pediatrics, University Hospital, Syracuse NY

Send response to journal:
Re: Re: Source for 99+% accuracy claim

I don't think any ELISA has been claimed to be 99% accurate, except through those trying to disprove the claim! The ELISA however has been shown in several studies (of various assays) to be around 95-100% sensitive, depending on how old the paper is (older tests were less good). I have certainly seen nigh on 100% accuracy for combined HIV ELISA/Western Blot when compared with culture and PCR in several studies. Depending on how you judge the data, there were no inconsistencies (at least two assays agreed) or the culture/PCR was less than 100% sensitive (they missed occasional samples) or the serology was less than 100% specific (they mislabelled HIV negatives as positive). For the small series I've looked at these errors are under 1.3% of the sampleset. [1]

Seropositive: 409. Positive by culture AND PCR 402. Positive by culture OR PCR 409.

Seronegative: 131. Positive by culture OR PCR 0.

These give a 98.7% correlation between serology, virus culture and PCR detection. A startling alignment that is superior to many (if not most) other infectious disease diagnostic tests.

Many of the tests currently marketed are run against reference samples and the accuracy claims are usually verifiable in the package insert (or on the product website!). Such standardisations typically quote high-90's to 100% senstivitity and specificity.

One thing that should be noted about the global variaton in Western Blot interpretation, as mentioned by Dr Flegg. The same WB test is NOT used worldwide (if indeed they use the WB, most confirmatory tests these days are similar to but not actually Western Blots). If the test contains different analytes then one cannot expect it to display the same patterns as every other test. The interpretation MUST be made on a per-test basis, and since each country prefers to use different test kits the interpretation will have to vary.

For any one test it should be relatively easy to pull the necessary accuracy data from the web, and for several of the "false positive" reactions listed by Mr Tyler I would personally be happy to sign off on them having done the relevant research to check into the claims. With a decent patient history and exam many if not all of those other factors should be excludable as well, saving me a trip to the medical library to verify the source! Will this mean the test is fool-proof? Of course not! Dr Flegg neatly points this out, but as regards the alleged "false positive" causes of the HIV test, a rare handful are actually meaningful, if even true.

Nick Bennett

njb35@cantab.net

1. Jackson et al J Clinical Mole Bio 1990 pp 16-19

Competing interests: None declared