Slashdot Mirror
FDA Approves HIV Home-Use Test Kit
Hugh Pickens writes "The LA Times reports that the Food and Drug Administration has approved the first over-the-counter HIV test kit, allowing people to test themselves in private at home and get preliminary results in less than 30 minutes. The test, which works by detecting antibodies in a swab from the gums, should not be considered final — in trials, the test failed to detect HIV in 1 in every 12 patients known to be infected, and returned false positives in 1 in 5,000 cases. The new at-home test, called OraQuick, will be sold in supermarkets and pharmacies and manufacturer, OraSure, has not said how much the test will cost, only that it will be more than the $18 cost for the professional kit. The federal Centers for Disease Control and Prevention estimates that of the 1.2 million people in the U.S. with HIV, 1 in 5 is not aware of the infection and that a disproportionate number of the 50,000 new cases of HIV each year is linked to people who have not been tested. Chip Lewis, a spokesman for Whitman-Walker Health, which provides AIDS care in Washington, says at-home testing could reach some people who didn't want to go to a clinic but removing medical professionals from the process could cause problems. 'It's not like a home pregnancy test,' says Lewis. 'You need really a lot of information about how to read the test, how to use the test properly.'" Back in May, we reported that a panel of FDA experts recommended approval of an over-the-counter HIV test.
... lest you guys start thinking that this kit is a heavenly sent, that you guys will be 100% protected ...
This test kit is only 92% accurate
While 8% does not seem to be a big number, it still matters in this case for AIDS is still incurable
The test's accuracy is much higher than 92%. The test has 92% recall (it will correctly detect 92% of the true positives). In determining the accuracy, you need to take into account all the people who don't have HIV (which it will correctly detect 99.98% of the time). Based on the CDC's numbers, about 1 in 250 people in the U.S. have HIV, so the accuracy of this test would be (249/250)*99.98% + (1/250)*92% = 99.95%. The precision here (the probability that a positive returned by the test is a true positive) is the probability of a true positive detection over the total probability of a positive test result, or (1/250)*92% / ((1/250)*92% + (249/250)*0.02%) = 95%. In other words, if the test says you have HIV, there's a 95% chance it's correct. Doing the same for a negative result, you'll find that a negative result is correct 99.6% of the time.
Your point that the test fails to correctly detect 8% of the true positives is a reasonable one, but accuracy is not the metric you should be using to evaluate. To better illustrate why accuracy is a terrible metric to use, consider a test that always returns "no". Since 99.6% of people do not have HIV, the test is 99.6% accurate, yet totally useless (0% recall and undefined precision due to no positive results). Precision and recall are what you should care about.
TL;DR: