A chart on last week's Cutting Edge page may have left the impression that the military's screening program for AIDS is less accurate than it actually is. The number of positive readings expected to be false-positives was based on the average proficiency of laboratories across the country. In the program actually used by the military, 0.5 percent of positive findings in male recruits, and 1.3 percent in female recruits, are false-positives. (Published 11/3/87)
Laboratories testing blood for evidence of AIDS have such a high error rate that in some low-risk groups, nine out of 10 positive findings would probably be wrong, a new government analysis has found.
Such a high rate of these so-called "false positives" -- in which people without the AIDS virus wrongly appear to have it -- would cast grave doubts on the reliability of massive screenings.
The analysis was done by the congressional Office of Technology Assessment and presented last week at a hearing of the subcommittee on regulation and business opportunities of the House Small Business Committee. It was based on recent results of proficiency testing of approximately 700 U.S. laboratories by the College of American Pathologists (CAP).
Rep. Ron Wyden (D-Ore.), chairman of the subcommittee, said in an interview that he found the labs' error rates "mind boggling."
In addition to the false positives, the data suggest that labs may be reporting falsely negative blood test results on as many as 10 percent of individuals who are really infected with the virus.
"Both the false positives and the false negatives have frightening social implications," Wyden said. "False negatives are people who think they're well and spread the virus. False positives have the potential of just causing social chaos," because uninfected individuals might base decisions about marriage, childbearing and careers on an erroneously positive test result.
The standard blood tests currently available to detect infection with the AIDS virus are the enzyme immunoassay (EIA or ELISA) and the Western blot. The EIA, performed initially as a screening test, detects protein antibodies produced by an infected individual's immune system against particles of the virus, known as human immunodeficiency virus, or HIV.
The more accurate and expensive Western blot is done as a confirmatory test if the EIA is positive. Poor-quality Western blots are the main reason for the laboratories' high false-positive and false-negative rates.
Because donated blood is discarded even if positive only on the initial EIA test, these findings do not suggest that infected blood is any more likely than previously believed -- 1 in 10,000 -- to pass through undetected.
When they are perfectly performed, both the EIA and the Western blot are highly sensitive, providing positive results in infected individuals in 99.6 percent of cases. They are also highly specific, giving negative results in uninfected persons in at least 99 percent of cases. But the pathologists' proficiency testing program, in which laboratories were sent a series of "unknown" blood samples to test, showed that the error rates in laboratories around the nation are considerably higher than these ideal figures, especially for the technically more complicated Western blot.
The CAP's proficiency testing of laboratories during the period from July 1986 through June 1987 showed that the overall false-positive rate for the Western blot was 4.7 percent. The theoretical false-positive rate, if the test is properly performed, is 0.5 percent. The laboratories' overall false-negative rate was 9.3 percent, compared with a theoretical false-negative rate, if the test is properly performed, of only 0.4 percent.
Some laboratories achieve performance standards close to the ideal rates, while others have even higher error rates than the averages cited in the CAP data. But consumers having the blood tests, and doctors ordering them, often have no way of judging an individual laboratory's performance, the subcommittee was told.
In large-scale AIDS testing programs, false-positive results make up a greater proportion of all positive results when the group being tested is at low risk of infection with the virus. In contrast, false-negative results assume greater importance when a high-risk group is tested. The OTA analysis used the laboratories' performance record to predict the outcome of testing in six different populations for whom some estimate of infection with the AIDS virus was available.
For example, in a program to test 100,000 high-risk patients at a sexually transmitted disease clinic, the OTA predicted that 984 of an estimated 10,000 infected individuals would be missed because of a falsely negative blood test, while 72 uninfected people would test falsely positive.
On the other hand, if 100,000 blood donors were tested in Peoria, Ill., a group with a very low frequency of infection, 80 of the 89 positive blood tests, or 90 percent, would be false positives, the OTA analysis predicted.
Other experts testifying before the subcommittee echoed the concerns raised by the OTA report. Different laboratories use different criteria to identify a positive Western blot, and national standards have not been established for the test's performance, interpretation or quality, said James R. Carlson, director of the AIDS Virus Diagnostic Laboratory at the University of California at Davis.
Of 19 commercial laboratories sent "blind" samples by the U.S. Army to test their performance of the Western blot, 10 failed the test, some of them repeatedly, said Col. Donald S. Burke, who directs the Army's HIV screening program.
Most doctors who order AIDS blood tests assume they are accurate and know little about variations in laboratory performance, said Patricia Watson Martin, director of product development at Epitope Inc., a clinical laboratory and diagnostics firm based in Portland, Ore. "Because the opportunities for making profit from HIV antibody testing are so vast . . . we must act quickly to put controls into place to insure the quality of testing," she told the subcommittee.
Wyden said he plans to press for stronger federal oversight of the rapidly growing HIV-testing industry. "You can be pro or con testing, but as a prerequisite to a national testing policy, we're going to have to deal with this accuracy issue," he said.