By Dana Hawkins

Drug tests don't lie, people do. That's what Michelle Dunson used to think back when she administered urine tests for a temp agency near Toledo, Ohio. Sometimes when she told applicants they'd failed, the response would be wide-eyed silence, then tears and denial. She offered them tissue, but little sympathy.

Dunson later took a job with Whirlpool, where she was injured two years ago. When she returned, she tested positive for an opiate and was fired–even though a note from her doctor outlined her prescriptions for a nonnarcotic painkiller, which she believes threw off the test. She has a wrongful-termination suit pending against Whirlpool, which stands by its testing. "I feel tremendous guilt now when I think of those who came to me nearly hysterical, saying they did not do drugs," says Dunson. "I think: My God, at least a few of them were probably telling the truth."

Dunson can be forgiven for believing drug test results were the gospel truth. The tests are often heralded as infallible, and many private and government employers, along with school principals and judges, put their faith in them. Half of major U.S. firms now test their employees, and more than 500 school districts have screening programs. But reliably picking up drug traces that linger days after a user's last high, while ignoring contaminants and similar-looking compounds in medicine and food, is a tall order for even the best technology. In the real world, technical glitches can mean mistakes–so-called false positives.

Wide net. For employers who test to deter drug use, which they say leads to a safer, more productive workplace, so-so accuracy might be enough. But it comes at a high cost in reputations and livelihoods for those falsely accused. "Innocent people are being mislabeled because of unreliable products designed to cast a wide net," says Steven Karch, a medical researcher and author of Karch's Pathology of Drug Abuse.

The error rates are likely to rise. While the traditional method–lab-based urine testing–can be highly reliable when done carefully, the fastest-growing segment of the drug-testing industry consists of newer and often more error-prone tests. On-site urine testing appeals to private employers because it gives results in minutes and can cost as little as $3 per screening. Yet studies of the cheapest of these tests, designed to pick up a single drug type such as amphetamines, show they give the right answer as little as one third of the time. Other new tests, which rely on lab analysis of hair and sweat, can be a powerful probe of drug use–but can also be fooled by stray drug molecules from the environment. All these tests may get a major boost from the proposed new federal guidelines for drug testing, expected to be published later this year. The rules, which serve as the model for many state and private employers as well, will allow new tests, including on-site urine screens and lab-based sweat patch and hair tests, to be used for screening 10 million federal workers and contractors.

Currently, laboratory urine testing–the "gold standard" test–is the only type allowed for federal employees. Samples are probed with antibodies for THC (marijuana), cocaine, opiates, PCP, and amphetamines, and checked to be sure they aren't doctored or too dilute. When a worker tests positive, the federal program mandates a second, more accurate confirmation test called gas chromatography-mass spectrometry, GC-MS, along with an interview with a medical review officer to be sure that another substance isn't causing a false positive. The sample's chain of custody is carefully documented as well.

Even GC-MS is not perfect, however. The test works by extracting and heating molecules from a sample and using an electric field to separate and identify them. Kent Holtorf, a physician and expert on drug-testing accuracy, says he got 1,500 calls last year from people asking for help with what they said were lab errors. He says that when a lab uses GC-MS to identify the entire range of molecules, it is 95 percent to 99 percent accurate. But Holtorf says labs don't always use the equipment to its full advantage. "Accuracy rates are going down, not up, as employers contract with the lowest-bidding lab," he says. It's cheaper to use GC-MS to look only for a few fragments of the drug molecules, which raises the risk of mistaking legitimate medicines, herbs, and foods like poppy seeds for illegal drugs.

The drug-testing industry dismisses these concerns. "People always come up with cockamamie stories about how their test result is wrong," says Paul Rust, a vice president at Quest Diagnostics. "But the GC-MS process is 100 percent accurate." Not quite, says Michael Martin, who is HIV-positive and takes the prescription drug Sustiva. The Worcester, Mass., resident says he had to take a pre-employment urine test after he was offered a job at Sears. He presented a doctor's letter saying his medication could cause him to test positive for pot. When just that happened, Sears rescinded the offer. "I want to work so badly I can taste it," says Martin. "But I can't stop taking my prescription just to test clean."

Sears, which is investigating Martin's case, says its policy is to confirm positive results with a second test. But at many companies, employees who fail the initial screen get no second chance. "Most employers, even those who use a lab, don't confirm positives on pre-employment tests," says David Evans, director of the National On-Site Testing Association. "It wouldn't be a financially smart thing to do." And, in most cases, it's perfectly legal not to. There are no regulations, other than a smattering of state laws, regarding testing of nongovernment workers.

Nor has there been tough scrutiny of the newer drug-testing technologies. "The lack of scientific studies on these devices is most surprising," says Amanda Jenkins, coauthor of the new book On-Site Drug Testing. The U.S. Food and Drug Administration will approve a test based solely on a company's own studies. Still, drug-testing veterans, some independent studies, and even the makers' own data suggest that three leading "alternative technologies" can be disturbingly error prone.

On-site testing. Like lab assays, these tests work by combining urine with antibodies. If the test is negative, a line appears on a test strip. But seeing the lines can be a judgment call, and some products are especially hard to read. "We stopped using a few tests because some people would see a line where others wouldn't," says Brian Walters of Premier Drug Testing in Russell Springs, Ky. In one study, 17 common on-site tests were used to analyze specimens; most contained either no drugs or trace amounts. The accuracy ranged from 52 percent to 82 percent.

Sweat patch. In this test, a Band-Aid-like patch attached to the skin collects sweat for up to seven days and is later lab-tested for drug residue. It's tough to cheat–if the patch is removed, it can't be reattached–and the test is often used in parole, probation, and child-custody cases to determine whether a user has been rehabilitated. But studies by the U.S. Naval Research Lab and the patch manufacturer itself showed that drug molecules from outside sources–such as clothes or other people–can penetrate the patch and trigger a false positive. The sole maker, PharmChem, says both studies exposed the patch to unrealistically high levels of contamination. "Those were just not real-world situations," says Neil Fortner, PharmChem's chief scientific officer. But a federal court in New York recently ruled that the sweat patch "is susceptible to outside contamination."

Sheryl Woodhall thinks that's what cost her custody of her two youngest children. She wore the patches to prove she was no longer a methamphetamine addict. But after she flunked seven of eight tests, her kids were placed permanently in foster care, and she was forbidden to contact them. During the same period, Woodhall says she tested negative on dozens of urine screens, done under observation so she couldn't cheat.

Hair tests. Lab analysis of 1.5-inch long strands of hair cut near the scalp can give a drug history covering 90 days, compared with only a few days for most drugs in urine tests. Psychemedics, the largest hair tester, says 140 schools and 2,300 corporate clients use its services. But hair testing is also the most controversial of the new technologies because of concerns that it is discriminatory and can be thrown off by contaminants. Several studies by the National Institute of Drug Abuse show that some drug molecules, whether ingested or picked up from the environment, have an affinity for the pigment melanin and bind more strongly to dark hair than light. "If two employees use cocaine, the blond might barely test negative, and the other will get caught," says Robert Stephenson of the Substance Abuse and Mental Health Services Administration.

Raymond Kubacki, president of Psychemedics, dismisses concerns about discrimination and says a series of washes removes contaminants from hair samples. "You could be in a crack den and you won't test positive." But Ronnie Jones, a Boston police officer for 20 years, blames contamination for a marginally positive result for cocaine he got on a hair test last March. Jones–nicknamed "the deacon" for his Bible reading and clean living–submitted a second hair sample the same day, which tested negative. Although the company downplays the contamination risk, Kubacki says it has begun testing for drug metabolites, the fin- gerprints left after drugs are processed by the body, rather than just the parent substances.

So far there have been few technical challenges to another new test, which relies on analysis of a saliva sample. Because drugs show up in saliva more quickly than in urine, hair, and sweat, the test may reveal whether a person is currently high, making it useful for post-accident testing. But it is just beginning to catch on. And like the other tests, it will get its real trial in the real world, where jobs and reputations will hinge on the right answer.