How reliable is drug testing? Millions of urinalysis tests are administered every year -- by the military, civil aviation police, companies, newspapers -- but debate about their accuracy is just beginning.

Critics -- civil libertarians, labor unions and others -- say the tests are unreliable and an invasion of privacy.

The manufacturers of the urinalysis tests claim accuracy rates as high as 95 percent, but that figure has been disputed by physicians and scientists who say the actual rate varies depending on the size and nature of the group being tested and the handling of the test samples.

Critics also say the two most widely used urine tests are capable of registering false positives, that is, indicating a drug is present when it isn't. In one study cited by the civil liberties union, 25 percent of urine tests showing the presence of cocaine were found to be false after further testing.

The tests may also confuse one compound with another. Poppy seeds have shown up as opiates, and certain cold medicines as amphetamines. Antibiotics have registered as cocaine, aspirin as marijuana.

The tests cannot distinguish between occasional users and addicts. Nor can they measure job impairment. And because drugs are metabolized at different rates in different people, a heavy marijuana user may test positive for months after quitting; amphetamines, by contrast, pass through the body so quickly that they may not show up one day later.

Finally, even the best testing procedures can be dangerous in the hands of sloppy or overworked lab technicians. (On at least one occasion, according to sworn testimony by an Air Force major, test results at an Army lab had to be discarded because technicians were discovered to have been heating up breakfast tacos in a glassware drying oven used for tests.) In 1984, Army and Air Force labs were found to have improperly processed 35,000 urine specimens.

At Georgia Power, pre-employment urinalysis tests are required for all job applicants, from stock clerks to managers; anyone who fails may reapply in six months. People already on the job are also tested for fitness for duty. This is done at a supervisor's discretion, and may be prompted by a hot-line call, or an employe who is clearly impaired.

Does everyone who is turned in on the hot line at Plant Vogtle take a urine test? Yes, says Charles Whitney, unless a manager can make a convincing argument otherwise. "I'd want them to tell me, 'Chuck, I've known this person 20 years, we're next-door neighbors, I go to church with him and I know for a fact that he doesn't use drugs.' "

The company says internal polls show most employes support the testing program. The program, says Whitney, is rigorous because it has to be. "We decided, 'We're going to do whatever it takes to provide a drug-free environment,' " including the use of drug-sniffing dogs, urinalysis, searches of employes' lunch boxes.

Jerry L. McHan, the toxicologist in charge of the suburban Atlanta laboratory that processes many of the Georgia Power samples, says his company follows procedures designed to make the tests almost foolproof.

The lab has carefully controlled the chain of custody, meaning the sample is taken, sealed, signed by the donor, and held in a secure area until picked up by a courier. Any positive test results, he says, are verified with two additional backup tests, using different techniques and equipment. The chances of an error occurring in such a system, he said, is 1 in 10,000.

Even then, there is an additional safeguard. Georgia Power keeps one-half of the original urine sample in reserve so that the second half of the sample may be shipped off to a different lab for additional testing.

"This is not an absolutely foolproof system," says McHan. "But if you've got something that's going to cure a certain disease in 90 out of 100 people, but in the other 10 percent it makes them ill, are you going to use it? That's what we're talking about with drug testing. You don't just throw the whole thing away."