Nationwide premarital testing for infection with the AIDS virus would cost more than $100 million a year to detect about 1,200 infected individuals, far less than one percent of the estimated 1 1/2 million infected Americans, according to a study by researchers at the Harvard School of Public Health.

Each year, false positive results would also show up in about 380 of those tested, leading them to believe they were infected when they actually were not. And about 120 infected individuals would incorrectly test negative, according to the report published today in the Journal of the American Medical Association.

The Harvard study sharply questions the value of a testing strategy that has become a popular political response to the AIDS epidemic. Eighty bills to require premarital testing have been introduced in 35 states. And three states -- Illinois, Louisiana and Texas -- have passed laws that require some form of mandatory testing, according to Caitlin Ryan of George Washington University's Intergovernmental Health Policy Project.

"People perceive premarital screening as an accurate, harmless, low-cost, effective way of stopping the spread of the virus," said Paul D. Cleary, an associate professor of social medicine and health policy at Harvard and the study's principal author. "In fact, those assumptions don't hold up.

"I think compulsory premarital screening is an example of a technological 'quick-fix' that may distract us from development of a comprehensive national program" to control acquired immune deficiency syndrome, he added.

To estimate the cost and effectiveness of nationwide premarital testing, the Harvard researchers made calculations based on the prevalence of infection with the AIDS virus found in male and female blood donors, the accuracy and cost of commonly used blood tests, and the number of Americans who marry each year.

The effectiveness of premarital testing programs depends greatly on how many couples have had sexual intercourse before marriage and on how easily the virus is passed during intercourse, both of which are uncertain. The report pointed out that of the 1,445,000 first-born children born in 1980, only 63 percent were conceived in wedlock.

The researchers estimated that if premarital screening were implemented nationally, more than 3.8 million people would be tested each year at a cost of more than $100 million in laboratory fees and counseling. About 9,000 would test positive on an initial blood test, but only about 15 percent of them would actually be infected.

A second test called the Western blot, administered to those with positive initial results to identify those truly infected, would pick up about 1,200 infected people annually but would be falsely negative in more than 100 other cases, the researchers estimated. In addition, it would be falsely positive for about 380 uninfected people who would then be told they were carrying the virus.

The most common blood tests for AIDS detect chemicals called antibodies, which are manufactured by the immune system of an infected person. Most infected individuals begin to manufacture antibodies within three to 12 weeks after exposure to the AIDS virus, but occasionally a person may not produce antibodies for more than a year after becoming infected and therefore continue to test negative.

A new blood test that detects proteins from the virus may help identify infected individuals who are not yet producing antibodies, but it has not yet been licensed.

The Harvard report criticized premarital testing as an expensive, inefficient policy that diverts funds and attention from more effective measures, such as educational programs.

"Searching for the AIDS virus in the premarital population today is like fishing for shrimp in Lake Erie: You're not likely to find that many," said Dr. Harvey Fineberg, dean of the Harvard School of Public Health. "I favor more widespread testing than we currently have, but in a sensitive, reasonable way that takes advantage of what we know about the disease."