TWO YEARS AGO, the Texas Office of the Occupational Safety and Health Administration discovered 18 cases of brain cancer at the same plant. Investigators found that in the general area of the plant this relatively rare form of cancer appeared at the same rate as in the United States as a whole, but it appeared twice as often among the plant's workers. Here, however, the clues stopped. No pattern emerged from the types of jobs the dead worker had held. The plant's inventory included at least 10 known or suspected carcinogens, but none could be identified as the responsible one. The 18 deaths were spread evenly over 25 years.

Meanwhile, other plants were being investigated by government scientists and by employees of the petrochemical companies, many of which have active and well-established departments of epidemiology. By mid-summer of 1980, when the first news stories of an "epidemic of brain cancer" appeared, two directly conflicting sets of results had been obtained. Four different government-sponsored studies found that petrochemical workers had twice the risk of contacting brain cancer as did non-workers.Four other studies -- by Dow Chemical, Du Pont, Gulf Oil and a consortium of British refineries -- found no evidence of increased risk. None of the studies pointed toward any type of job or any particular chemical.

In an effort to reduce the confusion, OSHA asked a leading epidemiologist to organize a meeting that would includeboth sides as well as the best available neutral experts. The group could not find a single simple explanation for the conflicting results, but could find many possible sources of the confusion and dozens of issues to argue over.

This is a perfect illustration of the toughest problem facing health and safety regulation and those who would reform the regulators: the weakness of the only available tool -- epidemiology -- in detecting health threats, even potentially devastating ones. It took nearly a decade, after all, for a concensus to emerge that smoking causes lung cancer, despite the fact that there were millions of smokers to study.

In all likelihood, it will be years before anyone can perform a cost-benefit study of any kind of strict analysis of occupational brain cancer. In the meantime, workers are understandably scared and their representatives will, rightly, demand some action. Industry critics are likely to argue that all the effort would be better directed to other, larger risks, and they could well turn out to be right -- or wrong. And government regulators -- despite every desire for objectivity -- will be forced to rely on instinct, values and personal judgement.

A great deal can be done to make government regulation more sensible, less expensive and more effective. But sad truth is that, for many years to come, uncertainty and confusion will necessarily remain a large part of the process. There is just no way around it.