If anyone remakes "Erin Brockovich," this is a scene I want to see. A scientist launches a study to determine the toxicity of hexavalent chromium, the drinking-water contaminant at the center of the lawsuits Brockovich spearheaded. The study will be a meta-analysis, combining existing individual studies to, he says, produce more-authoritative conclusions. Some of the earlier studies measured rates of lung cancer among pigment-factory workers exposed to airborne chromium, so it makes sense to include them. But the scientist is working for industry, so he chooses his other studies carefully: he includes those that assessed all forms of cancer among residents who drank chromium-laced water. Only the workers, not the residents, had increased rates of lung cancer. No surprise there: only inhaled—not ingested—chromium can cause lung cancer. Since there are many more residents than factory workers, the data showing no rise in lung cancer swamp the large numbers of lung cancers in the workers. Thanks to this sleight of hand, the study—which happened in real life, not a movie—concludes that chromium "is only weakly carcinogenic for the lungs," giving the chemical a nice coat of whitewash.

That science can be bought is hardly news to anyone who knows about tobacco "scientists." But how pervasive, effective and stealthy this science-for-hire is—as masterfully documented by David Michaels of George Washington University in his new book, "Doubt Is Their Product: How Industry's Assault on Science Threatens Your Health"—will shock anyone who still believes that "science" and "integrity" are soulmates. In studies of how toxic chemicals affect human health, Michaels told me, "It's quite easy to take a positive result [showing harmful effects] and turn it falsely negative. This epidemiological alchemy is used widely."

The alchemy is all in how you design your study and massage the data. Want to show that chemical x does not raise the risk of cancer? Then follow the exposed population for only a few years, since the cancers that most chemicals cause take 20 or 30 years to show up. Since workers are healthier than the general population, they start with a lower death rate; only by comparing rates of something the chemical is specifically suspected of causing—a particular lung disease, perhaps—can you detect a problem. Or, combine data on groups who got a lot of the suspect chemical, such as factory workers, with those who got little or none, perhaps their white-collar bosses. The low disease rates in the latter will dilute the high rates in the former, making it seem that x isn't that toxic. All these ruses have been used, delaying government action on chemicals including benzene, vinyl chloride, asbestos, chromium, beryllium and a long list of others that cause cancer in humans. "Any competent epidemiologist can employ particular tricks of the trade when certain results are desired," Michaels writes.

The strategy of manufacturing doubt was pioneered, of course, by the tobacco industry. Tobacco-funded studies threw up enough red herrings—"Psychosocial, Familial Factors May Have Role in Lung Cancer," "Lung Cancer Rare in Bald Men"—to confuse consumers and delay the warning label on cigarettes. Doubt is still a crucial product for any industry fighting regulations that would protect workers and consumers from exposure to toxic compounds. Now some industries argue that it would be wrong to reduce allowable exposure limits until science can prove exactly what level is safe. Better to keep the limit at, say, 50—even though that level kills workers—than cut it to 10, when 12 might be OK.

This is all very big business. "Product-defense firms" have sprung up to spin the science and manufacture doubt—proudly. One boasted on its Web site of persuading the Food and Drug Administration to let an unnamed drug stay on the market for "10 additional years of sales" before the FDA banned it for harming people. This science-for-hire is aided and abetted by what Michaels calls "vanity journals," established by the firms. By naming like-minded consultants as the journal's peer reviewers (who decide what gets published), they can get anything that advances their cause into "a peer-reviewed journal"—magic words to jurors, regulators and reporters.

Make no mistake: raising doubt has run up the body count. By the early 1980s, for instance, studies had shown that children who took aspirin when they had a viral infection such as chickenpox were at greater risk of developing Reye's syndrome, which damages the brain and liver and is fatal in about one case in three. Desperate to protect their market, aspirin makers claimed the science was flawed, called for more research (a constant refrain), and ran public-service announcements assuring parents, "We do know that no medication has been proven to cause Reye's." The campaign delayed by years the requirement that aspirin carry a warning label about children and Reye's. In the interim, thousands of kids developed Reye's. Hundreds died.

Michaels's argument would be stronger if he included cases in which early studies overstated risk, to show that industry is sometimes right about how dangerous a compound is. (Sodium nitrite in hot dogs?) Still, his book should leave no doubt that science, and scientists, can be bought—the public welfare be damned.