Roger Bate, Ed. Butterworth-Heinemann: Oxford, UK. 329 pp. Cloth (1997): ISBN 0-7506-3810-9. $56.95. Paper (1999): ISBN 0 7506 4228 9. $29.95.

A train carrying radioactive waste had begun its trip in New York and was close to its destination in California. As it stopped, the engineer called to a bystander, "Congratulations." "What for?" said the man. "You get to die. We calculated that each person along the route would receive one-millionth of the lethal dose of radioactivity. No one has died yet and you are the millionth person." "But I have received only one-millionth of the lethal dose." "That doesn't matter, it's a question of statistics." (This story is paraphrased from Rockwell's piece in The Scientist, March 16, 1998, p 7.)

What Risk? contains 15 chapters (by 19 authors) arranged in five categories: methodology, science, science policy, commentaries, and perception. It deals in different ways, broadly speaking, with the problems raised by this anecdote. It would make a splendid textbook for high-school students or college undergraduates for a course dealing with pitfalls in extrapolation, unexpected variables, the proper use of statistics, political correctness and absolute safety, evaluation of the scientific literature, and the interplay of science and politics. Each article has an extensive reference list.

Among the specific risks discussed are asbestos, benzene, environmental (secondhand) tobacco smoke, dioxin, ionizing radiation, and carcinogens.

Some general principles emerge. (i) Since all organisms have repair mechanisms against environmental damage, there are thresholds for all damaging agents. Therefore, extrapolation from high dose rates to very low levels does not make sense. (ii) Doses and dose rates should not be confused. (iii) There are very large species differences in response to damaging agents. (iv) Unrecognized variables lurk everywhere. (v) The costs of enforcing demonstrably false standards are huge.

Here are some illustrations. Nilsson's article on environmental tobacco smoke (ETS) concludes that the dangers are about one order of magnitude less than those currently used for regulatory purposes. The errors arise from misclassification of smoking status, inappropriate controls, confounding factors having to do with lifestyle, and, possibly, heredity. Looked at another way, a child's intake of benzo[a]pyrene during 10 hours from ETS is estimated to be about 250 times less than the amount ingested from eating one grilled sausage.

Munby and Weetman's article on benzene and leukemia concludes that the risk of leukemia from nonindustrial exposure is probably zero. The slope of the hypothetically linear dose-effect curve currently in use is too large, the effect at low doses is overestimated, and the linear extrapolation to zero is not justified. The current standard for air quality is about six orders of magnitude below human toxicity levels.

Ames and Gold, in the chapter Pollution, Pesticides and Cancer Misconceptions, give a fine summary of the difficulties with animal cancer tests. "Rodent carcinogens are not rare. Half of all chemicals tested in standard high dose animal cancer tests, whether occurring naturally or produced synthetically, are 'carcinogens'. There are high dose effects in these rodent cancer tests that are not relevant to low dose human exposures... Though 99.9 percent of the chemicals humans ingest are natural, the focus of regulatory policy is on synthetic chemicals." For example, more than 1000 chemicals have been identified in coffee: 27 have been tested and 19 are rodent carcinogens at the high levels at which these tests are carried out.

Dioxin has been called the most toxic chemical known to man. Máller shows that this is not true by any measure. Part of the confusion is based on the fact that guinea pigs are killed by doses thousands of times less than those which affect humans. The chief symptom of dioxin exposure in humans is acne.

The chapter that most surprised me was that by Jaworowski on ionizing radiation. First, the extrapolation of data on the survivors of the Hiroshima and Nagasaki bombings involves dose rates on the order of 5000 mSv/year. For these dose rates, the effects are well established. The average natural dose rate (from the unperturbed environment) is about 2.4 mSv/year. Average additional levels resulting from the Chernobyl accident in Central Europe were about 0.01 mSv/year. So, are there measurable effects at these low dose rates? The linear extrapolation model says yes. But there is no evidence to support this model. Indeed, the author refers to a large body of literature (more than 1000 publications) which is said to show that not only are these low dose rates not harmful, but they are actually beneficial. Examples: people in houses with higher than average radon levels show a lower mortality from lung cancer. The number of birth defects in Hungary in the two years following Chernobyl was smaller than in the years preceding it. At low dose rates, the incidence of neoplasms in irradiated mice is lower than in nonirradiated controls. There are other examples. This literature should be critically examined.

Then there is the question of cost. Funds are limited. Are we spending our money wisely? Ames and Gold give some numbers that suggest not. The average toxin control program costs 60 times more per life-year saved than an injury prevention program and 150 times more than a health care program.

Chemical educators could do much for humanity by encouraging study of the material in this book.