Document Lg3rvVy1ZEypM0vLRE4rvjom3

`There is no good way to tell when there are Is Risk Assessment Conservative? overestimates or serious underestimates of the risk. If we could tell, we would fix the risk assessments." of Among regulators and the regulated, the estimate of the number of tumors likely to Crouch adds a couple of caveats. First, the `g dogma is that the risk assessment procedures arise from a given dosc. When they repeated findings are for the mid-dose range, and not le typically used by the Environmental Protec- the procedure, this time using low-dose for the extremely low doses that regulators .Y tion Agency (EPA) and other federal agen- data, the analysis yielded a cancer potency are typically intcrested in. Nonetheless, the If tics arc conservative-that is, they overesti- factor nine times higher than the first analy- authors conclude that "there is no reason to mate the risk a chemical poses, ohen by a sis. Thus, the model would substantially assunic that the uiiderestimation is any less I t rubstantial amount. EPA officials have made underestimate the actual cancer risk. frequent at doses much closer to zero." Nor ie ruch assertions repeatedly in the past few Is the one-hit model this misleading for does the study address the nagging question .e weeks in the debate over the pesticide Alar. other chemicals?To find out, they analyzed of whether it makes sense to extrapolate 11 The bias is intentional. Risk assessment is data from 1212 bioassays of 308 different linearly to low doses. But ifyou are going to Y IO uncertain, based as it is on extrapolation chemicals. They fitted the one-hit model to use that model, says Crouch, "the paper says e born animal data to humans and from high the zero and high doses and then looked at here are some things to watch out for." C closes to low, that federal regulators decided how well it predicted the observed cancer "The message to regulators is they should 1 to err on the side of safety to protect the incidence at the mid-dose range. modify what they say when they talk about r public health. That approach has nonethe- In a small but significant number of risk. I think they should be much less certain r Iss been roundly criticized by those who say cases-more often than would be expected in their language than they are now." it carries protection to the extreme, usually by chance-the model underestimates risk, Bailar hopes the paper will halt efforts to f at the expense of industry. they found. This underestimation occurs in make risk assessnientprocedures less conser- ) Now a study challenges that dogma. It about 2.5 to 4% of die cases. The model also vative, as some people are advocating. "I concludes that the cancer risk assessment overestimates risk in about 5 to 7% of the would prefer to see a move in the other f model typically used to estimate effects a t cases. The upshot, says Crouch of Harvard direction, or at least, to hold the current hw doses-known as the "one-hit" model- University, is that "in some fraction of cases line." Granted, he adds, there is conserva- I in most cases is not conservative at all. we may be making a substantial underesti- tism built into other risk assessment steps, Moreover, in a small number of cases, it mate of the risk, and we have no idea how but there is "anticonservatism" as well. `We substantially underestimates risk. The paper, big that underestimate is." just do not have evidence that our risk by John Bailar, Edmund Crouch, Rashid Nor is it clear when this will occur, adds assessments for chemical hazards are sub- Shaikh, and Donna Spiegelman, was pub- Bailar of McGill University and the U.S. stantially conservative." lished in the December issue of Risk A d y - Department of Health and Human Services. m LESLIE ROBERTS rir. What tipped off the investigators was vinyl chloride. Unlike most chemicals, there pulsar, PuIsar, Where A t Tho% Pulsar? is an abundance of experimental data on the cancer-causing effects of vinyl chloride at a It has been 2 months now since an interna- pulsar since 18 January. Presumably, of wide range of doses. This allowed them to tional team of astronomers announced their course, the pulsar is sitting in the middle of a xc how well the model predictions stack up detection of a furiously rotating pulsar at the rapidly expanding shell of debris from the against actual experimental data. The an- heart of Supernova 1987A-and no one has original explosion. So perhaps it was shining swer, in this case, is not well. seen it since. Supernova watchers are lek through a thin spot on 18 January, and has Most chemicals have been tested at just feeling as skeptical as they are intrigued. been lost behind the clouds again ever since. two or three doses: at close to the maximum On the one hand, the original observa- But 2 months? `There will be fewer excuses tolerated dose, at about half the maximum tions have held up well under ouuide scruti- as time goes by," says Kirshner. [olerated dose, and at zero. Risk analysts ny. "It really looks like a signal," says Robert Be that as it may, the theorists have lost then rely on a model, often the one-hit model, to estimate the effects at the very low doses to which humans are likely to be exposed. Both the one-hit model and EPA's Kirshner of the Harvard-Smithsonian Center for Astrophysics. `There's no `if' about it." Extensive computer analysis of the data shows that, on the night of 18 January, the no time in speculating about the implications of the pulsar if it is real. For example, its rotation rate puts it on the ragged edge of flyingapart, so its very existencewould force linrwized multistage model, which the in- supernova was flickering ever so faintly at vestigators lump in the same category, as- 1968.629 rimes per second, or more than sume that there is no threshold below which twice as fast as any other pulsar ever seen. a carcinogen docs not act and that the Equally astonishing, the rate gently rose and rclationship between dose and response is fell as though the pulsar were being tugged approximately linear at low doses. back and forth by the gravity of some kind Bailar, Crouch, and their colleagues tried the standard risk assessment procedure on vinyl chloride, for which the actual dosercsponse curve is known: it rises sharply at low doses and then flattens out near the high doses. They looked at how many tumors arose from die high and moderate doses, plugged those data into a computer, and canie up with a "cancer potelicy factor," an of companion object. Indeed, the most re- .. $.cent analy.s_es show .that I :% * i e p y p newborn pulsar, then the pulsar could have plausibly acquired its incredible spin rate by conservation of angular momentum. Some of the infalling material may also have condensed to form the mysterious companion. Unless, of course, the object is not rotating at all, but pulsating in and out. Stay tuned. M. MITCHELL WALDROP y M A R C H 19x9 ,. -