“EPA concluded that benzene satisfied the definition of a

“hazardous air pollutant” under Section 112 of the Clean Air
Act.
~~

THE STANDARD-SETTING PROCESS
Carciaogens and Section 112 .

;

.

t-

fo

Be

qualitative evaluation of the evidence to determine whether
@ substance should be considered a human carcinogen for

reguiatory purposes. As described earlier, this was done in
the case of benzene before the chemica) was listed as a
hazardous air pollutant in 1977. The next stage is quantitative: how large is the risk of cancer at various levels of

exposure” The result of this examination is a dose-response

function which gives the lifetime risk per unit of exposure

Once a substance has been listed as a hazardous air
pollutant, Section 112 of the Clean Air Act requires EPA to

(or “potency.”) The mext stage is to estimate bow many

ty” to protect the public health. However, neither the lan-

response function to obtain estimates of the risk caused by
emissions of the pollutant, im this case benzene, into the

publish standards which provide an “ample margin of safe-

guage nor the legislative history of Section 112 reveals any
apecific Congressional intent as to how to apply the phrase,
“ample margin of safety” to protect the public health from

pollutants like benzene.

in some cases, scientific evidence indicates that a given
chemica! is hazardous at high levels of exposure but has no

effect below a certain level. For most carcinogenic chemi-

cals such as benzene, however, scientists are unable to
identify such a threshold below which no effects take place;
moreover, to the extent scientists understand the process of
carcinogenesis, there is some reason to believe thresholds
may not exist. For such substances, EPA and other Federal
agencies have taken the position that any level of exposure
may pose some risks of adverse effects with the risks
increasing as the exposure increases.
Since any given environmental! carcinogen is responsible
for at most a smal! fraction of a community's overall cancer
incidence, with current statistical techniques it is virtually
impossible to directly link actual human cancers with actual

ambient air exposure to chemicals such as benzene. Conse-

quently, EPA relies on mathematical modeling techniques
to estimate these human bealth risks. These techniques —
“quantitative risk assessment” — are used to assess the risk
of adverse health effects from exposure to benzene in the
ambient environment by mathematically extrapolating
those effects found at the higher occupationa! exposure
levels down to lower concentration levels that more nearly
reflect the exposure of people from the ambient air around
industrial sources of benzene.

“Quantitative risk assessment" (described below) couples
the mathematical dose-response models with estimates of
population exposures to describe the magnitude of the risk
posed by sources of carcinogens such as benzene. It is an
attempt to synthesize and apply available scientific know)edge about carcinogens to predict the effects of environmenta! exposures. At best, quantitative risk assessment gives us
an estimate of bow severe the health problem could be.
What to do about the risks-what controls, if any, to require
— constitutes “risk management.” Risk assessment, then,

provides information that is important, but it alone is insuf-

ficient to make risk management decisions. That is, in
addition to information on health risks, any risk management policy also requires information on control techno}-

ogies, their effectiveness and costs.'

°

ss

Risk Assessment

EPA's approach to risk assessment for suspected carcino-

gens may be divided into several steps. The first is a
‘For a discussion of the important distinction between rish aseessment and risk management and their role in government decision-making. see “Science, Risk, and Public Policy” by Wilham D.
Ruckelshaus, presented at the National Academy of Sciences, June
22, 1983, reprinted in Science, September §, 1983.
12-23-83

people are exposed to the substance, and at what levels.
exposure estimates then are combined with the dose-

environment.
All stages of the process are subject to uncertainties
because of gaps in scientific knowledge and data limitations.
One step that has great uncertainty is estimating the doseresponse function The fundamental! problem is in extrapo-

lating from data on the relatively high doses in the epidemi-

ological or animal toxicological studies to the far lower
exposure leveis found in the environment. In the case of
benzene. the data showing increased risk are based on
workers exposed to many parts per million, bul most environmental exposures for the general public are no higher
than several parts per billion. In other words.it is necessary
to extrapolate to doses a thousand or more times lower than
those at which increased cancer rates have been observed.

Scientists have proposed many different mathematical

models for low-dose extrapolation. EPA generally rehes on
the linear, no-threshold model. which assumes that risk is

proportional to dose. This model is chosen because it has

some biological justification. With this model. decreasing
the dose by a factor of 1000 also reduces the risk by a factor
of 1000. Most of the other models predict much smaller risks
at low doses. The linear mode! generally yields a higher
estimate of potency than other models and most scientists
accept it as giving a plausible upper-limit estimate for a
chemical's potency at low levels of exposure. In other
words, the potency of a substance is unlikely to be higher
than estimated using the linear model. and could be substantially lower. Use of the linear model reflects EPA's decision
to err on the side of caution in the face of uncertainties The

Gnal result is a “unit-risk factor,” which gives the estiruated
upper-limit lifetime risk per unit of exposure.

Exposure levels for each specific source categories are

derived using emissions estimates, dispersion modeling. and
population data. For any given level of emissions. dispersion
models predict concentrations at different distances from
the emission source. By combining those estimated concentrations with census data on population densities, the number of people exposed at different levels can be estimated.

Severa! factors suggest that actual exposure levels wil] be

lower than those estimated. In estimating exposure, the
most exposed individuals are hypothetically subjected to the
Maximum annua! average concentration of the emissions
for 24 hours every day for 70 years (roughly a lifetime). This
does not take into account indoor vs. outdoor air, for instance, or the fact that most people in their daily routines
move in and out of the specific areas where the emussions
concentrations are the highest.
The final risk estimates are the product of the exposure
levels and the estimated unil-risk factor. Two summary

measures are of particular interest: “maximum individual

risk” and “total population impact.” The former refers to

the estimated increased lifetime risk from a source that is
faced by an individual who spends his or her entire life at

Enwronment Reporter

Select target paragraph3