in class” study of Vermont granite workers (with no co-expo-sure to anything) was null through 1987. This study kept IARC
from pulling the trigger. A 1995 publication reported a significant and more than three-fold excess of lung cancer among
long seniority, long latency workers—against a 20 percent nonsignificant excess in the full cohort. A 1988 publication by
the same authors found significant excess in the same cohort.
The results of this study would have been known long before
publication. (In future, I will count the number of studies finding an association published before 1987. This would constitute
a “no-effect” level for studies finding an association with lung
cancer before the epidemiologists will call it.)
I’ve also told people that the epidemiological record wasn’t
enough for silica to make the carcinogen list without an animal
bioassay, and that the bioassays of silica were thrown-in studies
using silica as a positive control for oil shale dust and Mt. St.
Helen’s dust. I couldn’t find documentation for that story without more digging; the decisive inhalation bioassay quoted was
published in 1983, finding excess tumors in rats at 12 mg/m3.
A later study found an excess at 1 mg/m3. Several studies
alerted to a cancer risk by other routes, but not by inhalation.
The main lessons are that one of the oldest occupational dis-ease-causing chemicals was not bioassayed by inhalation until
1983, and that the bioassay made a difference. The oil shale
and Mt. St. Helen’s are a nice touch, but not necessary to the
Giving In to Quantitative Risk Assessment
Starting with the Supreme Court’s benzene decision, and for my
tenure on the National Academy of Science committee that
wrote the Red Book (not the Little Red Book), I’d argued that
the most appropriate course of action was to ban carcinogens
rather than set PELs, and the most appropriate exposure limit
for a carcinogen was the lowest feasible exposure or limit of
detection, whichever was lowest.
Then I was sitting on the NTP Board of Scientific Counselors
subcommittee deciding the listing for silica to be published in
1991. There’s silica in dirt. Can you ban dirt? Do you label the
sandbox? There’s silica on the beach. Should lifeguards wear
respirators? On the latter question of lifeguard risk, I’m not
sure. Are there silica measurements for a windy day in Coney
Island? I’m not aware of a potency assessment for inhalation
of silica. One will exist when OSHA issues its risk assessment,
but the RA is waiting to be peer reviewed by a panel not yet
named. Colleagues on the BSC in 1990 voiced similar con-
cerns. Anyway, we limited the probable call to “silica of res-
On a related note, if silica inhalation is prevalent, and if to-
bacco smoking potentiates the effects of silica, how many of the
cancers attributed to smoking are really attributable to silica?
Finding a Limit
The best historical account of silica limits I found was the TLV
documentation, which is definitely worth reading now that silica is in play at OSHA. It looks to me like the limit was reduced
to 0.25 million particles per cubic foot (mppcf) for pure silica in
1971, from the limit of 5 million extant since 1961. That was a
20-fold reduction. It’s unclear when the current 10 mg/m3 (
percent respirable quartz plus 2) used by OSHA as a PEL came into
effect. The OSHA general industry PEL is equivalent, more or
less, to 100 µg/m3 for respirable silica. Thus, our PEL for silica
was set in 1971 and not much improved; only the units
The construction PEL is 250 mppcf (percent silica plus 5). I’m
guessing that the mass limit was adopted to be comparable to
the particle count limit back in 1970. Nobody does particle
counts anymore. The filter sample weighed on an analytical
balance followed by X-ray diffraction was much more efficient
than the impinger followed by the microscope, supported by
area high-volume samples for silica content. However, real-time
monitoring instruments available now will count particles in
specified ranges; a strategy of high-volume gravimetric silica
samples combined with a particle counter would work, and
would yield real-time data on exposure sources and job segments. Maybe particle count—now feasible—is a better measure
Some principles in setting a PEL would include recognizing
that the no observed effect level in a laboratory study is really
a 1 in 10 risk rate. People are more diverse than laboratory animals, so the extrapolation factor must take into account the
least resistant people. Rats are pretty resistant to inhaled particles, while mice and hamsters are refractory as far as lung cancer goes.
When using human studies to set a PEL, we should recognize
the following lessons:
• Cohort mortality studies in working populations are system-
atically biased against finding excess mortality compared to
the general population by the “healthy worker effect” and the
“healthy worker survivor effect.”
• Workers with shorter duration of employment generally suf-
fer adverse mortality experience compared to workers with
longer employment duration, because of health-related ter-
mination of employment. Therefore, cumulative or average
exposure is biased against an association with disease.
The underlying rate of lung cancer among American men is
about 5 percent. Therefore, an observed doubling in lung cancer
mortality is equivalent to a 5 percent risk rate.
Two paradigm shifts should be recognized in response to silica.
First, what we’ve been taught about silica health effects was superseded in at least 1983 (with a bioassay), certainly from 1987
(with IARC probable designation). Silica has been “known” to
be a human carcinogen since 1995. Second, the older PELs, in
this case going back to 1971, were aimed at no-effect levels or
even at then-acceptable effect levels. A health-based PEL must
be many factors below those levels.
Sciences in New York. He can be reached at (212) 481-7651 or