Risks from 1-A
Feinstein points out that since
World War II epidemiologists, who
study the patterns and natural his-
tory of diseases, have achieved
dramatic successes such as pre-
venting polio, eradicating smallpox
and demonstrating that rubella in
early pregnancy can cause birth
defects. In addition, associations
between cigarette smoking and
lung cancer or the protective den-
tal effect of fluoridated water are
now well established.
"These splendid achieve-
ments, however, have also been
accompanied by major uncertain-
ties and controversies in other ...
studies, particularly for cause-ef-
fect relationships," he wrote. He
cites studies that directly contra-
dict each other in at least 56 cases
where a specific "menace" has
been alleged to have caused a dis-
One of the problems research-
ers routinely face is that human
studies are much more difficult to
conduct and interpret than those
done in laboratories or with experi-
"Groups of people cannot be
assembled and studied as easily as
captive animals or inanimate mate-
rial," he wrote. "Data about nutri-
tion, medical exposures and life-
style are difficult to check for sci-
entific quality and the results often
receive statistical analysis with
methods that are unfamiliar and
// f/ Feinstein argues that epidemi-
ologists often violate basic princi-
pies of statistics and fail to identity
clearly what they are seeking to
achieve, thereby permitting fuzzy
interpretations of data.
The people used in studies are
often not well screened. In the
case of 200,000 women studied to
determine possible causes of
breast cancer, for example, inves-
tigators did not examine the wom-
en before the study and, therefore,
could not be sure that some did not
already have the disease.
He also notes the widespread
reliance on household surveys and
death certificates often produces
The association between alco-
ho! and breast cancer provides an
example of the difficulty assessing
environmental causes for disease.
Two studies in 1987 reported an
association between drinking and
increased risk of breast cancer.
Together, nearly 200,000 women
were followed and information
about diet and other factors were
obtained from them and analyzed.
The results suggested that women
who drink alcohol even in mod-
eration have an increased
chance of developing breast can-
But Feinstein points out that in
one of the studies the key results
were based upon voluntary re-
sponses given by women about
how much alcohol they had drunk
"over the last year." Researchers
have often complained that people
rarely give accurate reports about
how much they drink.
He suggests that fewer people
should take the results of these
studies as gospel.
Food for thought
Scientist says shoddy statistics
scare people about food dangers
By MICHAEL SPECTER
WASHINGTON Medical scare stories linking
common foods and drugs to maladies have become a
pervasive part of American life, but a prominent research-
er has charged that many of those reports are based on
faulty science and cause needless fear.
Rarely a week goes by without the widely heralded
release of a study that purports to show how the food we
eat, the air we breathe or the drugs we use cause cancer.
Almost every medical journal has at least one article in
each issue examining the relationship between something
as common as coffee and cancer or heart disease.
In the current issue of Science, however, Dr. Alvan
Feinstein, professor and director of clinical epidemiology
at Yale University School of Medicine, charges that the
statistical tools used to develop those relationships are
based on mrnthods that rarely stand up to scrutiny.
"People are clearly more afraid than they need to be
about the risks they encounter in their daily lives," he said
in an interview. "We are often shouting fire in a crowded
---- ... ,i