The potential of epidemiological studies,
and the likelihood of their misinterpretation, was driven home to me in a very
disturbing, yet amusing, way when I was a bit younger.
After having been married for a few years,
my wife and I decided to expand our family. When Jill was pregnant with our
first daughter, we decided to move from our apartment in the city to a house in
the suburbs. We spent innumerable weekends with our Real Estate agent looking
at houses and finally found one that we liked and could afford in a nice little
community north of Chicago. We embarked on the buying process and eventually
had a lot of our time and all of our money invested in that house.
Jill was 8 months pregnant when we moved in.
As a new father-to-be I was hit with a fit of protectiveness for my small but
growing family. Right from the start, life in the suburbs was wonderful. May in
Chicago is usually pretty sunny and the signs of Spring bursting forth all over
are very welcome after another long, cold winter. I spent much of my time at
the new house outside in the front lawn cutting the grass, planting flowers,
and generally just enjoying the setting.
One fine afternoon a neighbor stopped by to
introduce herself. Here was another affect of the suburbs that I expected to
delight in, friendly neighbors. The older woman from across the street
introduced herself and asked the usual questions. After taking in my general
air of joy at being there, she decided to truly welcome me to the neighborhood.
She started,
"You know
that you’ve moved onto THE CANCER STREET?"
I thought her hair looked a little strange;
it was a wig that could have been covering up some of the aftereffects
of chemotherapy. She proceeded to point out all the houses on the other side of
the street, hers included, where at least one of the occupants had come down
with the dreaded disease.
At this time I was well on my way to a
doctorate degree in Biomedical Engineering and knew something about statistics.
My laboratory was also in a children’s hospital and I daily witnessed the
pathetic sight of small children fighting cancer. Needless to say, I was
concerned. Maybe irrationally, but nonetheless a bad feeling came over me.
I left that conversation in a decidedly more
somber mood than when I had entered it. I told my wife about the neighbor
visiting but I did not mention the content of the conversation. In her state,
eight months pregnant with all sorts of hormonal imbalances, that was the last
thing that she needed to hear.
During the next few days I educated myself
about epidemiology. I spent time in the university library and read textbooks
and journals that dealt with interpreting the meaning of the occurrence and
distribution of disease. I came away from this short period of self-education
having learned three valuable lessons:
After a little education, I felt much better
about my new street in the suburbs. I resumed my daily routine of tending the
lawn and gardens after returning home from work. A short time later, I met some
other new neighbors. A younger couple was out one day taking a walk. They
noticed a new face and came over to say hello. They lived at the end of the
street and had been there for a number of years. We talked a bit about the
neighborhood and then I decided to broach the subject of my previous
conversation with a neighbor. I asked them what they knew about all the people
who had contracted cancer on our street. They asked me what I meant and I
related what I had heard the other day.
Upon hearing this tale, the husband looked
at his wife and, after a short pause, they both nodded. Their faces were somber
and I feared the worst. The husband turned back to me and drew a deep breath. I
girded myself for what I was sure would be more bad news. He spoke,
"Yeah, I
think she was probably right. You made a big mistake by moving to this
street."
My fears were confirmed and my heart slowly
rose into my throat. He continued,
"What you
really should have done was to find a house on the next street over."
I managed to murmur out a, "Why?"
to which he answered,
"Well, that’s
the wife swapping street!"
Expected Disease
It is unproductive, and even dangerous, to
jump to conclusions and misinterpret isolated instances of disease. If you are
a ham radio operator who belongs to a club with 200 members and two of them are
diagnosed with brain tumors, you may begin to worry that there is a link
between the use of radios and brain tumors. (Also, consider that there are many
different types of brain tumors that are not related to each other in any way
except that they all occur inside the head. For the sake of simplicity, let’s
assume that these two are the same type). An incorrect statistic that you could
present is that 1% of your club came down with brain tumors this year. While
this is unacceptably high, consider these figures:
The incidence rate of primary malignant
brain tumors is 6.1 per 100,000 (0.0061% of the population). (Central Brain
Tumor Registry of the United States. 1995 Annual Report, 1996)
The population of the United States (U.S.
Census Bureau, January 1, 1995) is about 261,638,000.
This means that there were approximately 15,960
people in the United States who were diagnosed with primary malignant brain
tumors in 1995.
There are more than 650,000 licensed Amateur
Radio operators in the United States.
From these statistics, we would expect
approximately 40 licensed amateur radio operators in the United States to be
diagnosed with primary malignant brain tumors each year.
Thus, your two club members consist of a
brain tumor cluster that could be an entirely normal incidence of this disease,
irrespective of their use of radio.
Proving a cause-and-effect between radio
waves and cancer is even more difficult than just this simple number-crunching.
Given a group of Amateur Radio operators in which all people have contracted
the same disease, even if the incidence of this disease is higher than what
would be expected in the general population, we still must consider the
following issues:
Epidemiology
Epidemiology is a complex science that has
the value of being able to recognize possible risks of causality for diseases
among groups with a population. Epidemiological studies demonstrate
associations between two sets of variables, often referred to as the "Risk
Factor" and the "Disease." However, an association, no matter
how strong, does not necessarily imply causality. For instance, in the brain
tumor example stated previously, there appears to be a 1% risk of getting a
brain tumor if you belong to that ham radio club; such a conclusion is
incorrect and dangerous. Let me present a more absurd example of the statistics
just to show how they can be misused. All (100%) of the people who get brain
tumors breath air (a very strong association, indeed). Does this mean that
breathing air causes brain tumors? Demonstration of causality requires, in
addition to a strong association, consistency, a dose-response relationship (as
the dose increases, the response should also increase), a temporal relationship,
and a reasonable theory as to the mechanism that leads to the relationship.
Most epidemiological studies do not contain
the rigor that will rule out every variable in the lives and habits of those
included in the statistics. It is for this reason that most epidemiological
results do not prove, but merely suggest, possible links between factors in our
environment and disease. However, an incomplete study that shows a
"statistically significant" increase in disease incidence may still
be purely coincidental due to the lack of information about other factors that
could affect the disease.
To deal with ambiguities in statistics and
in the collection of information about the people used in an epidemiological
study, epidemiologists use a measure that indicates the strength of the
positive results, the Risk Ratio (RR). The RR takes statistical significance
one step further and attempts to estimate the likelihood of contracting a
disease. It is a simple measure that is the ratio of the percentages of people
with and without a given risk factor that contract a given disease. A value of
RR less than 1 implies that the risk factor is beneficial (it protects people
from the disease), an RR of 1 says that there is no effect of the presumed risk
factor (there is no risk), and an RR greater than 1 says that the risk factor
may play a role in causing the disease. It is possible that there is a
statistically significant increase in the number of people in a given group
contracting a disease yet the chances of this happening are so small that we
will all die of old age first. A very large RR tells us that there really is
something to be concerned with. For instance, most studies on cigarette smoking
have shown the RR for death from lung cancer among smokers to be greater than
10; smokers are 10 times more likely to die of lung cancer than non-smokers. In
contrast, the very few studies that have shown statistically significant
increases in disease rates for people exposed to radio waves, such as Amateur
Radio operators, finds that Risk Ratios are very small (less than 1.1).
Conclusions
We have been bombarded by epidemiological
results in the popular press. According to what we get from the news, just
about everything in our lives can kill us prematurely. This illustrates the
danger of epidemiology. It is not enough to do the number-crunching and come up
with "statistical significance." It is crucial that these numbers be
tempered by the conditions of the study (just how much of the data about a
person’s life and habits was included in the study?) and that false inferences
be recognized (such as my breathing example). Unless these things are
specified, such results cannot be trusted and should not be used to cause
concern.
Copyright © 1998-2002 by Gregory D. Lapin