Commentary
David Ogilvie takes you through the paper and discusses what it
means
Epidemiologists
investigate the causes of disease. This paper describes how a classical
technique—the cohort study—was used to investigate whether
a person's risk of early type 2 diabetes and/or obesity is
increased if their mother smoked during
pregnancy.
What is a cohort
study?
A cohort is a group of people that are
followed through time to see what happens to them. This study is based
on a large cohort. Each member of the group was born during a single
week in 1958. At birth, data were collected about their mothers. Data
about their health was collected again when the offspring reached the
ages of 7, 16, and 33.
In a cohort
study two types of data need to be collected: which of the group have
ended up with the condition being researched, and which of the group
were exposed to the relevant risk
factors.
To find out whether a
person has ended up with the condition (outcome) you are interested in,
you might ask them (“a personal interview at age 33 years asked
about diabetes”), examine them (“interviewers measured
height and weight”), or do other tests. The idea is to divide
the cohort into people who have the condition (cases) and people who do
not (controls). It is best to have a clear case definition—for
example, obesity is defined as having a body mass index greater than
30.
In this study, to find out
whether people were exposed to the relevant risk factors, exposure
information collected included maternal smoking, social class, and
birth weight.
The analysis is
intended to test the hypothesis that exposure to certain risk factors
is associated with the outcome of
interest.
Why do a cohort
study?
Cohorts can be a rich source of
information. You may come across other studies on other diseases which
have been carried out using data from this 1958 birth
cohort.
A cohort study like this,
however, is a huge job—thousands of people have been followed up
for over 30 years—and you have to wait a very long time to get
your results. Any epidemiology textbook will contain a list of the
advantages and disadvantages of cohort studies, but an obvious one in
this example is that the exposure data were collected at the time, long
before the people became diabetic or obese. This reduces the likelihood
of bias in estimating people's exposure to the risk factors. If
you were interviewed at age 33 would you or your mother be able to say
how much she smoked when she was pregnant? How would your answer be
affected if you thought that you might have become ill as a
result?
How did they
analyse the data?
The top left part of the
table shows that people with diabetes were more likely to have had a
mother who smoked during pregnancy. Thirty two per cent of mothers of
people with diabetes were heavy smokers; compare this with 11%
of mothers of people without
diabetes.
The authors used a
technique called “multiple logistic regression” to estimate
how much greater the odds of diabetes are in people whose mothers
smoked. This technique allows an adjustment to be made for other risk
factors, such as birth weight, that might be relevant (confounding
factors). The odds ratios in the “adjusted” columns
represent the estimated odds ratios after taking all the confounding
factors into account. You will also see that they have stratified
(categorised) exposure to smoking by considering medium, variable, and
heavy smokers separately. If this all sounds a bit complicated refer to
a previous commentary that has covered the basics of odds ratios in a
much simpler
study.1
The
top right part of the table shows that the adjusted odds ratio for
diabetes increases from 1.01 (for medium smokers) to 4.02 (for heavy
smokers). These estimates are compared with 1.00 for
non-smokers. The implication is that the heavy smokers are at
increased risk of having a child who becomes diabetic, and the medium
smokers are not.
The other analyses
shown in the paper follow the same
principles.
So
what?
There has been much interest in recent
years in how our early life experiences, both before and after birth,
affect our health later in life. The findings of this study offer yet
another reason for helping people not to smoke, or to cut down their
smoking—especially during
pregnancy.
David Ogilvie, specialist registrar in public health medicine, Hamilton Lanarkshire
Email: david.ogilvie@lanarkshire.scot.nhs.uk
studentBMJ 2002;10:45-88 March ISSN 0966-6494
- Cullington
HE. Light eye colour linked to deafness after meningitis
[commentary by D Ogilive]. student BMJ 2001;9:152.
