Ecological studies

Welcome to this short lecture on ecological studies.

1
 Learning objectives
At the end of this lecture, students should be able to:
• Understand the key features of an ecological study
• Understand the strengths and limitations of ecological studies

Image by Carlos Ribeiro from Pixabay

At the end of the lecture you should be able to understand the key features of an
ecological study, the strengths and the limitations.

2
 Recap of key concepts
• Outcome: a disease or health state – e.g. death, mortality
rate, disease status, incidence rate, prevalence, length of
survival etc
• Exposure: a factor that may influence an individual’s risk of
developing a particular disease. Exposures can be
modifiable or unmodifiable.
• Causation: if an exposure causes an outcome, the exposure
is responsible for the subsequent occurrence of the disease
• Association: if there is an association between an exposure
and an outcome, the two are linked, although this
relationship may not be causal

Before we start, let’s just recap on the meaning of some important terms which you will
hear again and again over the next few weeks. In epidemiology, an outcome means a
disease or a health state. Examples of outcomes that you may come across in
epidemiological studies are death, death rates, disease status, incidence rates,
prevalence, length of survival, etc.

An exposure is any factor that may influence an individual’s risk of developing a

particular disease. For example, smoking is an exposure which influences an individual’s
risk of developing lung cancer. Another term for an exposure is a risk factor. An exposure
is said to be protective if it reduces an individual’s risk of developing a disease. For
example, the consumption of a diet rich in fruit and vegetables is protective against
bowel cancer. Exposures can be modifiable or unmodifiable.

A modifiable exposure is one that can be changed. Diet, body mass index and smoking
status are all examples of modifiable exposures.

An unmodifiable exposure is one that cannot be changed. Examples of unmodifiable

exposures are sex, age or having a particular genotype. In epidemiology, you will often
come across the terms association and causation and it is very important that you
understand the difference between these terms.

3
If an exposure causes an outcome, this means that the exposure is responsible for the
subsequent occurrence of the disease. For example, smoking causes lung cancer.
Causation is very hard to prove, so often in epidemiology you will come across the term
association.

There is said to be an association between an exposure and an outcome if there is a

relationship between them. For example, if there is an increase in hospitalisation for
respiratory disease that coincides with an increase in recorded levels of air pollution, it
would be reasonable to conclude that the two are linked. The evidence may not be strong
enough to conclude that the increase in air pollution caused the respiratory disease
(there may be other explanations that we are unaware of) but there is certainly a
relationship between these two events.

3
 Ecological/correlational studies
• Observational studies

• The unit of study is a population (NOT an individual)

• Exposure and outcome data are derived from different

populations (often from routine statistics) and the
association is compared

One of the simplest epidemiological study designs is the ecological study.

You may also come across the term correlation study, which is another name for the
same thing.

An ecological study is an observational study. It is an analytic study because it attempts

to analyse and understand the association between an exposure and an outcome.

The basic idea behind an ecological study is to look for evidence of an association
between an exposure and an outcome by comparing different populations.

Ecological studies often use routine statistics so they are relatively straightforward to
carry out.

4
 Comparisons include
• Geographical

• Time trends (secular trends)

• Migrant populations (vs non-migrants or native

populations)

• Occupational groups

Comparisons that might be used within an ecological study might be:

- geographical (comparisons between different countries or between different regions

within a country),

- trends over time (also known as secular trends, e.g. infant mortality rates before and
after the introduction of routine childhood immunisation),

- migrant populations (either comparing migrants to a new country with non-migrants

from the country of origin or comparing migrants to a country with the native
population of that country). An example of this would be comparing an Indian migrant
population in England and Wales either with their brothers and sisters who remain in
India or to the native population of England and Wales itself.

There are also frequently studies done of occupational groups, comparing mortality
between different populations based on occupation.

5
 Example: Raised cholesterol &
breast cancer incidence

Ref: Grasgruber P, Hrazdira E, Sebera M, Kalina T. Cancer incidence in Europe: an

ecological analysis of nutritional and other environmental factors. Front Oncol.
2018;8:151.

This ecological study by Grasgruber et al (2018), looked at the incidence of 24 types of

cancer in 39 European countries. The example on this slide from that study analysed the
correlation between raised cholesterol (which can come from a high fat diet) and the
incidence of breast cancer in women.

International statistics on food intake were collected from the FAOSTAT database of food
supply, which considers the total quantity of foodstuffs produced in a country, and the
quantity imported.

The study found that countries with raised cholesterol in women tended to have higher
rates of breast cancer incidence (a positive, statistically significant correlation). You can
probably imagine that there are lots of other differences between these populations
other than raised cholesterol that may be relevant to the risk of breast cancer incidence.
Ecological studies can mask complexities of exposure and outcome associations.

In any case, this ecological study is an interesting starting point to look further at the
association between this dietary-type element and cancer outcome.

6
 Example: Migrant Data

Here is an example of an ecological study based on migrant data.

The image shows age-standardised prostate cancer mortality rates per 100,000
population from 1990 – 2004 among native Germans and migrants from the former
Soviet Union.

The purpose of this study was to compare rates of prostate cancer in immigrant
populations living in Germany with rates of prostate cancer in native German men.

The study found that rates of prostate cancer were significantly lower in the immigrant
populations than in the native German population. They also found that this difference
did not change over time. If lifestyle factors (such as diet, smoking, levels of physical
activity etc) were associated with prostate cancer risk, you would expect to observe a
convergence of cancer risks over time. In other words, as time goes on and migrants
have lived for longer in Germany, you would expect their rates of cancer to become
similar to the rates in the native Germans.

The authors suggest that the fact that this has not happened and that cancer rates in
migrants remain significantly lower than in native Germans, suggests that genetic factors
are more important than environmental factors in prostate cancer risk. But there may
well be other factors that we do not know about that are behind the difference in cancer

7
rates between Germans and immigrants. Perhaps there is some lifestyle factor that is
more common in Germans than in immigrants. This is what we call a confounder or
confounding factor.

As we will discuss further, this possible confounding by other factors is one of the major
limitations of ecological studies.

7
 Example: Occupational Groups

This is an example of an ecological study comparing outcome rates among different

occupational groups.

The image shows standardised incidence ratios for cancer of the fallopian tubes among
different occupational categories of women in Scandinavian countries. The far right
column shows the average incidence ratio for all occupational groups.

This study used cancer registry data to compare the risks of fallopian tube cancer among
different occupations.

The study found that farmers and those who were economically inactive had lower than
average risks, while nurses, artistic workers, clerical workers, shop workers, smelting
workers, packers and hairdressers all had an increased risk.

Again, this does not prove that being a smelting worker or an artist causes fallopian tube
cancer – there may be some other confounding factor that we are unaware of. Perhaps
there is some other factor that causes fallopian tube cancer and perhaps this factor is
more common in smelting workers than in farmers.

It is also important not to make the mistake of drawing individual conclusions on the
basis of population-level data. It would be quite wrong, for example, to conclude on the

8
basis of this study that an individual smelting worker will develop fallopian tube cancer.

Ecological studies reveal trends at the population level but cannot be used to make
predictions at the level of the individual

8
 Strengths of ecological studies
• Useful when individual level data on exposure and outcome
not available (but data available on population level)

• Quick and cheap to do as data generally already available

• Good for generating hypotheses

• Good for studying rare outcomes and rare exposures

• Good for studying multiple outcomes and multiple

exposures

A strength of ecological studies is that they can be performed when individual level data
on outcomes and exposures are not available but where data are available at the
population level. So it is very useful at the early stages of investigating an association
between an exposure and an outcome to see if there are any patterns in ecological
studies.

Another strength of ecological studies is that they are cheap to do, as the data are
usually already available so the researcher doesn’t have to go out and collect extra data.

So ecological studies are good for generating hypotheses but it is at the individual level
where hypotheses on the association between exposure and outcome need to be
tested.

There are 2 other strengths of ecological studies that are worth mentioning. Firstly, it
can be quite difficult for epidemiologists to study rare diseases or rare exposures
because of the very large sample sizes needed to produce meaningful data.

Because ecological studies are conducted at the level of the population, they are good
for studying rare outcomes and rare exposures. This was illustrated by the fallopian tube
cancer study we have just looked at.

9
Secondly, ecological studies can be used to investigate multiple outcomes or multiple
exposures at the same time. For example, in the immigrant study we have just looked at,
differences in mortality rates for a whole range of diseases could have easily been looked
at as part of the same study.

9
 Limitations of ecological studies
• Major risk of confounding – i.e. not a good way to test
aetiological hypotheses

• Results may not be applicable at the individual level

(risk of ecological fallacy)

• May mask complex relationship between exposure and

outcome

Because of the risk of confounding that I’ve mentioned in several of the examples
previously, ecological studies are not a good way to test hypotheses: there is a major risk
of confounding and providing misleading information.

Another consideration to take into account with ecological studies is that results that are
found at a population level may not be applicable at the individual level. So, taking the
earlier example of the relationship between raised cholesterol and breast cancer
incidence at the population level, this does not mean that an individual who reduces
their fat intake will necessarily reduce their risk of breast cancer.

You may come across the term “ecological fallacy”. This refers to the mistake of drawing
individual conclusions on the basis of population-level data.

Finally, ecological studies may mask complex relationships between the exposure and
the outcome, whereby different levels of exposure may have different implications for
the outcome

10
 Summary of key points
 The unit of study is a population rather than individuals

 Ecological studies are an efficient way to generate hypotheses

 Findings of these studies need to be considered carefully as there is a signicant

risk of error

In summary, in ecological studies, the unit of study is a population or group of people

rather than individuals.

Ecological studies are an efficient (in terms of both time and expense) way to generate
hypotheses.

Despite their efficiency, these study results need to be considered carefully as there is a
significant risk of error.

11