An ecological study is an observational study in which at least one variable is measured at the group level. An ecological study is especially appropriate for initial investigation of causal hypothesis. Let's look at an example to understand what a group-level variable is.
Example 6-1: Section
Results from an ecological study examining diet and sunlight as risks for prostate cancer mortality.
Colli JL, Colli A. International comparison of prostate cancer mortality rates with dietary practices and sunlight levels. Urologic Oncology 2006:24;184-194.
In Figure 1 (reproduced from the Colli paper), the x-axis indicates per capita sugar calories consumed per day. The y-axis represent the age-adjusted mortality rate for prostate cancer. Each dot represents a country, plotting the per-capita sugar consumption and the age-adjusted prostate cancer mortality rate for that particular country. Per-capita sugar consumption is an ecological variable because it is the average measure of exposure to sugar for all the people in the country. It does not mean that every person in the country ate exactly the same specified amount of sugar. Similarly, the mortality rate is a group-level variable because it represents the country's experience, not any individual person's experience in the country. Neither of these variables are at the individual-level. The line in the figure is a fitted regression for the dots, showing a standard statistical procedure for ecological studies. (However, statistical procedures for ecological studies can be quite complex!) Do the data support the hypothesis that increased sugar consumption is associated with increased prostate cancer mortality? - YES! Do the data determine that sugar consumption causes prostate cancer death? NO!
The study also considered the effects of other factors on prostate cancer mortality. In Table 3, the authors consider the effect of a country's latitude and ultraviolet (UV) index on prostate cancer mortality rates. The data support the hypothesis that sunlight exposure is associated with prostate cancer death.
Finally, the investigators used a stepwise multiple regression model to show that ultraviolet index, cereals, sugar and onion consumption provided the best fit when ambient sunlight exposure was included as a factor. What conclusion do we draw?
Come up with an answer to this question and then click on the button below to reveal the answer.
What conclusion do we draw? Does a lack of onions and cereal in the diet, overindulging in sugar along with low UV index cause prostrate cancer death?
No, causation is not established by any individual study not matter how strong the findings or how strong the study design. In this study, there was also a strong relationship between animal fat consumption and prostrate cancer; however, animal fat consumption was collinear with UV index and thus if UV index was in the model, animal fat consumption was not.