Risk factors and correlations
What makes a good study?
All studies rely on sampleIn a survey, a sample is a small selection of the population. Samples are less reliable but quicker, cheaper and easier to question than the entire population.. It is very important that there is no biasBias is unfairness, for example, in a survey or a question. A biased die may be loaded to favour one particular number. in choosing the people in the study. It must be a representative sampleA representative sample is one that accurately represents the whole of the group. so that the conclusions it draws based on the sample can be reliably applied to the whole population.
Matched groups are studies which compare two groups of people who are very similar to each other apart from one factor, for example smoking. Some factors which should be matched are age, sex and weight. Differences in outcomes are therefore more likely to be able to be applied to the whole population.
Correlation and lung cancer
Scientists look for patterns in data.
Research has established links between cancer and various lifestyle factors, chemicals produced in the body or that enter the human body, and chemicals in the environment.
Results of studies are published in medical journals. This means other scientists can carry out similar studies to see if the results were reproducibleData are reproducible when other researchers have found the same or similar results under similar conditions. We can have more confidence in data that are reproducible. - meaning they get similar results. This process is called peer reviewA process where expert scientists check research written by others before it is published in a scientific journal..
Scientists have established several causal mechanisms for risk factors.
Below is an example of how risk factors have been found. This has come from studies on smoking cigarettes and lung cancer.
Historically, in the USA, a pattern can be seen in the amount of cigarette use and the incidence of lung cancer.
As the number of cigarettes smoked has increased over the years, the incidence of lung cancer has increased also. Note that there is a time lag, but cancer usually takes some years to develop. There is a clear association, called a correlationA relationship between two sets of data, such that when one set changes you would expect the other set to change as well., between the variables.
With cancer and other non-communicable diseases, scientists have found correlations.
Correlation and cause
If there is a correlation between a particular factor and an outcome, it does not mean that the factor necessarily causes the outcome. Scientists must look for a possible mechanism by which the factor could be the likely cause.
In the case of lung cancer, analyses of cigarette smoke have shown that at least 70 of the chemicals present in smoke will cause cancer in laboratory animals. This result establishes a causal link.
Correlation and effects of smoking on babies
For mothers who smoke during pregnancy:
- smoking increases the risk of miscarriageThe spontaneous or unplanned expulsion of a fetus from the womb before it is able to survive independently.
- the babies and children are more likely to suffer from respiratory infections and an increased risk of asthmaA disease that affects the respiratory system.
- the long-term physical growth and intellectual development of the baby/child is affected
- there is an increased risk of birth defects
- the birthweight of the baby is reduced
- when parents smoke, there is a greater risk of cot death - known as sudden infant death syndrome
Question
The bar chart shows that when mothers under 20 smoke, the birth weight of their babies is reduced.
Suggest two pieces of additional information that would need to be collected before drawing firm conclusions about the effect of mothers smoking on the birth weight of babies.
- Carry out a similar study of mothers of different age groups to see if similar results could be obtained.
- Check to ensure that other factors - for instance, related to lifestyle factors of the experimental group - could not have affected birth weight.