Problem
An article claims that “taking one aspirin a day reduces the risk of heart attack by 25%”. The study followed 1000 people for 10 years. (a) Which confounding variables should be controlled? (b) What does “control group” mean? (c) Why is a double-blind experiment even better? (d) Discuss the difference between correlation and causation in this case.
Solution
(a) Confounding variables to control. The risk of heart attack depends on very many factors besides aspirin. To attribute the effect specifically to the drug, one needs to control (or distribute equally between groups) at least: age, diet (fats, salt, calories), smoking habits, physical activity, genetic predisposition and cardiovascular family history. If these factors are not balanced, they alone could explain the observed difference.
(b) Control group. This is a group of people who do not receive the treatment under study — that is, they do not take aspirin (or they take a placebo, an inert pill identical to aspirin) — but who are comparable in every other respect to the treated group. It serves as a term of comparison: only by comparing treated and control can the effect of aspirin alone be isolated, separating it from the natural course of the disease.
(c) Why double-blind is better. In a double-blind experiment neither the patient nor the doctor evaluating the outcome knows who is taking the real drug and who the placebo. This removes two biases:
- the patient’s placebo/expectation effect, who might feel better simply because they believe they are being treated;
- the evaluator’s bias, i.e. the (even unconscious) tendency of the doctor to interpret the data in favour of the expected hypothesis. This way the comparison between the two groups remains as objective as possible.
(d) Correlation and causation. Observing that “those who take aspirin have fewer heart attacks” establishes a correlation, i.e. a statistical association. On its own it does not prove causation: there could be a confounding variable (for example, people who take aspirin may already be more health-conscious, exercise more and eat better). Only a controlled, randomised experiment — in which assignment to the treated or control group happens by chance — makes it possible to neutralise confounding variables and conclude that it is indeed aspirin that causes the reduction in risk.
Links
Topics: Metodo sperimentale