Answer

d—Randomisation means that all individuals have the same probability of being allocated to either treatment. Providing the number of research participants was large enough, randomisation at entry to the trial would ensure no systematic differences between treatment groups in potential confounding factors. Randomisation ensures that any differences between treatment groups at the end of the study are only due to differences in treatment and chance associations with confounders. The larger the sample size, the smaller will be the differences due to chance. Confounding would exist if the effect of treatment on outcome was distorted by an intervening factor, leading to a spurious effect estimate. To confound the relation, the intervening factor must be related to the outcome and unevenly distributed between treatment groups. For example, biological sex could act as a confounder if women generally reported more pain than men. If the oral ibuprofen group contained a greater proportion of women than the topical treatment group, the study would produce the spurious estimate that topical treatment was more effective than it actually was because of the large proportion of men in the topical treatment group. Other potential confounding factors include demographics (for example, age or ethnicity), prognostic factors (for example, clinical history or disease severity), and factors influencing the likelihood of someone participating in a trial.

Answer a is false because double blinding is used to minimise bias in the measurement of effect on behalf of both the participants and researchers.

Answer b is false since intention to treat analysis is used to address concerns over changes to treatment after randomisation has taken place.

Answer c is false because although matching by age and sex would reduce confounding, the reduction would apply only to these variables and would not reduce the effect of all potential confounding factors.

Cite this as: BMJ 2009;338:b2381