Evaluating the evidence about therapies. What the clinician needs to know about statistics

Study design is the most important determinant of whether the study conclusions are likely to be biased or misleading. Clinical practice should be based on the most rigorously designed studies to address a particular question. If a study is well designed and sufficiently powered to give an unbiased and precise estimate of treatment effect, the statistical analyses are usually straightforward and evaluating the statistical methods is not difficult. The statistical methods used in observational studies are more complex and much more difficult to evaluate, but study design (selection of subjects and controls, selection and measurement of confounders) is still a critical determinant of bias in the study results. When clinical practice is based on observational studies, extreme caution must be exercised in concluding that associations imply a causal relationship between the treatment or exposure and the outcome.