Rev Med Suisse Romande. 1998 Jun; 118(6): 547-52.
[Statistic pitfalls or how should we interprete numbers in the evaluation of a new treatment]
Are cholesterol lowering drugs useful? Do they increase life expectancy? Do third generation oral contraceptives increase the risk of venous thromboembolism? Is there a worldwide decline in semen quality over the last 50 years? Do vitamin supplements improve your child's IQ? Does homeopathy work better than placebo? These questions illustrate some statistical problems and some bias encountered during clinical studies, which can lead to erroneous results. Type I and II errors, surveillance, prescription or publication bias as well as the healthy user effect are described. Problems of regression to the mean, limits of meta-analysis validity and other statistical problems are discussed.