The presence of hypotheses in the medical literature

Although hypotheses are believed to be crucial to the scientific method, they are not always reported in clinical studies. We hypothesized that the presence of a hypothesis in the medical literature would be associated with study type.

All articles published in four major medical journals in the first half of 2002 were reviewed and abstracted (n = 332).

Hypotheses were explicitly stated in 23% and implied or inferable in 13% of studies; another 17% of studies without hypotheses reported sample size calculations; 46% of studies stated neither hypotheses nor reported sample size calculations. Randomized controlled trials reported hypotheses and sample size calculations 50% and 80% of the time, respectively, while other study types reported them much less often. Adjusting for journal, randomized controlled trials had 19.1 times (95% confidence interval: 9.4-39.1) greater odds of having stated hypotheses or sample size calculations than other study types.

Major medical journals have been very successful in encouraging authors to include sample size calculations in the reports of randomized controlled trials by making their presence a structural requirement for submission. However, stated hypotheses are rarer, and hypotheses and sample size calculations are even less prevalent in papers using other study designs. For some study designs, lack of these study elements may be related to vague guidelines. For others, such as observational and correlational studies, omission may be related to foundational disagreements about the scientific method.