In clinical trials and cost-effectiveness analyses, particular interest might focus on particular subgroups of patients defined by characteristics such as gender, race, age, study centre, country, comorbidity or disease risk factors. Subgroup analysis is the epidemiological and economic analysis of such groups. It can degenerate into data-mining. Many confounders such as sex, age, race, centre, smoking status, stage of disease or coexistent disorders can affect outcome. When these are examined post hoc, the risk of false positives and false inferences is high: there may be statistically significant differences in outcome between subgroups even when neither arm of the study receives any intervention. In one study (the Second International Study of Infarct Survival) there was found to be a slight adverse impact of aspirin therapy on patients born under the star signs Gemini and Libra.
332 Subjective Uncertainty
Was this article helpful?