Uncontrolled pearls, controlled evidence, meta-analysis and the individual patient.

Medicine has been dominated by uncontrolled data, often of unproven validity and insufficient to answer clinically important questions pertaining to individual patients. Controlled clinical trials, when designed and conducted rigorously, offer advantages over uncontrolled data, but they cannot be done for everything and often cater to the interests of sponsors rather than medical knowledge. With such sparse evidence, clinical research is doomed to look at main effects across populations rather than diversity of effects among individuals. By accumulating data from a large number of studies, meta-analysis provides a unique opportunity to address individual- and study-level heterogeneity. Diversity may be due to biases or may be real. Both sources must be scrutinized and meta-analysis may find a prime role in dissecting these components of diversity. Concurrent progress in basic sciences revolutionizing our predictive power for disease outcomes will heighten the importance of considering individual heterogeneity.