Assessing prognosis from nonrandomized studies: an example from brain arteriovenous malformations.
Raymond J., Naggara O., Guilbert F., Altman DG.
Two recent publications from Helsinki and Toronto that investigated the natural history of brain AVMs are the background topic for reviewing some principles and pitfalls of prognostic studies. Multivariable prognostic research involves 3 steps: developing the prognostic model, validating its performance in other individuals, and assessing its clinical impact on patients' outcomes. Unfortunately, the predictive ability of the model can be poor when it is applied to a new population, and clinical impact studies are rarely performed. Models that have not been validated should not be used to inform clinical decisions. Unfortunately, for rare outcomes in rare diseases, clinical data are limited. Although the 2 studies on brain AVMs may represent the best data currently available, they still included few patients with events and there are several methodologic concerns undermining the reliability of results. The estimates of risk of rupture per year are uncertain. Multiplying those uncertain numbers by the life expectancy of individuals can inflate error beyond control. Hence relying on these estimates to make clinical decisions may be dangerous.