Unruptured intracranial aneurysms: their illusive natural history and why subgroup statistics cannot provide normative criteria for clinical decisions or selection criteria for a randomized trial.

There is currently no evidence that treatment of unruptured aneurysms is beneficial. Confronted with the uncertainty, many clinicians are attracted by an individual calculus of risks using numbers extracted from subgroup statistics of observational studies or natural history data. The so-called natural history of unruptured aneurysms refers to a purely man-made ratio of events divided by the number of untreated patients identified by imaging, a ratio heavily influenced by referral patterns and arbitrary clinical decisions. Available studies lacked prespecified hypotheses, exposing all analyses to sampling error and bias, and sample sizes were too small to provide reliable subgroup statistics. Far from being "natural kinds" of aneurysms, subgroups were post-hoc creations. Resulting data-driven statistics can only be exploratory, the error too uncontrollable to serve for clinical decisions. A randomized trial is in order, but selection according to fixed size criteria is ill-advised, given the imprecision of imaging, the influence of other factors such as location, previous history, multiplicity of lesions, risks of treatment, age and the danger of arbitrarily excluding from a long trial a large segment of the population with aneurysms for whom the research question is most pertinent.