Estimating Rupture Risk of Intracranial Aneurysms: What We Know, What We Do Not Know, and What We Need

Management of unruptured intracranial aneurysms requires balancing the risk of aneurysm rupture against the risk of procedural complications. Estimates of rupture risk stem from a few landmark natural history studies whose findings differ substantially, creating uncertainty for clinical decision-making. This review appraises these studies, highlighting areas of agreement and contradiction to inform future directions. Across studies, short-term rupture risk is low and increases with aneurysm size. The magnitude of risk varies (0.20%-1.85% at 1 year). These discrepancies likely arise from methodological challenges inherent to natural history research, including selection, crossover, incomplete follow-up, and regional variation. The effects of these factors are difficult to disentangle due to confounding. Rupture risk is highest in Finnish studies, followed by Japanese, then other international cohorts. This geographic pattern is reversed for the treatment rate. Rupture risk also shows a strong inverse relationship with treatment rate (P=0.008, R²=0.79). This makes it impossible to know whether rupture risk reflects treatment or geography, and which estimates apply clinically. Studies have short follow-up (mean 2.8 years) and require substantial extrapolation to estimate lifetime risk (mean age at diagnosis 42-66 years). Small differences in short-term estimates produce large variations in long-term projections. Moreover, the underlying assumption that risk remains constant with time has not been formally evaluated. Half of the data sets are consistent with this, but half suggest it declines with time. The effects of key aneurysm rupture predictors vary between studies. This includes age, sex, hypertension, smoking, prior subarachnoid hemorrhage, family history of intracranial aneurysms, and aneurysm location, multiplicity, and size thresholds. It is unclear whether this reflects regional variation, overfitting, or other factors. Meta-analyses are most representative, but remain constrained by limitations of contributing data sets. Larger multicenter studies with longer follow-up, fewer losses, and deeper phenotyping are still needed, despite their practical challenges.