Objectives Develop a model to identify incidental meningiomas at risk of observation failure.

Design Retrospective study (2007–2015).

Subjects 385 patients. Mean age 62.6 years.

Methods Observation failure was defined as: new symptoms, MRI progression (absolute growth rate 2 cm³/year or absolute growth rate 1 cm³/year +relative growth rate 30%/year) or loss of treatment options. A prognostic model was developed using MRI and patient comorbidity.

Results Over a median of 36 months, 44 (10.9%) meningiomas failed observation. Median time to failure was 33 months. Model parameters were based on statistical and clinical considerations and included: increasing tumour size (HR=2.17 [95% CI=1.53–3.09], p<0.001), peritumoural signal change (HR=1.58 [95% CI=0.65–3.85], p=0.313), FLAIR/T2 hyperintense meningiomas (HR=10.6 [95% CI=5. 39–21.0], p<0.001) and proximity to neurovascular structures (HR=1.38 [95% CI=0.74–2.56], p=0.314). Patients were stratified based on the model into low, medium and high-risk groups and rates of failed observation at 5 years were 3%, 28% and 75% respectively. Low-risk patients had small meningiomas, free of all risk factors. After 5 years of follow-up the probability of failure plateaued in all risk groups. Older patients with comorbidities were 15-times more likely to die than to receive intervention at 5 years following diagnosis, regardless of risk group.

Conclusions Most meningiomas remain clinically and radiologically stable. Stratifying follow-up according to risk-group has the potential to reduce the cost to healthcare.