Temporal trends and short term prediction of relative survival in ependymoma using model based period analysis

Background Ependymoma is a rare central nervous system tumor with substantial heterogeneity in anatomical site, histology, and clinical outcome ^1–3 . Population-based evidence on temporal changes in survival remains limited, and up-to-date estimates are particularly needed ^4,5 . We therefore used model-based period analysis to evaluate temporal trends in relative survival and to generate short-term forecasts in patients with ependymoma ^6,7 . Methods Data were obtained from the Surveillance, Epidemiology, and End Results (SEER) database ⁸ (Incidence – SEER Research Data, 17 Registries, Nov 2024 Sub). Patients diagnosed during 2000–2019 with ICD-O-3 histology/behavior codes 9391/3, 9392/3, 9393/3, and 9394/3 and primary sites restricted to C71.* and C72.0–C72.1 were included. Five-year relative survival (RS) was estimated using period analysis ^9,10 ; expected survival was derived using the Ederer II method ¹¹ , and standard errors were calculated using Greenwood’s formula. Diagnosis periods were grouped as 2000–2004, 2005–2009, 2010–2014, and 2015–2019. Temporal trends were assessed across diagnosis-period mid-years on the absolute survival scale. Given that only four observed diagnosis periods were available, Joinpoint analysis was used only to assess whether an overall linear trend was present, rather than to identify multiple slope changes. Short-term forecasts for 2020–2024 were generated using inverse-variance-weighted models based on the observed period estimates. Results The baseline descriptive cohort included 4,070 patients, and the RS analytic cohort included 3,770 patients. Overall 5-year RS increased from 81.7% in 2000–2004 to 89.5% in 2015–2019. Period-specific estimates supported a significant upward linear trend over time. Non-anaplastic ependymoma showed consistently high 5-year RS, increasing from 86.1% to 94.4%, whereas anaplastic ependymoma showed lower but improving RS, increasing from 55.1% to 74.1%. Temporal improvement was more pronounced in anaplastic disease than in non-anaplastic disease. Model-based short-term forecasts suggested that overall 5-year RS may continue to improve in 2020–2024. Spinal/cauda equina tumors showed persistently favorable survival, whereas infratentorial and other intracranial tumors had lower RS with gradual improvement. In supplementary stage analyses, localized disease consistently showed better RS than non-localized disease. Conclusions Model-based period analysis showed a steady improvement in 5-year relative survival for ependymoma from 2000–2004 to 2015–2019. Short-term forecasts suggested that this favorable pattern may extend into 2020–2024, although prognostic heterogeneity remained evident across major clinicopathological subgroups.