Introduction Device-aided therapies including deep brain stimulation and continuous infusion therapies are effective for medication-refractory Parkinson's disease, yet longitudinal data on when patients undergo these interventions remain limited. We aimed to estimate cumulative incidence and identify predictors of device-aided therapy timing using survival analysis. Methods We analysed 1454 participants with Parkinson's disease from the Parkinson's Progression Markers Initiative (PPMI), including 1079 with sporadic disease. Kaplan-Meier analysis estimated cumulative incidence of device-aided therapy from diagnosis. Multivariable Cox proportional hazards models identified baseline predictors in sporadic Parkinson's disease, with a separate model examining genetic and biomarker effects across cohorts. Results During follow-up (median 3.0 years), 144 participants (9.9%) underwent device-aided therapy, predominantly deep brain stimulation (87.5%). In sporadic Parkinson's disease, cumulative incidence was 2.6% at 5 years and 15.3% at 10 years, with 10% reaching device-aided therapy by 8.6 years. Younger age at diagnosis (HR 0.92 per year, p < 0.001), higher tremor subscore (HR 1.16, p < 0.001), and higher postural instability/gait difficulty subscore (HR 1.68, p = 0.016) independently predicted earlier initiation. LRRK2 (HR 3.41) and GBA (HR 2.73) carriers had significantly earlier device-aided therapy compared to sporadic disease. Notably, no sporadic participant with negative CSF alpha-synuclein seed amplification assay (SAA) underwent device-aided therapy during follow-up. Site-level variation was not significant after adjusting for patient characteristics. Conclusions This study provides Kaplan-Meier-derived reference estimates for device-aided therapy in sporadic Parkinson's disease. Younger age, motor phenotype and genetic status were associated with earlier therapy initiation. The absence of device-aided therapy among SAA-negative sporadic participants warrants further investigation.
Ledingham et al. (Mon,) studied this question.