Factors associated with mortality in early stages of parkinsonism

Prognosis of patients with parkinsonism varies greatly between the various parkinsonian syndromes. However, it is often difficult to distinguish the different forms, particularly in early stages. We examined predictors of mortality and functional outcome in patients with recent-onset parkinsonism with an initially uncertain diagnosis (n = 156). Patients were recruited between 2003 and 2006, comprehensively investigated, and followed prospectively (up to 15 years, mean 7 years). A final clinical diagnosis was established after follow-up. Independent predictors of mortality were investigated with multivariable Cox regression and combined into a simple prediction model. Model performance to predict 5- and 10-year mortality and functional outcome after 3 years was evaluated and externally validated in a second cohort of 62 patients with parkinsonism with an initially uncertain diagnosis. Ninety-one patients died (58%). Orthostatic hypotension, impaired cognition, abnormal tandem gait, and elevated neurofilament light chain concentration in serum or CSF were associated with mortality. A simple model that combined these factors showed excellent performance for prediction of functional outcome after 3 years and mortality after 5 and 10 years (c-statistic similar to 0.90 for all models). Model performance was confirmed after external validation: prediction of functional outcome after 3 years (c-statistic 0.89, 95% CI 0.80-0.98) and mortality after 5 years (c-statistic 0.91, 95% CI 0.84-0.99) were comparable to the results in the discovery cohort. These findings help clinicians to estimate a patient's prognosis, irrespective of the specific diagnosis.

Automated summary

Prognosis of patients with parkinsonism varies greatly between different syndromes and it is difficult to distinguish the different forms, especially in early stages. This study identified predictors of mortality and functional outcome in patients with recent-onset parkinsonism and established a simple prediction model. The model showed excellent performance in predicting functional outcome and mortality, and the findings were confirmed in an external validation cohort.