Composite Measures of Prodromal Parkinson’s Disease and the Probability of Phenoconversion at 3 Years
Mario Torres1, Katherine Hughes2, Kjetil Bjornevik1, Marianna Cortese1, Albert Hung3, Michael Schwarzschild3, Alberto Ascherio1
1Harvard T.H. Chan School of Public Health, 2OPTUM, 3Massachusetts General Hospital
Objective:
To assess the ability of composite measures of prodromal Parkinson’s disease (PD) to predict phenoconversion at 3 years.
Background:
Features suggestive of prodromal PD are individually insufficient for early disease detection. Using composite measures may improve predictive validity and capture a more representative patient population.
Design/Methods:
The study population comprised 19,757 individuals without PD at baseline (2014-2015). Two composite measures were created: i) the co-occurrence of probable REM sleep behavior disorder (pRBD), constipation and hyposmia, 3 major features of prodromal PD, and ii) a probability score which incorporated information on 8 risk (e.g., sex, physical inactivity) and 7 prodromal markers following the MDS research criteria. For each composite measure we calculated the positive predictive value (PPV) of being diagnosed with PD within 3 years. We additionally calculated the PPV for pRBD, constipation and hyposmia independently, and that of all 3 features along with parkinsonism. We repeated analyses restricting to participants <70 years old.
Results:
Between 2014-15 and 2017-18, we identified a total of 86 incident PD cases. The PPV of being diagnosed with PD within 3 years was 2.48% for those with the 3 prodromal features, 2.62% for those with possible prodromal PD (score≥0.5) and 4.55% for those with probable prodromal PD (score≥0.8). The PPV of individual features was lower than those of the composite measures. Parkinsonism plus the co-occurrence of 3 features resulted in the highest PPV (11.8%). Measures based on the presence of 3 features performed somewhat better among men than women. The PPVs of measures based on the MDS score or parkinsonism were higher among those aged <70 years.
Conclusions:
The ability to predict phenoconversion at 3 years was comparable between the two composite measures evaluated. The performance of composite measures was superior to that of individual features. Longer follow-up may be needed to better understand the validity of these measures.