Evaluation of the PREDIGT score’s performance in identifying newly diagnosed Parkinson’s patients without motor examination

Abstract

Abstract Several recent publications described algorithms to identify subjects with Parkinson’s disease (PD). In creating the “PREDIGT Score”, we previously developed a hypothesis-driven, simple-to-use formula to potentially calculate the incidence of PD. Here, we tested its performance in the ‘De Novo Parkinson Study’ (DeNoPa) and ‘Parkinson’s Progression Marker Initiative’ (PPMI); the latter included participants from the ‘FOllow Up persons with Neurologic Disease’ (FOUND) cohort. Baseline data from 563 newly diagnosed PD patients and 306 healthy control subjects were evaluated. Based on 13 variables, the original PREDIGT Score identified recently diagnosed PD patients in the DeNoPa, PPMI + FOUND and the pooled cohorts with area-under-the-curve (AUC) values of 0.88 (95% CI 0.83–0.92), 0.79 (95% CI 0.72–0.85), and 0.84 (95% CI 0.8–0.88), respectively. A simplified version (8 variables) generated AUC values of 0.92 (95% CI 0.89–0.95), 0.84 (95% CI 0.81–0.87), and 0.87 (0.84–0.89) in the DeNoPa, PPMI, and the pooled cohorts, respectively. In a two-step, screening-type approach, self-reported answers to a questionnaire (step 1) distinguished PD patients from controls with an AUC of 0.81 (95% CI 0.75–0.86). Adding a single, objective test (Step 2) further improved classification. Among seven biological markers explored, hyposmia was the most informative. The composite AUC value measured 0.9 (95% CI 0.88–0.91) in DeNoPa and 0.89 (95% CI 0.84–0.94) in PPMI. These results reveal a robust performance of the original PREDIGT Score to distinguish newly diagnosed PD patients from controls in two established cohorts. We also demonstrate the formula’s potential applicability to enriching for PD subjects in a population screening-type approach.