Background:
Parkinson’s disease (PD) is characterized by a wide range of motor and non-motor symptoms. Broader biological definitions of PD are discussed and receive increasing attention, going beyond the current motor-centered
PD definition. The heterogeneity of non-motor PD symptoms poses a challenge for early and accurate diagnosis of PD.

Objectives:
The main objective of this study is to evaluate systematically whether non-motor symptoms affect the timing of PD diagnosis. This is accomplished by modeling disease progression in large-scale longitudinal data.

Question:
The project aims to determine whether specific non-motor symptoms of people with PD systematically delay or hasten the diagnosis compared to the typical time point of PD diagnosis.

Methods:
This study utilized data from three large PD cohorts and analyzed it through a latent time joint mixed-effects model (LTJMM). This approach allows an alignment of disease trajectories of individual people with PD on a common disease time scale, and subsequently the determination of whether diagnoses were made earlier or later than the cohort’s average diagnosis time. Initial clinical symptoms at the typical diagnosis time were estimated using several mixed-effects models, depending on the scales of the outcomes. Non-motor scores were grouped into 12 distinct non-motor domains and pooled estimates were calculated across all three cohorts using three-level meta-analyses with random effects. P-values were corrected for multiple testing using Benjamini-Hochberg procedure.

Results:
The analysis included 1,124 individuals diagnosed with PD. Several non-motor symptoms were found to contribute to a diagnosis later than the average: anxiety (p=0.0043), autonomic dysfunction (p=0.0019), depression (p=0.0004), fatigue (p=0.012), pain (p=0.0085), sleep disturbances (p=0.0043), and a higher overall burden of non-motor symptoms (p=0.0006, Fig. 1). In contrast, impulsivity (p=0.12), REM sleep behavior disorder (p=0.28), apathy (p=0.32), hyposmia (p=0.79), and hallucinations (p=0.09) did not impact diagnostic delay.

Conclusions:
Through statistical modeling of initial clinical presentations of people with PD and a model-based estimation of diagnostic delay, the study successfully identified several non-motor symptoms that impact the timing of PD diagnosis without requiring direct clinical observations at the point of typical diagnosis. These findings support a biological conceptualization of PD that considers early non-motor symptoms, highlighting the need for diagnostic criteria that reflect the whole disease’s heterogeneity.