Objective: To explore the use of mediation analysis to incorporate L-dopa equivalent daily dose (LEDD) for disease progression of Parkinson’s Disease (PD) in a simulation study.

Background: Slower progression on the MDS Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) has been noted in PD patients treated with dopaminergic medications compared to untreated (Holden 2017). Additionally, using PPMI data, slower progression on MDS-UPDRS Part III has been observed with greater LEDD, even in the “OFF” state, suggesting that greater use of PD medications may confound population-level disease progression estimates.

Method: For patients on PD medications, we simulated longitudinal MDS-UPDRS data with the following assumptions: 1) at each visit, the observed MDS-UPDRS score reflects the effect of the prescribed LEDD, and a latent, unobserved underlying disease condition without the effect of PD medications; 2) physicians prescribed the next LEDD based on the current underlying disease condition; and 3) a disease-modifying therapy would slow down the underlying disease progression while the LEDD would improve the observed MDS-UDPRS. A baseline group variable represented different treatment arms in clinical trials. We analyzed the data with 1. a linear mixed model with LEDD as a time-varying covariate and 2. mediation analysis through a structural equation model (SEM) that explicitly modeled the effect of the current underlying disease condition on the next LEDD. We compared the estimates with the true values in the simulation setting.

Results: Under the simulation setting, a linear mixed model with LEDD as a time-varying covariate generated biased estimates for the effect of LEDD on the observed MDS-UPDRS and biased estimates of the treatment effect on disease progression. Mediation analysis through SEM explicitly modeled the effect of the underlying disease condition on the LEDD and generated unbiased estimates.

Conclusion: In our simulation, the underlying disease condition is a time-varying confounder for the association between the LEDD and the observed MDS-UPDRS. If our simulation setting is plausible, using mediation analysis through SEM could be a viable solution to incorporate this time-varying confounder and estimate the investigational drug’s effect on the disease progression apart from the impact of the PD medications.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/using-mediation-analysis-to-account-for-the-effect-of-parkinsons-disease-medications-on-disease-progression/