Analysis of Genetic and Non-genetic Predictors of Levodopa Induced Dyskinesia in Parkinson's Disease

Abstract

Background: Levodopa (L-Dopa), representing the therapeutic gold standard for the treatment of Parkinson disease (PD), is associated with side effects like L-Dopa induced dyskinesia (LID). Although several non-genetic and genetic factors have been investigated for association with LID risk, contrasting results were reported and its genetic basis remain largely unexplored. Methods: In an Italian PD cohort (N = 460), we first performed stepwise multivariable Cox Proportional Hazard regressions modeling LID risk as a function of gender, PD familiarity, clinical subtype, weight, age-at-onset (AAO) and years-of-disease (YOD), L-Dopa dosage, severity scores, and scales assessing motor (UPDRS-III), cognitive (MoCA), and non-motor symptoms (NMS). Then we enriched the resulting model testing two variants-rs356219 and D4S3481-increasing the expression of the SNCA gene, previously suggested as a potential mechanism of LID onset. To account for more complex (non-linear) relations of these variables with LID risk, we built a survival random forest (SRF) algorithm including all the covariates mentioned above. Results: Among tested variables (N = 460 case-complete, 211 LID events; total follow-up 31,361 person-months, median 61 months), disease duration showed significant association (p < 0.005), with 6 (3-8)% decrease of LID risk per additional YOD. Other nominally significant associations were observed for gender-with women showing a 39 (5-82)% higher risk of LID-and AAO, with 2 (0.3-3)% decrease of risk for each year increase of PD onset. The SRF algorithm confirmed YOD as the most prominent feature influencing LID risk, with a variable importance of about 8% in the model. In genetic models, no statistically significant effects on incident LID risk was observed. Conclusions: This evidence supports a protective effect of late PD onset and gender (men) against LID risk and suggests a new independent protective factor, YOD. Moreover, it underlines the importance of personalized therapeutic protocols for PD patients in the future.

Keywords: D4S3481; Parkinson disease; SNCA; levodopa induced dyskinesia; rs356219; α-synuclein.

FIGURE 1

Cox curves of genetic model tested modeling incident LID risk vs (A) additive genetic model of rs356219 and vs (B) pseudo-recessive genetic model of D4S8134. Genotype classes were defined as (A) AA (black), AG (red) and GG (green) genotypes for rs356219 and (B) 263 bp allele carriers (black) vs all other genotypes (red) for D4S8134.

FIGURE 2

Relative variable importance (%) of the LID risk predictors tested within the survival random forest (A) excluding and (B) including genetic (SNCA) features.