Genetic Risk Score Aids in Parkinson’s Diagnosis

A new, noninvasive model significantly outperforms any single classifier in identifying patients with Parkinson's disease, according to a study published Aug. 11 in Lancet Neurology. The model is able to discriminate patients without evidence of dopaminergic deficit, typical of Parkinson's disease, from those patients with aetiologically typical disease, which the authors say, could improve detection of preclinical disease.

For complex progressive diseases such as Parkinson's disease, preclinical diagnosis and low error rates in diagnosis are crucial both not only for clinical care, but also for correctly classifying patients participating in clinical trials.

"Unlike dopamine transporter positron emission tomography [DAT] scanning, our model is portable and can be administered remotely at a fraction of the cost: our model costs around US$100 per sample versus DAT scanning, which can cost thousands of dollars per patient," write the authors led by Mike Nalls, Ph.D., from the National Institute on Aging (Bethesda, Md.). "This model might be useful as part of a diagnostic path towards more accurate preclinical detection of Parkinson's disease."

The researchers participating in the Parkinson's Disease Biomarkers Program and Parkinson's Progression Marker Initiative developed a model for disease classification using data from the Parkinson's Progression Marker Initiative study (n= 367 patients with Parkinson's disease and 165 controls). Factors selected to contribute to the model included: olfactory function, disease-specific genetic risk score (based on 30 risk factors replicated in one or more studies), family history of Parkinson's disease, age, and gender. The model was then tested using data from 825 patients with Parkinson's disease and 261 controls from five independent cohorts.

Each of five factors included in the model made significant contributions to the content informing the integrative predictive model (smell score, 63.1 percent of the explained variance; genetic risk score, 13.6 percent; family history, 11.4 percent; gender, 6.0 percent, and age 5.9 percent). The researchers found that using the discovery data, the model correctly distinguished patients with Parkinson's disease from controls with an area under the curve (AUC) of 0.923 and with high sensitivity (0.834) and specificity (0.903) at its optimum AUC threshold (0.655). External validation showed good classification of Parkinson's disease, with AUCs ranging from 0.894 to 0.998 across the different cohorts. Furthermore, four of 17 participants who had scans without evidence of dopaminergic deficit and who the model classified as having Parkinson's disease, converted to Parkinson's disease within 1 year.

The authors say that in the future the accuracy of the model can be improved by identifying more disease-specific biomarkers and genetic risk loci. Current resequencing of known loci are expected to generate more accurate estimates of genetic risk.

"Ninety-three percent (N=28/30) of the genetic risk variants that we used to create our genetic risk score are from genome-wide association studies and are therefore probably surrogates for true functional variants because of the inherent nature of these imputation- based studies," write the authors. "Identification of the true functional variants within loci would improve our algorithmic classification of Parkinson's disease."

Takeaway: An algorithm may improve identification of patients with Parkinson's disease, even prior to clinically significant disease.