Liu · The Lancet. Neurology 2017 · Multicenter longitudinal cohort study · n=2482

Prediction of cognition in Parkinson's disease with a clinical-genetic score: a longitudinal analysis of nine cohorts.

Cited 190 times in the scientific literature.

Level 3 - non-randomized controlled study

Non-randomized multicenter longitudinal cohort study developing and validating a prognostic algorithm

PubMed 28629879 · doi:10.1016/S1474-4422(17)30122-9 · record verified 2026-08-30

What was done

Researchers developed and validated a clinical-genetic risk score to predict global cognitive impairment (defined as Mini Mental State Examination [MMSE] ≤25) within 10 years of disease onset in patients with Parkinson's disease. Using data from nine longitudinal cohorts from North America and Europe evaluated between 1986 and 2016, 3,200 patients were assessed for eligibility. After excluding 235 with baseline MMSE ≤25, 135 with first visits over 12 years after onset, and 334 with missing covariates, a discovery population of 1,350 patients across six cohorts (median follow-up 2.8 years, IQR 1.6-4.6) was used to select predictors via backward eliminated Cox proportional hazards analysis. The model included age at onset, baseline MMSE, years of education, motor exam score, sex, depression, and β-glucocerebrosidase (GBA) mutation status. Validation was performed in an independent replication population of 1,132 patients (14 excluded for missing covariates) across three cohorts (median follow-up 6.5 years, IQR 4.1-7.2), as well as across 10,000 randomly resampled training and test sets.

What was found

The cognitive risk score predicted cognitive impairment within 10 years of disease onset with an area under the curve (AUC) greater than 0.85 in both the discovery population (95% CI 0.82-0.90) and the replication population (95% CI 0.78-0.91). Patients in the highest score quartile had an increased hazard for cognitive impairment compared to those in the lowest quartile (hazard ratio 18.4, 95% CI 9.4-36.1). At a predefined cutoff of 0.196, the score predicted dementia or disabling cognitive impairment with an AUC of 0.88 (95% CI 0.79-0.94) and a negative predictive value of 0.92 (95% CI 0.88-0.95).

Why it matters

This algorithm provides a validated tool to stratify Parkinson's disease patients by long-term risk of cognitive decline. It can inform clinical prognosis and enrich participant selection in clinical trials testing cognitive therapies.

Limits

Cognitive impairment was defined primarily using MMSE, which is a coarse screening instrument rather than a comprehensive neuropsychological battery. A total of 348 patients were excluded across discovery and replication cohorts due to missing covariates, potentially introducing selection bias. The sample was restricted to cohorts from North America and Europe, limiting generalizability to other populations.

Cited by