Li Yan, McDonald-Webb Millie, McLernon David J, Counsell Carl E, Macleod Angus D
Institute of Applied Health Sciences, University of Aberdeen, Polwarth Building, Foresterhill, Aberdeen, UK.
School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Polwarth Building, Foresterhill, Aberdeen, UK.
NPJ Parkinsons Dis. 2025 Aug 29;11(1):266. doi: 10.1038/s41531-025-01112-x.
Predicting outcomes for people with Parkinson's (PwP) can enable better information provision, personalised treatments, and enhanced trial design. It is unclear what prognostic models are optimal for use. We systematically reviewed previously published prognostic models for PwP, assessed quality, and made recommendations. We searched MEDLINE and EMBASE for studies developing/validating models predicting clinical outcomes in PwP. We assessed risk of bias and applicability using the PROBAST tool. We screened 1024 references and identified 25 studies (41 prognostic models). The most common outcomes were falls (11 studies), dementia (7) and motor complications (4). Most models made short-term predictions (60% ≤2 years). All studies had concerns about bias, e.g., inadequate population details (n = 16), suboptimal methods for missing data (n = 21), and no external validation (n = 22). 13 models had sufficient information to be used in practice. Further development and validation of prognostic models is needed which follows existing guidelines to reduce risk of bias.
预测帕金森病患者（PwP）的预后可以提供更好的信息、个性化治疗并优化试验设计。目前尚不清楚哪种预后模型是最适合使用的。我们系统回顾了先前发表的针对PwP的预后模型，评估了其质量并给出了建议。我们在MEDLINE和EMBASE中检索了开发/验证预测PwP临床结局模型的研究。我们使用PROBAST工具评估偏倚风险和适用性。我们筛选了1024篇参考文献，确定了25项研究（41种预后模型）。最常见的结局是跌倒（11项研究）、痴呆（7项）和运动并发症（4项）。大多数模型进行短期预测（60%≤2年）。所有研究都存在偏倚问题，例如，人群细节不足（n = 16）、处理缺失数据的方法欠佳（n = 21）以及缺乏外部验证（n = 22）。13种模型有足够的信息可用于实际应用。需要按照现有指南进一步开发和验证预后模型，以降低偏倚风险。
