Fondation FondaMental, Créteil, France; Sorbonne Universités, UPMC Univ Paris 06, UMR_S 1136, Institut Pierre Louis d'Epidémiologie et de Santé Publique; INSERM, UMR_S 1136, Institut Pierre Louis d'Epidémiologie et de Santé Publique, F-75013 Paris, France.
Fondation FondaMental, Créteil, France; AP-HM, Aix-Marseille Univ, Faculté de Médecine - Secteur Timone, EA 3279: CEReSS, 13005 Marseille, France.
Prog Neuropsychopharmacol Biol Psychiatry. 2019 Jun 8;92:226-234. doi: 10.1016/j.pnpbp.2019.01.003. Epub 2019 Jan 10.
Existing staging models have not been fully validated. Thus, after classifying patients with schizophrenia according to the staging model proposed by McGorry et al. (2010), we explored the validity of this staging model and its stability after one-year of follow-up.
Using unsupervised machine-learning algorithm, we classified 770 outpatients into 5 clinical stages, the highest being the most severe. Analyses of (co)variance were performed to compare each stage in regard to socio-demographics factors, clinical characteristics, co-morbidities, ongoing treatment and neuropsychological profiles.
The precision of clinical staging can be improved by sub-dividing intermediate stages (II and III). Clinical validators of class IV include the presence of concomitant major depressive episode (42.6% in stage IV versus 3.4% in stage IIa), more severe cognitive profile, lower adherence to medication and prescription of >3 psychotropic medications. Follow-up at one-year showed good stability of each stage.
Clinical staging in schizophrenia could be improved by adding clinical elements such as mood symptoms and cognition to severity, relapses and global functioning. In terms of therapeutic strategies, attention needs to be paid on the factors associated with the more stages of schizophrenia such as treatment of comorbid depression, reduction of the number of concomitant psychotropic medications, improvement of treatment adherence, and prescription of cognitive remediation.
现有的分期模型尚未得到充分验证。因此，在根据 McGorry 等人（2010 年）提出的分期模型对精神分裂症患者进行分类后，我们探讨了该分期模型的有效性及其在一年随访后的稳定性。
我们使用无监督机器学习算法将 770 名门诊患者分为 5 个临床阶段，阶段越高表示病情越严重。采用方差分析比较每个阶段在社会人口统计学因素、临床特征、合并症、正在进行的治疗和神经心理学特征方面的差异。
通过细分中间阶段（II 期和 III 期）可以提高临床分期的准确性。IV 期的临床验证指标包括伴有并发重度抑郁发作（IV 期为 42.6%，IIa 期为 3.4%）、认知功能更严重、药物依从性较低和使用>3 种精神药物。一年随访显示每个阶段的稳定性良好。
通过将情绪症状和认知等临床元素添加到严重程度、复发和整体功能中，可以改善精神分裂症的临床分期。在治疗策略方面，需要关注与精神分裂症更多阶段相关的因素，如合并抑郁的治疗、减少同时使用的精神药物数量、提高治疗依从性以及开具认知矫正治疗。
