Medical Institute of Developmental Disabilities Research, Showa University, Tokyo, Japan.
Department of Neuropsychiatry, Keio University School of Medicine, Tokyo, Japan.
Neuroimage Clin. 2021;32:102852. doi: 10.1016/j.nicl.2021.102852. Epub 2021 Oct 7.
One-third of patients with schizophrenia are treatment-resistant to non-clozapine antipsychotics (TRS), while the rest respond (NTRS). Examining whether TRS and NTRS represent different pathophysiologies is an important step toward precision medicine.
Focusing on cortical thickness (CT), we analyzed international multi-site cross-sectional datasets of magnetic resonance imaging comprising 110 patients with schizophrenia (NTRS = 46, TRS = 64) and 52 healthy controls (HCs). We utilized a logistic regression with L1-norm regularization to find brain regions related to either NTRS or TRS. We conducted nested 10-fold cross-validation and computed the accuracy and area under the curve (AUC). Then, we applied the NTRS classifier to patients with TRS, and vice versa.
Patients with NTRS and TRS were classified from HCs with 65% and 78% accuracies and with the AUC of 0.69 and 0.85 (p = 0.014 and < 0.001, corrected), respectively. The left planum temporale (PT) and left anterior insula/inferior frontal gyrus (IFG) contributed to both NTRS and TRS classifiers. The left supramarginal gyrus only contributed to NTRS and right superior temporal sulcus and right lateral orbitofrontal cortex only to the TRS. The NTRS classifiers successfully distinguished those with TRS from HCs with the AUC of 0.78 (p < 0.001), while the TRS classifiers classified those with NTRS from HCs with the AUC of 0.69 (p = 0.015).
Both NTRS and TRS could be distinguished from HCs on the basis of CT. The CT pathological basis of NTRS and TRS has commonalities, and TRS presents unique CT features.
三分之一的精神分裂症患者对抗非氯氮平类抗精神病药物（TRS）治疗无反应，而其余患者则有反应（NTRS）。检验 TRS 和 NTRS 是否代表不同的病理生理机制是迈向精准医学的重要一步。
我们专注于皮质厚度（CT），分析了包含 110 名精神分裂症患者（NTRS=46，TRS=64）和 52 名健康对照者（HCs）的国际多地点横断面磁共振成像数据集。我们利用具有 L1-范数正则化的逻辑回归来找到与 NTRS 或 TRS 相关的脑区。我们进行了嵌套的 10 倍交叉验证，并计算了准确性和曲线下面积（AUC）。然后，我们将 NTRS 分类器应用于 TRS 患者，反之亦然。
NTRS 和 TRS 患者可从 HCs 中以 65%和 78%的准确率和 0.69 和 0.85 的 AUC （p=0.014 和<0.001，校正）进行分类。左侧颞上回（PT）和左侧前岛叶/额下回（IFG）对 NTRS 和 TRS 分类器均有贡献。左侧缘上回仅对 NTRS 有贡献，右侧颞上回和右侧外侧眶额皮层仅对 TRS 有贡献。NTRS 分类器成功地以 AUC 为 0.78（p<0.001）将 TRS 患者与 HCs 区分开来，而 TRS 分类器则以 AUC 为 0.69（p=0.015）将 NTRS 患者与 HCs 区分开来。
基于 CT，可将 NTRS 和 TRS 与 HCs 区分开来。NTRS 和 TRS 的 CT 病理基础具有共性，而 TRS 则具有独特的 CT 特征。
