Wong Sandra Chi Yiu, Lo Heidi Ka Ying, Chau Anson Kai Chun, Ng Mary Chung Mun, Chan Joe Kwun Nam, Chu Ryan Sai Ting, Fang Catherine Zhiqian, Hui Christy Lai Ming, Chan Sherry Kit Wa, Lee Edwin Ho Ming, Lui Simon Sai Yu, Chang Wing Chung
Department of Psychiatry, School of Clinical medicine, LKS Faculty of Medicine, The University of Hong Kong, Kowloon Tong, Hong Kong.
State Key Laboratory of Brain and Cognitive Science, the University of Hong Kong, Kowloon Tong, Hong Kong.
Eur Arch Psychiatry Clin Neurosci. 2025 May 21. doi: 10.1007/s00406-025-02022-y.
Reinforcement-learning (RL) impairment is an important determinant of functional outcome in bipolar-disorder (BD). This study examined RL in 38 euthymic bipolar-I-disorder (BD-I) individuals aged 16-40 years who were treated within three years from first-episode mania with psychosis and 40 demographically-matched healthy-controls using a computerized RL-paradigm, which investigated rapid and gradual learning, and reward-driven and punishment-driven learning. Symptom severity and cognitive functions were assessed. Our results showed that BD-I individuals displayed lower lose-shift scores than controls (p = 0.03). There were no group differences in other rapid RL measures. Regarding overall RL, a repeated-measures ANOVA revealed main effect of group (F1,76 = 6.5, p = 0.03; controls performed better than patients), block (F2.87,218.45 = 43.7, p < 0.001; performance improving over time) and probability (F1,76 = 15.6, p < 0.001; better performance in 90% than in 80% condition). Post-hoc analysis revealed that controls performed better than BD-I individuals on loss-avoidance stimuli (p = 0.02). Better performance of controls relative to BD-I individuals on gain stimuli approached statistical significance (p = 0.06). No correlations of RL measures with symptoms, cognition or antipsychotic dose were observed. In conclusion, this study is among the few to examine RL impairment in euthymic BD-I with history of psychosis at a relatively early illness stage, and indicates that BD-I individuals displayed punishment-driven learning (i.e., negative RL) deficits compared with controls. Diminished punishment-sensitivity may indicate vulnerability to maladaptive behaviors, particularly in response to negative events or stress, and poorer functional impairment. Further research is required to clarify longitudinal trajectory of punishment-insensitivity and its relationship with psychosis and functional outcomes in the early-stage of BD.
强化学习（RL）障碍是双相情感障碍（BD）功能转归的重要决定因素。本研究使用计算机化RL范式，对38例年龄在16 - 40岁、首次躁狂发作伴精神病性症状后三年内接受治疗的双相I型障碍（BD-I）心境正常个体，以及40名人口统计学匹配的健康对照者进行了RL研究，该范式考察了快速和渐进学习，以及奖励驱动和惩罚驱动学习。评估了症状严重程度和认知功能。我们的结果显示，BD-I个体的损失转移得分低于对照组（p = 0.03）。在其他快速RL测量中，两组之间没有差异。关于总体RL，重复测量方差分析显示组间主效应（F1,76 = 6.5，p = 0.03；对照组表现优于患者）、区组（F2.87,218.45 = 43.7，p < 0.001；随着时间推移表现改善）和概率（F1,76 = 15.6，p < 0.001；90%条件下表现优于80%条件）。事后分析显示，在避免损失刺激方面，对照组表现优于BD-I个体（p = 0.02）。在获得刺激方面，对照组相对于BD-I个体的更好表现接近统计学显著性（p = 0.06）。未观察到RL测量与症状、认知或抗精神病药物剂量之间的相关性。总之，本研究是少数在相对早期疾病阶段对有精神病性症状史的BD-I心境正常个体的RL障碍进行研究的之一，表明BD-I个体与对照组相比存在惩罚驱动学习（即负性RL）缺陷。惩罚敏感性降低可能表明易出现适应不良行为，特别是对负面事件或压力的反应，以及功能损害更严重。需要进一步研究来阐明惩罚不敏感的纵向轨迹及其与BD早期精神病性症状和功能转归的关系。
