This cross-sectional study was conducted in the inpatient units of a psychiatric hospital.

An experimenter-designed questionnaire was used to collect disease information, including gender, age, marital status, education level, family history of psychiatric disease, onset age, and history of mental stimuli.

All data were analyzed by SPSS 26.0 (IBM Corp., Armonk, NY, USA). Continuous variables (age, education level, CPT, WCST, etc.) were analyzed by one-way ANOVA. Categorical variables (occupation, family history, etc.) were analyzed using chi-square tests. The least significant difference (LSD) method was used for post-hoc tests. Additionally, analysis of covariance was applied to control for education level, a factor that may affect cognitive function. Spearman correlation analysis was used to examine the correlations between cognitive indicators and clinical variables. Results were expressed as (mean $\pm$ SD) or frequency (percentage), respectively. Power analysis and Bonferroni correction were conducted in order to ensure that the sample size, interpret the research results and control the multiple comparison error. p 0.05 was considered statistically significant.

The BD patients showed significantly lower educational level (p = 0.003) and higher unemployment rates (p 0.001) than did the healthy controls. The depressive-onset subgroup demonstrated earlier onset age [(21.5 $\pm$ 6.1) vs. (25.3 $\pm$ 3.9), p = 0.013], more frequent hospitalizations [(3.5 $\pm$ 2.5) vs. (1.7 $\pm$ 1.1), p 0.001] The details were showed in Table 1.

The ANOVA results showed that there were 18 cognitive indicators, 13 cognitive indicators have statistically significant differences between groups (p 0.05). Nearly half of cognitive indicators had a “large effect” ($\eta$²$\ge$ 0.14), indicating that the “grouping” (manic group/depression group/healthy control group) had a statistically significant impact on cognitive functions. The details were showed in Table 2.

There were significant differences in large effect indicators (such as CPT3, DST-B, and DST-T). The score of the manic group was higher than the depressive group. The cognitive differences of the two groups mainly lie in the complex cognitive executive functions (such as working memory, DST-B, and memory function). The depression group was significantly worse than the manic group and healthy control group in core cognitive functions (attention, working memory, cognitive accuracy) (All p 0.01 or 0.001). The details were shown in Table 3.

The family history of mental disease was likely to be associated with patients’ executive function (WCST-CC: r = 0.278, p = 0.007; WCST-TE: r = –0.206, p = 0.048; WCST-PCR: r = 0.257, p = 0.013). Additionally, a lower educational level and an early onset of BD could potentially have impaired memory functions (DST-T: r = 0.214, p = 0.041; DST-F: r = 0.356, p = 0.005) (Table 4). The details of Digit Span Test was showed in (Table 5).

Compared to the manic-onset group, the depressive-onset group of BD patients exhibited a distinct profile of “earlier onset age—higher admission times—lower occupation rate”. Early onset age may reflect neurodevelopmental vulnerabilities during sensitive periods. For instance, the prefrontal cortex (PFC) is the chief executive officer of the brain, controlling the highest-level cognitive and emotional processes. Dysregulation of PFC synaptic transmission is strongly linked to affective disturbance, social deficits, and memory loss in mental disorders (Solé et al., 2017). Recurrent hospitalizations (3.5 $\pm$ 2.5 episodes) suggest a relapsing-remitting course of the depressive-onset BD patients, and potentially exacerbated neuroprogressive damage through a “relapse–cognitive decline–functional deterioration” cycle (Yan and Rein, 2022). The lower occupation rate in the depressive-onset group suggested severe social dysfunction, likely mediated by deficits in working memory and sustained attention, which impaired occupational performance. Prolonged unemployment may serve as an indicator of severe cognitive impairment.

The FEP depressive group exhibited significantly lower performance on the sustained-attention tasks (CPT-2, CPT-3, CPT-4) and the working-memory tasks (DST-F, DST-B, DST-T) than did the healthy control group (Table 2). However, they performed better in long-term memory retrieval tasks, such as recall efficiency and recall correctness, showing a dissociation pattern of “impaired attention control and relatively preserved memory retrieval”. This phenomenon can be explained by dual-process theories in cognitive neuroscience (Spunt, 2015): (a) Attention control networks impaired: sustained attention relies on top-down regulation by the dorsolateral prefrontal cortex (DLPFC). Glucose metabolism reduced in DLPFC during depressive states diminishes inhibitory control over some stimulus factors, increasing error rates in CPT tasks (Meng et al., 2025). (b) Dissociation between memory encoding and retrieval: enhanced response accuracy in the depressive-onset group may relate to preferential processing of negative stimuli. Depressive-state patients exhibited heightened attention bias toward sad faces, leading to the wrong encoding of emotional information. Focusing on sad faces was associated with recent negative life events or other factors (Klawohn et al., 2020).

In contrast, the manic-onset group showed minimal cognitive differences from healthy controls, with slight advantages only in DST-B and CPT-2. This may reflect dopamine-system hyperactivity during mania, enhancing memory encoding, especially memory of novel events (Wang et al., 2023). However, no significant correlation was found between BRMS scores and cognition impairment, suggesting that the hypomania state affects cognition only in specific dimensions, possibly modulated by symptom severity.

Our research found that there were differences in cognitive function among patients with BD of different FEPs. The various indicators of CPT and DST, as well as CC of WCST in the depression-onset group and the mania/hypomania-onset group, were all lower than those in the normal control group. Those indicators in the depression-onset group were all lower than those in the mania/hypomania-onset group, suggesting that the degree of cognitive-function impairment in the depression-onset group was more severe than that in the mania/hypomania-onset group, especially in attention function and memory function. We also found similar characteristics of cognitive impairment in an earlier study (Wang et al., 2023).

A significantly higher family history of mental disorders (30%–31.7%) in patient groups supports possible genetic factors affecting cognitive impairment in BDs. BD has high heritability, so unaffected first-degree relatives and offspring of patients with BD present with certain cognitive dysfunctions (de la Serna et al., 2016). Genetic studies have also indicated that environmental contexts are risk factors for the onset, course, and cognitive impairment of bipolar-spectrum disorders.

Higher education level correlated positively with DST-B and DST-T, suggesting that higher education may enhance synaptic density and reduce cognitive impairment (Jansen et al., 2021). Notably, occupation status negatively correlated with CC, possibly indicating that the demands of the occupational state help maintain cognitive function. One study found that cognitive ability is robustly positively correlated with higher occupational status and enhanced job performance. Cognition (which encompasses processing speed, reasoning, and working-memory capacities) is essential for performing complex and cognitively demanding tasks that entail novel problem-solving and rapid learning (Aichele et al., 2024). This point supports combined “social intervention - cognitive rehabilitation” strategies, in which vocational training could delay cognitive decline. Of course, we believe that the first-episode polarity as a predictor of cognitive impairment in BD may be a clinical hypothesis.

It is important to set up the cognitive screening (e.g., CPT and DST) to identify attention and working-memory deficits, enabling early interventions by cognitive behavior rehabilitation (CBR) to reduce the degree of the cognitive impairment (Razavi et al., 2024; Tsapekos et al., 2023). It is a better intervention plan for improving cognitive impairment by drug combination with physical therapy. Some antidepressants with minimal cognitive side effects and adjunct Repetitive Transcranial Magnetic Stimulation (rTMS) (to enhance prefrontal excitability) may improve working memory. Lifestyle modifications (diet, physical activity, sleep and stress management) (Simjanoski et al., 2023), psychological and behavioral intervention therapy (cognitive remediation therapy, CRT) may alleviate cognitive impairment, enhance cognition, and functioning (Strawbridge et al., 2021).

The sample size of this study was relatively small, and the sample was limited to only one psychiatric hospital in China. Further validation of the clinical characteristics of cognitive impairment in BD patients is required in larger cohorts. The difference in sample size among subgroups affects statistical power and limits generalizability. We did not conduct a power analysis for the sample size. All samples were inpatients (no outpatients); this is a potential recruitment bias. Due to the limited sample size, not all potential confounding factors (e.g., age, number of hospitalizations) could be controlled, and future studies are needed for further validation. The cross-sectional data cannot distinguish between disease state, episode period, and remission period. The correlation results only reflect the co-occurrence relationship between variables and do not indicate the causal direction. In the future, longitudinal studies should be used to track cognitive function level recovered during remission are required. The structural relationships among variables can be further explored through longitudinal studies or Path analysis.