Association between serum non-enzymatic antioxidants and cognitive impairment in drug-naïve patients with schizophrenia Authors: Zhenning Feng, Yuying Qiu, Yanyan Wang, Meijuan Li, Xiaoxiao Sun, Qiao Su, Zaimina Xuekelaiti, Shuo Wang, Ying Xu, Yonghui Zhang, Jie Li Published: 2026 ABSTRACT BackgroundCognitive impairment (CI) is a core feature of schizophrenia (SZ) and is strongly associated with poor functional outcomes. Oxidative stress and impaired antioxidant defense are implicated in the pathogenesis of CI across psychiatric disorders. But evidence in drug-naïve SZ remains limited.MethodsSerum levels of non-enzymatic antioxidants, including high-density lipoprotein cholesterol (HDL-C), albumin (ALB), total bilirubin (TBIL), and uric acid (UA), were measured in drug-naïve patients with SZ and healthy controls (HCs). Cognitive function was assessed using the MATRICS Consensus Cognitive Battery (MCCB), and psychiatric symptoms using the Positive and Negative Syndrome Scale (PANSS). Multivariable logistic regression models were used to evaluate associations between antioxidants and CI, with interaction and stratified analyses based on demographic and clinical variables. Linear regression analyses were used to evaluate the relationships between antioxidants and specific cognitive domains.ResultsA total of 219 drug-naïve patients with SZ and 101 HCs were included. CI was identified in 115 SZ patients (52.5%). Serum levels of HDL-C (p = 0.004) and ALB (p = 0.001) were significantly lower in patients with SZ than in HCs, whereas UA (p = 0.001) was higher. Within the SZ group, patients with CI showed lower levels of HDL-C (p = 0.019) and ALB (p = 0.033) compared with those without CI. Logistic regression analyses showed that higher HDL-C (OR:0.30, 95% CI: 0.11–0.82) and ALB (OR:0.91, 95% CI: 0.84–0.99) levels were independently associated with lower odds of CI. In addition, linear regression analyses revealed that HDL-C (p = 0.025) and ALB (p = 0.008) were positively associated with overall cognitive performance, and ALB was also associated with verbal learning (q = 0.007).ConclusionThis study suggests that higher serum HDL-C and ALB levels are associated with lower odds of CI and better overall cognitive performance in drug-naïve patients with SZ. Background Schizophrenia (SZ) is a severe and chronic psychiatric disorder characterized by profound disturbances in cognition, emotional regulation, and perception of reality [1]. Recent epidemiological surveys of mental disorders in China estimate the lifetime prevalence of SZ to be approximately 0.6% [2]. Despite advances in pharmacological and physical therapies, only a minority of patients achieve clinical recovery, whereas most experience persistent functional impairment and long-term disability, imposing a substantial burden on families and society [3, 4]. Cognitive impairment (CI) is a core feature of SZ, characterized by deficits in multiple domains of neurocognition, including memory, attention, executive function, processing speed, and social cognition [5]. It has been reported that approximately 80% of individuals with SZ experience varying degrees of CI, which contributes to poor treatment adherence and difficulties in returning to family and social life [6, 7]. Consequently, improving cognitive function has become a major focus in the development of novel pharmacological treatments and rehabilitation strategies for SZ [8]. However, the pathophysiological mechanisms underlying CI in SZ remain incompletely understood. Several mechanisms have been proposed, including alterations in cortical thickness in the frontal and temporal regions [9], reduced function of N-methyl-D-aspartate receptors (NMDAR), and impaired synaptic plasticity [10, 11]. Immune dysregulation has received increasing attention in psychiatric disorders [12, 13]. One proposed consequence of immune dysregulation is the induction of oxidative stress, as immune activation in the central nervous system (CNS) often leads to excessive production of reactive oxygen species (ROS) [14], which can damage macromolecules such as DNA, proteins, and membrane lipids, ultimately resulting in neuronal injury and impaired neurotransmission, processes that have been implicated in a variety of psychiatric conditions [7, 15, 16]. The antioxidant defense system, which counteracts excessive ROS, comprises both enzymatic and non-enzymatic components. Enzymatic antioxidants include superoxide dismutase, catalase, glutathione peroxidase, and other heme protein peroxidases [17]. Non-enzymatic antioxidants include high-density lipoprotein cholesterol (HDL-C), albumin (ALB), uric acid (UA), and bilirubin [18, 19]. Previous studies have demonstrated that the total antioxidant capacity in both peripheral and the CNS is reduced in patients with SZ [20]. Impairment of the antioxidant defense system can lead to excessive generation of free radicals, resulting in oxidative damage to neuronal membranes, subsequent neuronal dysfunction, and disruption of neurotransmission, thereby contributing to the pathophysiology of SZ [15, 21]. Multiple studies in Alzheimer’s disease (AD), Parkinson’s disease, and idiopathic REM sleep behavior disorder indicate that higher circulating levels of non-enzymatic antioxidants—including bilirubin, ALB, HDL-C, and UA—are linked to preserved cognitive function [22–24]. However, studies examining the relationship between non-enzymatic antioxidants and cognitive function in individuals with SZ remain limited. Therefore, the present study aimed to investigate the association between serum non-enzymatic antioxidant levels and cognitive performance in Chinese drug-naïve patients with SZ. We hypothesized that: (1) higher levels of non-enzymatic antioxidants would be associated with lower odds of CI; and (2) non-enzymatic antioxidant levels would correlate with performance across multiple cognitive domains. Methods Study subjects and enrollment In this study, patients with SZ were recruited from inpatient and outpatient services, while healthy controls (HCs) were recruited from surrounding communities. Inclusion criteria for the patients were: (1) diagnosis of SZ or schizophreniform disorder per Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5) criteria; (2) age 18–65 years; (3) at least primary school education; (4) the Positive and Negative Syndrome Scale (PANSS) total score ≥ 60; (5) Virtually antipsychotic-naïve, defined as < 28 cumulative days of total exposure and < 14 days of recent continuous use. Exclusion criteria were: (1) other DSM-5-defined psychiatric disorders; (2) severe or unstable physical illnesses; (3) prior physical interventions (e.g., rTMS or ECT); (4) pregnancy or lactation; (5) history of alcohol/substance abuse. Inclusion criteria for the HCs group were as follows: (1) age 18–65 years; (2) no history of mental illness. Exclusion criteria were: (1) Family history of mental illness; (2) serious physical, neurological, immune system diseases, or mental retardation; (3) alcohol and drug dependence; (4) pregnant or breastfeeding women. Laboratory and physical examinations Based on previous reports of their association with CI, four non-enzymatic antioxidants—HDL-C, ALB, UA, and total bilirubin (TBIL)—were selected for measurement [22–24]. Blood samples were collected from all participants via the antecubital vein after an overnight fast. Serum levels of HDL-C, ALB, TBIL, and UA, as well as plasma glucose (GLU), were measured using the Toshiba TBA120FR system (Canon, Tokyo, Japan). Height and weight were measured by trained clinical nurses using standardized instruments, and body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Clinical measures Demographic information, including age, sex, marital status, education level, and smoking status, was collected from all participants using a standardized questionnaire. For the assessment of symptoms and cognitive function in the patient group: The PANSS was used to assess psychiatric symptoms [25], while the MATRICS Consensus Cognitive Battery (MCCB) was employed to evaluate cognitive function [26]. The MCCB primarily assesses seven cognitive domains: speed of processing (SOP), attention/ vigilance (AV), working memory (WM), verbal learning (Verb), visual learning (Vis), reasoning and problem solving (RPS), and social cognition (SC). Raw scores for each MCCB item were first converted into standardized T-scores using a computerized scoring program, with subsequent adjustment for sex, age, and education level to generate corrected T-scores. The Chinese version of the MCCB has been demonstrated to possess satisfactory reliability and validity [27]. In this study, a corrected MCCB total T-score of 30 was used as the cut-off. In studies of the Chinese population, the normative mean for healthy adults is 50 with a standard deviation (SD) of 10; therefore, scores below 30 (2 SD below the mean) are considered indicative of significant CI, reflecting widespread cognitive deficits [27]. Participants scoring below 30 were classified as having CI, whereas those scoring 30 or above were classified as non-cognitive impairment (NCI). To ensure consistency and reliability, all evaluators underwent standardized training in the administration of the PANSS and MCCB. Interrater reliability for the PANSS was assessed through repeated ratings and showed an intraclass correlation coefficient greater than 0.80. Statistical analysis For group comparisons of demographic and clinical characteristics, t-tests were used for continuous variables, chi-square tests for categorical variables, and Mann-Whitney U tests for non-normally distributed variables. Normally distributed continuous variables are expressed as mean ± standard deviation (SD), while non-normally distributed continuous variables are presented as interquartile range (IQR, Q1-Q3). For the analysis of covariance (ANCOVA), these variables are presented as mean ± standard error (SE). Categorical variables are displayed as frequencies and percentages. For comparing serum levels of non-enzymatic antioxidants between the HC group and the SZ group, as well as between the NCI and CI groups, ANCOVA was applied, with age, sex, education, marital status, BMI, smoking status, and GLU as covariates. In addition, given the potential influence of diet and physical activity on non-enzymatic antioxidant levels [28, 29], patients were hospitalized during the assessment period, during which diet and physical activity were relatively standardized and generally consistent with those of the general population, although detailed dietary and activity assessments were not available. Notably, because the corrected MCCB T-scores were already adjusted for age, education, and sex, unadjusted T-scores were used in the logistic and linear regression analyses to avoid double adjustment. Multivariable logistic regression models were used to evaluate the association between each non-enzymatic antioxidant and the odds of CI. Both unadjusted and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. For continuous variables, ORs were calculated per 1 mmol/L increase for HDL-C, per 1 g/L increase for ALB, per 10 µmol/L increase for UA, and per 1 µmol/L increase for TBIL. Confounding factors were selected based on their association with the outcome of interest or their ability to change the effect estimate by more than 10%. To examine the associations, we conducted a series of sequential regression models. Model 1 was unadjusted. Model 2 adjusted for sociodemographic characteristics, including age, sex, education, marital status, and smoking status [30]. Model 3 built on Model 2 by additionally adjusting for physical health factors (blood GLU and BMI) [31]. Model 4 served as the fully adjusted model, further including psychological status as measured by PANSS negative symptom score, which has been linked to cognitive performance in drug-naïve patients with SZ [32]. Since negative symptoms could be part of the causal pathway to CI, we conducted a sensitivity analysis excluding PANSS negative symptom score to prevent overadjustment. In addition, interaction and stratified analyses were conducted across subgroups defined by age (less than 30 years versus 30 years or older), sex, marital status (single, married, or divorced), education (less than 16 years versus 16 years or more), current smoking status (yes versus no), and BMI (less than 23 kg/m² versus 23 kg/m² or higher). Heterogeneity between subgroups was evaluated using multivariable logistic regression, and interactions between subgroups and non-enzymatic antioxidant levels were tested with likelihood ratio tests. Associations between serum non-enzymatic antioxidant levels and cognitive function were further examined using multiple linear regression analyses. FDR correction for multiple comparisons was applied to the seven cognitive domains of the MCCB, whereas overall cognitive performance (total score) was analyzed without correction as it was defined a priori as the primary outcome. All statistical analyses were conducted using R software (version 4.3.0; R Foundation for Statistical Computing, Vienna, Austria), available at https://www.r-project.org, and IBM SPSS version 27.0 (IBM Corp., Armonk, NY, USA). Statistical significance was defined as two-tailed p < 0.05 for individual tests. For analyses involving multiple comparisons, the false discovery rate (FDR) was controlled, and q values < 0.05 were considered significant. Results Demographics and clinical characteristics A total of 219 patients with drug-naïve SZ and 101 HCs were included in this study (Fig. 1). Demographic comparisons between the SZ and HC groups are presented in Table 1. The SZ group had a significantly lower marriage rate (p = 0.001) and BMI (p = 0.001) compared with HCs. Demographic and clinical characteristics of CI and NCI patients are summarized in Table 2. Among the 219 SZ patients, 115 (52.5%) were classified as CI and 104 (47.5%) as NCI. No significant differences were observed between the CI and NCI subgroups regarding age, sex, education, marital status, smoking status, BMI, GLU, or PANSS negative score. Fig. 1Flowchart of participant screening and enrollment. Seven patients were excluded due to missing uric acid data because uric acid measurement was not routinely recorded in the laboratory system during the early stage of the study Flowchart of participant screening and enrollment. Seven patients were excluded due to missing uric acid data because uric acid measurement was not routinely recorded in the laboratory system during the early stage of the study Table 1Comparison of demographic characteristics between patients with SZ and HCsVariablesSZ (n = 219)HCs (n = 101)t / χ²/Z p Age (years)36.00 (28.50, 44.00)35.00 (30.00, 41.00)-0.430.664Education (years)12.00 (9.00, 16.00)15.00 (9.00, 16.00)-1.400.161Sex0.530.467Female86 (39.27)44 (43.56)Male133 (60.73)57 (56.44)Marital status43.330.001**Single151 (68.95)30 (29.70)Married68 (31.05)71 (70.30)Currently smoking1.290.256No175 (79.91)75 (74.26)Yes44 (20.09)26 (25.74)BMI (kg/m2)22.06 (19.81, 24.73)23.73 (21.30, 27.46)-3.550.001**GLU (mmol/L)5.14 ± 1.345.28 ± 1.210.880.381Abbreviations: BMI: body mass index; GLU: Glucose; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Table 1Comparison of demographic characteristics between patients with SZ and HCsVariablesSZ (n = 219)HCs (n = 101)t / χ²/Z p Age (years)36.00 (28.50, 44.00)35.00 (30.00, 41.00)-0.430.664Education (years)12.00 (9.00, 16.00)15.00 (9.00, 16.00)-1.400.161Sex0.530.467Female86 (39.27)44 (43.56)Male133 (60.73)57 (56.44)Marital status43.330.001**Single151 (68.95)30 (29.70)Married68 (31.05)71 (70.30)Currently smoking1.290.256No175 (79.91)75 (74.26)Yes44 (20.09)26 (25.74)BMI (kg/m2)22.06 (19.81, 24.73)23.73 (21.30, 27.46)-3.550.001**GLU (mmol/L)5.14 ± 1.345.28 ± 1.210.880.381Abbreviations: BMI: body mass index; GLU: Glucose; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Comparison of demographic characteristics between patients with SZ and HCs Abbreviations: BMI: body mass index; GLU: Glucose; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Table 2Comparison of demographic characteristics between the CI group and the NCI group in patients with SZVariablesNCI group (n = 104)CI group (n = 115)t / χ²/Z p Age (years)37.00 (29.75, 46.00)34.00 (28.00, 41.501.700.090Education (years)12.00 (9.00, 16.00)12.00 (9.50, 16.00)-0.830.406Sex2.910.088Female57 (54.81)76 (66.09)Male47 (45.19)39 (33.91)Marital status0.810.667Unmarried51 (49.04)61 (53.04)Married32 (30.77)36 (31.30)Divorced21 (20.19)18 (15.65)Currently smoking0.140.709No82 (78.85)93 (80.87)Yes22 (21.15)22 (19.13)BMI (kg/m2)21.94 (20.06, 24.49)22.15 (19.61, 24.84)-0.080.940GLU (mmol/L)5.18 ± 1.485.12 ± 1.210.330.739PANSS negative score19.84 ± 6.2920.91 ± 6.47-1.250.214Note: Participants were categorized based on a cutoff score of 30 on the MCCB total score, with scores < 30 defined as the CI group and scores ≥ 30 as the NCI groupAbbreviations: BMI: body mass index; t: t-test; CI: cognitive impairment; GAF: Global Assessment of Functioning scale; GLU: Glucose; NCI: non-cognitive impairment; SZ: schizophrenia; PANSS: Positive and Negative Syndrome Scale Table 2Comparison of demographic characteristics between the CI group and the NCI group in patients with SZVariablesNCI group (n = 104)CI group (n = 115)t / χ²/Z p Age (years)37.00 (29.75, 46.00)34.00 (28.00, 41.501.700.090Education (years)12.00 (9.00, 16.00)12.00 (9.50, 16.00)-0.830.406Sex2.910.088Female57 (54.81)76 (66.09)Male47 (45.19)39 (33.91)Marital status0.810.667Unmarried51 (49.04)61 (53.04)Married32 (30.77)36 (31.30)Divorced21 (20.19)18 (15.65)Currently smoking0.140.709No82 (78.85)93 (80.87)Yes22 (21.15)22 (19.13)BMI (kg/m2)21.94 (20.06, 24.49)22.15 (19.61, 24.84)-0.080.940GLU (mmol/L)5.18 ± 1.485.12 ± 1.210.330.739PANSS negative score19.84 ± 6.2920.91 ± 6.47-1.250.214Note: Participants were categorized based on a cutoff score of 30 on the MCCB total score, with scores < 30 defined as the CI group and scores ≥ 30 as the NCI groupAbbreviations: BMI: body mass index; t: t-test; CI: cognitive impairment; GAF: Global Assessment of Functioning scale; GLU: Glucose; NCI: non-cognitive impairment; SZ: schizophrenia; PANSS: Positive and Negative Syndrome Scale Comparison of demographic characteristics between the CI group and the NCI group in patients with SZ Note: Participants were categorized based on a cutoff score of 30 on the MCCB total score, with scores < 30 defined as the CI group and scores ≥ 30 as the NCI group Abbreviations: BMI: body mass index; t: t-test; CI: cognitive impairment; GAF: Global Assessment of Functioning scale; GLU: Glucose; NCI: non-cognitive impairment; SZ: schizophrenia; PANSS: Positive and Negative Syndrome Scale Comparison of serum non-enzymatic antioxidant levels among different groups. Serum levels of non-enzymatic antioxidants in SZ patients and HCs are shown in Fig. 2, with ANCOVA results summarized in Table 3. Compared with the HC group, SZ patients had significantly lower serum levels of HDL-C (p = 0.004) and ALB (p = 0.001), while their serum level of UA (p = 0.001) was significantly higher. Fig. 2Distribution of serum ABL, UA, HDL-C, TBIL between patients with SZ and HCs. Abbreviations: ALB: albumin; HC: Health controls; HDL-C: high-density lipoprotein cholesterol; SZ: schizophrenia; TBIL: total bilirubin; UA: uric acid; ** p < 0.01 Distribution of serum ABL, UA, HDL-C, TBIL between patients with SZ and HCs. Abbreviations: ALB: albumin; HC: Health controls; HDL-C: high-density lipoprotein cholesterol; SZ: schizophrenia; TBIL: total bilirubin; UA: uric acid; ** p < 0.01 Table 3Covariance analysis of serum non-enzymatic antioxidants levels between patients with SZ and HCsVariablesSZ (n = 219)HCs (n = 101)F p HDL-C (mmol/L)1.31 ± 0.021.43 ± 0.038.240.004**TBIL (µmol/L)12.56 ± 0.4013.04 ± 0.610.400.527ALB (g/L)43.63 ± 0.2446.82 ± 0.3649.790.001**UA (µmol/L)321.44 ± 6.65268.80 ± 10.2117.190.001**Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariatesAbbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; TBIL: total bilirubin; UA: uric acid; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Table 3Covariance analysis of serum non-enzymatic antioxidants levels between patients with SZ and HCsVariablesSZ (n = 219)HCs (n = 101)F p HDL-C (mmol/L)1.31 ± 0.021.43 ± 0.038.240.004**TBIL (µmol/L)12.56 ± 0.4013.04 ± 0.610.400.527ALB (g/L)43.63 ± 0.2446.82 ± 0.3649.790.001**UA (µmol/L)321.44 ± 6.65268.80 ± 10.2117.190.001**Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariatesAbbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; TBIL: total bilirubin; UA: uric acid; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Covariance analysis of serum non-enzymatic antioxidants levels between patients with SZ and HCs Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariates Abbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; TBIL: total bilirubin; UA: uric acid; SZ: schizophrenia; HCs: health controls; ** p < 0.01 Serum levels of non-enzymatic antioxidants in the CI and NCI groups are shown in Fig. 3, with results summarized in Table 4. ANCOVA results indicated that the CI group had significantly lower serum levels of HDL-C (p = 0.019) and ALB (p = 0.033) than the NCI group. Fig. 3Distribution of serum ABL, UA, HDL-C, TBIL between CI group and NCI group in patients with SZ. Abbreviations: ALB: albumin; CI: cognitive impairment; HDL-C: high-density lipoprotein cholesterol; NCI: non-cognitive impairment; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Distribution of serum ABL, UA, HDL-C, TBIL between CI group and NCI group in patients with SZ. Abbreviations: ALB: albumin; CI: cognitive impairment; HDL-C: high-density lipoprotein cholesterol; NCI: non-cognitive impairment; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Table 4Covariance analysis of serum non-enzymatic antioxidants levels between the CI group and the NCI group in patients with SZVariablesNCI group (n = 104)CI group (n = 115)F p HDL-C (mmol/L)1.37 ± 0.031.27 ± 0.025.570.019*TBIL (µmol/L)13.23 ± 0.6012.03 ± 0.572.060.153ALB (g/L)44.09 ± 0.3443.09 ± 0.324.600.033*UA (µmol/L)323.73 ± 9.84309.77 ± 9.351.050.308Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariatesAbbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; NCI: non-cognitive impairment t; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Table 4Covariance analysis of serum non-enzymatic antioxidants levels between the CI group and the NCI group in patients with SZVariablesNCI group (n = 104)CI group (n = 115)F p HDL-C (mmol/L)1.37 ± 0.031.27 ± 0.025.570.019*TBIL (µmol/L)13.23 ± 0.6012.03 ± 0.572.060.153ALB (g/L)44.09 ± 0.3443.09 ± 0.324.600.033*UA (µmol/L)323.73 ± 9.84309.77 ± 9.351.050.308Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariatesAbbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; NCI: non-cognitive impairment t; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Covariance analysis of serum non-enzymatic antioxidants levels between the CI group and the NCI group in patients with SZ Note: Covariance analysis was performed with sex, age, education, marital status, smoking status, BMI, and GLU as covariates Abbreviations: ALB: albumin; CI: cognitive impairment; GLU: Glucose; HDL-C: high-density lipoprotein cholesterol; NCI: non-cognitive impairment t; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Association between serum non-enzymatic antioxidant levels and CI in drug-naïve patients with SZ Table 5 presents the associations between serum non-enzymatic antioxidant levels and the incidence of CI in drug-naïve patients with SZ, as determined by multivariable logistic regression analyses. In the unadjusted Model 1, HDL-C (OR: 0.35, 95% CI: 0.14–0.87 per 1 mmol/L increase) and ALB (OR: 0.93, 95% CI: 0.86–0.99 per 1 g/L increase) levels were significantly associated with CI, while the association between serum TBIL levels and CI was marginally significant (OR: 0.96, 95% CI: 0.92–1.00 per 1 µmol/L increase). After adjusting for sex, age, education, marital status, and smoking status in Model 2, the associations of HDL-C (OR: 0.29, 95% CI: 0.11–0.79 per 1 mmol/L increase) and ALB (OR: 0.92, 95% CI: 0.85–0.99 per 1 g/L increase) levels with CI remained statistically significant. Further adjustments for GLU and BMI in Model 3 showed that HDL-C (OR: 0.30, 95% CI: 0.11–0.83 per 1 mmol/L increase) and ALB (OR: 0.91, 95% CI: 0.84–0.99 per 1 g/L increase) levels maintained their significant associations. In the fully adjusted Model 4, higher serum HDL-C levels were significantly associated with lower odds of CI (OR = 0.30, 95% CI: 0.11–0.82 per 1 mmol/L increase), and higher serum ALB levels were also significantly associated with lower odds of CI (OR = 0.91, 95% CI: 0.84–0.99 per 1 g/L increase), whereas the association between serum TBIL levels and CI was marginally significant (OR = 0.96, 95% CI: 0.91–1.01 per 1 µmol/L increase). In both univariate and all multivariable logistic regression models, the association of serum UA levels with the incidence of CI was not statistically significant. Table 5Multivariable logistic regression analysis evaluating the associations between serum non-enzymatic antioxidants levels and the incidence of CI in patients with SZVariablesModel 1Model 2Model 3Model 4OR (95%CI) p OR (95%CI) p OR (95%CI) p OR (95%CI) p HDL-C (mmol/L)0.35 (0.14–0.87)0.024*0.29 (0.11–0.79)0.016*0.30 (0.11–0.83)0.020*0.30 (0.11–0.82)0.019*TBIL (µmol/L)0.96 (0.92–1.00)0.0770.97 (0.92–1.01)0.1340.97 (0.92–1.01)0.1470.96 (0.91–1.01)0.076ALB (g/L)0.93 (0.86–0.99)0.042*0.92 (0.85–0.99)0.042*0.91 (0.84–0.99)0.032*0.91 (0.84–0.99)0.027*UA (µmol/L)0.99 (0.96–1.01)0.2070.99 (0.96–1.02)0.4330.99 (0.96–1.01)0.3080.99 (0.96–1.01)0.328Notes: Model 1: Crude; Model 2: Adjust: Sex, Age, Education, Marital status, Smoking status; Model 3: Adjust: Model2་BMI, GLU; Model 4: Adjust: Model3་PANSS negative score; Abbreviations: ALB: albumin; CI: confidence interval; HDL-C: high-density lipoprotein cholesterol; CI: cognitive impairment; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Table 5Multivariable logistic regression analysis evaluating the associations between serum non-enzymatic antioxidants levels and the incidence of CI in patients with SZVariablesModel 1Model 2Model 3Model 4OR (95%CI) p OR (95%CI) p OR (95%CI) p OR (95%CI) p HDL-C (mmol/L)0.35 (0.14–0.87)0.024*0.29 (0.11–0.79)0.016*0.30 (0.11–0.83)0.020*0.30 (0.11–0.82)0.019*TBIL (µmol/L)0.96 (0.92–1.00)0.0770.97 (0.92–1.01)0.1340.97 (0.92–1.01)0.1470.96 (0.91–1.01)0.076ALB (g/L)0.93 (0.86–0.99)0.042*0.92 (0.85–0.99)0.042*0.91 (0.84–0.99)0.032*0.91 (0.84–0.99)0.027*UA (µmol/L)0.99 (0.96–1.01)0.2070.99 (0.96–1.02)0.4330.99 (0.96–1.01)0.3080.99 (0.96–1.01)0.328Notes: Model 1: Crude; Model 2: Adjust: Sex, Age, Education, Marital status, Smoking status; Model 3: Adjust: Model2་BMI, GLU; Model 4: Adjust: Model3་PANSS negative score; Abbreviations: ALB: albumin; CI: confidence interval; HDL-C: high-density lipoprotein cholesterol; CI: cognitive impairment; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Multivariable logistic regression analysis evaluating the associations between serum non-enzymatic antioxidants levels and the incidence of CI in patients with SZ Notes: Model 1: Crude; Model 2: Adjust: Sex, Age, Education, Marital status, Smoking status; Model 3: Adjust: Model2་BMI, GLU; Model 4: Adjust: Model3་PANSS negative score; Abbreviations: ALB: albumin; CI: confidence interval; HDL-C: high-density lipoprotein cholesterol; CI: cognitive impairment; TBIL: total bilirubin; SZ: schizophrenia; UA: uric acid; * p < 0.05 Interaction and stratified analyses based on additional variables We additionally performed interaction and stratified analyses to examine potential effect modifications in the associations between serum non-enzymatic antioxidant levels and CI in drug-naïve patients with SZ. The relationships of HDL‑C and ALB with CI are illustrated in Figs. 4 and 5. The associations identified in the multivariable logistic regression analyses remained consistent across subgroups, and no significant interactions were observed for age, sex, education level, marital status, smoking status, or BMI. Fig. 4The relationship between serum HDL-C levels and CI in patients with SZ. The p-value for interaction represents the likelihood of an interaction between the variable and HDL-C levels. Each stratification was adjusted for sex, age, education, marital status, smoking status, and BMI (excluding the stratification variable). Abbreviations: BMI: body mass index; HDL-C: high-density lipoprotein cholesterol; SZ: schizophrenia; CI: cognitive impairment The relationship between serum HDL-C levels and CI in patients with SZ. The p-value for interaction represents the likelihood of an interaction between the variable and HDL-C levels. Each stratification was adjusted for sex, age, education, marital status, smoking status, and BMI (excluding the stratification variable). Abbreviations: BMI: body mass index; HDL-C: high-density lipoprotein cholesterol; SZ: schizophrenia; CI: cognitive impairment Fig. 5The relationship between serum ALB levels and CI in patients with SZ. The p-value for interaction represents the likelihood of an interaction between the variable and ALB levels. Each stratification was adjusted for sex, age, education, marital status, smoking status, and BMI (excluding the stratification variable). Abbreviations: ALB: albumin; BMI: body mass index; SZ: schizophrenia; CI: cognitive impairment The relationship between serum ALB levels and CI in patients with SZ. The p-value for interaction represents the likelihood of an interaction between the variable and ALB levels. Each stratification was adjusted for sex, age, education, marital status, smoking status, and BMI (excluding the stratification variable). Abbreviations: ALB: albumin; BMI: body mass index; SZ: schizophrenia; CI: cognitive impairment Sensitivity analyses In sensitivity analyses, PANSS negative score was removed from the fully adjusted multivariable logistic regression model (Model 4). The associations of HDL-C (OR: 0.30, 95% CI: 0.11–0.83 per 1 mmol/L increase) and ALB (OR: 0.91, 95% CI: 0.84–0.99 per 1 g/L increase) with CI in drug-naïve patients with SZ remained statistically significant. Associations between serum non-enzymatic antioxidants levels and cognitive function in drug-naïve patients with SZ To explore domain-specific cognitive effects, we performed multiple linear regression analyses to assess the associations between serum non-enzymatic antioxidant levels and the seven cognitive domains of the MCCB in drug-naïve patients with SZ, adjusting for sex, age, education, marital status, smoking status, PANSS negative score, and BMI. As summarized in Table 6, higher HDL-C levels were significantly associated with better overall cognitive performance (p = 0.025), and higher ALB levels were significantly associated with overall cognitive performance (p = 0.008). In addition, higher ALB levels were significantly associated with Verb after FDR correction for multiple comparisons (q = 0.007). Table 6The associations between serum non-enzymatic antioxidants levels and the all dimensions of the MCCB in patients with SZVariablesHDL-C (mmol/L)TBIL (µmol/L)ALB (g/L)UA (µmol/L)β p q β p q β p q β p q SOP4.460.1130.1580.230.1060.7420.300.2250.2250.010.7370.737AV4.830.0720.1260.160.2280.5320.360.1270.191-0.010.6110.737WM5.150.045*0.1050.120.3450.5320.320.1470.1910.010.1690.574Verb6.990.015*0.1050.100.5000.4940.790.001**0.007**0.010.2600.574Vis0.920.7710.7710.130.3910.5320.380.1640.1910.020.0750.525RPS1.340.5840.6810.100.4230.4940.420.045*0.1580.010.6830.737SC3.770.030*0.1050.110.2230.7420.240.1120.191-0.010.3280.574TS6.390.025*-0.220.115-0.650.008**-0.010.337-Note: Multiple linear regression analysis was performed to examine the association between serum non-enzymatic antioxidant levels and cognitive function after adjusting for sex, age, education, marital status, smoking status, PANSS negative score, and BMIAbbreviations: ALB: albumin; AV: Attention/Vigilance; HDL-C: high-density lipoprotein cholesterol; RPS: reasoning and problem solving; SC: social cognition; SOP: speed of processing; TS: MCCB total score; TBIL: total bilirubin; UA: uric acid; Verb: verbal learning; Vis: visual learning; WM: working memory; * p or q < 0.05; ** p < 0.01 Table 6The associations between serum non-enzymatic antioxidants levels and the all dimensions of the MCCB in patients with SZVariablesHDL-C (mmol/L)TBIL (µmol/L)ALB (g/L)UA (µmol/L)β p q β p q β p q β p q SOP4.460.1130.1580.230.1060.7420.300.2250.2250.010.7370.737AV4.830.0720.1260.160.2280.5320.360.1270.191-0.010.6110.737WM5.150.045*0.1050.120.3450.5320.320.1470.1910.010.1690.574Verb6.990.015*0.1050.100.5000.4940.790.001**0.007**0.010.2600.574Vis0.920.7710.7710.130.3910.5320.380.1640.1910.020.0750.525RPS1.340.5840.6810.100.4230.4940.420.045*0.1580.010.6830.737SC3.770.030*0.1050.110.2230.7420.240.1120.191-0.010.3280.574TS6.390.025*-0.220.115-0.650.008**-0.010.337-Note: Multiple linear regression analysis was performed to examine the association between serum non-enzymatic antioxidant levels and cognitive function after adjusting for sex, age, education, marital status, smoking status, PANSS negative score, and BMIAbbreviations: ALB: albumin; AV: Attention/Vigilance; HDL-C: high-density lipoprotein cholesterol; RPS: reasoning and problem solving; SC: social cognition; SOP: speed of processing; TS: MCCB total score; TBIL: total bilirubin; UA: uric acid; Verb: verbal learning; Vis: visual learning; WM: working memory; * p or q < 0.05; ** p < 0.01 The associations between serum non-enzymatic antioxidants levels and the all dimensions of the MCCB in patients with SZ Note: Multiple linear regression analysis was performed to examine the association between serum non-enzymatic antioxidant levels and cognitive function after adjusting for sex, age, education, marital status, smoking status, PANSS negative score, and BMI Abbreviations: ALB: albumin; AV: Attention/Vigilance; HDL-C: high-density lipoprotein cholesterol; RPS: reasoning and problem solving; SC: social cognition; SOP: speed of processing; TS: MCCB total score; TBIL: total bilirubin; UA: uric acid; Verb: verbal learning; Vis: visual learning; WM: working memory; * p or q < 0.05; ** p < 0.01 Discussion In this cross-sectional study, several important findings were observed in drug-naïve patients with SZ. First, compared with HCs, SZ patients exhibited significantly lower peripheral levels of HDL-C and ALB, along with significantly higher serum levels of UA. And compared with SZ patients without CI, those with CI had significantly lower serum levels of HDL-C and ALB. Second, serum HDL-C and ALB levels showed significant negative associations with the odds of CI, and these associations remained robust after adjusting for potential confounders and in sensitivity analyses excluding PANSS negative score. Lastly, serum HDL-C levels were positively correlated with overall cognitive function, and serum ALB levels were positively correlated with both overall cognitive function and Verb. First, our findings indicate that 52.5% of drug-naïve SZ patients exhibited CI, closely aligning with previous reports that 50–70% of SZ patients experience moderate-to-severe cognitive deficits [33]. Another key finding of this study was that patients with SZ had lower serum HDL-C levels than HCs, consistent with previous reports [34]. Notably, evidence has also shown that in patients experiencing a first episode of SZ, treatment with antipsychotic medications that do not induce weight gain is associated with increased serum HDL-C levels, which in turn correlate with improvements in psychotic symptoms [35]. Extending these findings, the present study further demonstrated that, among drug-naïve patients with SZ, those with CI had significantly lower serum HDL-C levels than those without CI. Moreover, after adjusting for potential confounding variables, higher serum HDL-C levels remained independently associated with reduced odds of CI in drug-naïve patients with SZ and were also linked to better overall cognitive performance. This observation is consistent with accumulating evidence suggesting a positive relationship between HDL-C levels and cognitive function in both healthy individuals and patients with psychiatric disorders [36, 37]. Mechanistically, HDL-C exerts neuroprotective effects primarily by buffering oxidative stress and promoting neuroplasticity [38]. For oxidative defense, HDL-C and its associated enzyme paraoxonase 1 degrade neurotoxic ROS and lipid peroxides, shielding neuronal membranes from oxidative damage [39]. Regarding neuroplasticity, HDL-C regulates localized cholesterol efflux via ABCA1 and SR-B1 receptors, a process essential for synaptic membrane remodeling [40]. Furthermore, HDL-C activates intracellular cascades (e.g., PI3K/Akt and MAPK), which directly upregulate Brain-Derived Neurotrophic Factor expression and enhance dendritic spine density [40]. Given that serum HDL-C levels are associated with HDL-C concentrations in the CSF [41], the observed associations between serum HDL-C and cognitive function in our study may reflect underlying CNS mechanisms. Previous research has shown that serum ALB levels are lower in patients with SZ compared with HCs [42], which is consistent with our findings. Furthermore, among drug-naïve SZ patients, those with CI exhibited significantly lower serum ALB levels than those without CI. Importantly, in the full cohort of drug-naïve SZ patients, higher serum ALB levels were independently associated with reduced odds of CI (p = 0.027 in the fully adjusted model) and with better overall cognitive performance, particularly in the Verb domain. Evidence from both human studies in AD populations and animal experiments indicates that increasing ALB levels—either via intravenous administration or pharmacological interventions—can improve cognitive performance, including memory and verbal abilities [43–45]. ALB protects cognition by coupling antioxidant defense with synaptic plasticity. Primarily via its cysteine-34 residue, ALB neutralizes ROS [46]. This scavenging prevents oxidative damage to NMDARs, whose redox-induced hypofunction would otherwise disrupt long-term potentiation—the foundation of neuroplasticity and memory [30]. Consistently, NMDAR antagonists precipitate profound spatial learning impairments in mouse models [47, 48]. Given the strong correlation between serum and CSF ALB levels [49], our peripheral ALB findings reliably reflect underlying CNS mechanisms that buffer against CI. Bilirubin possesses potent antioxidant properties [50]. Yin et al. reported a positive correlation between serum TBIL levels and immediate memory among patients with SZ [51]. However, in our study, neither significant group differences in serum TBIL levels nor significant associations between TBIL and cognitive function were observed. This discrepancy may be partly explained by differences in sample size, cognitive assessment methods, and disease stage between the two cohorts, as Yin et al. employed the RBANS scale and included medicated SZ patients, whereas the present study used the MCCB and focused on drug-naïve SZ patients. UA is an important endogenous antioxidant that can reduce oxidative stress in the CNS and protect the brain from damage [52]. In the present study, serum UA levels were significantly higher in patients with SZ than in HCs, consistent with previous findings [18]. However, among SZ patients, no significant differences in serum UA levels were observed between the CI and NCI groups, and no significant associations were detected between UA levels and cognitive function. This lack of association, despite the elevated UA levels in the SZ group, suggests that the endogenous antioxidant capacity of UA may simply be insufficient to buffer against the overwhelming neuroinflammatory and developmental cascades that drive CI during the early stages of the illness [53]. Furthermore, the relatively restricted variance of UA levels within our specific drug-naïve cohort may limit the statistical power to detect subtle linear correlations. The brain is highly vulnerable to oxidative stress [54]; thus, maintaining redox homeostasis is of critical importance. In SZ, immune dysregulation is increasingly recognized as a key pathogenic mechanism [12, 13], and immune activation and cell infiltration may further exacerbate oxidative stress [14]. Accordingly, regulation of immune dysregulation may also be of relevance. Consistent with previous reports, patients with SZ in our cohort exhibited significantly lower marriage rates compared with HCs, reflecting impaired social functioning [55]. Although BMI was significantly lower in SZ patients than in HCs, the median BMI (22.06 kg/m²) remained within the normal range for Chinese adults [56]. Therefore, to minimize potential confounding effects, both variables were included as covariates in all relevant analyses. The Chinese and English versions of the MCCB are based on the same standardization system; therefore, T-score-based thresholds are generally applied in a consistent manner across studies conducted in different populations, including Chinese and Western samples [57, 58]. We selected a cutoff score of 30 on the MCCB total T-score (approximately 2 SD below the normative mean), a level typically associated with widespread CI [27]. Previous studies have reported that approximately 59% of patients with SZ fall below this level, accompanied by significant real-world and occupational functional impairments [58]. Our study also has several limitations. First, we only performed a cross-sectional analysis, which precludes causal inferences regarding the relationships between serum HDL-C, ALB, and CI; longitudinal studies are needed to further clarify these associations. Second, only four specific non-enzymatic antioxidants were included, which represent just one component of the antioxidant defense system. Future research should examine a broader range of antioxidants to better elucidate the role of the antioxidant defense system in cognitive function. Third, since MCCB requires high cooperation, excluding unstable or non-compliant patients may have introduced selection bias. Future research should utilize more diverse cohorts and alternative assessment strategies to enhance generalizability. Fourth, although patients received guidance and supervision regarding diet and daily physical activity to minimize variability, the potential influence of these factors on non-enzymatic antioxidants such as HDL-C and ALB cannot be fully excluded due to the lack of detailed assessments. Future studies should include routine and standardized assessment of dietary intake and physical activity. Fifth, to preserve statistical power for hypothesis-driven logistic regression and interaction analyses, multiple comparison corrections were omitted, necessitating future independent replication. Finally, given that previous studies have reported non-linear relationships between cognitive function and non-enzymatic antioxidants such as UA [59], future investigations utilizing larger sample sizes are warranted to systematically explore these complex non-linear dynamics. Conclusion This study suggests that higher serum HDL-C and ALB levels are associated with lower odds of CI and better overall cognitive performance in drug-naïve patients with SZ.