Introduction

Schizophrenia is a severe mental disorder with a global prevalence of approximately 1%. Symptom expression is highly heterogeneous, but core symptoms include disturbances in thought, perception, affect, and behavior that markedly interfere with education, social life, and career aspirations1. Early-onset schizophrenia (EOS), defined as schizophrenia occurring before 18 years of age, accounts for 12.3% of all cases, and 3% of all cases manifest before 14 years of age2,3. Compared to adult-onset schizophrenia, EOS typically presents with more severe symptoms, including greater cognitive impairment and social dysfunction, higher risk of educational disruption, higher life-time disease burden4,5, and generally poorer long-term outcome6,7,8. Despite significant advances in our understanding of schizophrenia genetics and effects on central nervous system (CNS) function, it remains unclear if certain cases emerge earlier with greater disease burden due to acceleration of core pathologies or due to unique pathological processes.

The platelet-derived growth factor (PDGF) family comprises four subtypes, PDGF-A, PDGF-B, PDGF-C, and PDGF-D9, that regulate a plethora of essential biological processes by forming homodimers or heterodimers (such as PDGF-AA, AB, BB, CC, and DD)10. These isoforms activate multiple downstream signaling pathways, including PI3K/AKT, MAPK/ERK, and JAK/STAT pathways, by binding to two tyrosine kinase receptors, PDGFR-α and PDGFR-β10,11. In the CNS, PDGF and its receptors are widely distributed in neurons and glial cells, and participate in crucial processes such as neuronal development, survival, synaptic plasticity, and myelination9,12. Abnormalities in PDGF signaling pathways are associated with various neuropsychiatric disorders, including schizophrenia13 (Fig. 1). Further, genetic variations in PDGF and its receptors are schizophrenia risk factors14. In model mice as well, knockout of the PDGFR-β gene reduced social behaviors, disrupted pre-pulse inhibition, and decreased levels of parvalbumin-immunoreactive neurons in key brain regions15. Clinical studies have also found elevated serum PDGF-BB levels in schizophrenia patients, suggesting that dysregulation of PDGF signaling contributes to disease pathogenesis and symptom expression16. Notably, PDGF is an essential neurotrophic factor during early neurodevelopment, suggesting unique pathogenic contributions to EOS17,18.

Fig. 1: Flowchart illustrating how PDGF signaling pathway dysfunction may contribute to the pathogenesis of schizophrenia.
figure 1

PDGF signaling abnormalities may lead to the inhibition of downstream pathways, including PI3K/AKT, MAPK/ERK, and JAK/STAT, potentially altering neural cell survival and differentiation or dysregulating immune responses. Simultaneously, reduced PDGF levels may decrease neurotrophic support, leading to impaired synaptic development. These abnormalities may ultimately converge to cause neural network dysfunction, potentially manifesting as schizophrenia symptoms. Conversely, the emergence of schizophrenia symptoms may also trigger PDGF abnormalities, potentially exacerbating disease progression through a feedback loop.

Full size image

Oxidative stress, resulting from an imbalance between reactive oxygen species (ROS) production and scavenging by endogenous antioxidant defense systems, is also implicated in the development of schizophrenia19,20,21. Among endogenous antioxidants, superoxide dismutase (SOD) serves a vital function by converting highly reactive superoxide anions to hydrogen peroxide and oxygen22. Multiple studies have reported abnormal SOD activity in schizophrenia patients, suggesting that diminished antioxidative capacity may contribute to disease onset and progression23,24. Furthermore, altered SOD activity has been reported in association with cognitive deficits in schizophrenia patients25,26, and these deficits were generally worse in EOS. Therefore, it is possible that more severe oxidative stress and accelerated accumulation of oxidative damage may contribute to cognitive dysfunction in the early stages of schizophrenia.

Several previous studies have investigated PDGF or SOD abnormalities in adults with schizophrenia14,16,25,26,27, but there has been limited study of the immune and oxidative stress biomarker profiles among adolescents with psychotic disorders, including EOS28. We hypothesized that the levels of PDGF isoforms and SOD isoenzymes would be reduced in the peripheral blood of EOS patients, and that these abnormalities would be associated with symptom severity. If so, PDGF isoform concentrations and SOD isoenzyme activities in peripheral blood could be accessible biomarkers for EOS risk and prognosis. The current study aimed to (1) compare serum PDGF isoform (PDGF-AB, PDGF-BB, PDGF-CC, and PDGF-DD) concentrations and plasma SOD isoenzyme (T-SOD, CuZn-SOD, Mn-SOD) activities between EOS patients and age-matched controls, (2) assess the associations among clinical symptoms, serum PDGF isoform concentrations, and plasma SOD isoenzyme activities, and (3) evaluate the contributions of PDGF and SOD abnormalities to EOS risk.

Results

Demographic and clinical characteristics of study participants

There was no significant difference in age (t = 1.373, P = 0.172), sex ratio (χ2 = 0.294, P = 0.588), years of education (t = 0.657, P = 0.512), and BMI (t = −0.286, P = 0.765) between groups (Table 1). In the patient group, 11 subjects (11.1%) reported a family history of mental disorders. The mean age at onset was 14.09 ± 1.79 years, and the median duration of illness was 12.0 months (interquartile range: 3.0–26.0 months). Among EOS patients, the mean PANSS total score was 80.65 ± 15.85, and factor scores were 18.70 ± 6.78 for positive symptoms, 13.84 ± 3.69 for negative symptoms, 13.11 ± 3.13 for cognitive symptoms, 12.44 ± 4.50 for excitement/hostility, and 8.55 ± 2.57 for anxiety/depression (Table 1).

Table 1 Demographics of the EOS and healthy control (HC) groups.
Full size table

Serum PDGF isoform concentrations and plasma SOD isoenzyme activities in EOS patients versus healthy controls (HCs)

There were significant between-group differences in serum PDGF isoform concentrations and plasma SOD isoenzyme activities between EOS and control groups according to MANCOVA with diagnostic group as the fixed factor and age, sex, education years, and BMI as covariates (Wilks’ Lambda, F = 6.173, P < 0.001), with patients demonstrating significantly lower serum PDGF-AB (F = 16.050, P < 0.001, q < 0.001, Cohen’s d = 0.73), serum log-PDGF-BB (F = 18.195, P < 0.001, q < 0.001, Cohen’s d = 0.84), and plasma Mn-SOD activity (F = 13.419, P < 0.001, q < 0.001, Cohen’s d = 0.66) according to ANCOVA adjusted for age, sex, education level, and BMI (Table 2). In addition, plasma T-SOD activity was significantly lower in EOS patients compared to controls (F = 9.037, P = 0.003, q = 0.005, Cohen’s d = 0.53). All findings remained significant after FDR correction (Fig. 2). In contrast, PDGF-CC (F = 0.441, P = 0.508, q = 0.711), PDGF-DD (F = 0.035, P = 0.852, q = 0.852), and CuZn-SOD (F = 0.053, P = 0.818, q = 0.852) did not differ significantly between groups.

Fig. 2: Concentrations and plasma SOD activities between EOS patients and healthy controls.
figure 2

A serum PDGF-BB concentration. (B), plasma total superoxide dismutase activity (T-SOD). (C), plasma manganese superoxide dismutase activity (Mn-SOD). (D), All four factors were lower in early-onset schizophrenia patients than healthy controls (HC).

Full size image
Table 2 Differences in serum PDGF isoform concentrations and plasma SOD isoenzyme activities between EOS patients and HCs.
Full size table

Correlations among serum PDGF isoform concentrations, plasma SOD isoenzyme activities, and clinical symptom severity scores

Pearson’s analysis revealed a positive correlation between serum PDGF-AB concentration and plasma Mn-SOD activity (r = 0.267, P = 0.007, Fig. 3A), while Spearman’s analysis revealed a negative correlation between serum PDGF-BB concentration and PANSS cognitive factor subscore (r = −0.406, P < 0.001, Fig. 3B). In addition, plasma T-SOD activity was negatively correlated with PANSS excitement/hostility factor subscore (Pearson’s r = −0.354, P < 0.001, Fig. 3C), while plasma CuZn-SOD activity was negatively correlated with PANSS excitement/hostility factor subscore (Pearson’s r = −0.224, P = 0.025). No significant correlation was observed between PANSS scores or subscores and serum PDGF-CC, serum PDGF-DD concentrations, and plasma Mn-SOD activity (all P > 0.05). After FDR correction, however, only the negative correlations between PDGF-BB and cognitive symptom subscore and between T-SOD and excitement/hostility factor subscore remained significant.

Fig. 3: Correlations between PDGF concentrations, SOD activities, and PANSS clinical symptom scores in EOS patients.
figure 3

Association between (A) Mn-SOD activity and PDGF-BB concentration, (B) PANSS cognitive factor and SOD activity, and (C) PANSS excitement/hostility factor and PDGF-BB.

Full size image

Furthermore, stepwise multiple regression analysis controlling for age, sex, BMI, illness duration, and age of onset revealed a significant association of PANSS cognitive symptom subscore with serum PDGF-BB concentration (B = −4.703, 95% CI: −6.838 to −2.568, t = −4.371, P < 0.001; adjusted R2 = 0.156, F = 19.108, P < 0.001) and a significant negative association of PANSS excitement/hostility factor subscore with plasma T-SOD activity (B = −0.308, 95% CI: −0.472 to −0.144, t = −3.729, P < 0.001; adjusted R2 = 0.116, F = 13.906, P < 0.001).

Independent variable predictive of EOS

Based on the correlation between PDGF-AB and Mn-SOD, we constructed an interaction term (PDGF-AB × Mn-SOD). Because these biomarkers were lower in the EOS group compared to HCs, we stratified all participants into low-expression groups (less than the 67th percentile, coded as 1) and high-expression groups (greater than 67th percentile, coded as 0) for PDGF-AB, PDGF-BB, T-SOD, Mn-SOD, and PDGF-AB × Mn-SOD, and conducted modified Poisson regression analyses with adjustment for sex, age, BMI, and education level as potential confounding factors to evaluate the predictive value of each variable for EOS risk. The results revealed that low PDGF-AB (B = 0.581, RR = 1.788, 95% CI: 1.226–2.608, P = 0.003), low PDGF-BB (B = 0.564, RR = 1.758, 95% CI: 1.208–2.558, P = 0.003), and the PDGF-AB × Mn-SOD interaction (B = 0.378, RR = 1.460, 95% CI: 1.044–2.042, P = 0.027) were independent risk factors for EOS, while T-SOD (B = 0.243, RR = 1.275, 95% CI: 0.931–1.748, P = 0.130) and Mn-SOD (B = 0.218, RR = 1.243, 95% CI: 0.927–1.667, P = 0.146) did not show statistical significance. Attributable risks were 44.1% for low PDGF-AB, 43.1% for low PDGF-BB, and 31.5% for the PDGF-AB × Mn-SOD interaction, while corresponding population attributable risks were 34.5%, 33.7%, and 23.6%, respectively.

Discussion

The main findings of this study are as follows: (1) serum PDGF-AB and PDGF-BB concentrations, as well as plasma T-SOD and Mn-SOD activities were significantly lower among EOS patients than well-matched HCs; (2) serum PDGF-AB concentration was positively correlated with plasma Mn-SOD activity in patients; (3) low serum PDGF-BB concentration was associated with more severe cognitive symptoms (higher PANSS cognitive factor subscore), while low T-SOD activity was associated with more severe excitement/hostility symptoms; (4) low serum PDGF-AB, low serum PDGF-BB, and the PDGF-AB × Mn-SOD interaction were independent risk factors for EOS; and (5) the attributable fractions (AFs) for PDGF-AB, PDGF-BB, and the PDGF-AB × Mn-SOD interaction were 44.1%, 43.1%, and 31.5%, respectively, while the population AFs were 34.5%, 33.7%, and 23.6%, respectively. To our knowledge, this is the first report simultaneously documenting abnormalities in serum PDGF subtype concentrations and plasma SOD isoenzyme activities among EOS patients and their associations with specific clinical symptoms.

In contrast to these results, Boiko and colleagues found that serum PDGF-AA concentration was elevated in adult schizophrenia patients13. This discrepancy may reflect the differential regulation and function of PDGF isoforms during adolescence compared to adulthood, or alternatively distinct pathogenic functions of PDGF isoforms at various stages of the disease. Specifically, Boiko et al. studied chronic schizophrenia patients with a mean age of 42 years, suggesting age and disease chronicity may differentially impact PDGF regulation13. Our findings of reduced plasma T-SOD and Mn-SOD activities are also inconsistent with a study by Zeng et al. reporting significantly increased SOD activity among 45 adult schizophrenia patients exhibiting residual negative symptoms following a 24-week intervention with sulforaphane, suggesting potential benefits of antioxidant interventions29, but consistent with our previous study of male chronic schizophrenia patients24. The reasons for this inconsistency require further study, but could reflect differences in disease stage, current disease status, sex (our sample included both males and females), and (or) detection methods. Notably, our findings align with meta-analytic evidence from Fraguas et al., who concluded that total antioxidant capacity is generally reduced in early-onset psychosis30, and of Flatow et al., who similarly found generally reduced SOD activity in schizophrenia patients31. These meta-analyses provide support for antioxidant system dysregulation as a major pathophysiological mechanism in schizophrenia, particularly in early-onset cases.

PDGF signaling contributes to neuronal development, survival, synaptic plasticity, and myelination9,32,33, so reduced serum concentrations may reflect broad neurodevelopmental abnormalities leading to EOS. Similarly, reduced SOD activity suggests that earlier oxidative stress and greater accumulation of oxidative injuries may accelerate symptom development and result in earlier disease onset34. Furthermore, we found that serum PDGF-AB concentration rose in parallel with plasma Mn-SOD activity among EOS patients, the first such report in a clinical setting. Mn-SOD is primarily localized to mitochondria, where it serves as a first line of defense against superoxide anions generated by oxidative phosphorylation35. The deficient PDGF-AB reduces antioxidant capacity and leads to accelerated neural circuit dysfunction from oxidative stress during the pathogenesis of schizophrenia, ultimately resulting in earlier symptom onset. Similar to the correlation between BDNF and SOD reported by Xiu et al. in adult schizophrenia26, our results supported a functional association between neurotrophic factor signaling and endogenous antioxidant efficacy. However, evidence for direct regulation of antioxidant systems by PDNF signaling is still required.

Alterations in markers of PDGF signaling and antioxidative function have also been detected in the CSF of patients with schizophrenia. Coughlin et al. reported significantly reduced SOD levels in CSF samples from recent-onset schizophrenia patients36,37, consistent with our observation of reduced SOD activity in the peripheral blood of EOS patients. In addition, Nation et al. found that elevated CSF PDGFRβ levels were associated with cognitive decline38, suggesting blood–brain barrier (BBB) disruption. However, as is the case for peripheral blood markers, it is essential to clarify the associations between CSF markers and CNS pathology.

Both PDGF-BB and PDGF-AB are essential signaling factors regulating neurodevelopment and are further implicated in psychiatric disorders. Kajizuka et al. reported increased serum PDGF-BB levels in male children with autism that correlated with stereotyped behavior patterns39. PDGF-BB promotes cortical neuronal dendritic growth and enhances BBB integrity, while PDGF-AB regulates both neuroprotection and neurogenesis through actions on PDGFR-α and PDGFR-β38. The negative association between serum PDGF-BB concentration and PANSS cognitive symptom scores reported in the current study suggests that deficient PDGF signaling contributes to cognitive impairments in EOS. A study by de la Monte et al. also reported a similar association between PDGF-BB and cognitive dysfunction in Alzheimer’s disease patients40. Furthermore, lower T-SOD activity was associated with more severe excitement/hostility. Yang et al. similarly reported an association between T-SOD and excitement/hostility symptoms among chronic schizophrenia patients24, consistent with a major contribution of oxidative stress to emotional instability. This association in younger patients further suggests that insufficient antioxidant capacity contributes to symptom expression (e.g., emotional instability) in schizophrenia across different age groups. However, longitudinal studies are needed to establish a causal relationship.

Low serum PDGF-AB and PDGF-BB concentrations, as well as the PDGF-AB × Mn-SOD interaction terms, were identified as independent risk factors for EOS. These findings are consistent with Funa et al., who reported that a deficiency in PDGF may lead to neurodevelopmental disorders9. However, Boiko et al. found elevated rather than decreased serum PDGF-AA concentrations in adult schizophrenia patients13, suggesting a potential difference in pathomechanism between early- and adult-onset schizophrenia or distinct pathogenic functions at different disease stages. The association with the PDGF-AB × Mn-SOD interaction term aligned with the theoretical framework proposed by Rawani et al. regarding the synergistic action of multiple biological mechanisms in psychotic disorders19, including neurotransmitter dysregulation, neuroinflammation, oxidative stress, and mitochondrial dysfunction. Furthermore, the relative risk (RR) values suggested strong contributions to EOS risk, especially considering the low incidence of schizophrenia in adolescents. These factors may be valuable biomarkers for the identification of high-risk individuals.

The AFs for PDGF-AB, PDGF-BB, and PDGF-AB × Mn-SOD interaction (44.1%, 43.1%, and 31.5% respectively) and population AFs (34.5%, 33.7%, and 23.6%, respectively) further underscored the potential importance of PDGF isoforms and SOD isoenzymes in the etiology of EOS. Smith et al. noted that AFs are effective tools for quantifying the contributions of specific risk factors to schizophrenia onset41. A meta-analysis by Dragioti et al. identified numerous potentially modifiable risk factors for mental disorders42, including a population AF of 37.8% for childhood adversities in schizophrenia spectrum disorders. In our study, PDGF-AB and PDGF-BB reached population AFs of 34.5% and 33.7%, respectively, suggesting that these neurotrophic factors were as important for EOS pathogenesis as known environmental risk factors. Jomova et al. emphasized the critical contribution of Mn-SOD as a first-line antioxidant defense enzyme35, and the high AF of the PDGF-AB × Mn-SOD interaction (23.6%) found in the current study further supported the theory of synergistic actions among multiple biological mechanisms in psychotic disorders as proposed by Rawani et al.19. However, Rückinger et al. cautioned that potential confounding factors and methodological limitations should be considered when interpreting AFs43. The current findings thus require further validation in prospective studies.

This study had several limitations. First, as a cross-sectional study, we could not determine whether the observed biomarker changes were the cause or result of EOS. Further, this single-point measurement cannot reveal the dynamic pattern of biomarker levels throughout disease progression. Second, the sample size was relatively small, particularly the HC group, due to practical difficulties in collecting blood samples from healthy adolescents in China, which limits statistical power and generalizability to other EOS populations. Third, the proportion of female patients in our study exceeded that of males due primarily to more male patients failing to complete the full baseline assessment protocol. This gender distribution bias may further limit the generalizability of our findings to the broader EOS population. Fourth, our measurements of PDGF in serum and SOD in plasma, while based on established methodological considerations, may introduce within-subject variability influencing the associations examined. Future studies should consider examining these markers in the same blood component. Finally, although all patients were drug-naïve and first-episode, variations in illness duration may still have influenced the results. Future studies should use a prospective longitudinal design, expand the sample size, include multi-ethnic populations, and explore underlying mechanisms.

Conclusion

In conclusion, serum concentrations of PDGF isoforms PDGF-AB and PDGF-BB, as well as plasma T-SOD and Mn-SOD activities, were abnormally low in EOS patients. Serum PDGF-AB concentrations were also correlated with plasma Mn-SOD activities, suggesting that this neurotrophic factor signaling pathway may influence antioxidant capacity and reduce oxidative damage associated with EOS. Lower serum PDGF-BB concentration was associated with more severe cognitive symptoms and lower T-SOD activity with more severe excitement/hostility, suggesting direct contributions of these abnormalities to disease pathogenesis or expression. Lower concentrations of PDGF-AB and PDGF-BB, and the interaction effect between PDGF-AB and Mn-SOD were identified as strong independent risk factors for EOS (with AFs of 44.1%, 43.1%, and 31.5% respectively), underscoring the potential clinical significance of these factors as EOS biomarkers.

Methods

Subjects

This study enrolled 99 patients with EOS (12–18 years of age) from the Fourth People’s Hospital of Lianyungang between June 2020 and June 2024. All patients were diagnosed with EOS according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV). The inclusion criteria for the patient group were Han Chinese ethnicity, first-episode and drug-naïve status, primary school or higher education, and the ability to complete all assessment questionnaires. A self-designed demographic questionnaire was used to collect information on age, sex, height, weight, family history of mental disorders, educational level, age at onset, and duration of illness. Additionally, 40 HCs (HCs) matched for age, sex, education, and body mass index (BMI) were recruited from the same geographical area.

Exclusion criteria for all participants included substance abuse/dependence and alcohol abuse/dependence (based on DSM-IV criteria, self-reported), antibiotic use within the previous 3 months, current acute infection or chronic infection, inflammatory disorders (e.g., rheumatoid arthritis, inflammatory bowel disease), autoimmune diseases (e.g., systemic lupus erythematosus, multiple sclerosis), hepatic or renal dysfunction, metabolic diseases (e.g., diabetes, severe obesity), consumption of antioxidant supplements (e.g., vitamins C and E, coenzyme Q10) within the previous three months, use of non-steroidal anti-inflammatory drugs or corticosteroids within the previous 3 months, history of major surgery or trauma within the previous 6 months, blood transfusion within the previous 3 months, and hematological disorders that could affect platelet generation or function. The HC group was additionally excluded for psychiatric disorders according to DSM-IV Axis I criteria, family history of psychiatric disorders, or current use of medications that could affect CNS function.

The study protocol was approved by the Ethics Committee of the Fourth People’s Hospital of Lianyungang (approval number: 2021LSYYXLL-P11), and all procedures were conducted in accordance with the Declaration of Helsinki. Given that the study participants were minors (12–18 years of age), detailed information about the study’s purpose, methods, expected outcomes, potential risks, and benefits was provided to all potential participants and their legal guardians before enrollment. Written informed consent was subsequently obtained from all participants and their legal guardians. Participants were informed of their right to withdraw from the study at any stage without prejudice. All collected data were de-identified and kept confidential, with appropriate technical and organizational measures implemented to ensure data security. Only authorized research personnel had access to the data, and results are presented in aggregate form without any personally identifiable information.

Clinical assessment

Symptom severity in EOS patients was evaluated using the Positive and Negative Syndrome Scale (PANSS)44,45. Based on the five-factor model46,47,48, PANSS items P1, P3, P5, P6, and G9 were used to evaluate positive symptoms, N1, N2, N3, N4, N6, and G7 to evaluate negative symptoms, P2, N5, G5, G10, and G11 for cognitive symptoms, P4, P7, G8, and G14 for excitement/hostility, and G1, G2, G3, and G6 for anxiety/depression. The PANSS total score was calculated based on all 30 scale items. All clinical assessments were independently conducted by two senior attending psychiatrists who had received standardized PANSS training. Inter-rater reliability was evaluated by calculating intraclass correlation coefficients, and values exceeded 0.8 for all scores. All patients completed the PANSS assessment within 3 days of hospital admission to ensure that symptom evaluation reflected the drug-naïve state.

Fasting blood sample collection and biochemical analyses

Venous blood samples were collected from all participants in the early morning (between 07:00 and 09:00) after an overnight fast (at least 8 h without food and caloric beverage intake). Samples were drawn into procoagulant tubes for serum separation and anticoagulant tubes for plasma isolation, followed by centrifugation at 3000 rpm for 15 min at 4 °C and storage at −80 °C until further analysis. Serum PDGF-AB, PDGF-BB, PDGF-CC, and PDGF-DD concentrations were determined using Luminex liquid suspension chip technology strictly according to the manufacturer’s guidelines (R&D Systems, Minneapolis, MN, USA). The laboratory personnel performing these assays were blinded to the clinical status of the subjects. Total (T)-SOD activity, copper-zinc (CuZn)-SOD activity, and manganese (Mn)-SOD activity (expressed in U/mL) were assessed utilizing commercially available kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the supplier’s instructions49. The intra- and inter-assay coefficients of variation for all measured blood parameters ranged from 1.4% to 7.2%.

Statistical analysis

All continuous variables were first assessed for normality using the Shapiro–Wilk test. Based on the results, PDGF-BB data were natural log-transformed before parametric testing, while other raw variables met normality assumptions. Sample size calculation and effect size estimation were conducted using G*Power 3 software (version 3.1.9.7). For sample size estimation, we employed the ‘difference between two independent means’ analysis method, with a large effect size (d = 0.8), α error probability set at 0.10, and target statistical power of 90%. The calculation indicated that a total sample of 56 participants would be required. Continuous variables were compared between groups by independent samples t-tests and categorical variables by chi-square tests. Continuous variables are presented as mean ± standard deviation (SD) and categorical variables as counts and percentages. Normally distributed biomarkers (including log-transformed PDGF-BB) were first compared by multivariate analysis of covariance (MANCOVA), with diagnostic group as a fixed factor, and age, sex, years of education, and BMI as covariates. Individual biomarkers were then compared by analysis of covariance (ANCOVA), controlling for the same covariates. The false discovery rate (FDR) method was applied to the q-value for multiple comparison correction. Associations among biomarkers and clinical variables were examined using Pearson’s or Spearman’s rank correlation method as indicated. To determine the predictive relationships between biomarkers and clinical symptom severity, stepwise multiple linear regression analysis was performed with an entry criterion of P < 0.10 and removal criterion of P > 0.05, while controlling for potential confounding factors.

Based on univariate analysis results, PDGF-AB, PDGF-BB, T-SOD, Mn-SOD, and their interaction terms were dichotomized into high-expression groups (above the 67th percentile, coded as 0) and low-expression groups (below the 67th percentile, coded as 1). Modified Poisson regression models were employed to estimate RRs while adjusting for sex, age, BMI, and education level as confounding factors. The AF and population attributable fraction (PAF) were calculated to assess the contributions of decreased biomarker levels to EOS risk using the formulas AF = (RR-1)/RR and PAF = P₁(RR − 1)/[1 + P₁(RR − 1)], where P₁ represents the proportion of the low-expression group (>67%). All statistical analyses were conducted using SPSS 25.0 software (IBM Corp., Armonk, NY, USA), with a two-tailed P < 0.05 considered statistically significant.

Ethics approval and consent to participate

We declare that all experiments on human subjects were conducted in accordance with the Declaration of Helsinki and that all procedures were carried out with the adequate understanding and written consent of the subjects or their legal guardians. All experimental protocols were approved by the Ethics Committee of Lian Yun Gang Fourth People’s Hospital. Informed consent was obtained from all participants and/or their legal guardians (approval number: 2021LSYYXLL-P11). All procedures were conducted in accordance with relevant guidelines and regulations.