INTRODUCTION
Attention-deficit/hyperactivity disorder (ADHD) is a common neurodevelopmental disorder, typically diagnosed in childhood, and often accompanied by behavioral, emotional, and cognitive dysfunctions [1,2]. Clinically, ADHD is classified into three major subtypes based on symptom presentation: inattentive type (ADHD-I), hyperactive/impulsive type (ADHD-HI), and combined type (ADHD-C) [3-5]. ADHD diagnosis typically relies on behavioral rating scales such as the ADHD Rating Scale-IV (ADHD-IV) and the Child Behavior Checklist (CBCL), which are inherently subjective and based on caregiver or teacher reports [6-8]. The absence of objective biological markers continues to hinder the accurate diagnosis and differentiation of ADHD subtypes.
Although significant research has been conducted on the neurobiological basis of ADHD in recent years, reliable and clinically applicable biomarkers remain elusive. Electroencephalography (EEG), a non-invasive technique that reflects realtime brain electrical activity, has been widely used to explore the neurophysiological characteristics of ADHD [9-11]. Alterations in specific EEG frequency bands, particularly increased theta (θ) power and elevated θ/β ratios, have been frequently associated with attentional impairments in ADHD [12,13]. Furthermore, event-related potentials (ERPs), especially the P300 wave, offer important insights into cognitive processing and have been suggested as potential indices for ADHD-related executive dysfunction.
In addition to electrophysiological measures, serum neurotrophic factors such as brain-derived neurotrophic factor (BDNF), nerve growth factor (NGF), glial cell line-derived neurotrophic factor (GDNF), and neurotrophin-3 (NTF3) are believed to play critical roles in neural development, synaptic plasticity, and cognitive function. Abnormal levels of these factors have been reported in ADHD populations, suggesting their potential as peripheral biological markers [14-16]. However, studies investigating the combined diagnostic value of EEG features and neurotrophic factors in differentiating ADHD subtypes—especially among pediatric patients—remain limited.
The innovation of this study lies in the integrated analysis of EEG characteristics and serum neurotrophic factor levels to assess their joint diagnostic value in subtype classification of ADHD. Compared to traditional behavioral scales, the use of objective biomarkers may offer improved accuracy in differentiating ADHD-I, ADHD-HI, and ADHD-C. Moreover, the incorporation of P300 wave parameters with EEG spectral data introduces a novel perspective in ADHD biomarker research. This study addresses a gap in the literature by focusing on children aged 6-12 years and provides important clinical insights for early identification and individualized treatment planning.
METHODS
Study design
This study is a retrospective cohort study aimed at exploring the diagnostic value of EEG features and serum neurotrophic factors (such as BDNF, NGF, GDNF, NTF3) in differentiating ADHD subtypes. The study included data collected from January 2023 to December 2024 at Changzhou Children’s Hospital. The study subjects were children aged 6 to 12 years diagnosed with ADHD, and all diagnoses were made according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria. All procedures were carried out in accordance with the ethical principles of the Declaration of Helsinki and approved by the Changzhou Children’s Hospital of Nantong University institutional ethics committee (No. NT-KY-021). All data are used for research purposes only and are kept strictly confidential, with no personal information being disclosed.
Study subjects
A total of 322 children with ADHD were included, with a mean age of 9.5±1.8 years. The children were divided into three groups according to DSM-5 criteria: ADHD-I (106 cases), ADHD-HI (109 cases), and ADHD-C (107 cases). The diagnoses for all children were confirmed by two clinical doctors according to DSM-5 criteria, and the severity of symptoms was assessed using the ADHD-IV and the CBCL (Supplementary Table 1). Inclusion criteria: 1) aged between 6 and 12 years; 2) diagnosed with ADHD according to DSM-5 criteria, with clear subtype classification; 3) no history of severe neurodevelopmental disorders, brain diseases, or drug dependence; and 4) no known diseases affecting the nervous system (such as brain trauma, epilepsy, etc.). Exclusion criteria: 1) co-occurrence of schizophrenia or other psychiatric disorders, 2) major physical diseases (such as brain tumors, cerebrovascular diseases, etc.), and 3) failure to sign informed consent.
Study materials
EEG Equipment: EEG signals were recorded using a Neuroscan brand EEG machine, with a sampling frequency of 1,000 Hz. EEG signals were recorded using the standard international 10-20 electrode system to record brain activity in the frontal, central, and parietal regions.
Serum neurotrophic factor measurement: serum levels of BDNF, NGF, GDNF, and NTF3 were measured using ELISA kits. All kits were provided by R&D Systems, and the procedures strictly followed the kit instructions.
EEG signal collection
EEG recordings were conducted in a quiet environment for all study participants. To minimize external interference, children were asked to remain quiet and as relaxed as possible before EEG recording. During the EEG recording, children kept their eyes closed to avoid the impact of excessive attention on the EEG signals. EEG signals were recorded using the standard international 10-20 electrode system, with a total of 19 electrodes placed in the frontal, central, and parietal regions. The EEG signal was recorded for 5 minutes, and the children were asked to avoid vigorous movements during the recording. The collected data were analyzed using Fast Fourier Transform to calculate the power in different frequency bands (δ, θ, α, β waves). The relative power of each frequency band was calculated, and statistical analysis was performed. Additionally, the P300 wave, elicited using an “oddball task,” was analyzed to assess cognitive function. Both the latency and amplitude were measured, and the P300 amplitude area—defined as the integral of the positive waveform within the 250-500 ms window—was selected as it provides a comprehensive measure of cortical activation by capturing both the magnitude and duration of the response [17].
Serum neurotrophic factor measurement
After the EEG recordings, all participants underwent blood sampling. The blood samples were collected in the morning using venipuncture, immediately centrifuged, and stored as serum samples. All serum samples were stored at -80°C until ELISA analysis. Neurotrophic factors were measured using ELISA kits provided by R&D Systems, which measured the levels of BDNF, NGF, GDNF, and NTF3 in the serum. The concentration of neurotrophic factors in each sample was calculated using a standard curve and measured three times, with the average value used as the final result.
Quality control
To ensure the high quality and reliability of the experiment, all EEG recordings and serum measurements were conducted by professionally trained technicians. Prior to and after EEG signal collection, all technicians performed standardized procedures to ensure that electrode placement and signal collection followed international standards. All ELISA experiments strictly adhered to the kit instructions, and three repeated measurements were performed to ensure the accuracy and stability of the data. To avoid analysis bias, all data were analyzed using a double-blind experimental design.
Data analysis
All data were statistically analyzed using SPSS 26.0 software (IBM Corp.). First, the Shapiro-Wilk test was used to assess the normality of the data. Homogeneity of variance was evaluated using Levene’s test. Variables that followed both normality and homogeneity assumptions were expressed as mean± standard deviation and analyzed using one-way analysis of variance followed by Tukey’s post hoc test. For variables violating normality or homogeneity assumptions, appropriate data transformations (e.g., logarithmic transformation) were performed. If assumptions remained unmet after transformation, non-parametric alternatives such as the Kruskal-Wallis test were applied to verify the robustness of results. Non-normally distributed data are expressed as median and interquartile range. A p-value of less than 0.05 was considered statistically significant.
To evaluate the role of EEG features and serum neurotrophic factors in differentiating ADHD subtypes, multivariate regression analysis was used to identify independent predictors. Multivariate logistic regression analysis was performed to identify independent predictors. Candidate variables were initially screened based on univariate analysis (p<0.05) and their clinical relevance as reported in prior studies. A forward stepwise selection method was then applied to construct the final model with the most informative predictors.
To further validate the diagnostic performance, receiver operating characteristic (ROC) curve analysis was conducted. The area under the curve (AUC), sensitivity, and specificity were calculated. An AUC value closer to 1 indicates better diagnostic performance of the model.
RESULTS
Differences in serum neurotrophic factor levels
Serum neurotrophic factor levels exhibited significant differences among ADHD subtypes. As shown in Table 1, ADHD-I patients had significantly lower serum levels of BDNF (15.2±5.6 ng/mL) and GDNF (3.1±1.2 ng/mL) compared to ADHD-HI patients (22.3±6.8 ng/mL and 5.6±1.4 ng/mL, respectively; p<0.01). Conversely, NGF and NTF3 levels were significantly higher in ADHD-HI than in ADHD-I (p<0.01). The ADHD-C group displayed intermediate values across all four neurotrophic factors, with levels closer to ADHD-HI. These findings suggest that serum neurotrophic profiles differ by subtype and may serve as candidate biomarkers for ADHD classification.
EEG feature differences among ADHD subtypes
EEG analysis showed significant differences between ADHD subtypes in terms of EEG frequency band (δ, θ, α, β waves) power and θ/β ratio. As summarized in Table 2, whole-brain averaged θ power was highest in ADHD-C (19.2±4.6 μV2), followed by ADHD-I (16.5±3.2 μV2) and ADHD-HI (13.2±2.8 μV2) (p<0.01). However, regional analysis showed that frontal θ power was significantly elevated in ADHD-I compared to ADHD-C and ADHD-HI (p<0.05), consistent with its association with attentional deficits (Supplementary Table 2). In addition, ADHD-I patients exhibited the highest θ/β ratio (1.85±0.40), significantly greater than ADHD-HI (0.85±0.32) and ADHD-C (1.25±0.35) (p<0.05). In contrast, β wave power was significantly higher in ADHD-HI (23.6±5.4 μV2, p<0.01), especially in the parietal region. ADHD-C patients again showed intermediate values. These EEG findings suggest that increased θ/β ratios reflect inattention (more common in ADHD-I), while elevated β activity is associated with hyperactivity and impulsivity (more pronounced in ADHD-HI).
P300 wave latency and amplitude analysis
The P300 wave is an important indicator of cognitive response time and information processing ability. This study analyzed the P300 wave latency, amplitude, and amplitude area in different ADHD subtypes. The results in Table 3 show significant differences in P300 wave characteristics, especially between ADHD-I and ADHD-HI.
In ADHD-I patients, the P300 wave latency was significantly prolonged, with an average latency of 340.2±45.3 ms, and the amplitude was 5.6±1.1 μV, with the amplitude area significantly lower than that in ADHD-HI and ADHD-C (p<0.01). Compared to ADHD-I, ADHD-HI patients had a significantly shorter P300 latency, with an average latency of 295.4±40.7 ms, an amplitude of 8.9±1.8 μV, and a significantly larger amplitude area than ADHD-I (p<0.01) and ADHD-C (p<0.05). ADHD-C patients had a P300 latency of 310.8±42.1 ms and an amplitude of 7.2±1.3 μV, with P300 characteristics between those of ADHD-I and ADHD-HI. These findings indicate that the P300 wave latency and amplitude reflect the mixed features of ADHD-C, with cognitive response speed and information processing ability between ADHD-I and ADHD-HI. In summary, the significant differences in P300 wave latency and amplitude between different ADHD subtypes provide valuable insights into the neurobiological mechanisms of ADHD, especially the differences between ADHD-I and ADHD-HI, which can serve as biomarkers for distinguishing different subtypes.
Independent predictor analysis: the role of EEG features and neurotrophic factors
Multivariate regression analysis identified several independent variables that significantly contributed to the differentiation between ADHD-I and ADHD-HI. θ wave power (β=0.45, p<0.01) and θ/β ratio (β=0.38, p<0.05) were statistically significant EEG predictors (Table 4). In addition, NGF (β=0.32, p<0.05) and NTF3 (β=0.29, p<0.05) levels, as well as P300 amplitude area (β=0.41, p<0.01), were identified as significant serum-based or ERP predictors. These variables remained independently associated with subtype classification in the multivariate model, indicating that both electrophysiological and biochemical markers contributed to distinguishing ADHD-I from ADHD-HI.
ROC curve analysis: diagnostic performance of EEG features and neurotrophic factors
In this study, ROC curve analysis was used to evaluate the combined use of EEG features and serum neurotrophic factors in distinguishing ADHD subtypes. The results show that the combined use of EEG features and neurotrophic factors exhibited high diagnostic performance in differentiating ADHD subtypes (Table 5). Specifically, in distinguishing ADHD-I from ADHD-HI, the combined model had an AUC value of 0.90, demonstrating very high diagnostic sensitivity and specificity, indicating that the combination of EEG features and neurotrophic factors has a significant advantage in accurately distinguishing ADHD-I and ADHD-HI (Figure 1). In the differentiation between ADHD-I and ADHD-C, the combined model had an AUC value of 0.84, showing good diagnostic performance and effectively distinguishing these two subtypes. For the differentiation between ADHD-HI and ADHD-C, the combined model had an AUC value of 0.87, also demonstrating high diagnostic performance. In conclusion, the combined use of EEG features and neurotrophic factors shows significant advantages in distinguishing ADHD subtypes, especially in distinguishing ADHD-I and ADHD-HI, with extremely high diagnostic performance.
DISCUSSION
This study explored the diagnostic value of EEG features and serum neurotrophic factors in differentiating ADHD subtypes. The results demonstrated that distinct EEG and serum biomarker profiles exist across the subtypes, particularly between ADHD-I and ADHD-HI. ADHD-I patients exhibited increased frontal θ activity and a higher θ/β ratio, while ADHD-HI patients showed enhanced β wave activity and lower θ/β ratios. Differences in serum levels of BDNF, NGF, GDNF, and NTF3 were also observed among subtypes, with ADHDI showing lower levels of BDNF and GDNF, and ADHD-HI exhibiting higher NGF and NTF3 levels. Additionally, the P300 latency was significantly longer and amplitude smaller in ADHD-I, indicating differences in cognitive processing. These findings support the utility of combining EEG and neurochemical data for ADHD subtype classification, especially between ADHD-I and ADHD-HI.
The findings of this study are consistent with existing literature, which suggests that EEG features and neurotrophic factors may serve as potential biomarkers for ADHD subtype differentiation [18,19]. Mechanistically, the observed EEG differences reflect variations in brain activity patterns. Increased θ power and elevated θ/β ratios in ADHD-I are associated with underarousal and deficits in sustained attention and executive function [20,21]. In contrast, higher β power in ADHD-HI is typically linked to increased cortical activation and motor drive, which aligns with the behavioral symptoms of hyperactivity and impulsivity [22-25].
Regarding serum neurotrophic factors, the lower levels of BDNF and GDNF in ADHD-I patients may be associated with neuroplasticity and structural changes in the brain. BDNF is an important regulator of neurodevelopment and plasticity [26,27], and its reduced levels may affect neuronal growth and synaptic plasticity, leading to cognitive deficits in ADHD-I patients. GDNF plays a crucial role in neuroprotection and neuroplasticity [28], and lower GDNF levels may impact the neural recovery and repair capacity in ADHD-I patients. Conversely, ADHD-HI patients exhibited higher levels of NGF and NTF3, which may relate to stronger neural activity and plasticity in this subtype, particularly in fast responses and regulation of hyperactive behaviors, where the role of neurotrophic factors is more prominent. The changes in P300 wave latency and amplitude further support these mechanisms. The prolonged latency and lower amplitude in ADHD-I are likely linked to delayed cognitive processing and low efficiency, while the shorter latency and higher amplitude in ADHD-HI reflect faster information processing, consistent with their impulsivity and rapid response characteristics. These differences in EEG and biomarkers may reveal the essential neurobiological differences between ADHD subtypes, providing new evidence for accurate diagnosis.
Despite the significant findings of this study, there are some limitations. First, the retrospective cohort design may introduce bias in sample selection, especially in terms of the children’s clinical characteristics and treatment history, which could affect the external validity of the study results. Second, the relatively small sample size, particularly the sample difference between ADHD-I and ADHD-HI, might impact the reliability of the statistical results. Further prospective studies and larger randomized controlled trials are needed to validate these findings. Third, the EEG signal collection and serum sample processing in this study were performed once, without considering the long-term effects of follow-up on neurotrophic factor changes and EEG characteristics, which may impact the continuity and stability of the results in some cases. Additionally, the study only compared ADHD subtypes without including a healthy control group, which limits the interpretation of the biomarkers’ specificity. Finally, while EEG frequency band analysis and P300 wave features yielded significant results, other EEG bands (such as α and δ waves) and their relationship with ADHD subtypes were not further explored, which could be a future research direction.
This study offers preliminary evidence supporting the use of EEG features and neurotrophic factors as objective tools for ADHD subtype differentiation. Future research should consider integrating real-time EEG monitoring with repeated biomarker assessments, particularly in treatment trials, to evaluate the stability and predictive value of these markers over time. Machine learning approaches may also enhance classification performance using multimodal data inputs. Ultimately, the combined use of neurophysiological and biochemical markers may support precision diagnosis and the development of individualized intervention strategies in pediatric ADHD.
In conclusion, this study demonstrated that EEG spectral features and serum neurotrophic factors exhibit distinct profiles across ADHD subtypes. The combined analysis of θ power, θ/β ratio, P300 amplitude, and neurotrophic factor levels (BDNF, NGF, GDNF, and NTF3) showed high diagnostic accuracy in differentiating ADHD-I from ADHD-HI and moderate accuracy for the other subtype comparisons. These findings support the use of multimodal neurobiological markers as objective tools for ADHD subtype classification. Integrating EEG and biochemical data may enhance early identification and subtype-specific management in clinical practice. Such multimodal biomarkers offer promising potential for future integration into early diagnostic systems and subtype-specific treatment planning.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
