This is a noninterventional, case-control, cross-sectional, and assessor-blinded study, which is part of the development of a medical diagnostic test. This manuscript describes the latest version of the study protocol (HUPH-BP-BMCCEEG-CC22 version 4.0, approved on June 7, 2024, as Amendment 4), which was reviewed and approved by the appropriate ethics committee (see below). In order to allow for more complete and detailed reporting, we adhere to 2 acceptable reporting guidelines that match the study design: the Standards for Reporting Diagnostic Accuracy Studies (STARD; Checklist 2) and the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) 2025 checklist (Checklist 1) of items to address in a randomized trial protocol, though, sensu stricto, our study is not a randomized trial. Following the terminology of the STARD, the EEG and the algorithms used to analyze the EEG data will be regarded as the index test, and the executive control task during which brain connectivity metrics will be pursued will be regarded as the index task. The best estimated diagnosis based on clinical interviews and neuropsychological assessment will be regarded as the reference test. To minimize bias, the PI establishing the reference standard will not perform EEG analyses or access EEG results. Blinding to index test results will be maintained, as no diagnostic or therapeutic decisions will be based on these findings. Consequently, no clinical decision will be influenced by information from the index test. Likewise, the team in charge of EEG processing will not have access to the clinical diagnosis data during data collection for the study. Participants will be categorized into 3 groups, including ADHD predominantly inattentive, ADHD combined or hyperactive/impulsive, and a control group, based on clinical evaluation including the reference CPT-3.

A total of at least 165 participants will be recruited (Multimedia Appendix 1). The target population will consist of children who exhibit symptoms and signs suggesting ADHD who contact the health system and who may or may not have the disorder. It will be attempted to represent this population through a source population consisting of (1) children aged 7-12 years who either received a prior diagnosis of ADHD (cases; predominantly hyperactive/impulsive or combined presentation or predominantly inattentive) or (2) children attending routine pediatric check-ups at the Puerta de Hierro University Hospital in Majadahonda (HUPHM) who do not meet criteria for any significant psychiatric disorder (controls). Recruitment ended in December 2024. Accordingly, our study was a convenience series of participants.

A data monitoring committee was not considered necessary due to the low-risk nature of the study; oversight is ensured through standard monitoring procedures.

The index test (also known as the Bit task) is a go/no-go interference test consisting of 3 parts, preceded by a 4-minute resting-state EEG recording. Each part of the test contains 400 trials, with an interstimulus interval between 450 and 550 milliseconds long. The ratio of go and no-go trials changes across parts, including 25:75 (first part), 50:50 (second part), and 75:25 (third part). If participants make mistakes, a warning will be displayed on the screen reminding them of the instructions. This task will be executed using PsychoPy software (version 2022.1.3; Open Science Tools Ltd).

We will use the Waveguard net (ANT Neuro) HD-C68 Tyco with Ag rings, 2 HD-C68pin connectors, and 64 semidry electrodes arranged in an equidistant layout, compatible with eego processing software (version 1.9.2 or higher; ANT Neuro). The EEG eego adapter will be used for 6 type A channels, along with the eego amplifier EE-225 (ANT Neuro BV). The reference signal is placed at the 5Z position, while the ground signal is positioned at the 0Z coordinate.

The study will be conducted in the Department of Psychiatry at the HUPHM. Participants will be recruited from 2 distinct sources. First, cases will be identified by reviewing medical records from the hospital to locate participants with an existing diagnosis of predominantly inattentive, predominantly hyperactive/impulsive, or combined ADHD prior to the commencement of study procedures. Additionally, cases may be prospectively selected from patients routinely seen at the study site who meet the specified selection criteria. These diagnoses will be confirmed through clinical evaluation during the study reference test. Control participants will be recruited from individuals attending Pediatric Services at HUPHM.

Selected candidates will be contacted and invited to participate, with an in-person appointment scheduled to complete the study procedures. ADHD cases will be instructed to temporarily discontinue their medication for the disorder prior to the first visit, with a 24-hour cessation for those using methylphenidate or atomoxetine and 48-hour for those using lisdexamfetamine.

This clinical study focuses on the characterization and initial validation of an experimental medical diagnostic test under development. To avoid bias, the physicians involved in diagnosing and selecting the study groups will not participate in the analysis. Furthermore, missing data will not be imputed. In addition, indeterminate index test or reference standard results will not be allowed, as a positive or negative diagnosis will be required according to the previously defined cutoffs. Analyses will be performed on complete cases within each analysis set. The planned sample size (Multimedia Appendix 1) includes a margin for nonevaluable data (10% added for nonassessable data-complete-case analysis). No interim analyses or stopping rules are planned.

This study was approved by the Research Ethics Committee of the HUPHM on February 13, 2023 (PI 207/22). Informed consent was obtained from legal guardians and assent from minors, in accordance with the provisions of Organic Law 3/2018 of December 5 [72], on the Protection of Personal Data and Guarantee of Digital Rights. All procedures involving human participants were conducted in accordance with the ethical standards of the responsible institutional and national research committee and with the 1964 Declaration of Helsinki and its subsequent amendments. Monitoring of the clinical trial was carried out as required by applicable clinical research regulations. Privacy was always guaranteed. All participants had a secret code which was dissociated and saved separately from personal identification data. All participants entered into the study without compensation. All data generated or analyzed will be fully anonymized to ensure participant confidentiality (ISRCTN Registry ISRCTN12110752).

Any adverse events from the EEG recordings, index task, and neuropsychological tests will be recorded according to the clinical guidelines established in the legal framework. Adverse events will be recorded from the first visit to the last. The researcher (HB-F) will assess the association between the events and the assessment and report this to the research ethics committee.

Data will be collected using a pre-established standardized variable form and entered in a password-protected electronic database with role-based access control. Data quality will be ensured through predefined regular monitoring and periodic audits to identify missing or inconsistent entries. Access to the final dataset will be restricted to authorized members of the research team.

EEG has been extensively employed to explore various neurophysiological metrics in ADHD, such as power spectral densities and event-related potentials. For instance, Clarke et al [73] demonstrated that children with ADHD exhibit increased theta band activity, decreased alpha band activity, and an elevated theta/beta ratio (TBR) compared to controls. Furthermore, these differences persisted into adulthood, with individuals showing lower beta levels in specific regions and a reduction in the ratio in ADHD compared to controls in recordings with eyes closed [74]. Similarly, Jarret et al [75] found increased relative theta power in the T5 and F7 regions in individuals with ADHD, aligning with findings from Chutko et al [76], who observed elevated TBR in children with both ADHD and SCT. Despite these promising findings, the utility of metrics like TBR as a robust diagnostic tool has been questioned due to inconsistent results [77,78], although in recent years researchers have increasingly favored the use of TBR in combination with Bayesian Gaussian models [79].

Although EEG power metrics offer valuable insights, recent studies have shifted focus toward functional connectivity analyses, which may serve as more stable and reliable biomarkers for ADHD. McNorgan et al [80] identified impairments in functional connectivity within the default mode network, reinforcing the hypothesis that ADHD involves disrupted brain networks underlying attention and executive control. Michelini et al [81] further highlighted abnormalities in the corpus callosum and connectivity between the default mode network and frontoparietal/salience networks, suggesting that ADHD is fundamentally a disorder of brain connectivity and providing critical insights by examining brain connectivity across controls, individuals with persistent ADHD, and those in remission. Their findings revealed that individuals with ADHD exhibited less pronounced increases in theta band connectivity following cognitive control tasks, indicating deficits in modulating connectivity during transitions from rest to task-oriented states. Additionally, during the Flanker task, those with ADHD demonstrated global hyperconnectivity in theta, alpha, and beta bands when the task’s prestimulus cues were incongruent and only showed hyperconnectivity in the beta band during the processing of target stimuli when the stimuli were congruent, which correlated with poorer cognitive control and dysfunctional behavior [81]. These findings are consistent with previous research showing enhanced connectivity in the alpha and beta bands during attention tasks [36,82].

Despite recent advancements, significant limitations persist in the literature. Many studies rely on small sample sizes and predominantly male samples or fail to report sex distribution, limiting the generalizability of their findings [81,83,84]. Furthermore, individual biomarkers often explain only a small fraction of the variance in treatment responses, highlighting the need for multifactorial and integrative approaches.

In addition, several methodological issues remain unaddressed in previous studies. Most existing tasks used for biomarker discovery are not grounded in theoretical models of attentional functioning and do not incorporate inhibition as a core element of the pathology, despite its central role in ADHD-related executive dysfunction [7]. These studies also tend to overlook functional connectivity at the source level and instead rely on scalp-level analyses, which limits the spatial precision and neural interpretability of their findings [45]. Moreover, EEG-based investigations are rarely validated against higher spatial resolution techniques such as magnetoencephalography, which could serve as a useful benchmark for signal source validation [47]. Finally, many studies use binary group comparisons (ADHD vs controls), neglecting the added value of including subtypes or differential diagnoses such as SCT and dyslexia in a 3-group design, which would allow for more clinically informative models [44,85].

Our study targets to address these gaps by using functional connectivity patterns, derived from EEG signals in the frequency domain, as biomarkers for ADHD. In this study, which will include at least 165 participants, we seek to improve the diagnostic accuracy achieved in the clinic at the time of this writing and to differentiate ADHD subtypes or other comorbid conditions that might be potentially confused with ADHD, using EEG-based ML algorithms applied to a go/no-go task. This study’s strength lies in its comprehensive design and robust sample size, enabling the identification of ADHD with high accuracy.

By integrating EEG connectivity metrics with ML algorithms, we aim to uncover reliable biomarkers for ADHD, advancing the field and contributing to a deeper understanding of this neurodevelopmental disorder.

The study included primarily right-handed children (10% left-handed) aged 7‐12 years who could suspend medication and tolerate EEG recording sessions. The inclusion of left-handed children, up to 10% of eligible participants (Table 1), partly addressed this limitation within the study. Despite excluding children with ASD or other comorbidities, the sample of this study is fairly representative of most ADHD populations. However, the team is planning future studies to examine biomarker performance in more heterogeneous, clinically representative ADHD populations, including children with common comorbidities, such as ASD.

Another limitation was that the final clinical diagnosis made by the PI was used as the reference test, as well as to define the participant grouping for algorithm training, thus introducing some circularity that could inflate diagnostic accuracy. However, several measures were introduced in order to mitigate this issue: first, the PI established the diagnosis independently, following established clinical criteria and without knowledge of any algorithmic or EEG outputs, thereby reducing the risk of circularity; second, several patients were previously diagnosed by other professionals, and the PI only confirmed the ADHD diagnosis; and third, a contract research organization was hired and an external neuropsychologist independently evaluated all patients.

Despite the aforementioned limitations, the study also presents notable strengths. First, it benefits from a large sample size, which enhances the reliability of the findings and supports the generalization of the results and allows for comparisons between ADHD subtype groups. Additionally, it incorporates a comprehensive neuropsychological assessment that enables a detailed examination of participants’ cognitive profiles. This approach allows for a more accurate identification of the specific neurocognitive processes affected in children with ADHD, providing a robust dataset that can inform future research and support the exploration of other clinical and neurodevelopmental aspects of the disorder across different ages.