This review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure methodological rigor and transparency (Supplementary Materials-PRISMA 2020 Checklist) [49, 50]. A review protocol detailing the inclusion/exclusion criteria, search strategy, and data extraction procedures was developed prior to conducting the systematic search. This systematic review was not prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) or other systematic review registries. The review protocol was developed a priori, specifying inclusion and exclusion criteria, search strategy, data extraction procedures, and analytical framework prior to commencing the literature search. The complete protocol is available from the corresponding author upon reasonable request. We acknowledge that prospective registration would have further enhanced methodological transparency, and this represents a limitation of the current study. An initial pool of 447 records was identified through systematic searches across PubMed (https://pubmed.ncbi.nlm.nih.gov/), Scopus (https://www.scopus.com/), Web of Science (https://www.webofscience.com/), and PsycINFO (https://www.apa.org/pubs/databases/psycinfo) databases. After the initial screening process:

• 198 duplicate records were removed.

• 23 non-English language studies were excluded.

• 36 records were excluded for being published before 2020.

• 65 records were excluded based on irrelevant titles or non-P300 ERP focus.

This resulted in 125 studies eligible for full-text review. A total of 52 studies were identified as being suitable for inclusion in this systematic review after the inclusion criteria had been applied to include only those studies that are based upon randomized and controlled study designs which utilize clinical populations; these studies have been organized into a comprehensive database format to document study aims, paradigms and measurements used to obtain P300 data, demographic information regarding the participants, specific clinical population(s) studied, type of interventions utilized and study outcomes related to each of the six research questions.

The 52 studies identified were all empirical and employed either a randomized controlled trial (RCT), a controlled clinical trial, or a quasi-experimental design with an appropriately matched comparison group. In addition, the clinical populations examined in these studies included dementia spectrum disorders (Alzheimer’s disease, vascular cognitive impairment, mild cognitive impairment), acquired brain injuries and disorders of consciousness, mood and anxiety disorders, neurodevelopmental and attention disorders, psychotic disorders and addiction, and chronic neurological and medical conditions.

Additionally, P300 data from the studies were used to examine the relationships between cognitive processing (i.e., attention, context updating, working memory, and cognitive control) and symptomatology, disease progression, treatment response, and functional outcomes in the respective clinical populations. Finally, the results of these studies were synthesized qualitatively to identify consistent, reliable, and clinically applicable biomarkers for addressing each of the six core research questions; the systematic review process is summarized in Fig. 1.

Fig. 1.

PRISMA flow diagram of study selection process. RCT, randomized controlled trial; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

The search strategy was designed to capture studies examining P300 event-related potentials as cognitive biomarkers in neurological and neuropsychiatric populations. Key search terms included:

• “P300” OR “P3” OR “Event-Related Potentials” OR “ERP” OR “Electroencephalography”.

• “Cognitive Biomarker” OR “Neural Marker” OR “Electrophysiological Marker”.

• “Neurological Disorders” OR “Neuropsychiatric Disorders” OR “Psychiatric Disorders”.

• “Alzheimer’s Disease” OR “Dementia” OR “Mild Cognitive Impairment” OR “Vascular Cognitive Impairment”.

• “Traumatic Brain Injury” OR “Disorders of Consciousness” OR “Stroke” OR “Brain Injury”.

• “Depression” OR “Anxiety” OR “Schizophrenia” OR “ADHD” OR “Addiction”.

• “Treatment Response” OR “Clinical Outcome” OR “Diagnosis” OR “Prognosis”.

• “Randomized Controlled Trial” OR “Controlled Study” OR “Clinical Trial”.

Search strings were adapted for each database to ensure comprehensive coverage:

(“P300” OR “P3” OR “Event-Related Potentials”) AND (“Cognitive Biomarker” OR “Neural Marker”) AND (“Neurological Disorders” OR “Neuropsychiatric Disorders”) AND (“Dementia” OR “Brain Injury” OR “Depression” OR “Schizophrenia” OR “ADHD” OR “Addiction”) AND (“Treatment Response” OR “Diagnosis” OR “Prognosis”) AND (“Randomized” OR “Controlled Trial”).

The search was limited to peer-reviewed articles published in English between January 2020 and August 2025. Only studies reporting empirical P300 ERP data related to cognitive function or clinical outcomes in neurological and neuropsychiatric populations with randomized or controlled study designs were included.

A structured set of inclusion and exclusion criteria was applied during the screening and selection process to ensure the relevance, rigor, and applicability of the included studies to the research questions.

Bias in the 52 reviewed studies was assessed using an adapted form of the Cochrane Risk of Bias Tool (RoB 2.0; Cochrane Collaboration, London, UK; https://www.riskofbias.info/), designed to assess risk of bias in neuroimaging research across both clinical and cognitive neuroscience. This version was developed based on the design-specific methodological features of P300 ERP studies, including randomized controlled trials, controlled clinical trials, and quasi-experimental study designs. Six domains were assessed:

(1) Selection Bias (Random sequence generation and allocation concealment)

Low Risk: Randomized controlled trials using randomized or appropriately matched groups were common.

Moderate Risk: Quasi-experimental studies lacking a description of the process by which subjects were assigned to either group or how participants were matched within a study.

(2) Performance Bias (Blinding of participants and personnel)

High to Moderate Risk: Studies utilizing ERP or treatment-based studies with behavioral treatments, neurofeedback, cognitive rehabilitation, or neuromodulation (transcranial direct current stimulation [tDCS]/repetitive transcranial magnetic stimulation [rTMS]) often utilized unblinded participants and personnel.

(3) Detection Bias (Blinding of outcome assessors)

Low Risk: Most studies used objective P300-derived measures (amplitude/latency), standardized clinical assessment tools, and/or automated processing of ERP signals to minimize assessor bias. Although several studies failed to describe assessor-blinding protocols for measuring behavioral outcomes.

(4) Attrition Bias (Incomplete outcome data)

Moderate Risk: Longitudinal intervention studies often reported high dropout rates. While many studies employed statistical methodologies (intention-to-treat analysis/multiple imputation) to address missing data, few provided specific methodological descriptions.

(5) Reporting Bias (Selective reporting of outcomes)

Low Risk: While most studies clearly reported primary P300 and clinical outcomes, a small number omitted secondary results or exploratory analyses, potentially indicating selective reporting.

(6) Other Bias (Funding sources and potential conflicts of interest)

Moderate Risk: Several studies, especially those utilizing commercial EEG equipment, neurofeedback platforms, pharmaceutical interventions, or neuromodulation devices, failed to provide full disclosure of potential conflict of interest or funding influences.

Two independent reviewers assessed each study within each domain. When two reviewers disagreed on their assessment of a study, they discussed it until they reached an agreement. If the two reviewers could not reach an agreement by discussing the study, a third reviewer would provide their opinion. This process of three reviewers evaluating studies increased the objectivity, clarity, and consistency of their assessments and enhanced the credibility of their study-quality ratings. Overall, across all studies reviewed, the risk of bias ranged from Low to Moderate. It can be stated specifically that the Selection Bias and Detection Bias domains were strong. The strength of these domains was based upon the design of the studies being Randomized Controlled Studies, and the objective nature of the P300-based outcome measure. Thus, the risk-of-bias assessment identified variability across the six domains of the studies reviewed. Therefore, caution is warranted when evaluating the results of studies with unclear blinding practices, incomplete data, or unidentified commercial interests. The summary of the risk-of-bias assessment for all 52 studies is shown in Table 1 and visually represented in Fig. 2.

Fig. 2.

Risk of bias assessment across 52 studies.

The risk-of-bias assessment illustrated the variety of study features across the six studied domains. More specifically, selection bias was generally managed appropriately; approximately 71.2% of studies used either proper randomization or participant matching. Similarly, the detection bias domain also had a low risk of bias (approximately 75%) in most studies; this reflected the inherent objectivity of P300-derived outcome measures and the standardization of ERP analysis pipelines. However, performance bias was much more difficult to manage. Only 25% of the studies were classified as low risk of performance bias, primarily due to the practical constraints of blinding participants and personnel in treatment-based interventions that use neuromodulation, neurofeedback, or cognitive rehabilitation protocols.

Attrition bias showed mixed results: 48.1% of studies demonstrated low risk through complete outcome reporting and appropriate handling of missing data, while 34.6% had moderate concerns regarding dropout rates or unclear handling of incomplete datasets in longitudinal intervention studies. Reporting bias was generally well-controlled (67.3% low risk), though 5.8% of studies showed evidence of selective outcome reporting.

The “other bias” category was of particular importance because it focused primarily on whether there was a conflict of interest or what funded the research. In only about 43 percent of the studies, we were able to classify them as low risk. It is worth noting that almost 14 percent of studies were classified as having a high risk of “other bias” due to commercial affiliations (e.g., EEG equipment manufacturers, pharmaceutical sponsors, neuromodulation device companies), which could have affected how the research was conducted and reported.

Additionally, just over 11 percent of studies did not provide enough information for us to evaluate their risk. These findings suggest that although the overall methodology of the included studies was sufficient—largely because they were randomized, controlled, etc.—the three categories of performance blinding, commercial influence, and drop-out management are important to consider when evaluating their results. There was no doubt that the greatest confidence would be placed on studies that completely documented their methodologies and had zero risk across all categories; that represents less than one-third of the total number of studies.

The data extraction process was conducted systematically using an extraction form created specifically for this literature review (Supplementary Table 1 provides complete study characteristics, including population, design, and intervention details). Data extracted from each of the studies included in this literature review were: bibliographical information (author(s), year, publication title), study design features (randomization method, type of control group, use of blinded methodologies), sample features (size of sample, age of participants, gender distribution of participants, clinical diagnoses of participants, diagnostic criteria utilized), P300 paradigm characteristics (type of stimulus used, type of oddball paradigm used, task parameters), P300 recording parameters (electrode placement, sampling frequency, reference), P300 outcome variables (P300 amplitude, P300 latency, electrode site), clinical outcome variables (cognitive assessments, symptom rating scales, functional measures), intervention characteristics (pharmacologic interventions, neuromodulation interventions, rehabilitation interventions) and primary findings related to P300 as a biomarker.

Articles were grouped according to the six research questions and qualitatively synthesized for each group. A summary of the 52 research articles examined in the systematic analysis is presented in Table 2 (Ref. [51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102]) to highlight the diverse array of P300-based studies examining neurological and neuropsychiatric disorders. The distribution of the studies across research questions was: RQ1 (Dementia Spectrum Disorders) = 11 studies; RQ2 (Acquired Brain Injury and Disorders of Consciousness) = 10 studies; RQ3 (Mood, Anxiety and Stress-Related Disorders) = 7 studies; RQ4 (Neurodevelopmental and Attention Disorders) = 6 studies; RQ5 (Psychotic Disorders and Addiction) = 12 studies; RQ6 (Chronic Neurological and Medical Conditions) = 6 studies.

The qualitative synthesis was designed to identify common patterns in P300 results across studies within each diagnostic category, evaluate the strength of evidence supporting the use of P300 as a diagnostic, prognostic, or treatment-response biomarker, and assess the clinical relevance and translational potential of P300-based assessment tools. A meta-analysis was not performed due to substantial heterogeneity in study designs, P300 paradigms, clinical populations, and outcome reporting across the included studies.

Eleven studies examined P300 as a cognitive biomarker for diagnosis, progression monitoring, and treatment response in dementia spectrum disorders, including AD, VCI, vascular dementia (VaD), and MCI. The evidence consistently demonstrates that P300 abnormalities—particularly prolonged latency and reduced amplitude—serve as sensitive indicators of cognitive dysfunction in these populations, with significant potential for monitoring the effects of interventions.

Ten studies investigated P300 as an effective biomarker for predicting recovery of consciousness, effects of cognitive rehabilitation, and overall functional outcome in patients who suffered from an acquired brain injury, such as post-stroke cognitive impairment (PSCI) and post-stroke depression (PSD); traumatic brain injury (TBI); and a minimally conscious state (MCS). The results provide strong support that P300 is also a valuable prognostic indicator of the individual’s capacity to recover, and serves as a highly sensitive measure of the degree of change associated with interventions intended to promote cognitive improvement—particularly in disorders of consciousness where behavioral assessments are frequently limited.

Seven studies used the P300 to determine if it was a cognitive biomarker for the level of symptoms, how well treatments would work, and cognitive impairment in mood and anxiety/stress related disorders, which included: Major Depressive Disorder (MDD), Anxiety Disorders, Obsessive-Compulsive Disorder (OCD), Tourette Syndrome (TS), Body-Focused Repetitive Behaviors (BFRB), and Eating Disorders. The results from these studies support the conclusion that the P300 is useful as both a biomarker of current cognitive function and, uniquely among the disorders studied, a biomarker for predicting future treatment success and symptom progression.

Six research studies investigated whether P300 could be used as an indicator of cognitive function in terms of ADHD and autism spectrum disorder (ASD) attention issues, treatment effects, and neurodevelopmental outcomes in ADHD, ASD, and Developmental Dyslexia. Studies provide evidence that P300 is sensitive to certain pharmacological and neurofeedback treatments, although variability in effect sizes across treatment modalities limits its use as a treatment-response biomarker for individuals diagnosed with neurodevelopmental disorders.

Studies that evaluated twelve P300 as a cognitive biomarker for cognitive control deficits, craving, substance effects, and treatment monitoring in psychotic disorders and many other forms of addictive behaviors such as alcohol use disorder, cannabis use, Internet Gaming Disorder (IGD) smartphone addiction, and nicotine dependency were the most studied of all research questions areas of study; this is due to the substantial interest in using P300 as a biomarker for understanding and treatment of these frequently treatment resistant conditions. Research indicates P300 can be sensitive to both the cognitive effects of substances and to changes related to treatment; there are emerging uses for P300 in the field of behavioral addictions.

Six research studies have demonstrated that the P300 could serve as a cognitive biomarker in assessing cognitive dysfunction caused by a variety of chronic neurological and medical illnesses; such as epilepsy, multiple sclerosis (MS), vestibular disorders, hepatic encephalopathy, and chronic obstructive pulmonary disease (COPD). These six studies demonstrate the wide range of applications of P300 as a cognitive biomarker, extending far beyond primary neurologic and psychiatric disorders to conditions in which cognitive dysfunction is secondary to medical illness or its treatment. The data from this group of studies indicate that the P300 can detect both disease-related cognitive deficits and drug effects, as well as treatment-related positive changes in cognition across many different medical conditions studied.

The table below (Table 3) presents an overview of the primary P300 findings for each of the six diagnostic categories discussed throughout this systematic review (detailed study-level P300 findings are provided in Supplementary Table 2). In addition to demonstrating the consistency of P300 abnormality findings (i.e., prolongation of P300 latency and reduction of P300 amplitude) across these diagnoses, the results suggest that P300 is a useful marker of cognitive dysfunction in individuals experiencing neurological or neuropsychiatric disorders. In support of this conclusion, the data also indicate that P300 can be influenced by a wide variety of treatments; thus, it may serve as a clinically relevant biomarker of treatment response.

To compare the performance of the P300 biomarker across disorder categories, a radar plot presents normalized scores (0–10) for eight aspects of P300 utility by diagnostic category. Those eight aspects are: Latency Reduction, Amplitude Increase, Treatment Response Sensitivity, Prognostic Value, Diagnostic Utility, Clinical Correlation Strength, Intervention Sensitivity, and Study Quality. A visual review of the radar plots (Fig. 3) shows that Dementia Spectrum Disorders (RQ1) and Acquired Brain Injury (RQ2) have the greatest number of P300 utility aspects with high scores. Mood/Anxiety Disorders (RQ3) are best suited for predictive uses of the P300, while Psychotic Disorders/Addiction (RQ5) are best suited for diagnostic uses of the P300. Neurodevelopmental Disorders (RQ4) had the greatest variability in P300 utility due to the varying levels of evidence regarding P300 responsiveness in ADHD and other neurodevelopmental disorders.

Fig. 3.

Comparative radar plot of P300 biomarker findings across diagnostic categories. RQ, research question.

Table 4 identifies a consistent direction of effects (i.e., the same) for the P300 measures across all 52 studies considered in this analysis. Specifically, prolonged latencies and lower amplitudes were observed at baseline across all three diagnostic categories, while effective treatments resulted in consistent normalization towards control values. The above-identified consistent trend provides a useful basis for interpreting P300 results and for tracking patient treatment progress in clinical environments.

To quantify the overall treatment effect on P300 parameters across included studies, a random-effects meta-analysis was conducted for studies reporting P300 latency changes with intervention. Fig. 4 (Ref. [52, 53, 55, 56, 60, 63, 66, 67, 69, 70, 73, 79, 84, 86, 91, 93, 98, 100]) presents a forest plot displaying the standardized mean difference (SMD) with 95% confidence intervals for individual studies and the pooled effect estimate. The analysis included 18 studies with comparable outcome measures, yielding a pooled SMD of –0.72 (95% confidence interval [CI]: –0.89 to –0.55), indicating a moderate-to-large treatment effect favoring P300 latency reduction (improvement) with intervention. Substantial heterogeneity was observed (I² = 67.3%, $\tau$² = 0.15, Q = 42.8, df = 17, p 0.001), reflecting variability in effect sizes across intervention types and clinical populations. The prediction interval (–1.40 to –0.04) suggests that while the direction of effect is consistently favorable, the magnitude varies considerably across contexts. Subgroup analyses revealed differential heterogeneity by intervention type: neuromodulation studies demonstrated moderate heterogeneity (I² = 45.2%), cognitive training studies showed low-to-moderate heterogeneity (I² = 38.6%), while pharmacotherapy studies exhibited high heterogeneity (I² = 78.4%), reflecting the diversity of pharmacological agents examined. The evidence is particularly strong and consistent in dementia spectrum disorders (RQ1), where multiple studies demonstrated latency reduction following cognitive rehabilitation or neuromodulation, and in acquired brain injury (RQ2), where neuromodulation interventions showed significant effects. Neurodevelopmental disorders (RQ4) showed less consistent patterns, with variable responses to different intervention modalities. These findings highlight the importance of considering both disorder-specific and intervention-specific factors when interpreting P300 treatment response.

Fig. 4.

Forest plot of P300 latency changes across intervention studies (k = 18 studies, random effects model). Negative SMD values indicate P300 latency reduction (improvement). Square size reflects study weight; horizontal lines represent 95% confidence intervals. The pooled estimate (red diamond) shows SMD = –0.72 [95% CI: –0.89, –0.55] with 95% prediction interval [–1.40, –0.04]. Heterogeneity statistics: I² = 67.3% [46.2%, 80.5%], $\tau$² = 0.15, Q = 42.8, df = 17, p 0.001. Ref. [52, 53, 55, 56, 60, 63, 66, 67, 69, 70, 73, 79, 84, 86, 91, 93, 98, 100]. Color coding indicates intervention type: blue = neuromodulation; green = cognitive training; orange = pharmacotherapy; purple = neurofeedback; brown = rehabilitation; pink = creative therapies; yellow = light therapy. Reference numbers correspond to included studies in the systematic review. SMD, standardized mean difference; CI, confidence interval; PI, prediction interval.

The data presented in Table 5 shows P300 responsiveness patterns across eight intervention categories from studies reviewed (detailed intervention protocols and study limitations are provided in Supplementary Table 3). The neuromodulation methods (rTMS and tDCS) produced the most consistent effects on both latency and amplitude in numerous conditions, including disorders of consciousness, in which P300 emerged. Patterns of P300 normalization were reliably shown to correlate with functional improvements when comparing cognitive training/rehabilitation and pharmacotherapy. Overall, these results demonstrate the value of using P300 as a marker of treatment response across various diagnostic conditions.

Fig. 5 provides an overview of how the P300 response has been studied across intervention type and diagnosis, presented in a matrix format. In addition to providing information on effect direction and the strength of evidence, each cell in the matrix is also colored; green denotes a strong positive effect (reduction in latency and/or an increase in amplitude), yellow indicates moderate or mixed effects, red indicates no or negative effects, and gray indicates there were no studies examining that specific intervention-diagnosis combination.

Fig. 5.

P300 intervention response patterns: summary visualization. Symbols: $\uparrow$, increased; $\downarrow$, decreased. Lat$\downarrow$, latency reduction (faster processing); Amp$\uparrow$, amplitude increase (enhanced engagement); norm, normalization; NS, non-significant. Color coding indicates effect strength: green, strong positive effect; yellow, moderate/mixed effect; red, null/negative effect; gray, no studies available.

Upon reviewing the colored cells in the matrix, several patterns emerged. First, neuromodulatory interventions (tDCS, rTMS) had the strongest and most consistently positive effects when comparing all diagnostic categories, especially within the dementia spectrum disorders and acquired brain injuries. Second, the effects of pharmacotherapeutic interventions vary widely depending on the disorder being studied. Pharmacological interventions have demonstrated positive effects in Alzheimer’s Disease (Fosgonimeton), ADHD (Methylphenidate), and Hepatic Encephalopathy (LOLA); however, pharmacological interventions have demonstrated variable or no effects in many other diagnoses. Third, the effects of neurofeedback appear to be variable and may depend on an individual’s ability to learn, rather than solely on the type of neurofeedback intervention. Each of these observations has implications for clinicians selecting P300 as an outcome measure in future clinical trials and for interpreting P300 changes in their clinical practice.

The P300 has been demonstrated to be a reliable cognitive biomarker useful for diagnosis, outcome prediction, and treatment monitoring across the 52 studies reviewed (Supplementary Table 4 summarizes key findings by research question). It was found that the P300 latency parameter provided the most reliable information about cognitive processing speed, and the P300 amplitude parameter provided the most reliable information about the resources available to focus on different stimuli at any given time. Additionally, the P300 was found to respond to various types of therapy (e.g., pharmacological, electrical stimulation (tDCS, rTMS, tACS), cognitive rehabilitation, neurofeedback, and creative/behavioral therapies), which supports the translational potential of the P300 for the assessment of clinical outcomes. The correlations observed between changes in the P300 and standardized cognitive measures (e.g., MMSE, MoCA, CRS-R, BRIEF2) across multiple studies provide additional support for the clinical validity of the P300 as a neurophysiologic measure of cognition. Fig. 6 represents a conceptual framework illustrating the connections between the clinical populations (six research question categories) studied in this review, P300 parameters (latency, amplitude, topography), therapeutic interventions (electrical stimulation, pharmacology, cognitive training, behavioral), and clinical applications (diagnosis, prediction, monitoring treatment, drug development).

Fig. 6.

Conceptual framework of P300 as a cognitive biomarker across neurological and neuropsychiatric disorders. HE, hepatic encephalopathy; VR, virtual reality.

The conceptual framework illustrates that the P300 can serve as a transdiagnostic biomarker linking neurophysiology to clinical decision making; it can be applied to such varied uses as the identification of individuals who may have a disease or condition early on and/or for the differential diagnosis of one type of disorder from another type of disorder to monitor responses to treatments and to develop new drugs.

Our systematic review supports the assertion that P300 parameters demonstrate a high degree of correlation with cognitive dysfunction across multiple neuropsychiatric disorders and, to some extent, specific diagnoses. Additionally, while the degree of diagnostic specificity of P300 abnormalities varied across disorders, they exhibited consistent patterns that supported differential assessment. Therefore, the consistency of our findings across 52 studies employing disparate methodologies provides us with greater confidence in the neurophysiological validity of P300 as a cognitive biomarker.

The association between P300 abnormalities and disorder-specific cognitive patterns was also examined by determining if the same P300 abnormalities were evident within each disorder, with implications for differential diagnosis. In disorders of the dementia spectrum (RQ1), MCI was characterized by both prolonged latency and increased latency variability, and Bae et al. (2024) [51] reported that individuals diagnosed with MCI had significantly less beta-band synchronization of the P300 waveform than their healthy counterparts. Moreover, the event-related desynchronization component of the P300 waveform was absent in individuals diagnosed with MCI; these findings are consistent with theoretical models that describe P300 latency as an indicator of cognitive processing speed and therefore indicate that prolonged P300 latency reflects slowed stimulus evaluation in neurodegenerative disorders. Vascular cognitive impairment also exhibited similar P300 latency abnormalities; however, whereas researchers in their study [56] documented both P300 latency and amplitude abnormalities in individuals diagnosed with vascular cognitive impairment, the amplitude abnormalities correlated with performance on the MMSE and MoCA, providing additional evidence of the concurrent validity of P300 with other established cognitive measures.

P300 has been shown to possess unique prognostic capabilities in disorders of acquired brain injury and disorders of consciousness (RQ2). The presence or absence of P300 waveforms in patients diagnosed with a minimally conscious state provides objective indicators of remaining cognitive capacity, in addition to information obtained from behavioral assessments. Specifically, the study [70] demonstrated that parietal rTMS elicited P300 waveforms in previously unresponsive patients, suggesting that the presence of P300 waveforms may indicate preserved attention allocation capabilities amenable to modulation via rTMS. Similarly, studies [64, 65] demonstrated that P300 latency improvements following tDCS were associated with improved performance on the Coma Recovery Scale-Revised. These findings establish P300 as a critical tool for assessing consciousness in situations in which behavioral measures are unreliable.

Abnormalities in P300 were also found in mood and anxiety disorders (RQ3); however, the implications of these abnormalities were primarily predictive rather than diagnostic. For example, the study [78] demonstrated that baseline P300 amplitude during monetary incentive delay tasks predicted therapy completion in individuals diagnosed with major depressive disorder; this represents a new application of P300 that extends its utility beyond diagnosis to the prediction of treatment engagement. Similarly, the study [76] provided prospective evidence that reduced flanker P300 amplitude predicted the development of depression over two years in female adolescents; thus, P300 can serve as a vulnerability marker for targeting interventions. These predictive uses represent a unique contribution to the P300 biomarker literature by mood disorder researchers.

P300 was found to be variably sensitive to interventions in neurodevelopmental and attention disorders (RQ4). The study [84] demonstrated that methylphenidate decreased Nogo-P300 latency in children diagnosed with ADHD, with latency changes corresponding to improvements in executive function domains. In contrast, neurofeedback studies produced variable results: specifically, study [79] found significant non-specific effects across neurofeedback modalities, whereas researchers in study [82] did not observe P300 enhancements following tACS. These variable results suggest that P300 responsiveness in neurodevelopmental conditions may be influenced more by the intervention mechanism and successful learning than by the modality employed.

P300 was found to be both a trait and state marker in psychotic disorders and addiction (RQ5) that represented the largest category of disorders studied (n = 12). For example, Liu et al. (2020) [90] demonstrated that individuals with alcohol dependence had lower P3a and P3b amplitudes than their healthy counterparts, and that partial recovery of P3a and P3b amplitudes occurred following four weeks of abstinence; however, P3a and P3b amplitudes persisted at levels below those of their healthy counterparts, indicating the existence of trait-like abnormalities. In contrast, researchers in their study [95] demonstrated that cognitive control deficits in nicotine dependence remained stable and independent of the acute abstinence state, thereby distinguishing nicotine dependence from alcohol dependence based on state dependency. Furthermore, behavioral addictions were associated with P300 correlations with craving: For instance, researchers in their study [93] demonstrated that P300 at Pz correlated positively with current craving in internet gaming disorder, indicating that P300 may be used as a craving biomarker and therapeutic target.

Finally, chronic neurological and medical conditions (RQ6) demonstrated that P300 is applicable to a wide range of disorders beyond primary neuropsychiatric disorders. Specifically, researchers in their study [102] demonstrated that the three anti-seizure medications tested (topiramate, carbamazepine, and levetiracetam) differentially affected P300: topiramate and carbamazepine had adverse cognitive effects, as indicated by prolonged P300 latency, whereas levetiracetam had positive cognitive effects, as indicated by reduced P300 latency. Researchers in their study [99] also demonstrated that levetiracetam has favorable effects on cognition, as evidenced by a reduction in P300 latency relative to placebo in healthy adults. Therefore, the pharmacological effects of P300 have direct clinical implications for selecting medications when preserving cognition is a priority.

In quantitative terms, the 52 included studies have both advantages and disadvantages in assessing methodological rigor for establishing P300 as a clinically reliable biomarker. An advantage is that the majority of the included studies have used randomized controlled trial designs, which is an advantage compared with the broader EEG biomarker literature. On the other hand, there is a large degree of heterogeneity in the acquisition parameters, processing pipelines, and outcome specifications among these studies; this hinders cross-study validation and the integration of results into meta-analysis.

The paradigms employed by each study elicit varied P300 responses. There was an assortment of paradigms, including auditory and visual oddball tasks, Go/NoGo paradigms, monetary-incentive delay tasks, and emotion regulation tasks. Although this variety demonstrates P300’s application across cognitive domains, it makes comparisons of findings difficult. Auditory oddball paradigms were used extensively in dementia and consciousness research. Visual paradigms and specialized tasks (e.g., cue reactivity for addiction) were used across other diagnostic categories. Establishing standardization of the core paradigm parameters (i.e., target probability, interstimulus interval, and attentional instructions) will enhance reproducibility while enabling continued methodological innovation.

The locations of P300 measurement sites varied across studies, with Pz, Cz, and Fz being the most used electrode sites. Additionally, some studies employed high-density arrays to localize sources, whereas others used simple montage techniques (with as few as two channels, as seen in [51]), which are more compatible with clinical use. Reference schemes (including linked mastoids, average reference, and nose reference) can significantly affect P300 amplitude measurements; however, they are typically not consistently reported. As such, the technical decisions regarding the reference scheme can greatly influence the absolute values of P300 parameter measurements and limit direct comparisons of amplitude findings among studies using differing reference schemes.

The sample sizes in the 52 studies exhibited considerable variation, with some studies proposing sample sizes of 80–160 participants, and completed trials having participant samples of 20–60 participants per group. The smaller sample sizes in many studies raise concerns about Type II errors and inflated effect sizes, particularly in studies that detect statistically significant effects. In addition, variability in sample sizes affects the precision of treatment effect estimates and the reliability of correlational findings between P300 parameters and clinical measures. Therefore, future studies should perform formal power analyses based on the sizes observed in this review to ensure adequate statistical power.

Although test-retest reliability is critical to establishing a biomarker, it has been systematically assessed in most of the studies included in this review. For example, treatment-monitoring applications of P300 assume that changes in the measured P300 response reflect the treatment effects rather than measurement variability. Studies that demonstrate P300 changes over the course of treatment (typically 4–12 weeks) provide indirect evidence of measurement stability; if consistent differences in the P300 response from baseline to endpoint are observed across all intervention and control groups, the measurements appear reliable. However, explicit reliability data from the studies reviewed would increase confidence in the utility of P300 as a longitudinal monitoring tool.

Despite promising research on P300 biomarkers, numerous technical and practical issues limit their use in clinical settings. Some factors that will influence the potential for biomarkers developed through research to become part of clinical practice are: hardware requirements, availability of qualified personnel to apply electrodes and process signals, incorporation into existing clinical workflow, and cost-effectiveness.

Acquiring P300 biomarkers using EEG depends on obtaining EEG data, applying electrodes to the scalp, and processing the data. Although this technology is available in many clinical settings, it is unlikely that all settings have the necessary technical infrastructure to obtain high-quality EEG data. For example, traditional high-end EEG systems that utilize 32–64 channels produce higher-quality EEG signals than lower-channel-count systems. Lower-channel-count systems are easier to transport and less expensive than their counterparts. There have been recent developments in low-cost, portable EEG systems utilizing 2–16 channels. These types of systems were utilized by some of the authors who contributed to this review. For example, researchers in their study [51] validated a small, portable EEG system that used two EEG channels to assess mild cognitive impairment, whereas researchers in their study [101] used a standard clinical EEG system to measure multiple sclerosis-related fatigue. These approaches reduce the technical expertise required to administer P300 assessments and lower equipment costs, while maintaining adequate sensitivity in detecting certain biomarkers.

Administering P300 assessments in clinical settings also requires careful consideration of how they will be integrated into current clinical practices. Time spent administering the P300 assessment includes time required to prepare patients for the assessment, time spent conducting the assessment, and time spent interpreting the results. In general, obtaining EEG data using EEG acquisition techniques typically adds 20–45 minutes to a clinical protocol. Most of this time is spent preparing the patient’s scalp with electrodes and administering the P300 paradigm. Many of the authors referenced in this review have utilized very short P300 paradigms. For example, researchers in their study [55] used P300 assessment immediately after the patient received medication, whereas researchers in the study [100] demonstrated changes in P300 over a period of three days following metabolic interventions for hepatic encephalopathy. Therefore, P300 can be easily integrated into time-sensitive clinical decision-making when appropriate protocols are established.

Although there are challenges to implementing P300 assessments in clinical settings, several of the applications referenced in this review demonstrate near-term clinical feasibility. Treatment response monitoring was identified as the most supported application across the diagnostic categories assessed. Changes in P300 that correlated with clinical improvements in cognition have been demonstrated to occur in association with both cognitive training [53, 54, 63], and various forms of neuromodulation [64, 65, 70]. A correlation has also been shown between P300 normalization and standardized clinical measures (MMSE, MoCA, CRS-R, BRIEF2), supporting the use of P300 as an objective adjunct to subjective symptom reports and clinician ratings.

Treatment prediction represents an emerging application with considerable clinical value. Baseline P300 amplitudes have been shown to predict therapy completion in depression [78], thereby enabling optimized treatment allocation. P300 was also shown to predict future depression development [76], indicating utility in the prevention of depression through targeted preventive interventions. If demonstrated to be valid in prospective implementation studies, these predictive applications would likely reduce treatment failure rates and related health care costs.

P300 demonstrates clinical value in disorders of consciousness where behavioral assessment may be unreliable. The appearance of P300 waves after neuromodulation [64, 65, 70] provides an objective indication of residual cognitive capacity that may not be evident on clinical examination. The application of P300 to disorders of consciousness directly addresses a significant unmet clinical need to obtain prognostic information to inform treatment decisions and support family counseling in patients diagnosed as being in a minimally conscious state. The simple interpretation of the presence of P300 (i.e., preserved attention allocation) facilitates its clinical acceptance without requiring clinicians to possess specialized expertise in EEG interpretation.

Although P300 biomarkers are currently used primarily in research settings to assess the neural mechanisms underlying cognition, they also have great potential as a population-level tool for studying neuropsychiatric disorders. Compared with many other methods for studying neuropsychiatric disorders, P300 biomarkers are especially attractive as a population-level tool because of their relatively noninvasive nature, millisecond temporal resolution, and declining cost-to-performance ratio. The results of the studies discussed above suggest at least four ways P300 biomarkers could be applied to address the major public health challenges posed by neuropsychiatric disorders at the population level.

Population-level early detection of neuropsychiatric disorders through the measurement of P300 abnormalities in “prodromal” or “at-risk” populations may enable early intervention and potentially prevent the development of full-blown clinical syndromes. For example, researchers in the study [76] found that lower amplitude P300s measured in adolescents prior to the onset of depression were associated with the subsequent onset of depression. These findings demonstrate the possibility of detecting individuals at high risk of developing a disorder and intervening early to prevent its development. Researchers in the study [51] found that portable EEG could distinguish older adults with mild cognitive impairment from those without (based on P300 measures). Although this finding has yet to be replicated in a larger sample, it is possible that similar technology will become available for assessing cognitive function in older adults in primary care or community settings.

Another way in which P300 biomarkers may be useful as a population-level tool is in identifying which treatments are most likely to be effective for an individual, based on whether the individual’s P300s change in a manner consistent with the known effects of a particular treatment. Although no published studies have examined how well the effectiveness of treatments for neuropsychiatric disorders can be predicted using P300 biomarkers, there is considerable evidence that the P300 is sensitive to drug effects on the brain and that changes in the P300 occur rapidly after drug administration. For example, researchers in study [99] found that P300 latency decreased significantly after acute administration of levetiracetam, supporting the neural efficiency hypothesis.

Finally, because P300 biomarkers provide objective evidence of treatment success, they may be used as a measure of treatment outcome. Although the relationship between P300 and clinical improvement has been studied extensively in individual subjects, a few published studies have examined how well changes in the P300 correlate with changes in symptoms in groups of subjects receiving different treatments. However, one study did find that P300 normalization in subjects with schizophrenia was significantly correlated with improvements in both positive and negative symptoms.

Because portable EEG devices are becoming less expensive and more accessible, it is now possible to collect P300 data in settings beyond the traditional laboratory. For example, researchers in their study [51] compared the ability of a portable EEG device with only two channels to that of a laboratory-based EEG device to detect P300 abnormalities in older adults and found that both devices were equally effective. Therefore, it should be feasible to train non-specialists to collect P300 data using portable EEG devices and to expand the availability of cognitive assessment tools in community health settings, rural areas, and low-resource countries.

The fact that P300 biomarkers reflect changes in the brain over time provides another advantage of using them as a population-level tool. Because changes in P300 are detectable weeks before the completion of a treatment regimen, it should be possible to use the P300 to objectively evaluate the effects of long-term interventions. In addition, because P300 changes are detectable during various interventions, including cognitive training, neuromodulation, and pharmacotherapy, it should be possible to evaluate the relative efficacy of different interventions.

Key findings from an analysis of methodological limitations across the 52 studies included in this systematic review indicate a need for significant research advancement to develop P300 biomarkers into a useful tool for clinicians. These areas of advancement are largely associated with establishing standardized acquisition methods, harmonizing processing strategies, developing valid biomarker validation methods, and advancing processing methods. The diversity of P300 acquisition methods across studies restricts cross-comparisons and meta-analysis. The following areas of acquisition parameter standardization are of high priority:

(1) Specify the electrode montage used and provide a rational basis for any reference scheme;

(2) specify the paradigm parameters (e.g., target probability, interstimulus interval, and total trials);

(3) use sampling rates and filters sufficient to capture the frequency content of the P300;

(4) ensure that participants receive identical instructions regarding attentional focus; and

(5) minimize state-dependent variability due to environmental factors.

Any efforts to standardize P300 acquisition should aim to strike a balance between providing a basis for reproducibility and enabling researchers and clinicians to innovate or adapt their methods as they see fit. Single-center studies can serve as excellent initial validation, but they do not provide sufficient evidence of generalizability to support the use of biomarkers in clinical settings. A well-designed multi-site study using a harmonized protocol can determine whether there are differences across sites, identify P300 features that are robust enough to account for variations in equipment and procedures, and establish population-based normative values across diverse populations. The protocol studies identified in this review [58, 59, 61, 66, 71, 94, 97] represent a first step towards validating P300 as a biomarker on a larger scale. However, collaborative efforts among multiple centers will be required to validate the P300 as a biomarker.

P300 parameters can only be interpreted clinically in relation to age-stratified normative data. P300 peak latency is known to increase approximately 1–2 milliseconds/year over the course of adulthood, while amplitude peaks during development (early childhood and late adolescent/young adult periods) and decreases gradually throughout the remainder of the life cycle. Clinical interpretations of P300 parameters frequently rely on controls within each study rather than on population-based norms. The creation of large-scale normative databases that cover all developmental ages would enhance the clinical interpretability of the P300 and enable the examination of individuals against age-matched standards.

Computational innovations could significantly improve the ability to extract and interpret P300 biomarkers. For example, machine learning algorithms may be able to better reject artifacts and detect components in EEG data than existing traditional methods. Additionally, source localization techniques can identify which brain regions generate P300 activity that is abnormal in patients with specific disorders (for example, the study [72] differentiated OCD and TS based on the source location of the P300, even though both had similar amplitude reductions). Finally, deep learning models trained on large amounts of EEG data could identify P300 features that are not evident with traditional methods, such as amplitude and latency.

Combining the P300 with additional measures could also improve the understanding of its mechanisms and predictive accuracy. Several of the studies reviewed here have shown relationships between changes in the P300 and changes in other markets: Researchers in the study [71] found that improvements in the P300 correlated with reductions in inflammatory markers and BH4 metabolism in patients with post-stroke depression; Other researchers in the study [67] found that the P300 correlated with DMN connectivity following rTMS; Additionally, researchers in the study [56] found that the P300 and BDNF changed similarly following transcranial ultrasound stimulation. Ultimately, systematic approaches that combine the P300 with additional modalities will clarify its relationship to other biological processes and help create integrated biomarker panels.

A systematic review is presented that provides an overview of where P300 currently stands as a cognitive biomarker and how it can help reduce the clinical burden associated with various neurological and neuropsychiatric disorders. The review was based on a synthesis of 52 randomized controlled studies and indicates both the barriers to P300 as a biomarker and how P300 can be translated into neuropsychiatric care, shifting from symptom-based to neurophysiological-based approaches.

The P300 has been shown to serve two roles: as a diagnostic marker of cognitive dysfunction and as a treatment-response biomarker across a variety of disorders. Consistent findings regarding prolonged latency and decreased amplitude at baseline followed by normalization after successful treatment across dementia, traumatic brain injury, mood disorders, addiction, and chronic medical conditions support the use of P300 as a transdiagnostic marker of cognitive processing efficiency. Trans-diagnosis allows a common marker of cognitive function to be used across multiple disorders and, therefore, may limit P300’s utility as a diagnostic tool; however, it also increases P300’s utility as a universal treatment-monitoring tool applicable across various clinical contexts.

P300 shows differential responsiveness to different intervention modalities. Neuromodulation (tDCS and rTMS) consistently shows the most positive effects across most disorders studied (dementia, disorders of consciousness, depression, and addiction) and results in P300 normalization following the intervention. Cognitive training resulted in consistent P300 improvements in dementia spectrum disorders and stroke rehabilitation. Positive pharmacotherapy effects were found in Alzheimer’s disease (fosgonimeton), ADHD (methylphenidate), and hepatic encephalopathy (LOLA), whereas no effect was found in schizophrenia with roflumilast, and some antiepileptic medications (topiramate and carbamazepine) were found to decrease P300 activity. The variability in neurofeedback results suggests that its effectiveness may depend on whether the individual learns successfully, rather than on the type of intervention.

These patterns will aid in establishing clinical expectations and research priorities for P300 biomarker applications. Studies reviewed in this systematic review indicate that P300 changes may occur prior to or coincide with clinical improvement within a 2–6-week timeframe. Researchers demonstrated significant reductions in P300 latency within 3 days of LOLA treatment in patients with hepatic encephalopathy [100]. Acute P300 effects have also been observed following single-dose fosgonimeton administration in patients with Alzheimer’s disease [55]. These rapid changes in P300 activity suggest that P300 may serve as an early indicator of treatment response, enabling clinicians to adjust treatment plans before clinical endpoints are reached. This temporal sensitivity is a significant advantage over standard clinical assessments, which typically take 4–8 weeks to evaluate treatment efficacy.

Several research priorities are established based on this systematic review: (1) Large scale, multi-center validation studies using harmonized methods to validate the reliability of P300 biomarkers across sites and populations; (2) Prospective studies investigating the implementation of P300 guided treatment selection in clinical environments; (3) Development of standardized and portable P300 assessment protocols that can be used in primary care and community settings; (4) Studies assessing longitudinal changes in P300 activity relative to disease progression and treatment response; (5) Research studying the P300 subcomponents (P3a, P3b) and other ERP components (N200, MMN) that may provide additional diagnostic information; and (6) Cost-benefit analyses evaluating the cost-effectiveness of treatment planning strategies utilizing P300 versus standard clinical practices.

Applying principles from the GRADE framework, the certainty of evidence for key findings was assessed qualitatively based on risk of bias, consistency, directness, precision, and publication bias considerations. Table 6 summarizes the certainty ratings for the primary conclusions of this systematic review.

These certainty ratings highlight that while P300 demonstrates promise as a cognitive biomarker, the evidence base varies considerably across applications. The strongest evidence supports P300 as a marker of cognitive dysfunction and of treatment response to neuromodulation, while predictive applications and the effects of pharmacotherapy require further investigation. Strengthening the evidence will require standardized protocols, larger sample sizes, and multi-site validation studies before definitive clinical recommendations can be established.

The high incidence of neurological and neuropsychiatric disorders creates a need for fast and accessible ways to assess them. From a public health viewpoint, P300 biomarkers fill significant gaps in today’s clinical assessment of neurological/neuropsychiatric disorders with significant public health implications. There are over 55 million individuals affected by dementia globally; there are over 280 million individuals suffering from depression; therefore, a huge demand exists for assessment services related to cognition, which currently exceeds the clinical capacity to provide this service. Biomarkers such as P300 can help augment limited specialist resources by providing standardized cognitive assessments in primary care and community-based settings. Researchers in the study [51] provided evidence of the technological feasibility of utilizing simplified portable EEG systems for broad-based use of P300 technology. Traditional methods for diagnosing neuropsychiatric disorders rely primarily upon the clinical interview and behavioral observation, and both of these methods may fail to identify subtle cognitive impairments or lack an objective neurobiological basis for their diagnoses. The P300 provides a direct measure of neural processing speed and efficiency that complements the clinical evaluation of the individual being assessed. The studies included in this review demonstrate the concurrent validity of P300 relative to MMSE and MoCA measures of cognition, as well as its sensitivity to intervention effects, thereby supporting its use as an objective measure of treatment outcome. The cost of neuropsychiatric disorders includes both the direct costs of healthcare, as well as indirect costs resulting from disability, loss of productivity, and caregiver burden. Non-responsive treatment failures contribute significantly to the cost of treating neuropsychiatric disorders, including longer treatment duration and prolonged periods of recovery time. Potential reductions in treatment failure and associated costs can be achieved through P300-guided treatment selection, assuming that the efficacy of this method is supported by future implementation studies. Findings suggesting treatment prediction [78], along with early treatment response indicators from multiple studies, indicate the potential for improved treatment outcomes and greater treatment efficiency.

One major shortcoming of present-day P300 biomarker research is the high degree of heterogeneity across studies, which severely hampers direct comparisons between them. The heterogeneity observed in present-day P300 biomarker research stems from numerous aspects of study methodologies and reporting practices.

The paradigms used to produce P300 responses in the studies reviewed were quite variable, including auditory oddball, visual oddball, Go/No-Go, Monetary Incentive Delay, Cue Reactivity, and Emotion Regulation. The variety of paradigms shows that P300 can be applied across virtually all cognitive areas; however, it also poses serious obstacles to comparing P300 amplitude and latency values across studies. Study parameters that relate to the task, such as the probability of the target, the interval between stimuli, the duration of stimuli, and instructions regarding the focus of attention, have varied greatly across studies and are known to influence P300 characteristics. The variability in paradigms among the studies reviewed prevents a formal meta-analysis of effect size and reduces the precision of pooled effect size estimates.

Differences in EEG recording parameters were observed across the studies reviewed. The number of channels recorded in the studies reviewed ranged from two [51] to high density, and Pz, Cz, and Fz represented the most common but not universal electrode location(s). Reference schemes were inconsistently documented and varied across studies that used linked mastoid references, an average reference, or another scheme. Differences in sampling rate, filter settings, and artifact-rejection approaches introduced technical variability that could not be easily resolved post hoc. As previously mentioned, these methodological differences will restrict the comparability of P300 parameters across studies.

Differences in population samples were observed in diagnostic criteria, illness duration, medication status, comorbidities, and symptom severity across studies of clinical populations. Within the same diagnostic categories, differences in patient populations are likely to affect P300 measures. Developmental factors affected by age range were introduced, with the age range spanning from childhood (ADHD, Autism Studies) to older adulthood (Dementia, Stroke Studies) in the studies reviewed. Additionally, the lack of age-referenced normative data made it difficult to interpret absolute P300 values and treatment effects.

Design differences among studies with only randomized and controlled studies represent methodological advantages. However, design differences affect the interpretation of studies. For example, some studies used an active control condition (i.e., Sham Stimulation), whereas others used a waitlist or treatment-as-usual control. Treatment duration ranged from 1 session to 12 weeks. The duration of follow-up after treatment also varied among studies, ranging from immediately post-treatment to 6 months post-treatment. These design differences affect the comparability of treatment effects and the sustainability of P300 changes following treatment.

Moderate risk of bias has been determined in several reviewed studies due to limitations related to sample size, limited use of blinding procedures, and limited reporting of outcome measures. The challenge of blinding in neuromodulation studies exists because patients may experience stimulation sensations. Outcome reporting in many of the studies reviewed emphasized statistical significance and did not place emphasis on non-statistically significant results or effect sizes. Despite the challenges associated with blinding and outcome reporting, the dominance of randomized controlled designs and the use of objective EEG-based outcomes (rather than subjective self-reported outcomes) represent methodological advantages that increase confidence in the observed treatment effects.

While the intervention studies reviewed here employed pre/post designs to evaluate treatment effects, the broader P300 literature remains dominated by cross-sectional designs, limiting causal inference and highlighting the need for more longitudinal research.

The results of this systematic review reveal several high-priority pathways for the use of P300 biomarkers in both clinical and public health settings. Further large scale multi-site studies using standardized methodologies are required to quantify variability between sites and determine how reliably P300 biomarkers can be measured across different environments and populations; to establish normative values for P300 parameters that allow clinicians to provide personalized interpretations based on an individual’s performance relative to their peer group; and to define optimal methods for collecting and analyzing P300 data [103, 104, 105, 106, 107].

To translate P300 biomarkers from laboratory to clinical use, it is necessary to examine the factors involved in their implementation in clinical practice. For example, randomized controlled trials conducted in real-world clinical settings that evaluate the feasibility, acceptability, and effectiveness of P300-guided treatments compared with current practices would provide evidence on the ability to implement P300 biomarkers in routine clinical practice. In addition, cost-effectiveness analyses comparing P300-guided approaches to conventional treatment methods would help inform healthcare policy decisions. Finally, training programs for clinical professionals to assess and interpret P300 data would provide needed professional development opportunities. Finally, integrating P300 assessments with electronic health records and clinical decision support systems would facilitate their seamless integration into the clinical workflow [13, 28, 108, 109].

For the P300 biomarker to be used more broadly, continued development of low-cost, easy-to-use EEG equipment will be necessary. Specifically, research demonstrating that simplified EEG electrode configurations and automated data processing techniques can produce results equivalent to those obtained with laboratory-based EEG systems will be required to establish technical criteria for translating P300 into clinical practice. Additionally, cloud-based data analysis software platforms may enable researchers to conduct P300 assessments in locations without local processing resources and/or integrate P300 assessments with telemedicine platforms to provide cognitive assessments to individuals who are remote from centers that provide these services [110, 111, 112, 113, 114, 115, 116].

While this systematic review has been limited to the application of P300 as a biomarker of clinical status, further understanding of the neurobiological basis of P300 abnormalities and their normalization will provide a stronger theoretical foundation for clinical applications of the P300 biomarker. Furthermore, integrating P300 with other neurobiologically relevant measures, such as fMRI, structural magnetic resonance imaging (sMRI), biochemical markers, and genetic information, will provide a broader neurobiological framework for the P300 biomarker. Recent advances in mapping EEG metrics to human affective and cognitive models offer promising interdisciplinary frameworks for understanding the relationship between P300 parameters and underlying cognitive processes. Understanding why some interventions result in normalization of the P300 will inform the development of optimized treatment protocols [117, 118, 119, 120, 121].

Emerging technologies present transformative opportunities for P300 biomarker research and clinical translation. The integration of artificial intelligence with biomarker analysis, including digital twin cognition approaches that create personalized computational models of individual brain function, represents a frontier in biomimetic neuropsychology that could enable real-time, adaptive P300-based interventions. Similarly, machine learning algorithms applied to neuroimaging data, including EEG-derived P300 parameters, show considerable promise for predicting mental disorder outcomes and treatment responses with greater precision than traditional statistical approaches. These computational advances may facilitate the development of automated P300 interpretation systems that can assist clinicians in making diagnostic and prognostic decisions [122, 123, 124].

In addition to expanding the breadth of potential applications of P300 biomarkers by investigating their use in other clinical populations, it is also important to investigate why certain interventions result in normalization of the P300. Clinical conditions not well represented in this systematic review, including post-Coronavirus Disease 2019 (COVID-19) cognitive deficits, cognitive deficits secondary to Parkinson’s disease, cognitive deficits secondary to multiple sclerosis, and age-related cognitive decline without dementia, all warrant systematic study. Additionally, pediatric applications require developmentally sensitive P300 paradigms and normative data spanning childhood and adolescence [5, 20, 83, 84, 125, 126, 127].

Finally, as P300 biomarkers begin to be implemented in clinical practice, consideration of equity and access issues will be crucial. Research studies utilizing diverse populations will be needed to ensure that the P300 biomarker is valid across demographic groups. Development of affordable, usable P300 technologies for low-resource environments will be important for expanding access to the P300 biomarker. Strategies for increasing access to P300 biomarkers for underserved populations will be important to avoid widening existing disparities in neuropsychiatric care [128, 129, 130, 131].

This systematic review has several limitations that should be acknowledged. First, substantial methodological heterogeneity was observed across the 52 included studies, including variability in P300 paradigms (auditory oddball, visual oddball, Go/No-Go, monetary incentive delay), EEG acquisition parameters (2 to 64 channels, variable reference schemes), and population characteristics (age ranges from childhood to older adulthood, diverse diagnostic criteria, medication status, and comorbidities). This heterogeneity limits direct cross-study comparisons and contributes to statistical heterogeneity (I² = 67.3%). Second, risk of bias concerns were identified, with moderate to high performance bias in approximately 69% of studies due to challenges in blinding participants to neurostimulation or behavioral interventions. Sample sizes varied considerably, with some studies potentially underpowered for detecting reliable effects. Third, study design variability affected the interpretation of findings. Control conditions ranged from sham stimulation to waitlist controls, treatment duration varied from single sessions to 12 weeks, and follow-up periods ranged from immediately post-treatment to 6 months. Fourth, this review was restricted to English-language peer-reviewed articles published between January 2020 and August 2025, potentially excluding relevant studies published in other languages or timeframes. Finally, the absence of standardized normative data and established clinical cutoffs for P300 parameters limits the clinical translation of findings. Despite these limitations, the predominance of randomized controlled designs and objective EEG-based outcomes provides methodological strengths that increase confidence in the observed treatment effects.