Diagnostic Accuracy of Plasma Biomarkers for Mild Traumatic Brain Injury in Older Adults Authors: Gershon Spitz, Jake Mitchell, Beatrice Duarte Martins, Abigail Astridge, Charlotte Copas, William T. O’Brien, Georgia F. Symons, Meng Law, Terence J. O’Brien, Matthew P. Pase, Jennie L. Ponsford, Biswadev Mitra, Stuart J. McDonald, Sandy R. Shultz Published: 2026 ABSTRACT This cross-sectional study explores the diagnostic accuracy of 4 plasma biomarkers to identify mild traumatic brain injury and to improve diagnostic certainty in clinically ambiguous cases among adults older than 60 years. Introduction Mild traumatic brain injury (mTBI) is a substantial and escalating public health problem, particularly among older adults,1,2,3 who experience the highest incidence of TBI-related hospitalizations and deaths.4 Following an mTBI, older adults can face disproportionately worse outcomes, including slower recovery trajectories and higher mortality, compared with their younger counterparts.1,5,6 In older adults, the diagnostic process for mTBI in the emergency department is challenging.7 Diagnostic criteria for mTBI typically involve a history of a plausible mechanism of injury coupled with the presence of clinical signs, symptoms, or abnormalities on clinical examination or neuroimaging.8 The cardinal signs (eg, loss of consciousness, amnesia, and confusion) and symptoms (eg, headache, dizziness, feeling slowed down, vision problems, and mental fog) of mTBI are often subjective and nonspecific and overlap considerably with age-related changes and a host of neurological conditions common in this age group, such as syncope, delirium, or stroke.9 This clinical ambiguity is further compounded by preexisting cognitive impairments, polypharmacy, and concurrent medical events, which complicate both diagnosis and management.10,11 Consequently, there is an urgent need for objective measures to complement current diagnostic criteria. Blood-based biomarkers offer a promising solution by quantifying brain-specific proteins released into circulation following brain injury.12,13,14 Glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase-L1 (UCH-L1), markers of astroglia and neuronal injury respectively, have already demonstrated clinical utility. When measured within 24 hours of injury, they are approved to help rule out the need for head computed tomography (CT) scans in adults with suspected mTBI.15,16,17,18 Neurofilament light (NfL) is a promising biomarker for axonal injury following mTBI, particularly when measured subacutely.19,20 Total tau can rise after TBI, although findings are inconsistent in mTBI; new assays for brain-derived tau (BD-tau) aim to improve specificity, but its temporal profile and utility in mTBI are unknown.21 Under the latest American Congress of Rehabilitation Medicine (ACRM) mTBI diagnostic criteria, a plausible mechanism of mTBI combined with either clinical signs or neuroimaging abnormalities, or a plausible mechanism of mTBI with both acute symptoms and clinical examination or laboratory findings, can meet diagnostic criteria. Elevated biomarker concentrations are included under laboratory findings. Therefore, once validated and operationalized, they may facilitate diagnosis (eg, in suspected mTBI when clinical signs and neuroimaging abnormalities are absent or confounded but a plausible mechanism and symptoms are present). However, blood biomarker research has focused on younger patients with mTBI, and a critical knowledge gap persists regarding the diagnostic accuracy of biomarkers in older adults.1,18 For example, baseline plasma concentrations of key biomarkers in healthy cohorts, most notably GFAP and NfL, are significantly higher and more variable in older individuals.22,23 In addition, studies have largely focused on the day of injury utility. Given that many cases present later—and this may be more common in older adults due to delayed recognition or reporting24—an optimal biomarker would retain utility over a larger time frame. This study’s aim is to determine the diagnostic utility of 4 plasma biomarkers (GFAP, UCH-L1, NfL, and BD-tau) in older adults presenting to an emergency department within 72 hours of injury. We compared biomarker performance between diagnosed mTBI cases (meeting ACRM criteria) and healthy controls, established age- and sex-adjusted diagnostic thresholds, and explored whether multiple biomarker combinations improved diagnostic accuracy. As a principal second aim, we applied these thresholds to clinically ambiguous suspected mTBI cases, that is, cases involving individuals with plausible injury mechanisms and symptoms but lacking definitive clinical signs, to assess biomarker potential for improving diagnostic certainty. This objective was designed to directly address the diagnostic gray zone formalized by the 2023 ACRM criteria and to explore the clinical utility of biomarkers among patients whose condition is difficult to classify. Methods Study Design and Participants This was a single-center, cross-sectional study conducted at The Alfred Hospital Emergency Department, a tertiary adult trauma center in Melbourne, Australia. This study was approved by the Alfred Hospital Human Research Ethics Committee. All participants provided written informed consent prior to participation. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. Prospective recruitment of patients and community controls occurred between June 27, 2024, and August 5, 2025. The study was designed to capture the full spectrum of head injury presentations in older adults. All patients presenting to the emergency department with a suspected head injury were screened for eligibility. Following routine clinical evaluation and initial diagnosis by the treating medical team, enrolled patients were systematically classified according to the 2023 ACRM diagnostic criteria by trained research staff. This prespecified approach stratified patients into either a diagnosed mTBI group, defined by a plausible mechanism and at least 1 objective clinical sign of brain injury (eg, loss of consciousness, posttraumatic amnesia), or a suspected mTBI group, defined by a plausible mechanism and acute symptoms but a lack of definitive, observable clinical signs. A third group of healthy community control participants was also recruited. Controls were community-dwelling volunteers at least 60 years of age with no recent head injury or bodily trauma recruited from the local community via advertisement. Exclusion criteria for all groups included a history of moderate to severe TBI, major preexisting neurological disorders (eg, dementia, stroke, or multiple sclerosis), or major psychiatric disorders (eg, schizophrenia, bipolar disorder). Patients were also excluded if they required neurosurgical intervention, were under the influence of illicit substances, or were non-English speaking. Race and ethnicity data were collected to more fully characterize the cohort. Information was obtained by participant self-report at enrollment. Response categories comprised the following options: Arab (eg, Saudi, Egyptian); Black; East Asian (China, Hong Kong, Japan, Macau, Mongolia, North Korea, South Korea, and Taiwan); Indigenous or First Nations; Latin American; South Asian (Bangladesh, Bhutan, India, Pakistan, Nepal, and Sri Lanka); Southeast Asian (Brunei, Myanmar, Cambodia, Indonesia, Laos, Malaysia, the Philippines, Singapore, Thailand, and Vietnam); West Asian (eg, Afghan, Iranian); White; other (free-text specification); and prefer not to say. Participants were permitted to select a single category, and the 1 respondent classified as other self-described as Albanian. Procedures and Data Collection All assessments were performed within 72 hours of injury for the mTBI groups. During enrollment, demographic information, medical history, and details of the injury event (mechanism and Glasgow Coma Scale score) were collected. Glasgow Coma Scale scores range from 3 to 15, with higher values indicating milder injury. Postconcussive symptom burden was quantified using the Rivermead Post-Concussion Symptoms Questionnaire (RPQ), from which the number of positive symptoms (score ≥2) and the total severity score (range from 0 to 37, with higher values indicating greater number of symptoms) were derived. Venous blood samples were collected from all participants. Samples were processed to obtain plasma, which was stored at −80 °C until analysis. Plasma concentrations of 4 biomarkers (GFAP, BD-tau, NfL, and UCH-L1) were quantified using the Simoa Neurology 4-Plex D (N4PD) Advantage PLUS assay. Statistical Analysis All statistical analyses were performed using R, version 4.2 (R Project for Statistical Computing). A 2-sided P < .05 was considered statistically significant for all tests. Participant demographic and clinical characteristics were summarized for 3 groups: control, diagnosed mTBI, and suspected mTBI. To assess differences in plasma biomarker concentrations and RPQ scores between participant groups, robust linear models were fitted using the lmrob() function from the robustbase package. These models evaluated group status while controlling for age and biological sex as covariates. The diagnostic performance of each biomarker and RPQ score for discriminating between control participants and participants diagnosed with mTBI was evaluated using covariate-adjusted receiver operating characteristic (AROC) curve analysis. The semiparametric model (AROC.sp function) from the ROCnReg package was used, which adjusts the ROC curve for age and biological sex. The primary outcome was the adjusted area under the curve (AAUC), with 95% CIs calculated via bootstrapping. To determine age- and sex-specific diagnostic thresholds, the compute.threshold.AROC() function was applied to the fitted AROC models. The optimal cutoff for each biomarker was determined using the Youden index. To quantify the uncertainty around these thresholds, a robust, nonparametric bootstrap procedure with 1000 replications was implemented, with 95% CIs derived using the percentile method. These personalized thresholds were then applied to individuals in the suspected mTBI cohort. To assess whether combining multiple biomarkers could improve diagnostic accuracy beyond individual markers, we evaluated all pairwise and higher-order biomarker combinations using pooled standardized scores. Statistical comparisons between combination models and the best-performing individual biomarker were conducted using DeLong tests. Sensitivity analyses assessed biomarker performance stratified by timing of blood collection (<24 vs ≥24 hours after injury) and CT scan result (CT negative vs CT positive for mTBI). Results Participant Characteristics A total of 89 participants (mean [SD] age, 71.0 [6.5 years]; age range, 60.0-84.0 years; 41 [46%] female and 48 [54%] male; 82 [92%] White and 7 [8%] all other categories combined [as defined in the Methods]) were included: 35 healthy controls, 45 with diagnosed mTBI, and 9 with suspected mTBI (Table). Controls were slightly younger (mean [SD], 67.8 [5.2] years) than individuals in the diagnosed mTBI (mean [SD], 72.2 [6.8] years) and suspected mTBI (mean [SD], 75.6 [5.7] years) groups; age was controlled for in all analyses. In the injury groups, the most common injury mechanism was the head striking a surface, accounting for 42 (93%) of the diagnosed mTBI cases and 8 (89%) of the suspected mTBI cases. The majority of both injury groups presented with an initial Glasgow Coma Scale score of 15 (34 [76%] in the diagnosed mTBI group; 9 [100%] in the suspected mTBI group), with smaller percentages of the diagnosed mTBI group scoring 14 (8 [18%]) or 13 (3 [7%]). The mean (SD) time from injury to blood collection was 26.6 (15.8) hours for the diagnosed mTBI group and 17.1 (8.5) hours for the suspected mTBI group. Extracranial injuries were present in 25 individuals in the diagnosed mTBI group (56%) and 1 individual in the suspected mTBI group (11%). Table. Participant CharacteristicsCharacteristicParticipants, No. (%)Control (n = 35)Diagnosed mTBI (n = 45)Suspected mTBI (n = 9)Age, mean (SD) [range], y67.8 (5.2) [61.0-80.0]72.2 (6.8) [60.0-84.0]75.6 (5.7) [65.0-82.0]Sex Female17 (49)18 (40)6 (67) Male18 (51)27 (60)3 (33)Educational achievement