Accuracy of a rapid glial fibrillary acidic protein/ubiquitin carboxyl‐terminal hydrolase L1 test for the prediction of intracranial injuries on head computed tomography after mild traumatic brain injury

Abstract Objective The objective was to determine the accuracy of a new, rapid blood test combining measurements of both glial fibrillary acidic protein (GFAP) and ubiquitin carboxyl‐terminal hydrolase L1 (UCH‐L1) for predicting acute traumatic intracranial injury (TII) on head CT scan after mild traumatic brain injury (mTBI). Methods Analysis of banked venous plasma samples from subjects completing the Prospective Clinical Evaluation of Biomarkers of Traumatic Brain Injury (ALERT‐TBI) trial, enrolled 2012–2014 at 22 investigational sites in the United States and Europe. All subjects were ≥18 years old, presented to an emergency department (ED) with a nonpenetrating head injury and Glasgow Coma Scale score (GCS) 9–15 (mild to moderate TBI), underwent head CT scanning as part of their clinical care, and had blood sampling within 12 h of injury. Plasma concentrations of GFAP and UCH‐L1 were measured using i‐STAT Alinity and TBI plasma cartridge and compared to acute TII on head CT scan. Results Of the 2011 subjects enrolled in ALERT‐TBI, 1918 had valid CT scans and plasma specimens for testing and 1901 (99.1%) had GCS 13–15 (mTBI), for which the rapid test was intended. Among these subjects, the rapid test had a sensitivity of 0.958 (95% confidence interval [CI] = 0.906 to 0.982), specificity of 0.404 (95% CI = 0.382 to 0.427), negative predictive value of 0.993 (95% CI = 0.985 to 0.997), and positive predictive value of 0.098 (95% CI = 0.082 to 0.116) for acute TII. Conclusions A rapid i‐STAT–based test had high sensitivity for prediction of acute TII, comparable to lab‐based platforms. The speed, portability, and high accuracy of this test may facilitate clinical adoption of brain biomarker testing as an aid to head CT decision making in EDs.


INTRODUC TI ON Background
Head computed tomography (CT) remains the imaging modality of choice for the diagnosis of intracranial injury such as hemorrhage or edema for patients with mild traumatic brain injury (mTBI) treated in the emergency department (ED) during the acute postinjury period. 1 CT scan along with patient symptoms and physical findings are the main factors used to direct care of such patients. 2 It has long been suggested that CT head imaging for mTBI patients is overutilized in those patients at very low risk for acute intracranial hemorrhage. 3 This concept, in conjunction with factors such as radiation exposure, 4 excess ED resource usage, cost, 5 and others, has prompted a search for tools that can effectively and safely identify patients at very low risk for intracranial abnormalities and, therefore, not needing CT imaging.
Clinical decision rules (CDR) that utilize easily obtained patient historic and physical findings have been developed in attempt to reduce CT use in low-risk mTBI patients. 6,7 Despite considerable enthusiasm for CDRs, they have been difficult to widely implement 3,8,9 and possibly are less sensitive than originally described. 10

Importance
The results of the ALERT-TBI trial suggested that a test combining glial fibrillary acidic protein (GFAP) and ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1; brain trauma indicator [BTI]) done within 12 h of head injury could reduce unnecessary head CT scanning by 30%, 11 and data from this trial were used to support Food and Drug Administration (FDA) clearance for this test in February 2018. The BTI, when used in patients with mTBI in whom a head CT scan is felt to be clinically indicated, can, with a high degree of certainty, identify those at low risk for acute traumatic intracranial injury (TII).
The 4 h required to perform the lab-based BTI test, however, has precluded widespread clinical adoption for ED patients given the average wait time of over 3 h for a head CT scan in the ED. 12,13 The recent development of a rapid GFAP/UCH-L1 test measured on the i-STAT Alinity platform overcomes this hurdle. 14,15 The i-STAT TBI test measures both GFAP and UCH-L1 simultaneously and received FDA clearance on January 8, 2021.

Goals of this report
Our study's main objective was to determine the predictive accuracy of a test combining both GFAP and UCH-L1 measured using the i-STAT Alinity and TBI plasma cartridge among those with mTBI by analyzing banked plasma samples from the prospective ALERT-TBI trial and comparing results to the standard head CT scan. A secondary preplanned objective was to determine predictive accuracy among the following TBI subsets: those with a Glasgow Coma Scale (GCS) score of 14-15, a GCS of 15, and a GCS of 15 undergoing blood draw within 2 h of injury.

Study design and setting
This is a secondary analysis of banked venous plasma samples from subjects successfully completing Prospective Clinical Evaluation of Biomarkers of Traumatic Brain Injury (ALERT-TBI) trial. 11 GFAP and UCH-L1 were measured in available plasma samples using a rapid test, which were combined into a single test result that was compared to head CT results. The study design and reporting followed the STARD guidelines for diagnostic accuracy. 16 The study of plasma specimen testing with the i-STAT Alinity and TBI plasma cartridge was conducted in compliance with the clinical protocol. Institutional review board (IRB) approval was obtained by each testing site's IRB.

Measurements
Blood sample handling Specimens were thawed and tested using i-STAT TBI plasma cartridges and i-STAT Alinity instruments.

Rapid test
The i-STAT TBI plasma test is a panel of in vitro diagnostic plasma quantitative measurements of GFAP and UCH-L1 and a semiquantitative interpretation of test results derived from a combination of these measurements. The i-STAT TBI plasma test consists of a single-use test cartridge (i-STAT TBI plasma test cartridge) that functions with the i-STAT Alinity system, a portable in vitro diagnostic test system (Data S1, Figure S1, available as supporting information in the online version of this paper, which is available at http:// onlin elibr ary.wiley.com/doi/10.1111/acem.14366/ full). The i-STAT TBI plasma test cartridge consists of immunoassays for GFAP and UCH-L1, which are evaluated simultaneously from a single plasma sample. In this study, each plasma specimen was thawed, aliquoted, and centrifuged at 10,000 RCF for 10 min. Approximately 20 µL was then pipetted into the sample well of the i-STAT TBI plasma test cartridge, which was then inserted into the i-STAT Alinity system. Sample analysis takes 15 min and concentrations of the two biomarkers are displayed on the analyzer screen. The reportable range for GFAP is 30-10,000 pg/ml and for UCH-L1 is 200-3200 pg/ml.
For the FDA cleared i-STAT TBI test, the analyzer will not display a value for measurement beyond the reportable range. The estimated lower limit of quantification was 23 pg/ml for GFAP and 70 pg/ml for UCH-L1. For both assays, the interrun coefficient of variation was less than 10%.

Outcomes
The primary outcome was acute TII on head CT imaging as deter-

Data analysis
The GFAP and UCH-L1 concentrations measured by the i-STAT TBI plasma test were used to report a test interpretation of either "elevated" or "not elevated." On the i-STAT Alinity instrument, the test interpretation is displayed on the first page and the second page shows the quantitative results ( Figure S1). The i-STAT TBI plasma test reports elevated if the concentration of one or both proteins is at or above its respective cutoff. The i-STAT TBI plasma test reports not elevated if the concentrations of both GFAP and the UCH-L1 are below their respective cutoffs. The cutoffs were determined prior to the study and designed to maximize sensitivity and negative predictive value (NPV). The cutoff for GFAP was 30 pg/ml and for UCH-L1 was 360 pg/ml. Additional methods describing the derivation of cutoffs are available in Data S1. Dichotomous i-STAT TBI test interpretations (elevated/not elevated) were correlated to presence or absence of CT-detected intracranial injury to determine the primary indicators of accuracy, sensitivity, and NPV. The following indicators of accuracy were also determined: specificity, positive predictive value (PPV), likelihood ratio positive, and likelihood ratio negative. Confidence intervals (CIs) for sensitivity, specificity, NPV, and PPV were calculated using the Wilson score method, 20 while CIs for likelihood ratios were calculated using the Miettinen-Nurminen score method. 21 Risk ratios with Haldane's correction 22 were used to compare the proportion of CT-positive subjects in those with protein concentrations in the upper 25th, 10th, and 5th percentiles to those below the prespecified cutoff. All analyses were performed using SAS v. 9.4 (SAS Institute).
The total sample size of this study was constrained by the number of archived specimens that met the subject and specimen eligibility requirements. Minimum sample size estimates were determined using the Wilson score test based on the allowable width of the 95% CI for a proportion. Assuming 95% clinical sensitivity with lower bound of the Wilson score 95% CI no less than 90%, the sample size was estimated to be a minimum of 110 CT-positive subjects.

Characteristics of study subjects
Of the 2011 subjects enrolled in the ALERT-TBI study, 1936 had plasma specimens available for testing on the rapid assay platform.
Plasma samples from 75 subjects were unavailable because the subject withdrew, did not have blood drawn, or did not consent to future testing, as shown in Mild TBI subjects were predominantly Caucasian with a mean (±SD) age of 49.1 (±20.99) years and just over half of the cohort was male (Table 1). Subjects were most commonly injured in a fall or motor vehicle collision, with 42.2% reporting a loss of consciousness (LOC). Of the 1901 mTBI subjects, 1879 (98.8%) had a GCS of 14-15 (

Main results
Of 1901 mTBI subjects, 1176 (61.9%) had an elevated test and 725 had a not elevated test. Among those with an elevated test, 115 had a positive head CT scan, while among those with a not elevated test, five had a positive scan (Figure 1). The rapid test, thus, had a sensitivity of 0.958 (95% CI = 0.906 to 0.982), specificity of 0.404 (95% CI = 0.382 to 0.427) and NPV of 0.993 (95% CI = 0.985 to 0.997) for acute TII ( Figure S2, Table 2). Five subjects had falsenegative test results; four had a presenting GCS of 15 and three had GFAP levels within 20% of the 30 pg/ml cutoff (Table 3). Of the five false-negative subjects, three had small SAH, one had a small SDH, and one had a parenchymal contusion (Figure 2). None were neurosurgically manageable. Subjects with GFAP or UCH-L1 concentrations in top 5%, 10%, and 25% of tested subjects were more likely to be CT-positive compared to subjects with concentrations below their respective prespecified cutoffs for GFAP (30 pg/ml) and UCH-L1 (360 pg/ml), as shown in Table 4. For both GFAP and UCH-L1, median values were higher among CT-positive subjects than from CT-negative subjects (Figure 3).  Table 2, Figure S2, and Table S3). Subjects with GCS 13-14 had similar test performance (Table S2).

DISCUSS ION
The results of this study suggest that, when applied to mTBI patients (GCS 13-15) in whom a head CT is felt to be clinically indicated, a test combining GFAP and UCH-L1 concentrations measured on a portable, hand-held device can detect acute TII with a high sensitivity (95.8%). This test was also highly sensitive when applied to the subgroup of patients with a presenting GCS of 15 and to those with GCS of 15 having blood drawn within 2 h of injury. Importantly, the performance characteristics of the i-STAT TBI plasma test were similar to those for the lab-based and FDA-cleared BTI panel using serum and different cutoff values. 11 In that study, the sensitivity and specificity among those with a GCS 14-15 were 0.973 and 0.367, respectively. However, in contrast to the i-STAT Alinity, which is easy to use and provides rapid results, the BTI requires labor-intensive sample preparation, long processing times, and substantial logistical support; these characteristics limit the utility of this platform for acute evaluation of patients.
For patients seeking care for potential mild to moderate TBI in the ED, CT scan continues to be the imaging modality of choice. There is general agreement, however, that CT scans for mTBI are overutilized and many are avoidable.  Value is within 20% of the 30 pg/ml cutoff. b No head CT findings were neurosurgically manageable injuries (see Figure 2) for CT images.

ACK N OWLED G M ENTS
We thank the following for their assistance in analyzing plasma sam-

CO N FLI C T O F I NTE R E S T
JJB is on the advisory board and has research support from Abbott, is on the Q30 advisory board, and has research support from BrainScope. RDW has been a paid presenter for Abbott sponsored programs. CAJ, RC, TM, SAD, HZ, and BM are employees of Abbott.
SAD is a shareholder of Abbott.