Increased specificity of the “GFAP/UCH-L1” mTBI rule-out test by age dependent cut-offs

Study population

This prospective study was conducted between 2022 and 2023 at the KAT General Hospital in Kifissia, Greece. All procedures were compliant with the Hospitals standard protocol for the management of patient with TBI, and the admission and discharge of each patient was decided after clinical evaluation and analysis of CT head scan report by the attending neurosurgeon. CT head scans were performed on all subjects with suspected TBI upon admission at the Emergency Department (ED) by an experienced attending radiologist using a Philips ingenuity 5,000 CT scanner. All clinicians who evaluated the patients at the ED were blinded to mTBI test results throughout the study (Figure 1).

Figure 1:

Diagram showing the flow and evaluation of patients in our study.

In addition, healthy control subjects matched to the patients (age and sex) were additionally recruited from a population of healthy volunteers participating in a study on reference ranges. The control subjects had the same exclusion criteria as the mTBI subjects.

For each enrolled patient, a K₂-EDTA blood sample was collected and sent to Clinical Biochemistry department. Samples were centrifuged immediately and plasma aliquoted and stored at −80 °C until tested. Blood samples were collected as close as possible to the time of evaluation, but no later than 12 h after the reported time of head injury.

All participants or legal guardians were granted written informed consent to participate. The study was approved by the KAT-Hospital Scientific and Ethical Committee under the number 746/10-9-2021.

Inclusion and exclusion criteria were as follows:

Inclusion criteria

1. >18 years of age

2. GCS 13–15

3. Indication of brain CT head scan: neurological focal deficit; anterograde amnesia; GCS <15 after 2 h post-TBI; suspicion of vault depression fracture; fracture of the basal skull; persisting nausea, vomiting or headache; post-TBI seizures; pre-injury treatment with antithrombotic drugs; loss of consciousness or amnesia with age >65 years, fall >1 m or hit pedestrian.

4. Available CT head scan and CT report

Exclusion criteria

1. Children

2. GCS 3–12 on admission

3. Primary admission for non-traumatic neurological disorder (e.g., stroke, spontaneous intracranial hematoma) as well as subject diagnosed neurodegenerative disease or other neurological disorder including dementia, Parkinson disease, multiple sclerosis, seizure disorder, or brain tumors

4. Time of injury cannot be determined or >12 h

5. Primary diagnosis of ischemic or hemorrhagic stroke or transient ischemic attack (TIA) within the last 6 months) or history of neurosurgery procedure within the last 90 days

6. Blood collection not feasible (i.e., skin integrity compromised at the venipuncture sites, blood vessel calcification (i.e., IV drug users, advanced atherosclerosis) both upper limbs missing (congenital or amputee))

7. Participating in an interventional, or therapeutic clinical study that may affect the results of this study

8. Penetrating head trauma

9. Patient with mechanical ventilation

10. Administration of blood transfusion after head injury at the admission and prior to the study blood draw

11. The subject is a female who is pregnant or lactating

Laboratory measurements

We used the Abbott’s “GFAP/UCH-L1” mTBI test on the automated Alinity I system (Abbott Park, IL, USA) to evaluate all patients and normal controls. The “GFAP/UCH-L1” mTBI test is a panel of in vitro diagnostic chemiluminescent microparticle immunoassays (CMIA) used for the quantitative measurement of GFAPand UCH-L1, in human plasma and serum. Although each biomarker is calibrated separately and provides quantitative results in ng/L, the “GFAP/UCH-L1” mTBI test combines both results to provide a semiquantitative interpretation of the test results derived from these measurements. The measurement of both biomarkers was performed according to manufacturer’s instructions on plasma EDTA and the mTBI result was evaluated as follows and according to the manufacturer’s datasheet: The mTBI test is considered as negative if GFAP is below 35 ng/L and UCH-L1 below 400 ng/L. In case, either GFAP or UCH-L1 or both are above these cut-offs, the mTBI test is considered as positive. The analytical coefficients of variation (CVs), determined according to CLSI guidelines, were 3.9 % for UCH-L1 and 3.4 % for GFAP.

Statistical analysis

Variables were assessed for normality using the Shapiro–Wilk test and plots. Since none of the continuous variables showed a normal distribution, nonparametric tests were used. For descriptive statistics, categorical data are reported as absolute numbers, whereas continuous variables are expressed as median, interquartile range (IQR). Given the usual clinical practice, we partitioned the mTBI cohort into patient younger or older than 65 years old when subgroups were investigated. Between groups comparison were performed using the Mann-Whitney test. Fitting curve in healthy subjects was drawn with LOESS smoothing with a span of 80 %.

Receiver operating characteristics (ROC) curves were drawn to calculate the area under the curve (AUC) and to define the new cut-offs. Sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) were calculated using cut-offs provided by Abbott and our own cut-offs.

For all statistical tests, the level of significance was fixed at 0.05. Medcalc statistical software (Medcalc, Belgium) was used for all statistical analyses.

Healthy controls

To evaluate the potential need for age-dependent cut-offs, we first evaluated the GFAP and UCH-L1 trend according to age in a healthy control population. The fitting curve in healthy controls showed an inflection point around 50 years old upon visual inspicetion (Figure 2A and B). A statistically significant increase was found between subjects <50 years old, subjects comprised between 50 and 64 years old and subjects ≥65 years old for both UCH-L1 and GFAP (p<0.0001) (Table 1).

Figure 2:

Fitting curve of GFAP (A) and UCH-L1(B) according to age.

mTBI cohort description

The cohort recruitment scheme is presented in Figure 1 and patients consisted of 362 individuals, 38.4 % of whom were women and 47.2 % were older than 65 years old (Table 2). A total of 37.0 % of the cohort reported to have anticoagulant treatment.

A total of 68.8 % of individuals had a negative CT head scan, whereas 31.2 % of subjects were positive on CT head scan with a GCS ≥13.

Median GFAP was significantly higher in individuals with positive CT head scan compared to those with a negative CT head scan (p<0.001). Additionally, median GFAP was significantly increased in subjects ≥65 years old, not only in the overall population (p<0.001), but also in the CT-scan negative population (p<0.001) compared to their younger counterparts (Table 2). The median GFAP in subjects ≥65 years and with negative CT head scan was almost twice the manufacturer’s cut-off for GFAP.

Similarly, to GFAP, UCH-L1 was elevated in people ≥65 years old in both the overall population and the CT-scan negative population (both p<0.001). Still, median UCH-L1 for the negative CT head scan population was below the cut-off suggested by the manufacturer (Table 2).

mTBI test performances

In our population, the sensitivity of the “GFAP/UCH-L1” mTBI test was 99.1 % and the NPV was 99 % while both GFAP and UCH-L1 sensitivity were decreased when considered separately (Table 3). Only one subject was negative on the “GFAP/UCH-L1” mTBI test while being positive on the CT head scan (Table 3). For this 27 years old patient whose CT head scan was objectivating a skull fracture, the GFAP concentration was 31 ng/L and the UCH-L1 concentration was 312 ng/L.

On the other hand, the specificity of the “GFAP/UCH-L1” mTBI test was less than 50 % in our cohort. The test specificity was driven by age as the specificity was significantly decreased in individuals older than 65 years old (Table 3). To better understand the impact of age on the number of positive tests, we further partitioned the cohort by 10 years. Although the number of CT head scans was constant regardless of the age, the number of positive tests increased from 45.5 % in the youngest adults to reach 96.5 % after 80 years of age (Figure 3A and B). Additionally, these patients were positive for GFAP in 40.0 % of the first two age groups and 96.5 % in the last two age groups whereas for UCH-L1, it was 34.5 % for the youngest and 56.5 % for the oldest (Figure 3C and D).

Figure 3:

Positivity rate of the “GFAP/UCH-L1” mTBI test is age-dependent. Positivity rate of CT head scan (A), “GFAP/UCH-L1” mTBI test (B), GFAP (C) and UCH-L1 (D) when the cohort is partitioned by 10 years.

To evaluate the impact of anticoagulant use, we selected the subjects older than 65 years old since the test performance is age-related and these medications are predominantly prescribed to the elderly. In this subgroup, 118 subjects were taking anticoagulant drugs while 40 did not (and 12 have unknown status). Among these 158 individuals, positive CT head scans increased, as expected, from 15 % in those not on anticoagulant medication to 31.4 % in those on anticoagulant medication (Figure 4A). However, the percentage of positive “GFAP/UCH-L1” mTBI tests was identical in both cases (90 and 89.8 %, respectively) (Figure 4B). Additionally, sensitivity and NPV were of 100 % for subjects older than 65 years old with or without anticoagulant intake while specificity was 14.8 and 16.7 % and PPV was 34.9 and 28.6 % for subjects older than 65 years old with or without anticoagulant intake, respectively. Specificity and PPV were not statistically different in both groups.

Figure 4:

Positivity rate of the “GFAP/UCH-L1” mTBI test is not modified by anticoagulant use: Positivity rate of CT head scan (A) and “GFAP/UCH-L1” mTBI test (B) when the cohort is divided based on reported anticoagulant use.

Age-dependent cut-offs for mTBI

Given the impact of age on test accuracy, we evaluated whether age-dependent reference ranges could increase test specificity. Receiver operating characteristic (ROC) curves were generated for this purpose. GFAP had an area under the curve (AUC) of 0.904 while the AUC for UCH-L1 was 0.864 when the entire population was evaluated. When only individuals older than 65 years old were used to generate ROC curves, the AUCs were as good as for the entire population (GFAP AUC: 0.928 and UCH-L1 AUC: 0.842), indicating that the same diagnostic performance can be achieved in the oldest population. To do so, we selected a sensitivity of 96 % for GFAP and 90 % for UCH-L1 in the subset of individuals >65 years old according to the Youden index. By taking advantage of both sensitivities, we obtained a “GFAP/UCH-L1” mTBI test sensitivity of 100 % as well as a NPV of 100 % with an increased specificity and PPV in this specific group (Table 4). Overall, the test accuracy with the new age-dependent cut-offs was 99.1 % sensitivity and 48.2 % specificity (Table 4). By applying these cut-offs, an additional 19 unnecessary CT head scans were avoided, doubling the number of ruled-out patients in patients older than 65 years old.