Controlled and blinded biomarker measurement
Biomarkers were measured in a single experiment cycle by technicians blinded to clinical data, with coefficients of variation below 10% and all values above assay limits.
↳ Methods, Biomarker Analyses
Abstract
INTRODUCTION: Monitoring and timely clinical assessment are fundamental to treating patients with traumatic brain injury (TBI) in the intensive care unit (ICU). Proactive, individualised treatment is unavailable with the current methods.
RESEARCH QUESTION: How do outlying blood biomarker levels associate with clinical events or outcome in TBI?
MATERIALS AND METHODS: Glial fibrillary acidic protein (GFAP), neurofilament light (NfL), interleukin-10 (IL-10) and total tau (t-tau) were analysed from blood plasma samples of 70 ICU-treated patients, aged ≥18, with a clinical diagnosis of TBI and an indication for a head CT scan. The blood samples were collected on arrival and on days 1, 2, 3 and 7. The biomarkers were screened for outliers and biomarker values outside Q1-1.5 × interquartile range (IQR) or Q3+1.5 × IQR on any day post-injury were considered outliers. The outlier group was compared with the non-outlier group targeting clinically significant aspects such as high intracranial pressure, seizures/status epilepticus, or mortality/poor outcome. The Glasgow Outcome Scale Extended (GOSE) was evaluated between 6 and 12 months after the injury.
RESULTS: Difference was found in epileptic activity (n = 7 vs n = 0) and the number of performed decompressive hemicraniectomies (n = 6 vs n = 0) between the outlier and the non-outlier groups, p = 0.015 and < 0.0001, respectively. Patients in the outlier group were also more likely to have a poor outcome (GOSE 1-3) than patients in the non-outlier group, p = 0.011.
DISCUSSION AND CONCLUSION: Biomarker outliers seem to associate with clinical events and poor outcome in ICU-treated patients with TBI, possibly due to greater severity of TBI.
Key Claims
Primary findingTentative
Biomarker outliers seem to associate with clinical events and poor outcome in ICU-treated patients with TBI, possibly due to greater severity of TBI.
Fragile — the claim's weakest requirement is that the effect exists at all. Experimental. Every point below is traceable to a quoted passage, but this forecast has not yet been validated against known replication outcomes.
small cohort-relative analysis with severity imbalance and incomplete temporal identification
Abstract; Conclusions
SecondarySupported
All seizure-related events (n = 7) occurred in the outlier group
direct seven-versus-zero count with adjusted association but extreme imprecision
Results, EEG and ICP
SecondaryTentative
decompressive hemicraniectomies were performed only in the patients in the outlier group.
raw six-versus-zero count remains descriptive, but the reported regression significance is unreliable
Discussion, opening paragraph; Results, Neurosurgical Interventions
SecondarySupported
Patients in the outlier group were also more likely to have a poor outcome (GOSE 1-3) than patients in the non-outlier group, p = 0.011.
pooled association supports direction, though severity confounding and wide subgroup intervals limit interpretation
Abstract, Results; Results, Table 4
Derived from the full evaluation — not a separate score.
Strengths · Limitations
Strengths
Cautious severity-based final interpretation
The conclusion describes associations rather than causation and explicitly identifies greater TBI severity as a possible explanation. The discussion also treats the prognostic-model findings as hypothesis-generating.
↳ Discussion; Conclusions
Relevant longitudinal ICU research question
Repeated measurements across arrival and days 1, 2, 3, and 7 extend the outlier question beyond admission data and connect it to clinically relevant ICU events.
↳ Abstract, Materials and Methods; Introduction, final paragraph
Limitations
Severity confounds the central associations
The outlier group had higher total and cranial ISS and greater therapy intensity, while the hemicraniectomy and basic GOSE analyses do not credibly control this imbalance. The study therefore cannot establish that outlier status contributes information independently of injury severity.
↳ Results, Tables 2a–b and 4
Separated regression output is unreliable
Table 3 reports an odds ratio far above 100,000 with p<0.0001 for hemicraniectomy despite six procedures versus none, using an unpenalised multinomial model. This makes the abstract's reported strength of association unreliable even though the raw six-versus-zero count remains descriptive.
↳ Abstract, Results; Tables 2b and 3
Outlier definition limits clinical portability
Outlier status is derived from the same cohort across any of four biomarkers and five sampling days, with incomplete serial sampling. The resulting threshold is not externally validated and gives patients unequal opportunities to be classified as outliers.
↳ Methods, Statistical Analysis; Discussion, limitations paragraph
Nabu Commentary
The repeated biomarker measurements and documented assay controls support a useful exploratory analysis of ICU-treated TBI. The central associations remain difficult to interpret independently because the outlier group had higher injury severity and therapy intensity, while only the seizure analysis received a substantive severity-related adjustment. Table 3's estimates under complete separation require correction, as the reported p<0.0001 is not credible from the stated model. The cautious conclusion is better calibrated than parts of the abstract and prediction-model discussion, and the results remain hypothesis-generating rather than clinically actionable.
Check another paper
Author Note / Field Feedback
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Dimension breakdown
Study Quality
Limited2.8
Confidence mediumContribution
2.7
The study extends admission-only outlier work to repeated measurements during the first ICU week, but its main result remains a descriptive association between outliers, greater injury severity, and adverse events. The null incremental mortality result and non-significant favourable-outcome improvements limit the advance.
Introduction, final paragraph; Results, Figure 3; Discussion, outcome-prediction paragraphs
Addition of biomarker outlier status did not materially improve performance
Methodological Rigour
2.5
Assay quality control, blinded measurement, continuous monitoring, and Firth regression for seizures are competently described. Central surgery and outcome associations remain confounded by injury severity, while cohort-relative group construction, missing samples, and unstable separated estimates create substantial uncertainty.
Methods, Biomarker Analyses and Statistical Analysis; Results, Tables 2a–4
the patients in the outlier group being more severely injured
Reporting
3.5
The argument is followable and the conclusion explicitly qualifies the findings by greater TBI severity. Reporting is reduced by the unreliable hemicraniectomy p-value in the abstract, garbled CT counts, a GFAP unit inconsistency, and stronger language than warranted around non-significant AUC changes.
Abstract; Results, Tables 1, 2a, and 3; Discussion, prognostic-model paragraph
these findings should be interpreted as hypothesis-generating
Positioning
3.4
The discussion engages prior biomarker-outlier, seizure, ICP, neurosurgical, and IMPACT/CRASH literature and acknowledges several alternative explanations. It does not fully integrate the circular cohort-relative definition or the likely effects of incomplete serial sampling into interpretation.
Introduction; Discussion, literature and limitations paragraphs
all patients did not have a biomarker sample from all the collection dates
Reliability
Concerns4 of 4 checks
Internal CoherenceConcern
The neurosurgical regression output is internally inconsistent with complete separation, making the reported model estimates and the abstract's hemicraniectomy p-value unreliable. Other table anomalies reinforce the need for numerical and editorial correction.
Research ConductClean
Ethics approval, written consent, funding, and competing interests are declared. No affirmative data-availability, code-availability, or preregistration claim is made.
Flags: 2 declared / 5 total
Reference IntegrityClean
37 references in manuscript 35 of 35 checkable references found in an index 2 are books, websites or datasets — counted, but not index-checkable
Post-Publication RecordClean
No retraction notice found in Retraction Watch.
Last checked: 08 Oct 2026
Sources: Retraction Watch ✓
Readiness
Where this paper’s evidence sits on the path from initial observation to real-world use.
Mechanistic
The evidence is an initial single-centre observational signal rather than a validated monitoring tool. Incremental prognostic performance is null for mortality and non-significant for favourable outcome.
Methods, Statistical Analysis; Results, Figure 3
these findings should be interpreted cautiously given the small sample size
AI-generated, human-governed. Something look off? Contact us to request a review.