Admission Lactate Adds Little to Prognosis in Severe Traumatic Brain Injury

Severe traumatic brain injury remains one of the most lethal conditions encountered in neurocritical care, with deaths concentrated in the first days after the insult, when the primary mechanical damage is compounded by an evolving cascade of secondary injury processes including mitochondrial dysfunction, excitotoxicity, neuroinflammation, and oxidative stress. Because this early phase is both the most dangerous and the most amenable to intervention, clinicians have long sought rapidly measurable markers that can identify, at the moment of admission, which patients face the highest risk of early death. Among the candidates, serum lactate has been particularly attractive: it is inexpensive, near-universally available, and obtainable within minutes as a product of anaerobic metabolism and tissue hypoperfusion. Yet a persistent question has shadowed its use. Is admission lactate a genuinely independent prognostic marker, or is it simply a biochemical echo of injury severity that adds little once severity itself is properly measured?

A new two-center retrospective study from China, published in Frontiers in Neurology, set out to answer exactly that question, and its findings suggest that lactate deserves a more modest role than some earlier work implied. The research team enrolled adults with severe traumatic brain injury, defined as a Glasgow Coma Scale score of 8 or less, who arrived at the hospital within 12 hours of injury and had a venous lactate measurement taken immediately on admission. Patients treated at the Clinical Medical College of Yangzhou University between January 2018 and December 2022 formed the development cohort of 287 patients, while those treated at Qinghai Provincial People’s Hospital between January 2019 and December 2023 formed an independent external validation cohort of 185 patients. The two centers differ markedly in geography and healthcare setting, with the Qinghai hospital serving a vast catchment area across the Qinghai-Tibet Plateau where prehospital transport times are often prolonged, providing a meaningful test of whether the findings travel beyond a single population.

The investigators deliberately applied restrictive eligibility criteria to isolate the neurological contribution to lactate levels. Patients in cardiac arrest, with multiple systemic injuries, severe infection, shock, major hepatic or renal insufficiency, malignancy, repeated blood transfusion, or coexisting hypertension and diabetes were excluded, because each of these conditions can substantially alter lactate production or clearance independent of brain injury. The craniotomy-timing criterion, excluding surgery performed more than five hours after injury, was intended to restrict the cohort to patients managed within a comparable early treatment window. In the development cohort, 412 patients were initially screened and 125 were excluded; in the validation cohort, 268 were screened and 83 were excluded. Both centers used the same blood gas analyzer model with an enzymatic method for lactate quantification, ensuring analytical comparability of the measurements across cohorts.

The clinical picture at baseline was sobering. Thirty-day mortality was 42.2 percent in the development cohort and 36.8 percent in the external validation cohort, and in-hospital mortality was 39.7 percent and 35.1 percent respectively. Non-survivors were older and more severely injured than survivors, with lower admission Glasgow Coma Scale scores, more frequently impaired pupillary reactivity, and higher Rotterdam computed tomography scores. Consistent with these severity differences, admission lactate was significantly higher in non-survivors than in survivors in both centers, at 4.5 versus 3.0 mmol/L in the development cohort and 4.2 versus 2.6 mmol/L in the validation cohort. Withdrawal of life-sustaining treatment occurred in roughly 14 percent of patients in each cohort and was far more frequent among non-survivors, a pattern the investigators addressed in a prespecified sensitivity analysis.

On univariable analysis, the association between lactate and mortality looked impressive. Each 1 mmol/L increase in admission lactate carried an odds ratio of 1.50 for 30-day mortality, with a 95 percent confidence interval of 1.30 to 1.73. But the crucial test came when lactate was placed alongside the established severity markers of age, Glasgow Coma Scale score, pupillary reactivity, and Rotterdam CT score in a multivariable model. After adjustment, the odds ratio for lactate fell to 1.19 per mmol/L, with a confidence interval of 1.00 to 1.42 and a P value of 0.051, indicating a positive association whose precision is limited rather than evidence of no effect. Within the adjusted model, pupillary reactivity, with an odds ratio of 2.18, and the Rotterdam CT score, with an odds ratio of 2.50, were the dominant predictors, while the Glasgow Coma Scale score was no longer independently significant, consistent with the overlapping severity information shared among these neurological and imaging markers.

The discrimination analysis told the same story in quantitative terms. Used alone, lactate discriminated 30-day mortality only moderately, with an area under the receiver operating characteristic curve of 0.711 in the development cohort, well below the severity baseline model’s 0.887. Adding lactate to the baseline changed discrimination only marginally, to 0.892, corresponding to a change in AUC of just +0.005 with a DeLong P value of 0.31. The reclassification indices were correspondingly small, with a continuous net reclassification improvement of 0.243 and an integrated discrimination improvement of 0.009, both with wide confidence intervals reflecting limited precision. The same pattern was reproduced in the external validation cohort, where the baseline-plus-lactate model yielded an AUC of 0.836 compared with 0.821 for the severity baseline, a change of +0.014 that was not statistically significant. Across both cohorts, the prognostic signal carried by lactate was largely already contained within established severity assessment.

To translate the model into an individualized risk tool, the investigators constructed a nomogram incorporating all five variables, with a published formula that yields, for example, an estimated 30-day mortality probability of approximately 73 percent for a 60-year-old patient with a Glasgow Coma Scale score of 6, one fixed pupil, a Rotterdam CT score of 4, and an admission lactate of 4.0 mmol/L. Bootstrap internal validation showed minimal optimism, with an optimism-corrected C-index of 0.884 and good internal calibration. External validation, however, revealed attenuated transportability, with a calibration slope of 0.81 suggesting some degree of overfitting, although the calibration intercept remained close to zero and the mean predicted mortality of 35.5 percent closely matched the observed 36.8 percent. Decision-curve analysis confirmed the limited incremental value of lactate, with the baseline and baseline-plus-lactate models yielding almost identical net benefit across the clinically plausible threshold range. Restricted cubic spline analysis showed a graded increase in adjusted odds of mortality with increasing lactate, with no significant evidence of nonlinearity.

The study also included an exploratory biological dimension. Using undifferentiated human SH-SY5Y neuroblastoma cells exposed to hypoxia at 1 percent oxygen for 12 hours, the researchers tested lactate at 5 mmol/L, a concentration approximating the admission levels observed in non-survivors. Lactate alone and hypoxia alone each reduced cell viability and increased apoptosis relative to control, with the largest effects observed in the combined hypoxia-plus-lactate group. Western blotting showed corresponding changes in Bax, Bcl-2, Caspase-3, LDH-A, and HIF-1α expression, most pronounced under combined hypoxia and lactate. These observations suggest that lactate may exert biological activity in this in vitro system, but the authors are careful to note that they do not establish a causal role for circulating lactate in neuronal injury and should be considered hypothesis-generating rather than confirmatory.

The findings sit within a literature that has reached conflicting conclusions about lactate in traumatic brain injury. Several studies have reported significant prognostic associations between early lactate levels and mortality, including a large multi-center cohort in which admission lactate emerged as an independent risk factor for one-week mortality, and a recent prospective study of isolated severe traumatic brain injury in which lactate above 4 mmol/L independently predicted unfavorable outcomes. A likely source of the discrepancy is how thoroughly injury severity was accounted for: previous studies seldom adjusted simultaneously for pupillary reactivity and the Rotterdam CT score, two of the strongest determinants of outcome, and rarely tested their models in an independent cohort. When these major predictors are comprehensively accounted for, the additional prognostic information provided by systemic lactate appears relatively limited.

The broader context of prognostic biomarker research in traumatic brain injury reinforces this interpretation. The field is moving toward multidimensional characterization, exemplified by the recent CBI-M framework integrating clinical, biomarker, imaging, and modifier pillars, and by brain-specific blood biomarkers such as GFAP and UCH-L1, which in large cohorts like TRACK-TBI have added prognostic information beyond clinical models, precisely the independent value that systemic lactate lacks in these data. Invasive cerebral microdialysis demonstrates that the brain’s local metabolic state, tracked by the lactate-to-pyruvate ratio and its temporal dynamics, most closely follows neurological outcome. Time-weighted lactate indices outperform static admission values for intensive care mortality prediction, and even prehospital lactate measured in the ambulance carries prognostic value for early mortality, suggesting that the temporal trajectory of lactate rather than a single admission value may carry the information that static measurement misses. The authors conclude that admission lactate should be interpreted as a readily available adjunctive marker rather than a stand-alone prognostic factor, and they acknowledge limitations including the retrospective design, the selected adult population, the absence of detailed pre-analytical data, and the restriction of the adjustment set to neurological severity markers. Further prospective multicenter studies are warranted to evaluate whether dynamic or multimodal lactate assessment provides additional prognostic value in the injured brain.

Subject of Research: Prognostic value of admission serum lactate in severe traumatic brain injury

Article Title: Is admission lactate an independent prognostic marker or a reflection of injury severity in severe traumatic brain injury?

Article References: Yan, F., Liu, H., Liang, M., Gao, W., Dong, L., Wu, J., Zhang, J., & Yang, M. (2026). Is admission lactate an independent prognostic marker or a reflection of injury severity in severe traumatic brain injury?. Frontiers in Neurology, 17, Article 1920098. https://doi.org/10.3389/fneur.2026.1920098

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1920098

Keywords: severe traumatic brain injury, admission lactate, prognostic marker, 30-day mortality, Glasgow Coma Scale, Rotterdam CT score, pupillary reactivity, external validation, SH-SY5Y cells, hypoxia, neurocritical care, biomarkers