A Heme-Scavenging Protein in Brain Fluid May Flag Sepsis After Stroke Bleeds

When a blood vessel bursts deep inside the brain, the damage is only the beginning. Patients who survive the initial blow of a spontaneous intracerebral hemorrhage often face a second, quieter enemy: infection. Sepsis, the body’s runaway inflammatory response to infection, is one of the most feared complications in the neurocritical care unit, and it can quietly dismantle the chances of recovery even when the bleed itself has been brought under control. Now, a translational proteomics study and the scientific debate it has sparked are shining a spotlight on an unexpected molecule floating in the cerebrospinal fluid — hemopexin, the blood’s most avid scavenger of free heme — as a possible early warning signal for sepsis in these vulnerable patients.

The original investigation, published in the journal Neurocritical Care by Xie and colleagues, set out to answer a deceptively simple question: can the protein composition of the fluid bathing the brain reveal which patients with intracerebral hemorrhage will go on to develop secondary sepsis during their hospital stay? Using an exploratory proteomics approach, the researchers examined cerebrospinal fluid samples and identified hemopexin as a candidate biomarker, one whose levels appeared to track both the emergence of sepsis and the overall in-hospital outcome. If the finding holds up in larger and more diverse cohorts, it could give clinicians a biochemical head start on a complication that is far easier to treat when caught early.

To understand why hemopexin makes biological sense as a player in this drama, it helps to know what the protein actually does. Hemopexin is a plasma glycoprotein produced mainly by the liver, and it binds free heme — the iron-containing prosthetic group of hemoglobin — with the highest affinity of any known protein. After an intracerebral hemorrhage, red blood cells that escape into brain tissue lyse and release hemoglobin and heme in large quantities. Free heme is not an innocent bystander; it is a potent pro-oxidant and pro-inflammatory molecule that catalyzes the formation of reactive oxygen species, damages the blood-brain barrier, and amplifies the very inflammatory cascades that can spiral into systemic illness. Hemopexin mops up this heme and shuttles it to the liver for degradation, acting as a critical line of defense against heme-driven toxicity.

Earlier experimental work had already hinted that hemopexin sits at the fulcrum of recovery after brain bleeds. A 2018 study published in the Journal of Cerebral Blood Flow and Metabolism reported that increased brain hemopexin levels improved outcomes after intracerebral hemorrhage in animal models, supporting the idea that the heme-scavenging system is not merely a passive witness to injury but an active participant in the brain’s response. Against that backdrop, the new finding that cerebrospinal fluid hemopexin might also serve as a sepsis-specific signal adds an intriguing systemic dimension: the same protein that protects the injured brain locally may also register, or even mediate, the body-wide inflammatory storm that accompanies sepsis.

The clinical stakes are considerable. Spontaneous intracerebral hemorrhage accounts for a substantial share of stroke worldwide and carries high mortality and disability rates, which is why the American Heart Association and American Stroke Association issued updated management guidelines in 2022 covering everything from blood pressure control to surgical intervention. Within this framework, post-hemorrhagic sepsis remains a stubborn problem: it prolongs intensive care stays, worsens functional outcomes, and complicates decisions about the withdrawal or escalation of care. Current detection relies on nonspecific markers such as fever, white blood cell counts, and inflammatory indices like procalcitonin and C-reactive protein, none of which are specific to the neurocritical population. A cerebrospinal fluid biomarker that rises ahead of overt sepsis could, in principle, reshape surveillance on the neuro-ICU.

Yet the path from an exploratory proteomics signal to a clinically trusted biomarker is littered with statistical pitfalls, and this is precisely where the new letter to the editor enters the story. Humaiz Ahmed and Mahroo Niazi of Karachi Medical and Dental College published a correspondence in Neurocritical Care engaging directly with the original study, and their critique centers on the statistical architecture of the biomarker analysis rather than on the underlying biology. Their concerns are grounded in two classic methodological references that have shaped how clinical prediction research is judged for decades.

The first is the influential 1996 simulation study by Peduzzi and colleagues on the number of events per variable in logistic regression analysis. That work established a widely cited rule of thumb: logistic regression models need roughly ten outcome events for every predictor variable to produce stable, unbiased estimates. When the number of events is small relative to the number of candidate predictors — a situation that arises almost inevitably in exploratory proteomics, where hundreds or thousands of proteins are screened in modest patient cohorts — the resulting models are prone to overfitting, inflated effect sizes, and poor performance when applied to new patients. Ahmed and Niazi invoke this principle to question whether the hemopexin-sepsis association, however biologically plausible, can be reliably estimated from the sample at hand.

The second pillar of their critique draws on the 2020 guidance by Riley and colleagues, published in the BMJ, on calculating the sample size required for developing a clinical prediction model. That framework argues that prediction model studies should be designed with a target sample size determined in advance, based on the number of candidate predictors, the expected outcome proportion, and the desired precision of key parameters — not simply on whatever cohort happens to be available. Applied to the hemopexin study, the implication is that an exploratory translational analysis, however carefully conducted, may not have been powered to deliver a prediction model that clinicians can trust at the bedside. The letter thus functions less as a demolition of the finding and more as a call for the rigorous, pre-specified validation work that biomarker science demands.

The debate also unfolds against a rapidly evolving backdrop of sepsis prediction in intracerebral hemorrhage. In 2025, a multicenter retrospective study published in the Journal of Medical Internet Research identified and validated an explainable prediction model for sepsis in patients with intracerebral hemorrhage, demonstrating growing interest in machine-learning and clinically interpretable tools for this exact problem. In such a competitive field, any new biomarker — whether a single protein like hemopexin or a composite score — will be judged not only on its biological story but on whether it adds predictive value beyond existing clinical variables, and whether it survives external validation in independent, multicenter cohorts. The correspondence from Ahmed and Niazi is, in effect, a reminder that the bar for that evidence is high and well defined.

What emerges from this exchange is a picture of science working as intended. The original proteomics study offers a biologically coherent candidate: a heme-scavenging protein whose cerebrospinal fluid levels could plausibly reflect both the severity of intracranial injury and the systemic inflammatory state that predisposes to sepsis. The letter to the editor subjects that candidate to the statistical scrutiny that separates promising signals from clinical tools, citing the events-per-variable literature and prediction model sample size guidance as the relevant standards. The study authors have published a response, keeping the dialogue open. For patients and clinicians, the takeaway is one of cautious optimism: hemopexin may yet earn a place in the neurocritical care toolkit, but it will have to pass through larger, prospectively designed validation studies before a cerebrospinal fluid measurement becomes a routine part of sepsis surveillance after brain hemorrhage. Until then, the molecule that ferries heme out of the injured brain remains one of the most interesting leads in a field where every early warning counts.

Subject of Research: Cerebrospinal fluid hemopexin as a candidate biomarker for secondary sepsis and in-hospital outcome after spontaneous intracerebral hemorrhage

Article Title: Cerebrospinal Fluid Hemopexin as a Sepsis-Specific Signal After Intracerebral Hemorrhage

Article References: Ahmed, H., & Niazi, M. (2026). Cerebrospinal Fluid Hemopexin as a Sepsis-Specific Signal After Intracerebral Hemorrhage. Neurocritical Care. https://doi.org/10.1007/s12028-026-02664-3

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02664-3

Keywords: hemopexin, cerebrospinal fluid, intracerebral hemorrhage, sepsis, biomarkers, proteomics, neurocritical care, logistic regression, prediction models, heme scavenging, stroke, in-hospital outcome