Association between route of nutrition and serum I-FABP and MMP-9 levels in critically ill patients with acute ischemic stroke: a prospective cohort study

Abstract

BACKGROUND:

Gut barrier dysfunction and neuroinflammation are increasingly recognised as linked pathophysiological processes in acute ischaemic stroke (AIS). However, the influence of nutritional routes on these processes remains poorly characterised.

OBJECTIVE:

To compare the longitudinal effects of enteral nutrition (EN) versus parenteral nutrition (PN) on serum intestinal fatty acid-binding protein (I-FABP), a marker of intestinal epithelial integrity, and matrix metalloproteinase-9 (MMP-9), a marker of neuroinflammatory activity, in critically ill AIS patients.

DESIGN:

Prospective observational cohort study

SETTING:

A tertiary neurology intensive care unit (ICU) in Türkiye

PATIENTS AND METHODS:

Consecutive adults (≥18 years) admitted to the neurology ICU within 24 hours of a confirmed AIS diagnosis were enrolled between October 2024 and January 2026. Nutritional route was assigned by the treating clinician according to ESPEN guidelines. Serum I-FABP and MMP-9 concentrations were measured using commercial enzyme-linked immunosorbent assay (ELISA) kits at baseline and on day 7.

MAIN OUTCOME MEASURES:

Longitudinal changes in serum I-FABP and MMP-9 concentrations between baseline and day 7, and their correlations with neurological severity (NIHSS).

SAMPLE SIZE:

73 patients (40 EN, 33 PN)

RESULTS:

Median age was 73 years (interquartile range, 64–80) and 37 (51%) were female. At baseline, I-FABP and MMP-9 levels were comparable between groups (all P>.05). Over the 7-day observation period, I-FABP trajectories diverged between groups (EN: 7.22 (2.68) to 6.69 (2.78) ng/mL; PN: 6.74 (3.09) to 7.82 (2.97) ng/mL), as did MMP-9 trajectories (EN: 11.5 (1.45) to 10.9 (2.31) ng/mL; PN: 11.3 (2.41) to 11.8 (1.87) ng/mL). At day 7, both biomarkers correlated significantly with NIHSS score (ρ=0.385 and ρ=0.440, respectively; both P<.001). Repeated measures analysis of covariance, adjusted for baseline NIHSS score, confirmed significant time × nutritional route interactions for both biomarkers (I-FABP: F[1,70]=4.22, P=0.044, P_FDR=.044, η²p=.057; MMP-9: F[1,70]= 4.76, P=.033, P_FDR=.044, η²p=0.064).

CONCLUSION:

In critically ill AIS patients, EN was associated with more favourable 7-day trajectories of gut barrier integrity and neuroinflammatory activity compared with PN, after adjustment for baseline stroke severity.

LIMITATIONS:

Single-center design, biomarker sampling restricted to two time points, reliance on delivered rather than utilised nutritional measures, and non-randomised nutritional route assignment with residual potential for confounding by indication.

INTRODUCTION

Stroke is the second leading cause of death worldwide, accounting for 12.2 million incident cases and 6.55 million deaths in 2019.¹ Acute ischaemic stroke (AIS) comprises approximately two-thirds of all stroke events, and patients requiring intensive care unit (ICU) admission represent the most severely affected subgroup, facing high short-term mortality and functional dependence.² The acute phase is characterised by excitotoxicity, oxidative stress, blood-brain barrier disruption, and neuroinflammation, each contributing to infarct expansion and secondary neurological deterioration.³

Impaired consciousness, dysphagia, and aspiration risk frequently preclude adequate oral intake in this population, rendering nutritional support a fundamental component of acute management.⁴ The European Society for Clinical Nutrition and Metabolism (ESPEN) guidelines endorse early enteral nutrition (EN) as the preferred route in critically ill patients, given its capacity to preserve intestinal mucosal integrity, sustain the gut microbiome, and attenuate systemic inflammation.^5,6 Parenteral nutrition (PN) is reserved for patients in whom the gastrointestinal tract cannot be adequately used, and has been associated with gut disuse atrophy, increased intestinal permeability, and higher rates of infectious complications.⁷ Despite these recommendations, critically ill AIS patients with haemodynamic instability frequently receive PN in practice, and the consequences of nutritional route on gut-mediated inflammatory pathways in this population remain poorly defined. Clarifying this relationship requires biomarkers capable of directly quantifying intestinal barrier injury. Intestinal barrier dysfunction is a recognised complication of critical illness, facilitating bacterial translocation, systemic endotoxaemia, and inflammatory cascade amplification.⁸ Intestinal fatty acid-binding protein (I-FABP), expressed predominantly by small intestinal enterocytes, is rapidly released into the circulation upon epithelial injury and has emerged as a sensitive biomarker of enterocyte damage.⁹ Circulating I-FABP concentrations reflect the degree of mucosal injury across critical illness phenotypes, and elevated levels have been reported in AIS patients compared with healthy controls, suggesting that gut barrier injury occurs early as part of the stroke-associated systemic response.^8,10,11 A parallel, brain-specific injury pathway is provided by neuroinflammation.

Neuroinflammation constitutes a cardinal pathophysiological mechanism in ischaemic brain injury, with microglial activation and peripheral immune cell infiltration extending injury beyond the infarct core.³ Matrix metalloproteinase-9 (MMP-9), a zinc-dependent gelatinase, contributes through degradation of blood-brain barrier components, facilitation of leukocyte transmigration, and propagation of neurotoxic cascades.¹² Circulating MMP-9 is elevated in acute AIS and has been linked to infarct growth and haemorrhagic transformation in both experimental and clinical settings.^13–15 Notably, gut-derived endotoxins can upregulate MMP-9 expression, providing a plausible mechanistic link between intestinal barrier failure and neuroinflammatory amplification.¹⁶

The gut-brain axis is a bidirectional neuroimmunological network through which intestinal barrier integrity and gut immune activation exert remote effects on neuroinflammatory tone.¹⁷ In AIS, this axis is particularly vulnerable: acute ischaemia reduces mesenteric perfusion, post-stroke dysmotility impairs mucosal trophism, and nutritional deprivation further compromises enterocyte integrity.^17,18 Despite this biological plausibility, however, whether nutritional route modifies gut barrier injury and its downstream neuroinflammatory consequences in this setting has not been systematically investigated.

We hypothesised that EN, by preserving intestinal epithelial integrity to a greater degree than PN, would attenuate I-FABP release and suppress MMP-9-mediated neuroinflammatory activity during the first week of ICU admission. The present prospective cohort study was designed to compare the longitudinal effects of EN versus PN on serum I-FABP and MMP-9 concentrations over a 7-day observation period, with secondary objectives including characterisation of the temporal relationship between biomarker levels and neurological severity.

METHODS

Study design and setting

This was a prospective observational cohort study conducted in the Neurology ICU of Kayseri City Education and Research Hospital between October, 2024 and January, 2026. The study was approved by the Kayseri City Hospital Non-Interventional Clinical Research Ethics Committee (No: 185, Date:10.09.2024) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients or their legal representatives prior to enrolment.

Participants and sample size

Adult patients (≥18 years) admitted to the neurology ICU with a confirmed diagnosis of AIS within the first 24 hours of ICU admission were eligible for inclusion. AIS was diagnosed based on clinical presentation and confirmed by neuroimaging (computed tomography or magnetic resonance imaging) by a neurologist.

Patients were excluded if they met any of the following criteria: (1) age under 18 years; (2) APACHE II score ≥25 at ICU admission, given the very high predicted short-term mortality and consequent inability to complete 7-day follow-up in this subgroup; (3) known malignancy; (4) acute or chronic mesenteric ischaemia; (5) short bowel syndrome resulting from prior gastrointestinal surgery; or (6) coeliac disease.

Sample size was estimated using a repeated measures design with two biomarker measurements (baseline and day 7). Based on the NUTRIREA-2 ancillary trial,¹⁹ which demonstrated a significant time-dependent difference in plasma I-FABP between enteral and parenteral nutrition in critically ill patients (P=.005; Cohen's f≈0.44), a conservative estimate of f=0.25 was adopted (within-subject correlation 0.5, α=0.05, power 80%), yielding a minimum of 66 patients. Accounting for a 10% dropout rate, the target enrolment was set at 73 patients. No AIS-specific effect size data were available, so this general shock cohort was used only as a conservative proxy.

Nutritional protocol

Nutritional support was initiated within 24 hours of ICU admission in all patients. The route of nutrition — enteral (EN) or parenteral (PN) — was determined by the treating clinician based on the patient's clinical status, gastrointestinal tolerance, and swallowing function, in accordance with current ESPEN guidelines for critically ill patients.⁵ Enteral nutrition was the default and clinically preferred route in our ICU, consistent with ESPEN guidance; parenteral nutrition was instituted only when the treating clinician determined that EN could not be safely or adequately delivered (e.g., severely impaired consciousness with high aspiration risk, haemodynamic instability requiring escalating vasopressor support, or clinically evident gastrointestinal intolerance), reflecting a documented clinical decision rather than random assignment. Gastrointestinal intolerance was defined in accordance with ESPEN criteria as a gastric residual volume >500 mL, or the presence of vomiting, abdominal distension, or ileus.⁵ Parenteral nutrition was administered as a standardised all-in-one admixture.

Caloric targets were set according to ESPEN recommendations:^5,6 20–25 kcal/kg/day during the acute phase and 25–30 kcal/kg/day during the recovery phase. Protein delivery was targeted at 1.3 g/kg/day for all patients. For obese patients, adjusted body weight was used in calculation according to ESPEN Guidelines. Daily energy and protein delivery were followed in study days and expressed as a percentage of the prescribed target, calculated as (delivered kcal or g protein / prescribed kcal or g protein) ×100, throughout the 7-day observation period.

Clinical assessments

Baseline clinical and demographic data were collected at ICU admission, including age, sex, body mass index (BMI), and comorbidities. Illness severity was quantified using the APACHE II score, the Sequential Organ Failure Assessment (SOFA) score, the National Institutes of Health Stroke Scale (NIHSS), and the modified Nutrition Risk in the Critically Ill (mNUTRIC) score; each score was assessed by the relevant specialist as part of routine daily evaluation — NIHSS by the neurologist, mNUTRIC by the clinical dietitian, and APACHE II and SOFA by the ICU intensivist. Baseline laboratory parameters — including serum electrolytes, renal function tests, full blood count, inflammatory markers (C-reactive protein, interleukin-6, procalcitonin), and serum albumin — were recorded within the first 24 hours of admission.

Neurological status was monitored daily using the Glasgow Coma Scale (GCS) from day 1 through day 7. In addition, vasopressor requirement, need for mechanical ventilation were followed throughout the study period. Furthermore, ICU mortality was recorded.

I-FABP and MMP-9 measurements

Blood samples were collected at baseline (day 1) and on day 7 of ICU admission. Serum was separated by centrifugation and stored at −80°C until analysis. Both biomarkers were quantified using commercially available enzyme-linked immunosorbent assay (ELISA) kits (Reed Biotech Ltd, Wuhan, China) according to the manufacturer's instructions. According to the manufacturer, the intra- and inter-assay coefficients of variation were below 3.4% and 7.3% for I-FABP, and below 5.7% and 7.5% for MMP-9, respectively. The delta (Δ) values for both biomarkers were calculated as the difference between day 7 and baseline concentrations.

Statistical analysis

Statistical analyses were performed using R (version 4.x; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were tested for normality using the Shapiro-Wilk test. Non-normally distributed variables are presented as median (interquartile range [IQR]) and compared between groups using the Mann-Whitney U test, with between-group differences additionally quantified using the Hodges-Lehmann (H-L) estimator with 95% confidence intervals. Normally distributed continuous variables are presented as mean ± standard deviation (SD) and compared using the independent samples t-test. Categorical variables are expressed as frequencies and percentages and compared using the chi-square or Fisher's exact test, as appropriate.

Changes in I-FABP and MMP-9 concentrations between baseline and day 7 were evaluated using a two-way repeated measures analysis of covariance (ANCOVA), with time (day 1 vs. day 7) as the withinsubject factor and nutritional route (enteral vs. parenteral) as the between-subject factor. Baseline NIHSS score was pre-specified as a covariate, given that stroke severity can independently influence both intestinal permeability and neuroinflammatory signalling. To evaluate the robustness of this model to other dimensions of illness severity and clinical management, four sensitivity analyses were additionally conducted, substituting or adding the following covariates alongside or instead of NIHSS: baseline APACHE II score; both NIHSS and APACHE II simultaneously; cumulative vasopressor exposure (days on vasopressor support); and mechanical ventilation status. Variance inflation factors were calculated for each multi-covariate model to confirm the absence of meaningful multicollinearity. Effect sizes are reported as partial eta-squared (η²p), with values of 0.01, 0.06, and 0.14 interpreted as small, medium, and large, respectively, and the Greenhouse-Geisser correction was applied where sphericity was violated.

Associations between biomarker levels and clinical severity score (reflected by baseline NIHSS) were assessed using Spearman's rank correlation coefficient (ρ). Correlations were performed separately for baseline and day 7 values, as well as for delta values, to characterise the temporal evolution of these relationships. An exploratory cross-lagged regression analysis further examined the temporal relationship between the two biomarkers: day 7 concentrations of each were regressed on their own baseline value, the baseline value of the other biomarker, and nutritional group. To account for the multiplicity of comparisons, P values within four pre-specified analysis families — the co-primary ANCOVA interactions, biomarker–NIHSS correlations, daily nutritional delivery comparisons, and the cross-lagged regression models — were adjusted using the Benjamini-Hochberg false discovery rate (FDR) procedure; baseline comparability testing and sensitivity analyses were excluded as non-confirmatory. All tests were two-sided, and a P value of <.05 was considered statistically significant.

RESULTS

A total of 73 critically ill patients with acute ischemic stroke were enrolled, of whom 40 (55%) received EN and 33 (45%) received PN. Among patients allocated to PN, the most common clinical indication was severe dysphagia or a high risk of aspiration, observed in 14 patients (42%). This was followed by depressed level of consciousness in 6 patients (18%), gastrointestinal intolerance (including high gastric residual volumes, ileus, or vomiting) in 5 (15%), haemodynamic instability in 4 (12%), gastrointestinal bleeding in 2 patients (6%), and other indications—predominantly perioperative or surgical interventions—in 2 patients (6%). All patients completed the 7-day observation period, with baseline and day-7 biomarker measurements available for the entire cohort. The median age of the cohort was 73 years (IQR: 64–80), and 37 patients (51%) were female. The two groups were comparable with respect to age, sex, and body mass index (all P>.05). The NIHSS scores tended to be higher in the PN group (9.0 [IQR: 7.0–14.0] vs. 7.0 [IQR: 5.8–11.3], P=.060). Systemic thrombolysis was administered to 27 patients (37%) overall, with a comparable rate between the EN (18/40, 45%) and PN (9/33, 27%) groups (P=.332). ICU mortality was 18% in the overall cohort and did not differ significantly between groups (EN Group: 15% vs. PN Group: 21%, P=.490) (Table 1), with all deaths occurring after day-7 biomarker sampling had been completed.

Table 1.

Baseline demographic and clinical characteristics of the study sample.

	Overall (n=73)	Enteral nutrition group (n=40)	Parenteral nutrition group (n=33)	P
Age, years	73 (64–80)	73 (66.8–81.0)	74 (63.0–80.0)	.618
Sex, n (%)
Male	36 (49)	22 (55)	14 (42)	.285
Female	37 (51)	18 (45)	19 (58)
BMI, kg/m2	25.3 (21.4–27.7)	26.0 (21.3–30.4)	24.4 (22.0–26.4)	.103
Comorbidity, n (%)	21 (29)	13 (32)	8 (24)	.438
Diabetes Mellitus	36 (49)	22 (55)	14 (42)	.285
Hypertension	5 (7)	2 (5)	3 (9)	.491
Coronary artery disease	44 (60)	26 (65	18 (54)	.364
Cerebrovascular disease	3 (4)	2 (5)	1 (3)	.673
Chronic kidney disease	3 (4)	1 (2)	2 (6)	.446
Congestive heart failure COPD	2 (3)	1 (2)	1 (3)	.890
APACHE II score	17.0 (13.0–19.0)	15.0 (12.0–19.0)	17.0 (15.0–19.0)	.060
SOFA score	5.0 (3.0–6.0)	4.0 (3.0–6.0)	5.0 (4.0–5.0)	.132
NIHSS score	8.0 (7.0–13.0)	7.0 (5.8–11.3)	9.0 (7.0–14.0)	.060
mNUTRIC score	3.0 (2.0–4.0)	3.0 (1.0–4.0)	3.0 (2.0–4.0)	.395
Mechanical ventilation, n (%)	7 (10)	3 (8)	4 (12)	.505
Vasopressor requirement, n (%)	8 (11)	3 (8)	5 (15)	.298
Systemic thrombolysis, n (%)	27 (37)	18 (45)	9 (27)	.332
Sepsis, n (%)	7 (10)	3 (8)	4 (12)	.505
ICU mortality, n (%)	13 (18)	6 (15)	7 (21)	.490

BMI, body mass index; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment; NIHSS, National Institutes of Health Stroke Scale; mNUTRIC, modified Nutrition Risk in Critically Ill; COPD, chronic obstructive pulmonary disease. Significance at P<.05.

Baseline laboratory parameters were comparable between the EN and PN groups (Supplementary Table S1). Inflammatory markers, specifically CRP (7.7 [2.8–18.1] mg/L overall; EN 9.2 [2.8–22.6] vs. PN 5.6 [2.9–11.8], P=.325) and IL-6 (19.6 [12.7–32.2] pg/mL overall; EN 19.6 [8.8–30.8] vs. PN 23.2 [13.4–32.2], P=.499), also did not differ significantly between groups, with the exception of serum bicarbonate, which was slightly higher in the PN group (P=.009).

Daily energy and protein delivery, expressed as percentages of prescribed targets, are illustrated in Figure 1. Caloric and protein delivery were similar between the EN and PN groups over the first three days (all P>.05, P_FDR≥.226). From day 4 onwards, caloric delivery was nominally higher in the EN group (P=.017–.019), a difference that did not persist after FDR correction (P_FDR=.076 for all three days). Protein delivery showed no significant between-group difference at any time point throughout the 7-day observation period (all P≥.099, P_FDR ≥.226).

Figure 1.

Daily caloric and protein delivery adequacy over the 7-Day observation period. Figure legend: Data represent the daily percentage of prescribed caloric and protein targets achieved in the enteral and parenteral nutrition groups. Nutritional delivery data are presented for days 1–6; day 7 corresponds to the final biomarker sampling point. Values are presented as mean±standard deviation (SD). EN, enteral nutrition; PN, parenteral nutrition.

Figure 2 shows the daily GCS assessments of the two groups. Although baseline GCS scores were comparable between groups (median: 11.5 vs. 10.0, P=.169), EN group demonstrated significantly higher scores from day 2 through day 5 (all P-values <.05). This between-group difference attenuated progressively thereafter, with non-significant differences observed on days 6 and 7 (P>.05). The PN group had more intubated patients than the EN group throughout the study period (Figure 2).

Figure 2.

Daily Glasgow Coma Scale Scores over the 7-day ICU observation period in the enteral and parenteral nutrition groups. Figure legend: Data are presented as median values with interquartile ranges. Numbers shown above each time point indicate the number of intubated patients in the enteral (EN) and parenteral (PN) groups on that day, respectively. *P<.05 between groups at the indicated time points. EN, enteral nutrition; PN, parenteral nutrition.

Spearman correlation analysis revealed that baseline MMP-9 was weakly correlated with NIHSS (Rho=0.253, P=.031, P_FDR=.047). On day 7, both I-FABP and MMP-9 levels demonstrated significant positive correlations with NIHSS score (Rho=0.385, P<.001, P_FDR=.003 and Rho=0.440, P<.001, P_FDR=.003, respectively). Furthermore, delta I-FABP was also positively correlated with NIHSS score (Rho=0.269, P=.021, P_FDR=.042) (Figure 3).

Figure 3.

Spearman correlation analysis between NIHSS Score and Serum I-FABP and MMP-9 concentrations Figure legend: Correlations are shown for baseline (Day 1) values, Day 7 values, and delta (Δ) values representing the change from Day 1 to Day 7. NIHSS, National Institutes of Health Stroke Scale; I-FABP, intestinal fatty acid-binding protein; MMP-9, matrix metalloproteinase-9.

At baseline, I-FABP levels were comparable between the EN and PN groups (mean (SD): 7.22 (2.68) vs. 6.74 (3.09) ng/mL, P=.481), as were MMP-9 levels (mean (SD):11.5 (1.45) vs. 11.3 (2.41) ng/mL, P=.615). Over the 7-day observation period, mean I-FABP levels declined in the EN group (6.69 (2.78) ng/mL) while increasing in the PN group (7.82 (2.97) ng/mL). A similar pattern was observed for MMP-9, with a decline in the EN group mean (SD): 11.5 (1.45) to 10.9 (2.31) ng/mL and an increase in the PN group mean (SD): 11.3 (2.41) to 11.8 (1.87) ng/mL. Delta values were −0.380 vs. +0.786 ng/mL for I-FABP (Mann-Whitney U=478, P=.04, rank-biserial r=0.277; H-L difference −1.373 ng/mL, 95% CI −2.818 to −0.052) and −0.358 vs. +0.287 ng/mL for MMP-9 (Mann-Whitney U=484, P=.05, rank-biserial r=0.266; H-L difference −0.765 ng/mL, 95% CI −1.511 to 0.00004) (EN vs. PN, respectively). Repeated measures ANCOVA adjusting for baseline NIHSS score showed significant time × nutritional route interactions for both I-FABP (F[1,70]=4.22, P=.044, P_FDR=.044, η²p=0.057) and MMP-9 (F[1,70]=4.76, P=.033, P_FDR=.044, η²p=0.064) (Figure 4). Sensitivity analyses yielded consistent results across all covariate specifications. Substituting APACHE II for NIHSS produced significant time × nutritional route interactions for both I-FABP (F(1,70)=5.93, P=.017, η²p=0.078) and MMP-9 (F(1,70)=6.54, P=.013, η²p=0.085), as did a combined model including both covariates simultaneously (I-FABP: F(1,69)=5.67, P=.020, η²p=0.076; MMP-9: F(1,69)=6.47, P=.013, η²p=0.086). Models additionally adjusting for vasopressor exposure (I-FABP: F(1,69)=4.19, P=.044, η²p=0.057; MMP-9: F(1,69)=5.25, P=.025, η²p=0.071) and for mechanical ventilation status (I-FABP: F(1,69)=4.72, P=.033, η²p=0.064; MMP-9: F(1,69)=4.47, P=.038, η²p=0.061) likewise remained significant, with no significant time × vasopressor-days interaction for either biomarker (I-FABP: F(1,69)=0.04, P=.849; MMP-9: F(1,69)=3.09, P=.083) and no meaningful collinearity between covariates in any model (VIF=1.02 for NIHSS and vasopressordays) (Supplementary Table S2).

Figure 4.

Temporal changes in serum I-FABP and MMP-9 concentrations over the 7-Day observation period in the enteral and parenteral nutrition groups Figure legend: Values are expressed as mean±standard deviation. The time × nutritional route interaction was statistically significant for both biomarkers (I-FABP: P=.044; MMP-9: P=.033). Panels B and D show the mean change (Δ, Day 7−Day 1) in each group, with the Hodges-Lehmann (HL) estimate of the median between-group difference and its 95% confidence interval (CI), derived from the Mann-Whitney U test. EN, enteral nutrition; PN, parenteral nutrition; I-FABP, intestinal fatty acid-binding protein; MMP-9, matrix metalloproteinase-9

The exploratory cross-lagged regression analysis showed that baseline I-FABP was independently associated with day 7 MMP-9 concentrations (β=−0.171, P=.027, P_FDR=.036), while baseline MMP-9 was not independently associated with day 7 I-FABP concentrations (β=0.265, P=.078, P_FDR=.078). Nutritional group remained independently associated with day 7 concentrations in both models (β=0.975, P=.026, P_FDR=.036 and β=1.436, P=.014, P_FDR=.036 respectively).

DISCUSSION

The principal finding of this study is that nutritional route was associated with divergent trajectories in two independent markers of gut barrier injury and neuroinflammation, I-FABP and MMP-9, over the first week of critical illness. Together with the emergence of biomarker–NIHSS correlations by day 7, this pattern raises the possibility of a relationship between intestinal integrity and neurological status in this population, consistent with gut–brain axis involvement described in other critical illness settings. The time × nutritional route interaction remained significant across sensitivity analyses adjusting for illness severity, vasopressor exposure, and mechanical ventilation status.

The divergent I-FABP trajectories observed between the EN and PN groups are consistent with the established trophic role of luminal nutrient delivery in maintaining enterocyte viability and mucosal integrity. These findings align with the ancillary analysis of the NUTRIREA-2 trial, which reported a significant time-dependent reduction in plasma I-FABP among enterally versus parenterally fed critically ill patients with shock, independent of caloric delivery.¹⁹ In addition, Yang et al demonstrated that EN maintains intestinal epithelial integrity through preservation of tight junction protein expression, an effect absent in parenterally fed mice.²⁰ The clinical relevance of sustained I-FABP elevation is underscored by Tyszko et al, who reported persistently elevated I-FABP concentrations in critically ill COVID-19 patients and demonstrated their independent association with 28-day ICU mortality.²¹ These findings suggest that the relationship between nutritional route and gut barrier integrity may extend to neurological ICU patients.

A parallel divergence was observed for MMP-9, with declining concentrations in the EN group contrasting with a progressive increase in the PN group. Although no clinical study has directly examined the effect of nutritional route on circulating MMP-9 in AIS patients, the observed divergence is biologically plausible. Enteral nutrition has been associated with lower circulating TNF-α, IL-1β, and IL-6 in critically ill patients with pancreatitis,²² and these cytokines are known upstream inducers of MMP-9 expression.¹³ Elevated MMP-9 has, in turn, been linked to worse neurological recovery, infarct growth, and haemorrhagic transformation in AIS.^14,15 To our knowledge, this is the first clinical observation of a possible relationship between nutritional route and MMP-9 in AIS, and it warrants confirmation in studies incorporating concurrent measurement of inflammatory cytokines. The relationship between the two biomarkers themselves, however, was less straightforward than a simple upstream-downstream model would predict. In our cross-lagged models, higher baseline I-FABP was associated with lower, rather than higher, day 7 MMP-9 concentrations. This inverse association does not fit a simple model in which mucosal injury directly drives neuroinflammation, and may instead reflect I-FABP capturing a non-specific stress response to critical illness rather than an upstream driver of neuroinflammatory activity. Given the exploratory, two-wave design and modest sample size, this association should be interpreted with caution.

A notable secondary finding was the emergence of biomarker–NIHSS correlations by day 7, which were absent at baseline. Early biomarker elevations may reflect a non-specific stress response common to critical illness generally.⁸ The later association with stroke severity may instead indicate that gut barrier and neuroinflammatory changes become more closely linked to neurological status as illness progresses, consistent with prior reports of elevated I-FABP specifically in AIS patients.¹¹ This would be consistent with a self-reinforcing cycle involving cerebral ischaemia and peripheral immune activation, as reported in other critical illness populations.^3,18 Day 7 I-FABP was similarly correlated with APACHE II, SOFA, and mNUTRIC scores, in line with prior reports linking enterocyte injury to worse outcomes in critical illness.^8,10,21,23

Caloric and protein delivery did not differ significantly between the EN and PN groups after correction for multiple comparisons. This weakens the case for nutrient adequacy as a confounding explanation. The route of delivery itself — whether nutrients passed through the gut or bypassed it — appears to be the more plausible driver of the divergent I-FABP and MMP-9 patterns.

The EN group showed higher GCS scores between days 2 and 5, but this finding warrants caution. A similar, non-significant difference was already present at baseline, alongside a trend towards higher NIHSS scores in the PN group, pointing to pre-existing severity imbalance as a possible contributor. In addition, the GCS verbal component cannot be scored in intubated patients, capping their total score at 10 regardless of true consciousness level — and intubation was more common in the PN group throughout the study. Together, these factors may account for much of the observed difference, and the GCS findings should be regarded as hypothesis-generating rather than confirmatory. A related question is whether the favourable EN biomarker trajectory instead reflects faster intrinsic neurological recovery, rather than a gut-protective effect of feeding. This seems less likely, since nutritional route is typically fixed near admission rather than revised with recovery, and nutritional group remained associated with day 7 biomarkers even after adjusting for each patient's own baseline biomarker value.

Several limitations of this study warrant acknowledgement, alongside the steps taken to address them where possible. First, the observational design precludes definitive causal inference. To partially mitigate this, our ANCOVA models adjusted for baseline illness severity, and consistent results across multiple covariate specifications reduce the likelihood that a single measured confounder explains the findings; nonetheless, only a randomised design could establish causality. Second, biomarker measurements were restricted to two time points, precluding characterisation of the full temporal trajectory; more frequent sampling would clarify this in future studies. Third, the distinction between nutrient delivery and systemic utilisation could not be directly addressed, as functional markers such as nitrogen balance or indirect calorimetry were not incorporated into the protocol. Fourth, data on infarct size, and vascular territory were not collected. Baseline NIHSS and APACHE II captured overall severity but not these radiological and procedural details. Fifth, NIHSS scores were recorded only at baseline, with daily monitoring based on GCS instead, reflecting standard ICU practice but limiting longitudinal tracking of stroke-specific severity. Sixth, gastrointestinal dysfunction severity could not be incorporated as a covariate, as it is itself closely tied to the indication for parenteral nutrition; residual confounding by indication cannot therefore be fully excluded, though sensitivity analyses adjusting for related factors did not alter the primary finding. Notwithstanding these limitations, the present study has several strengths. It is, to our knowledge, the first prospective cohort study to simultaneously examine the effects of nutritional route on both gut barrier integrity and neuroinflammatory biomarkers in critically ill AIS patients. An adequately powered sample size and the use of a repeated measures design with a predefined primary interaction effect strengthen the temporal inference of the findings.

Clinical implications

These findings offer a possible biomarker-level perspective on the existing rationale for favouring enteral feeding when clinically feasible in critically ill AIS patients. Because the nutritional route is typically decided within the first days of ICU admission, the early time course of biomarker divergence observed in this study suggests that this decision window may carry relevance beyond immediate feasibility. Should the association with MMP-9 — not previously examined in relation to nutritional route in AIS — be confirmed in larger studies, serial biomarker measurement might eventually help identify patients in whom gut barrier dysfunction is progressing unfavourably. These implications remain exploratory at this stage.

In conclusion, in this cohort of critically ill AIS patients, nutritional route was associated with divergent trajectories of I-FABP and MMP-9 over the first week of ICU admission. Biomarker levels also became increasingly correlated with NIHSS score over this period, a link that was not present at baseline. Together, these observations may be compatible with a relationship between gut barrier integrity and neurological status in this population. Larger, prospective studies would be needed to establish whether this relationship is reproducible and clinically meaningful.

SUPPLEMENTS

Supplementary Table S1.

Baseline laboratory parameters of the study sample.

	Overall (n=73)	Enteral group (n=40)	Parenteral group (n=33)	P
Hco3	25.0 (23.2–26.6)	24.3 (22.5–25.6)	25.6 (24.5–27.0)	.009
Sodium	140 (136–141)	139 (136–141)	140 (139–142)	.202
Potassium	4.1 (3.8–4.3)	4.1 (3.7–4.4)	4.1 (3.9–4.2)	.705
Hematocrit	40.1 (35.8–42.1)	40.3 (36.3–42.3)	40.0 (35.2–42.1)	.478
Bilirubin	0.50 (0.40–0.70)	0.50 (0.30–0.70)	0.50 (0.40–0.70)	.224
Platelet	231 (192–281)	235 (194–278)	231 (191–281)	.557
Creatinine	0.88 (0.75–1.15)	0.89 (0.76–1.20)	0.88 (0.74–1.14)	.934
WBC	8.5 (7.5–10.9)	9.0 (7.4–10.9)	8.4 (7.5–10.8)	.773
IL-6	19.6 (12.7–32.2)	19.6 (8.8–30.8)	23.2 (13.4–32.2)	.499
Glucose	135 (111–176)	132 (105–174)	138 (117–176)	.606
GFR	72 (48–87)	74 (54–90)	61 (48–81)	.352
Albumin	41.0 (36.0–43.0)	41.0 (36.3–43.0)	40.0 (35.0–43.0)	.673
CRP	7.7 (2.8–18.1)	9.2 (2.8–22.6)	5.6 (2.9–11.8)	.325
Procalcitonin	0.05 (0.04–0.1)	0.06 (0.04–0.10)	0.05 (0.04–0.09)	.540

Supplementary Table S2.

Sensitivity analyses for the time × nutritional route interaction.

Model	Covariate(s)	I-FABP: F(1,df), P, η2p	MMP-9: F(1,df), P, η2p
Primary	NIHSS	F(1,70)=4.22, P=.044, η2p=0.057	F(1,70)=4.76, P=.033, η2p=0.064
Sensitivity 1	APACHE II	F(1,70)=5.93, P=.017, η2p=0.078	F(1,70)=6.54, P=.013, η2p=0.085
Sensitivity 2	NIHSS + APACHE II	F(1,69)=5.67, P=.020, η2p=0.076	F(1,69)=6.47, P=.013, η2p=0.086
Sensitivity 3	NIHSS + Vasopressor-days	F(1,69)=4.19, P=.044, η2p=0.057	F(1,69)=5.25, P=.025, η2p=0.071
Sensitivity 4	NIHSS + Mechanical ventilation status	F(1,69)=4.72, P=.033, η2p=0.064	F(1,69)=4.47, P=.038, η2p=0.061

Results shown are for the time × nutritional route interaction term from two-way mixed repeated-measures ANCOVA (time: day 1 vs. day 7; between-subject factor: nutritional route). All models n=73. The Primary model uses baseline NIHSS score as the pre-specified covariate. Sensitivity 1 substitutes baseline APACHE II score for NIHSS. Sensitivity 2 includes both NIHSS and APACHE II simultaneously as covariates (variance inflation factor <1.2 for all terms, indicating no meaningful multicollinearity). Sensitivity 3 includes both NIHSS and cumulative vasopressor exposure (number of days, of 7, on vasopressor support) simultaneously as covariates (variance inflation factor=1.02 for both terms, indicating no meaningful multicollinearity). Sensitivity 4 includes both NIHSS and baseline mechanical ventilation status simultaneously as covariates. I-FABP, intestinal fatty acid-binding protein; MMP-9, matrix metalloproteinase-9; NIHSS, National Institutes of Health Stroke Scale; APACHE II, Acute Physiology and Chronic Health Evaluation II.

Funding Statement

The study was supported by the Scientific Research Projects Department of Ağrı İbrahim Çeçen University Project No: SABF.25.006.