Acute ischemic stroke (AIS) is a major global threat to human health and has emerged as the leading cause of death and disability among residents in both urban and rural areas of China (1). Data from 2020 indicated that the prevalence of stroke in Chinese adults aged 40 years and older reached 2.6% (2). Around 40% of AIS cases are caused by large vessel occlusion (LVO), and these patients often experience severe cerebral infarction, associated with high rates of disability and mortality (3). In recent years, endovascular therapy (EVT) has become the cornerstone of treatment for LVO patients, significantly improving their prognostic outcomes (4). However, patients with LVO typically present with large infarct volumes, severe paralysis, and a high risk of complications, such as pneumonia (5). These factors pose challenges to clinical management and emphasize the critical importance of timely treatment of complications for ensuring favorable AIS patient prognosis.

Stroke-associated pneumonia (SAP) is a common complication of AIS, occurring in 5%−26% of patients and most frequently within the first 48 h after stroke onset (6, 7). SAP not only significantly prolongs hospitalization and increases healthcare costs but also contributing to higher short- and long-term mortality among AIS patients (8, 9). Studies indicate that the incidence of SAP is notably higher in patients with LVO undergoing EVT than in those not receiving EVT (5). Therefore, identifying risk factors for SAP in LVO patients undergoing EVT is crucial for developing effective prevention and treatment strategies. Although various early clinical prediction models for SAP have been proposed, most rely on scale-based assessments, which are inherently limited by rater experience and patient cooperation, NIHSS score in particular is susceptible to these factors (10, 11), and its accuracy may be further compromised in patients with severe strokes such as LVO, who typically present with NIHSS scores >10 (12). These limitations underscore the need for objective and robust biomarkers to enhance SAP risk prediction in this patient population.

Several risk factors for SAP have been identified in existing studies, including advanced age, atrial fibrillation, dysphagia, stroke severity, and immunosuppression (13). Of these factors, immunosuppression has been identified as an independent risk factor for SAP (14). After AIS, the release of immune-inflammatory mediators protects brain tissue by suppressing excessive inflammation. However, this response concurrently induces a state of immunosuppression, known as stroke-induced immunodepression syndrome (SIDS), which is characterized by reduced lymphocyte counts, impaired T helper cell function, and dysregulated monocyte activation. These changes can emerge within 12–24 h post-stroke, predisposing AIS patients to an increased risk of SAP (15, 16). Notably, the use of prophylactic antibiotics has not demonstrated benefits in reducing the incidence of SAP, improving long-term functional outcomes, or decreasing mortality (17). As a result, recent researches has shifted focus toward evaluating and modulating the immunosuppressed state as a novel SAP prevention strategy (18). Thus, the early identification of high-risk SAP patients via inflammatory and immune biomarkers, followed by timely immunomodulatory or targeted antibiotic therapy, may therefore offers a promising approach to improving the prognosis of AIS patients (19).

Although multiple inflammatory and immune factors have been demonstrated to be closely associated with the development of SAP, no single factor or combination of factors has been fully translated into clinically practical predictors (15). Previous studies have shown that plasma levels of interleukin-6 (IL-6), C-reactive protein (CRP), and procalcitonin (PCT) are significantly elevated in patients with SAP during the early phase of stroke, suggesting their potential utility as biomarkers for predicting SAP (15, 20). The monocyte system is critical for immune regulation in the context of AIS development, and perturbations in its function may act as potential biomarkers to reflect altered immune status in patients with AIS (21). In the monocyte lineage, the expression of human leukocyte antigen-DR (mHLA-DR) on the surface of monocytes plays a central role in the immune response by presenting both exogenous and endogenous antigenic peptides to CD4⁺ T cells. Thus, mHLA-DR is widely used as a key marker to evaluate monocyte activity (22). Previous evidence further supports that decreased mHLA-DR expression is a significant predictor of SAP, with stroke patients exhibiting low mHLA-DR levels facing a substantially increased risk of pneumonia (14, 23). However, it remains unclear whether these biomarkers can effectively predict the risk of SAP in LVO patients undergoing EVT and warrants further investigation.

Therefore, this study employs a prospective design to examine the dynamic changes in mHLA-DR expression and plasma levels of IL-6, CRP, and PCT in patients with acute anterior circulation large vessel occlusion (ACLVO) following EVT, further seeks to assess the potential predictive value of these markers for SAP, with the objective of contributing to earlier identification of high-risk patients.

Patients with ACLVO who underwent EVT were enrolled in this prospective cohort study, and several key findings were observed. First, patients who developed SAP exhibited persistently lower levels of mHLA-DR expression throughout the follow-up period compared to patients who did not develop SAP. In addition, plasma levels of IL-6 and CRP were significantly elevated on days 3 and 7 after stroke onset, while PCT levels also increased markedly by day 7. Second, imaging analyses revealed that patients with SAP had significantly larger cerebral infarct volumes than those without SAP. Finally, in terms of predictive performance, dysphagia and mHLA-DR expression level were identified as independent predictors of SAP in ACLVO patients receiving EVT. Moreover, DCA analysis demonstrated that, as a single predictor, mHLA-DR achieved greater net benefit than other clinical prediction models for SAP. Moreover, mHLA-DR combination with the A2DS2 score demonstrated greater net benefit and stability.

Currently, mHLA-DR expression levels are widely used to predict the risk of infection following AIS (15). Previous studies have shown that reduced mHLA-DR expression on day 1 after AIS onset is an independent predictor of SAP (14). In this study, focusing on ACLVO patients undergoing EVT, we observed that mHLA-DR expression levels were significantly lower in SAP patients compared to non-SAP patients starting from the first postoperative day. Multivariable binary logistic regression analysis further confirmed mHLA-DR expression as an independent predictor of SAP. ROC curve analysis showed that an mHLA-DR expression level below 55.75% on the first day after SAP onset had good predictive value for SAP. Decreased mHLA-DR expression in AIS patients has been reported to predict SAP and can occur as early as 3 h after onset, often preceding functional changes in monocytes, thus serving as an early indicator of monocyte dysfunction (14, 23, 28, 29). These findings are consistent with the results of our study. In summary, for ACLVO patients treated with EVT, early reduction in mHLA-DR expression is a reliable predictor of SAP.

The persistence of low mHLA-DR expression levels suggests that immunosuppression is being maintained (30). In this study, the dynamic monitoring of mHLA-DR levels over a period of 7 days in patients with ACLVO who were receiving EVT revealed that those with SAP had consistently lower expression levels throughout the follow-up period compared to those without SAP. ROC curve analysis showed that mHLA-DR < 55.15% on day 3 and < 62.90% on day 7 after onset had a good predictive value for the development of SAP. Furthermore, patients in the SAP group exhibited significantly larger cerebral infarct volumes than those in the non-SAP group. Previous studies have similarly reported persistently low mHLA-DR expression in AIS patients who develop SAP, with greater infarct volumes correlating with more pronounced mHLA-DR reduction (23, 31, 32). The decrease in mHLA-DR expression levels in peripheral blood of SAP patients can last for 1–2 weeks, and the degree of decrease is closely related to the prognosis of patients with acute stroke (23, 31, 32). These findings suggests that monocyte dysfunction is one of the key factors predisposing to secondary infection in AIS patients, while the expansion of infarct size may further aggravate immune dysfunction (32). Continuous immunosuppression is prevalent in the acute phase of stroke, and this pathophysiological feature may be an important cause of increased susceptibility to SAP, also indicating that the immune function status of patients is closely related to the occurrence of SAP (14). Therefore, targeted intervention for the immunosuppressive mechanism in the acute phase of stroke may provide new strategies for the prevention and treatment of SAP and inform future drug development (33).

In this study, the AUC of mHLA-DR for predicting SAP reached 0.914, demonstrating strong discriminatory ability. However, such high predictive accuracy is uncommon in real-world clinical research and warrants cautious interpretation. This outcome may be attributable to factors including the single-center design, high homogeneity of the study population, stringent inclusion and exclusion criteria, as well as standardized detection methods and outcome definitions. It should also be noted that the limited sample size increases the risk of model overfitting, and the absence of external validation further restricts the generalizability of the findings (34, 35). Therefore, these results should be regarded as a promising preliminary signal rather than evidence supporting immediate clinical application. Future validation through large-scale, multicenter, and external cohorts is necessary to confirm the true predictive utility of mHLA-DR.

Previous studies have indicated that IL-6 is a strong predictor of SAP in stroke patients, it may help assess infection risk early in the disease course—within hours—with levels correlating with patient mortality (20, 36). In addition, plasma CRP and PCT are also considered potential predictive markers of SAP and significant increases in their levels are seen within 24–48 h after stroke (37). However, the present study found no statistically significant differences in plasma levels of IL-6, CRP, or PCT between SAP and non-SAP patients in the early phase. These biomarkers did not show significant divergence until day 3 or day 7. In our interpretation, this finding does not diminish the value of these biomarkers but may instead reflect a distinct post-stroke immune-inflammatory response profile in the setting of EVT.

First, the median time to SAP development in this cohort was 2.36 days, whereas systemic inflammatory responses after AIS generally peak between days 3 and 5 (33). Early blood sampling, such as within 24 h post-EVT, may therefore have captured a pre-inflammatory state before SAP became clinically evident. The differences observed after day 3 likely reflect a pronounced inflammatory response triggered by SAP itself (14), suggesting that the predictive time window for traditional inflammatory markers may need to be redefined for patients with ACLVO treated with EVT.

Second, EVT may modulate early inflammation through recanalization-related immunomodulation. Successful recanalization is known to attenuate systemic inflammation by reducing brain tissue necrosis, thereby improving outcomes (38, 39). In this study, all ACLVO patients underwent EVT, and those who developed SAP had significantly larger infarct volumes than non-SAP patients, supporting the concept that greater infarct burden exacerbates immunosuppression, and SAP risk (32). These findings suggest that inflammatory and immune parameters in both groups may be influenced by EVT. Successful recanalization likely limits infarct size, partially mitigating post-stroke immunosuppression, and systemic inflammation (40).

Finally, reverse causality must be considered. More severe baseline infarct core or neurological deficits may directly increase SAP risk and subsequent inflammatory responses, rather than early markers driving SAP development—a perspective consistent with the present findings. Therefore, the limited early predictive performance of IL-6, CRP, and PCT may partly stem from their collinearity with infarct volume as the underlying etiology (41). After EVT rapidly alters disease progression, the independent predictive utility of these markers may be diminished. In summary, for EVT-treated ACLVO patients, IL-6, CRP, and PCT appear to have limited value in the early prediction of SAP, and may be more suitable as indicators for monitoring inflammation after SAP onset. Further research is needed to clarify the dynamic evolution and mechanisms of peripheral inflammatory markers before and after EVT in AIS patients.

Our findings indicate that patients with SAP had significantly higher A2DS2, ISAN, and NIHSS scores than those without SAP, and consistently higher NIHSS scores during the first 7 days post-stroke, consistent with prior studies (42). It is important to note that scale assessment is susceptible to assessor experience and patient cooperation and is particularly evident in patients with severe stroke (10, 11). We therefore used DCA to compare the predictive utility of the objective immune marker mHLA-DR with that of clinical scales. DCA revealed that mHLA-DR alone offered a greater net benefit than any single scale, indicating its potential to reduce subjective assessment bias as an objective biomarker. This is consistent with previous evidence that blood-based inflammatory markers are more accurate than clinical scales in predicting SAP (43). Moreover, the model combining mHLA-DR and the A2DS2 score yielded the highest net clinical benefit and was the most stable across risk thresholds. These findings indicate that integrating objective immune markers with clinical scales enhances prediction robustness and clinical utility for SAP (36). However, this single-center study had a limited sample size, and the clinical applicability of the combined model requires further validation in larger, multicenter prospective cohorts.

This study confirms that the combined mHLA-DR-based model demonstrates good predictive performance and clinical net benefit for SAP. However, its clinical translation requires consideration of practical implementation factors. Currently, mHLA-DR testing by flow cytometry is available in most tertiary hospitals in China and has a turnaround time of 2–3 h, comparable to conventional inflammatory markers and suitable for early stroke risk stratification (44). Although mHLA-DR testing costs more than CRP or PCT (approximately ¥200–300 per sample), targeted prevention guided by mHLA-DR risk stratification, such as early antibiotics or enhanced airway care, can yield substantial savings, as one SAP episode increases healthcare costs by roughly ¥6000–19000 (45). In addition, mHLA-DR specifically reflects stroke-induced immunosuppression rather than systemic nonspecific inflammation (22), offering a clearer pathophysiological basis for infection prediction and demonstrating good clinical applicability.

This study has several limitations. (1) The primary limitations of this study include its single-center design, limited sample size, and absence of external validation. These factors may restrict the model's generalizability, and external validation through multicenter studies is needed before clinical application. (2) The elevation of inflammatory-immune markers observed on day 1 in patients who were later diagnosed with SAP suggest that SAP may be initiated within the first day after stroke onset. This implies that SAP may be a trigger for immunosuppression rather than merely a consequence of it. Future studies should include ultra-early immune marker assessments within hours of stroke onset. (3) This study collected blood samples only after EVT, lacking preoperative baselines. Consequently, it remains unclear whether biomarker changes resulted from the stroke itself or the EVT procedure. Future research should systematically collect samples before, immediately after, and at multiple time points following EVT to better assess inflammatory-immune dynamics and their association with SAP.

In summary, this preliminary exploratory study indicates that early monitoring of mHLA-DR levels following EVT may help predict post-stroke pneumonia, particularly when integrated with the A2DS2 model. Changes in mHLA-DR occurred earlier than those in inflammatory markers such as IL-6, CRP, and PCT, suggesting its potential role as an early warning indicator. However, given the small sample size, single-center design, lack of external validation, risk of overfitting, and undefined cut-off values, these results remain hypothesis-generating and are not yet suitable for clinical use. Future multicenter, large-sample, prospective external validation studies are needed to confirm the predictive utility of mHLA-DR and to establish clinically robust, generalizable cut-off criteria, and monitoring strategies.