Acute ischaemic stroke (AIS) is a major cause of death and long-term disability in the world and poses a significant public health problem.[1] The World Stroke Organization (WSO) reports that in 2021, there were an estimated 11.9 million new strokes, of which 7.8 million were ischaemic strokes, worldwide.[2] Approximately 7 million individuals died of stroke, and over 160 million disability-adjusted life years (DALYs) were also caused by stroke, which was more concentrated in low- and middle-income countries.[3] Although treatment has improved over the years with intravenous thrombolysis and endovascular thrombectomy, some patients may still not respond to these interventions.[4] A major concern is early neurological deterioration (END), with higher death rates, higher disability, and poor functional outcomes.[5] Therefore, early identification of patients with a high risk of neurological deterioration in the hospital setting is crucial for better clinical monitoring and treatment decisions.

The deterioration of neurological function from the onset of an acute ischaemic stroke is called early neurological deterioration.[5] Although its definition varies across studies, it is commonly identified by an increase in the National Institutes of Health Stroke Scale (NIHSS) score during the first several days after stroke onset.[6] A prospective multicentre cohort study of 9,114 patients with AIS found that 1,286 patients (14.1%) developed END in the first 7 days after their admission.[5] Interestingly, 62.5% of these deterioration events happened in the first 24 hours, highlighting the impact of early risk identification.[5] In addition, END at 90 days was significantly associated with poor functional outcomes, with an adjusted OR of 1.74 (95% CI: 1.53–1.97).[5] The results underscore the importance of clinically useful, widely available, and accurate predictors that are able to detect vulnerable patients in advance of significant neurological decline.

The stroke Severity Scale (NIHSS) is a commonly used clinical tool used to evaluate neurological deficits and determine the severity of stroke.[7] It assesses various neurological functions: consciousness, motor activity, sensory deficits, language, visual field, and coordination.[8] Baseline NIHSS scores tend to be higher in patients with more severe neurological deficits and are linked to worse clinical outcomes.[9] There is some evidence that baseline NIHSS is an independent predictor of neurological deterioration after AIS, thereby supporting the clinical utility of its use in predicting the occurrence of END.[10] 

Neutrophil to lymphocyte ratio (NLR) is an emerging and cost-effective systemic inflammation marker defined as the ratio of the absolute neutrophil count to the absolute lymphocyte count.[11] Inflammatory responses after cerebral ischaemia lead to endothelial dysfunction, disruption of the blood–brain barrier, oxidative stress and secondary neuronal injury.[12] Neutrophil activity in the inflammatory response can be mediated by the production of proteolytic enzymes, inflammatory mediators, and reactive oxygen species, and a decrease in lymphocytes may be a consequence of immune system changes that occur after stroke.[13] Hence, a high NLR could reflect a state of immune activation versus immune regulation.[13]

A combined assessment of these two parameters could be useful in differentiating between the effects of primary neurologic injury and systemic inflammatory activity in the pathogenesis of END. Therefore, this study aimed to compare the predictive value of baseline NIHSS and NLR for early neurological deterioration among patients with acute ischaemic stroke. In particular, the study compared and contrasted their discriminatory accuracy in detecting patients who develop END, evaluated each parameter's association with neurological deterioration, and investigated whether there are any benefits of combining the two parameters. The results could help develop a clinically accessible, cost-effective, and practical strategy for early risk stratification, facilitating early neurological surveillance and potentially improving outcomes in AIS patients.

A prospective observational cohort study was conducted in the Department of Neurology at Central Park Teaching Hospital, Lahore, Pakistan. The hospital was adequate to recruit patients with acute ischaemic stroke and to evaluate neurological status and inflammatory markers. The study was carried out during 6 months between 1st November 2025 and 30th April 2026. Sample size was determined using the online OpenEpi sample size calculator for a single population proportion. The expected frequency of early neurological deterioration was considered to be 14.1%, based on a prospective multicentre cohort study by Liu et al. (2023), which reported END in 1,286 (14.1%) out of 9,114 patients with acute ischaemic stroke.[5] The calculation was performed at the 95% confidence level with an absolute precision of 5%. The minimum calculated sample size was approximately 187 patients. After adding 10% to account for incomplete data and potential loss to follow-up, the final estimated sample size was 208 patients. The sampling technique employed was non-probability consecutive sampling. Patients aged 18 years and above, irrespective of gender, were included if they had a confirmed diagnosis of acute ischaemic stroke based on clinical assessment and brain imaging through computed tomography (CT) or magnetic resonance imaging (MRI). Eligible patients were those who presented within 24 hours of symptom onset, were admitted to Central Park Teaching Hospital during the study period, and had baseline complete blood count parameters, which included absolute neutrophil and absolute lymphocyte counts, available. Informed consent, which included written informed consent, was obtained from patients/legally authorised representatives before enrolment. Those who presented with a haemorrhagic stroke, TIA or other non-ischaemic neurological condition were excluded. Active infections, septicemia, systemic inflammatory diseases, haematological disorders, active malignancies and diseases that affect neutrophil and lymphocyte counts were also excluded. Further, chemotherapy, immunosuppressive therapy or systemic corticosteroids were not included. Patients whose baseline laboratory tests were not performed, had inadequate clinical data, or who were transferred to another health care facility prior to completion of the seven-day follow-up were excluded. Eligible patients were recruited from the Department of Neurology of Central Park Teaching Hospital after obtaining approval from the institutional ethics review committee. Informed consent was obtained in writing from patients or their legally authorized representatives. Data were collected using a structured proforma designed specifically for the study. Demographic factors such as age and gender, as well as clinical factors, such as presentation, duration of symptoms prior to admission to hospital, blood pressure and medical history, were recorded. Other common established vascular risk factors, such as hypertension, DM, dyslipidaemia, smoking, past stroke, and AF, were also recorded. The National Institutes of Health Stroke Scale (NIHSS) was used to evaluate baseline neurological impairment at admission.[5] The scale evaluated different neurological domains, including consciousness, gaze, visual fields, facial palsy, motor function, sensory function, language, dysarthria and neglect. Neurological impairment was more severe in patients with a higher NIHSS score. Neurological evaluations were repeated at regular intervals (usually every 24 hours) for the 7-day period of observation. Assessments were performed by trained healthcare professionals using the standardised NIHSS protocol. Complete blood count was performed on blood samples collected within the first 24 hours of admission to hospital. Absolute counts of neutrophils and lymphocytes were determined from the laboratory reports, and the neutrophil-to-lymphocyte ratio was calculated as absolute neutrophil count (ANC) divided by absolute lymphocyte count (ALC). The baseline inflammatory factor was recorded as the NLR. Where available, other laboratory parameters measured were also recorded, such as haemoglobin, total white blood cell count, platelet count, and blood glucose. Reperfusion treatments (intravenous and mechanical thrombolysis) were documented as present, if applicable. The main endpoint was early neurological deterioration (END) within 7 days of admission. Adopted from Liu et al. (2023), the definition of END was defined as a change of at least 2 points on the total NIHSS score or at least 1 point on the NIHSS motor items compared with the baseline assessment.[5] The patients were then divided into two groups: patients with the development of END and patients who did not develop any neurological deterioration during the observation period. The baseline NIHSS and NLR values were analyzed between the two groups to assess the predictive value individually and combined. The collected data were analyzed using IBM Statistical Package for the Social Sciences (SPSS), version 27.0. Continuous variables were expressed as mean ± standard deviation for normally distributed data or median with interquartile range for non-normally distributed data. Frequencies and percentages were used to present categorical variables. Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed continuous data were compared between the END and non-END groups using an independent t-test, and non-normally distributed data were compared using the Mann–Whitney U test. The chi-square test and Fisher's exact test were used to compare categorical variables, depending on the sample size. Baseline NIHSS and NLR were analyzed for their predictive ability to END using an ROC curve. Both parameters were estimated by the area under the curve (AUC) and 95% confidence intervals (CIs). Differences in discriminating performance between them were tested by DeLong's test, and their AUC values were compared. The best cut-off points were determined by the Youden index, and sensitivity, specificity, positive and negative predictive values were calculated. Independent associations of NIHSS and NLR with END were determined by binary logistic regression analysis. The analysis was adjusted for age, gender, hypertension, diabetes mellitus, atrial fibrillation, and reperfusion therapy. Odds ratios (ORs) and their 95% CIs were reported. A two-tailed p-value of less than 0.05 was considered statistically significant.

A total of 208 patients with acute ischaemic stroke were included in the study. Of these, 29 (13.9%) had neurological deterioration within 7 days of admission, while 179 (86.1%) did not. The mean age of the participants was 62.4 ± 12.1 years, and 58.2% were male. The patients with END were significantly older than patients without END (p = 0.006). Baseline demographic and clinical parameters such as sex, hypertension, diabetes mellitus, dyslipidaemia, and smoking status did not significantly differ between the two groups (Table 1).

Neurological and laboratory parameters revealed that patients with END had significantly higher NIHSS at baseline than patients without END (p < 0.001). Similarly, NLR was significantly elevated in the END group (p < 0.001). Higher neutrophil counts and lower lymphocyte counts were also observed among patients experiencing END. Admission blood glucose and total leukocyte counts were significantly higher in the END group (Table 2).

Of the 29 patients who developed END, 18 (62.1%) deteriorated within 24 hours, 8 (27.6%) deteriorated within 24-72 hours, and 3 (10.3%) deteriorated from 72 hours to 7 days. Among patients with END, the median change in NIHSS score was +3 (IQR: 2 to +5). Eight patients died during the seven days of observation (3.8% of total patients) (Table 3).

ROC curve analysis demonstrated that baseline NIHSS had better discriminatory performance for predicting END than NLR, with AUCs of 0.82 (95% CI: 0.74–0.90) and 0.70 (95% CI: 0.60–0.80), respectively. The best cut-off values were ≥8 for NIHSS and ≥4.2 for NLR. The model that included the NIHSS and NLR had the highest AUC of 0.86 (95% CI: 0.79-0.93). The improvement over NIHSS alone was not significant by DeLong's test (p = 0.18), but the improvement over NLR alone was significant by DeLong's test (p = 0.002) (Table 4).

In multivariable logistic regression analysis, baseline NIHSS and NLR remained independently associated with END after adjustment for selected clinical and demographic variables. Severe stroke increased by 19% for every one-point rise in NIHSS (p < 0.001). Likewise, a one-unit rise in NLR was associated with a 13% increase in the odds of END (p = 0.034). Age was also independently associated with END, whereas the other adjusted variables did not demonstrate statistically significant associations (Table 5).

The combined NIHSS and NLR model showed the largest apparent discriminatory performance (AUC: 0.86), followed by NIHSS only (AUC: 0.82) and NLR only (AUC: 0.70). The combined model also demonstrated the lowest Brier score (0.098) and Akaike information criterion (169.8), suggesting better apparent predictive performance and model fit. However, the AUC difference between the combined model and NIHSS alone was not statistically significant, which would suggest that NLR provided additional predictive information beyond the NIHSS alone, but that this predictive value was not definitively established (Table 6).

Table 1. Sociodemographic and baseline clinical characteristics according to END status

Table 2. Baseline neurological and laboratory parameters according to END status

Table 3. Characteristics and timing of early neurological deterioration

Table 4. ROC curve analysis of NIHSS, NLR, and their combined model for predicting END

Table 5. Multivariable logistic regression analysis of factors associated with END

Table 6. Comparative performance of individual and combined predictive models

The present study compared the predictive value of the National Institutes of Health Stroke Scale (NIHSS) and neutrophil-to-lymphocyte ratio (NLR) for early neurological deterioration (END) among patients with acute ischaemic stroke. The findings demonstrated that 13.9% of patients developed END within seven days of admission. Baseline NIHSS scores and NLR values were significantly higher among patients who experienced neurological deterioration than among those who did not. In addition, the individual discriminatory performance of the NIHSS was greater than that of the NLR, and the combined model had the best apparent area under the receiver operating characteristic curve (AUC). The results indicate that neurological severity and systemic inflammatory activity both may play a role in the early clinical course of ischaemic stroke, but that both have some add-on and relative contribution to the prediction of stroke should be assessed. The observed END frequency of 13.9% was comparable to the findings of Liu et al. (2023), who reported END in 14.1% of 9,114 patients with AIS in a prospective multicentre cohort study.[5] Their investigation found that baseline NIHSS, neutrophil count, and lymphocyte count are independent predictors of neurological deterioration. About 62.5% of the deterioration events also took place during the first 24 hours. Likewise, most of the data reported in the present case of END events were in the first 24 hours (62.1%). This agreement underlies the significance of early neurological surveillance, but it is important to keep in mind that the incidence rates vary due to population differences and to the different definitions of END. The higher baseline NIHSS scores among patients who developed END were consistent with the findings of Liu et al. (2023), who identified baseline NIHSS as an independent predictor of neurological deterioration.[8] In the same study, Chen et al. (2024) analysed 303 patients with AIS and found NIHSS to be an independent predictor of poor 90-day functional outcomes. They showed that their predictive model had an improved AUC from 0.694 to 0.826 when they added NIHSS to the clinical model.[14] The findings suggest that a baseline neurological assessment is clinically significant in determining patients with poor outcomes. The present comparison, however, was directed specifically at END within 7 days and not functional disability. The association between elevated NLR and END observed in the present study was also supported by previous investigations. In 2021, Gong et al. investigated 1060 patients with IVT and found that the NLR was found to be independently associated with post-thrombolysis END with an AUC of 0.763.[15] Zhao et al. (2024) also conducted a study in 228 patients with anterior circulation stroke and observed significantly higher NLR in patients with END than in patients without END (3.8 vs 2.4, p < 0.001). Their multivariable analysis included the NLR and revealed that an NLR > 2.65 was an independent predictor of END.[16] This is consistent with the increased NLR seen in patients with neurological deterioration in the current study, although the varying definitions for END and the length of observation periods could account for some of the differences in predictive ability. The prognostic power of NLR was further supported by Xu et al (2023), who performed a prospective observational study in 341 patients with ischaemic stroke. They reported stroke progression in 56 patients (16.4%) and identified NLR as an independent factor associated with stroke progression and poor functional outcomes. Their ROC analysis showed an AUC of 0.612 predicting poor functional outcome.[17] Furthermore, Fakhari et al. (2024) conducted a prospective cohort analysis of 217 patients and demonstrated that neutrophil count and NLR were significantly correlated to functional outcomes and varied across stroke subtypes.[18] These observations correlated well with the present results and indicated that inflammatory cell changes could offer further data regarding the clinical course of AIS. However, the relatively small discriminatory performance reported in some studies suggests not using NLR as a stand-alone predictor. Zhao et al. (2022) investigated 102 patients who received IVT and found that NLR was also associated with stroke outcomes. They discovered that the NLR measured at 24 hours was independently associated with early neurological improvement, and NLR measured at 48 hours was associated with favourable functional outcome at discharge.[19] Likewise, Majid et al. (2021) found in a retrospective study of 98 patients undergoing thrombolysis that patients with an NLR > 2.39 had a worse functional outcome at 3 months (OR: 2.7, 95% CI: 1.11–6.39).[20] The results of these studies indicate a potential role for inflammatory markers as a prognostic marker after acute ischaemic stroke, although the results varied with respect to the outcome variables and measurements of NLR. The results were also similar to those of the systematic review done by Sarejloo et al. (2022) and Sharma et al. (2022). In a study by Sarejloo et al., they observed that higher NLR correlated with greater risk of neurological deterioration.[21] Sharma et al. reviewed 35 studies involving 10,308 patients receiving reperfusion therapy and found that higher admission NLR was associated with unfavourable functional outcomes, mortality and haemorrhagic complications.[22] These reviews offer additional evidence to support the link between systemic inflammation and poor stroke outcomes. However, their findings also highlight that NLR performance may vary according to treatment modality, timing of blood sampling, and outcome definition. A key point of this study was to compare NIHSS and NLR instead of analyzing them separately. The combined model showed a higher AUC; NIHSS had better discriminatory power than NLR. But this difference between the combined model and NIHSS alone was not statistically significant. Similar findings were obtained by Chen et al. (2024), who found that the AUC rose by just 0.003 to 0.829 when NLR was added to a model that already included NIHSS.[14] This implies that while NLR can give complementary data, it can only give a relatively small amount of information beyond what is already known from the clinical information. Therefore, the results are not proof of the superiority of the combination of the two parameters over NIHSS alone. Clinically, the results indicate that baseline NIHSS and NLR can be used as supplementary parameters during the initial clinical evaluation of patients with AIS. NIHSS, which assesses neurologic deficit, as opposed to NLR, which reflects the systemic inflammatory response. The calculation of NLR is possible without specific laboratory tests, using common laboratory procedures on a complete blood count, and it can provide other clinical information. However, the present results are to be interpreted with caution. The sample included only 208 patients, with 29 END events, which limited the number of variables that could be reliably included in multivariable regression analysis. Moreover, findings may not be generalizable to other settings due to the one-centre design. All of the results need to be confirmed with real patient information and an independent patient population before clinical implications can be drawn. Overall, the results were consistent with previous studies which showed a link between neurological severity, systemic inflammation, and adverse outcomes after acute ischaemic stroke. The combination of NIHSS and NLR showed a slight improvement, though not significant, compared with NIHSS alone; the individual predictive performance of NIHSS was stronger than that of NLR. Thus, NLR could be used in addition to the conventional risk factors in early stroke risk assessment, but larger, well-designed, multicentre, prospective studies are necessary to confirm its added predictive value.

The baseline NIHSS and NLR were predictive of early neurological deterioration after acute ischaemic stroke. NIHSS had better individual predictive performance than NLR, and their combination had the best discriminatory ability. However, the improvement achieved by combining both parameters over NIHSS alone was not statistically significant. The results suggest that NLR may be a useful complement to the assessment of stroke in the early phase, but further studies are needed to see if NLR offers any further predictive value. Future studies are needed to validate these results and confirm the clinical utility of combining inflammatory biomarkers with traditional neurological assessment, using larger, multicentre, prospective studies.