Stroke remains a leading cause of mortality and disability worldwide, with its incidence rising rapidly in low- and middle-income countries, posing a significant global health challenge (Prust et al., 2024). Ischemic stroke (IS) accounts for approximately 65.3% of all stroke events (GBD 2021 Nervous System Disorders Collaborators, 2024). Between 1990 and 2019, the absolute number of stroke cases increased by 70%, prevalence by 85%, and stroke-related deaths by 43% (GBD 2019 Stroke Collaborators, 2021). According to the Global Burden of Disease Study, stroke ranked as the leading cause of disability-adjusted life years (DALYs) globally from 2010 to 2021, and remains the foremost contributor in China (GBD 2021 Diseases and Injuries Collaborators, 2024).

Post-stroke cognitive impairment (PSCI) is one of the most common and disabling sequelae of stroke, affecting more than 70% of survivors (Rost et al., 2022). Approximately 38% of patients develop cognitive deficits within the first year after stroke (Sexton et al., 2019). Although some cases experience early recovery, up to one-third of stroke survivors progress to dementia within 5 years (El Husseini et al., 2023). In 2024, the projected cost of long-term care for individuals aged 65 years and older with cognitive impairment or dementia is estimated to exceed $360 billion in the United States alone (Alzheimer’s Association Report, 2024). These alarming statistics highlight the urgent need for effective secondary prevention strategies to preserve cognitive function in stroke survivors.

DL-3-n-butylphthalide (NBP), a compound derived from the seeds of Apium graveolens, has shown great potential in the treatment of IS (Tan et al., 2023). Its neuroprotective mechanisms include reducing oxidative stress, suppressing neuroinflammation, inhibiting apoptosis-related pathways, and promoting angiogenesis (Chen et al., 2020; Dai et al., 2023; Ge et al., 2024). Since its approval as an anti-ischemic agent in China in 2002, growing evidence has supported the role of NBP in improving cognitive function after stroke. A meta-analysis confirmed its beneficial effects on cognitive performance in patients with PSCI (Wang et al., 2024), and both short-term and long-term studies have demonstrated its efficacy in enhancing cognitive outcomes post-IS (Han et al., 2025; Wang et al., 2024; Yan et al., 2017).

Diabetes mellitus is a well-established risk factor for PSCI, independent of stroke severity and age (Filler et al., 2024). Elevated blood glucose levels in diabetes can lead to increased oxidative stress, which is known to disrupt the blood–brain barrier (BBB) and impair synaptic function in the hippocampus, a key region for memory and learning (Rom et al., 2019). This oxidative stress can also exacerbate neuroinflammation, further contributing to cognitive deficits. Additionally, diabetes-induced insulin resistance in the brain can disrupt insulin signaling pathways, which are crucial for maintaining neuronal health and synaptic plasticity (Ansari et al., 2023; Khalid et al., 2022). Notably, NBP may mitigate diabetes-related cognitive decline through mechanisms such as inhibiting caspase-3-mediated apoptosis and enhancing neurovascular protection (Tan et al., 2022). Preclinical studies have also demonstrated its neuroprotective effects in diabetic models (Tian et al., 2017). Accumulating evidence indicates that chronic hyperglycaemia undermines the pro-angiogenic and mitochondrial protective effects of NBP by intensifying oxidative stress, promoting advanced glycation end-product (AGE) deposition, and thickening the microvascular basement membrane (Khalid et al., 2022). Concomitantly, diabetes-induced injury within the fronto-striatal circuitry and parahippocampal region occurs early and is already extensive at the sub-acute stage, thereby narrowing the therapeutic window and attenuating the efficacy of NBP in diabetic patients (Moran et al., 2013). Furthermore, recent developments in nanotechnology-based delivery systems, such as solid lipid nanoparticles (SLNs), have shown promise in enhancing the brain targeting and bioavailability of neuroprotective agents like NBP (Bukke et al., 2024; Han et al., 2024). These novel formulations may help overcome pharmacokinetic challenges, especially in diabetic individuals who often exhibit altered drug metabolism and distribution.

Despite growing evidence supporting the use of NBP in IS and PSCI, a significant gap remains in the literature: no studies have conducted subgroup analyses based on diabetes status to elucidate the therapeutic effects of NBP in diabetic versus non-diabetic patients. Moreover, most existing research has been conducted in urban settings, with limited data available on rural populations, who often bear a higher disease burden and face greater barriers to post-stroke care. Diabetes mellitus was considered as a potential effect modifier in our study based on prior preclinical studies. Given these established pathophysiological links between diabetes and cognitive impairment, as well as the potential interactions with NBP mechanisms (e.g., inflammation, insulin resistance), we prioritized diabetes for subgroup analysis. This hypothesis-driven approach aims to elucidate whether glycemic status influences the efficacy of NBP in preventing PSCI.

Therefore, this study aimed to evaluate the differential effects of 12-month NBP therapy on cognitive function in diabetic and non-diabetic IS patients in a rural Chinese community. The findings will provide valuable evidence for optimizing NBP therapy and developing individualized strategies for the secondary prevention of cognitive impairment in stroke survivors.

The primary objective of this study was to investigate the differential effects of NBP on cognitive impairment in IS patients with and without diabetes, with the goal of better evaluating its therapeutic potential in diverse patient populations. To our knowledge, this is the first study to report the long-term (12-month) effects of NBP treatment on cognitive outcomes in IS patients stratified by glycemic status. Our findings suggest that NBP is more effective in preventing cognitive decline in non-diabetic patients compared to those with diabetes. Furthermore, the cognitive domains influenced by NBP appear to differ between the two groups: NBP primarily preserved orientation in diabetic patients, whereas it was beneficial to language function in non-diabetic patients. These results highlight NBP’s promise as a targeted neuroprotective therapy and suggest that its mechanisms of action may vary depending on metabolic context.

NBP confers neuroprotection by upregulating vascular endothelial growth factor (VEGF)/CD31 to enhance angiogenesis and by dampening oxidative/nitrosative stress and pro-inflammatory pathways, thereby reducing infarct volume, improving neurological function, cerebral perfusion and mitochondrial fusion, and optimizing stroke outcome (Chen et al., 2020; Dai et al., 2023; Zhu et al., 2024). Meta-analyses confirm its efficacy, alone or combined, in ameliorating post-stroke cognition, with clinical trials showing benefits within 1 month that persist for 12 months (Fan et al., 2021; Han et al., 2025; Wang et al., 2024; Yan et al., 2017).

Our findings are consistent with this prior evidence, showing a beneficial effect of NBP on ΔMMSE regardless of diabetic status. However, its ability to prevent binary-defined cognitive decline was more pronounced in non-diabetic patients. This discrepancy may reflect the greater vulnerability of diabetic patients to chronic metabolic injury, which may blunt the cognitive benefits of NBP over the 12-month treatment period. Diabetes-induced hyperglycemia impairs hippocampal function by disrupting insulin signaling pathways, leading to synaptic dysfunction and memory deficits (Ansari et al., 2023). Additionally, the accumulation of AGEs triggers oxidative stress and inflammation, further damaging neuronal structures and exacerbating cognitive decline (Khalid et al., 2022). Neuroinflammation also compromises the integrity of the BBB, creating a vicious cycle that amplifies memory loss (Rom et al., 2019). Microvascular damage, another common complication of diabetes, can reduce cerebral blood flow and oxygen delivery to the brain, further impairing cognitive function (De Silva and Faraci, 2016). These mechanisms may explain why NBP’s efficacy in preventing cognitive decline was attenuated in diabetic patients compared to non-diabetic patients in our study. While our study administered NBP for 12 months, it is possible that a longer treatment duration is needed to observe more significant cognitive benefits in diabetic patients. Notably, although we observed an improvement in MMSE score changes, the magnitude of the difference fell short of the established minimal clinically important difference, indicating only a therapeutic trend rather than confirmed clinical benefit. Importantly, on the binary outcome of cognitive decline, we detected a protective effect of NBP exclusively in non-diabetic participants, suggesting that individuals with fewer baseline metabolic disturbances derived greater benefit. These findings underscore the need for longer-term follow-up to validate the robustness of the observed protection and to determine whether the early functional gains translate into a sustained reduction in dementia risk. While our study provides valuable insights into the differential effects of NBP on cognitive outcomes in diabetic and non-diabetic ischemic stroke patients, it is important to note that detailed glycemic control data were not specifically collected during the 12-month follow-up period. However, our study design included quarterly follow-ups (every 3 months), during which patients received guidance on the use of NBP and instructions on their baseline medications. This approach aimed to ensure proper medication adherence and minimize potential variability in diabetes management. Despite these efforts, the absence of specific glycemic control measures limits our ability to fully assess the impact of diabetes management on cognitive outcomes. Future studies should incorporate detailed measures of glycemic control (e.g., HbA1c levels) to better understand this relationship and to validate our findings.

Few studies have examined how NBP affects distinct cognitive domains. Our previous work indicated significant improvements in orientation and language scores after 12 months of NBP therapy (Han et al., 2025). Similarly, Yan et al. reported that NBP improved naming, memory, attention, and language after just 1 month of treatment (Yan et al., 2017). However, these studies did not account for glycemic status. Given that diabetes is a critical covariate in cognitive decline, our stratified analysis provides novel insights. In diabetic patients, NBP appeared to mitigate decline in orientation, potentially due to its role in upregulating VEGF and inhibiting caspase-3-mediated apoptosis (Tan et al., 2022). Animal studies have also shown that NBP improves spatial memory and reduces escape latency in diabetic rats, supporting its positive effect on orientation-related functions (Tian et al., 2017).

Interestingly, prior studies have shown that diabetes-related cognitive impairment predominantly affects visuospatial ability, naming, language, and memory, with orientation and attention relatively preserved (Sumbul-Sekerci et al., 2025). Similarly, Gallucci et al. found that the most frequently affected domains in PSCI were language, episodic memory, and executive function, with less impact on spatial orientation (Gallucci et al., 2024). Our study is the first to report that in diabetic individuals, the cognitive benefit of NBP is concentrated in orientation, whereas in non-diabetic individuals, the improvement is observed in language. This phenomenon may be attributable to the fact that the MMSE language (9 points) and orientation (10 points) sub-scales yielded relatively low baseline scores in our low-education, rural cohort, while their wider score ranges provided greater scope for detectable change, making any improvement more readily observable. These findings are exploratory and require further mechanistic studies and larger sample sizes to confirm and elucidate the domain-specific effects.

Notably, the absence of significant changes in the memory and executive sub-domains observed in our trial was likely attributable to both the limited follow-up duration and the restricted score range of these subscales. Compared with the language and orientation items, the baseline scores of the memory and executive components were already close to the ceiling, leaving insufficient room to detect subtle intervention effects. In addition, diabetes-associated cognitive impairment preferentially affects the fronto-striatal circuitry and the parahippocampal region, manifesting as decreased executive efficiency and slowed information-processing speed (Fu et al., 2022; Moran et al., 2013). In contrast, hippocampal structure remains relatively preserved within the first 6 months after stroke unless chronic, severe hyperglycaemia is present (Chaturvedi et al., 2020). The neuroprotective effects of NBP are mediated by improved mitochondrial energy metabolism, up-regulation of VEGF dependent angiogenesis, and inhibition of caspase-3-mediated apoptosis (Abdoulaye and Guo, 2016). These mechanisms are most effective in salvaging the acutely compromised cortico-subcortical penumbra. However, executive dysfunction resulting from chronic small-vessel disease is characterized by extensive white-matter demyelination (Abdoulaye and Guo, 2016; Kalaria and Erkinjuntti, 2006). 12 months of oral NBP was insufficient to reverse established axonal damage, explaining the lack of demonstrable benefit in the executive sub-domains. Memory consolidation, which relies heavily on hippocampal plasticity, may require a more protracted or multi-modal intervention (e.g., combined cognitive training) before any measurable improvement emerges. In our cohort, stroke patients were enrolled during the sub-acute phase and the mean glycated hemoglobin of the diabetic subgroup at one-year follow-up was 7.2%. This degree of glycaemic control, together with the relatively short observation window, had probably not yet produced advanced microvascular hippocampal sclerosis; consequently, no significant additional deterioration was observed in the memory sub-domain, and a putative “rescue” effect of NBP could not be detected.

Beyond the effect of NBP treatment, age and education level are well-recognized predictors of cognitive decline. In our analysis, we found that age ≥60 years was associated with an increased risk of cognitive decline in non-diabetic patients, while each additional year of education was associated with a reduced risk. This highlights the importance of including these demographic factors in our regression models to ensure an accurate assessment of NBP’s efficacy. However, these results were not significant in the diabetic subgroup. This may be due to the complex internal environment associated with the disease state of diabetes, or it may be attributed to the smaller sample size, which limited our ability to fully detect specific relationships. BMI, hypertension, and hyperlipidemia are also important factors that may influence cognitive outcomes (Chen et al., 2022; Hou et al., 2019). Although these factors were included in our multivariate models, their specific impact on the efficacy of NBP in diabetic and non-diabetic patients warrants further investigation. Given the established role of diabetes in cognitive decline and its potential interaction with the mechanisms of NBP, we considered diabetes status as a potential effect modifier in our study. Our subgroup analysis showed differential effects of NBP in diabetic and non-diabetic patients, suggesting that diabetes may indeed modify the therapeutic response to NBP. Future studies should explore other potential effect modifiers, such as genetic predispositions, lifestyle factors (e.g., smoking, alcohol use), and comorbid conditions (e.g., cardiovascular disease), to provide a more comprehensive understanding of the factors influencing NBP’s efficacy.

This study has several limitations that should be acknowledged. First, a phased cluster sampling strategy was employed rather than a fully randomized design. Although this sequential design facilitated field implementation in a rural setting and minimized treatment contamination, it may introduce temporal trends and selection bias. Inclusion of “study period” as a covariate in univariate models showed no significant effect on cognitive outcomes; nevertheless, residual bias cannot be fully excluded, and future studies should employ cluster or individual randomization for definitive validation. Second, self-reported schooling may misclassify MMSE thresholds and inflate baseline risk. Given the rural, low-literacy setting, we applied education-specific Chinese norms and validated findings with a binary “cognitive impairment” outcome that aligned with continuous ΔMMSE. Third, our cognitive impairment assessment used only the MMSE, which may not fully capture functions of the frontal, parietal, and occipital lobes. Despite its limitations, the MMSE is widely used in community settings due to its feasibility, especially among populations with lower education. Future studies should consider incorporating more comprehensive neuropsychological assessment tools, such as the Montreal Cognitive Assessment (MoCA) and the Addenbrooke’s Cognitive Examination (ACE), to better capture executive and visuospatial functions. Fourth, cognitive-modifying factors such as diet, physical activity, social engagement, stroke severity as measured by the National Institutes of Health Stroke Scale (NIHSS), neuroimaging metrics, and mood disorders, together with concomitant medications not related to stroke, were not assessed. Given that participants were enrolled during the sub-acute phase (1–6 months post-stroke) and remained functionally independent, NIHSS shows only a weak correlation with long-term cognition; we therefore used the modified Rankin Scale (mRS), blood pressure, body-mass index and lipid profile as partial surrogates of vascular burden. In future work, we will employ standardized questionnaires (International Physical Activity Questionnaire, IPAQ; Food Frequency Questionnaire, FFQ) and quantitative neuroimaging to capture these domains comprehensively, and will include comorbidities such as atrial fibrillation and hepatic or renal insufficiency to explore potential drug–disease interactions that may modify the effect of NBP. Fifth, diabetes was ascertained at baseline using HbA1c ≥ 6.5%, fasting glucose, or prior diagnosis, with blood drawn in the subacute stroke phase to avoid acute stress hyperglycaemia misclassification. Absence of 12-month HbA1c data constrains causal inference. Although continuous monitoring was not mandated in this observational cohort, it is now designated as a required refinement for future studies to strengthen result robustness. Moreover, although the overall sample (n = 594) satisfied the a-priori power requirement, the number of incident cognitive-decline events within each diabetes stratum remained limited, diminishing statistical precision. Consequently, the reported subgroup differences should be interpreted as exploratory and must be validated in adequately powered, multicenter trials specifically designed around event-driven sample-size calculations. Finally, the study was conducted in a single rural cohort from Jizhou District, Tianjin; caution is needed when extrapolating the findings to other ethnicities or urban settings. Yet this focus also represents a deliberate strength—providing the first locally relevant evidence on NBP efficacy among underserved, high-risk rural populations. Multi-center, multi-ethnic trials are warranted to confirm these observations.

In this prospective cohort study, we found that NBP treatment over a 12-month period was associated with a reduced risk of cognitive decline in IS survivors, with more pronounced benefits observed in non-diabetic patients. Moreover, the cognitive domains affected by NBP varied by glycemic status: orientation ability was beneficial to in diabetic patients, whereas language function was better preserved in non-diabetic individuals. It should be noted that this study is exploratory in nature, aiming to generate hypotheses for future research rather than to confirm causal relationships. Although the study is observational and cannot determine a causal association, these findings still provide novel evidence for the differential efficacy of NBP across various patient subgroups and underscore the potential value of early intervention during the subacute phase of ischemic stroke, which may maximize the neuroprotective effects of NBP.

Our results underscore the importance of considering metabolic comorbidities, such as diabetes, when evaluating neuroprotective interventions. For non-diabetic patients with subacute ischemic stroke and no prestroke cognitive impairment, adding NBP for 12 months on top of standard secondary prevention may be considered to reduce the risk of cognitive decline. In stroke survivors with diabetes, NBP has not demonstrated a clear overall protective effect against global cognitive deterioration, although a modest benefit on orientation was observed. Optimizing glycemic and metabolic control should take priority, and the decision to add NBP should be individualized under fully informed consent. These recommendations are preliminary evidence-based suggestions, given the single-center, observational nature of this study.

These findings warrant further validation in larger, multicenter randomized controlled trials and an exploration of their underlying mechanisms. Such efforts will further clarify the role of NBP in cognitive protection and support its integration into tailored post-stroke rehabilitation programs.