Novel serum biomarkers, medication adherence, and a predictive nomogram in patients with type 2 diabetes and ischemic stroke

Background: Long-term management of pharmacotherapy in patients with type 2 diabetes mellitus (T2DM) and ischemic stroke is challenging due to comorbidities, neurological impairment and functional limitations. Although clinical factors are known to influence adherence, the contribution of circulating biomarkers of neurological injury and repair remains poorly characterised. To investigate the clinical and biomarker-related factors associated with suboptimal medication adherence, and to develop a biomarker-enhanced nomogram for predicting poor adherence in patients with T2DM complicated by ischemic stroke. Methods: A retrospective cohort study was conducted in which 100 patients were screened, 100 enrolled, and 0 were excluded based on predefined inclusion and exclusion criteria. All the patients were treated from March 2025 to July 2025. Demographics, comorbidities, medication classes, National Institutes of Health Stroke Scale, Barthel Index, rehabilitation guidance, and circulating biomarkers, including neuron-specific enolase (NSE), S100 calcium-binding protein B (S100B), brain-derived neurotrophic factor (BDNF), high-sensitivity C-reactive protein, and HbA1c, were collected. Medication adherence was evaluated at 6 months post-discharge using the Morisky Medication Adherence Scale. Patients were randomly assigned to a training cohort (n=70) and a validation cohort (n=30). Univariate and multivariable logistic regressions were performed to find factors associated with poor adherence, and a biomarker-enhanced nomogram was built. Model performance was evaluated by the concordance index, calibration curves, receiver operating characteristic curves and net reclassification improvement (NRI). For NRI analysis, patients were categorised into three clinically interpretable predicted-risk groups: low risk (&lt;30%), intermediate risk (30%–60%), and high risk (&gt;60%). Results: Forty-three patients showed poor adherence to medication. Independent predictors of poor adherence were age <span style="color: rgb(32, 33, 34); font-family: sans-serif; font-size: 16px;">≥</span>60 years, primary education or less, <span style="color: rgb(32, 33, 34); font-family: sans-serif; font-size: 16px;">≤</span>2 types of medication, NIHSS score <span style="color: rgb(32, 33, 34); font-family: sans-serif; font-size: 16px;">≥</span>4, Barthel Index <span style="color: rgb(32, 33, 34); font-family: sans-serif; font-size: 16px;">≤</span>60, no guidance for rehabilitation, high NSE (&gt;15 ng/mL), high S100B (&gt;110 pg/mL) and low BDNF (&lt;22 ng/mL). Discrimination of the nomogram, including the biomarker, was good, with C-indexes of 0.851 in the training cohort and 0.838 in the validation cohort. The area under the curve values were 0.873 and 0.851, respectively. Calibration plots showed little difference between predicted and observed probabilities. The inclusion of NSE, S100B, and BDNF in the clinical-only model improved risk classification of patients (NRI 0.324, 95% CI: 0.112–0.536) using the predefined low, intermediate, and high-risk categories. Conclusion: onclusion: Clinical characteristics, neurological function, rehabilitation guidance and serum neurological injury/ repair biomarkers were related to the medication adherence of the patients with T2DM and ischemic stroke. A nomogram combining NSE, S100B, and BDNF biomarkers improved the prediction of poor adherence. They may facilitate the identification of high-risk patients who need early, individualised adherence-support interventions.