Abstract / Summary
Background Inflammatory cascades and the disruption of protein homeostasis, characterised by diminished serum albumin concentrations, play a pivotal role in the pathophysiological progression of acute ischaemic stroke (AIS). However, previous comparisons across various composite biomarkers have frequently overlooked the mathematical coupling bias introduced by shared components. Furthermore, a quantitative understanding of their non-linear dose–response relationships and synergistic interactions remains elusive. The present study sought to deconstruct highly collinear prognostic indices into distinct pathophysiological axes, delineate their non-linear dose–response trajectories, and quantify the cumulative risk associations and potential interactive effects of joint exposure to systemic inflammation and diminished serum albumin concentrations on 1-year functional outcomes in patients with AIS. Methods This retrospective cohort study consecutively enrolled patients with AIS between 2021 and 2023. To circumvent mathematical coupling bias, nine composite indices were rigorously deconstructed into four independent pathophysiological axes. The primary endpoint was a poor functional outcome (modified Rankin Scale score >2) at 1 year after stroke onset. Representative biomarkers were selected using bootstrap-based stability inference. Restricted cubic splines (RCS) and modified Poisson regression with robust standard errors were used to assess non-linear trajectories and synergistic interactions, respectively. A final joint predictive model was constructed following dimensionality reduction via the bootstrap-LASSO algorithm, and the E -value was introduced to quantify the robustness of the findings against unmeasured confounding. Results We included 495 patients, among whom 111 (22.4%) experienced a poor functional outcome. Following rigorous validation procedures, the neutrophil-to-lymphocyte ratio (NLR), C-reactive protein-to-albumin ratio (CAR), albumin (ALB), and prognostic nutritional index (PNI) were identified as the optimal representative indices for their respective axes. RCS analysis revealed significant non-linear dynamics: the risk trajectory for high-sensitivity C-reactive protein (hsCRP) exhibited a plateau effect, whereas CAR demonstrated an inverted U-shaped pattern. Cross-classification analysis demonstrated that the co-occurrence of elevated NLR and diminished ALB significantly increased the risk of adverse outcomes (adjusted RR 2.45, 95% CI 1.33–4.49), with an absolute risk of 36.9% in the joint exposure group (absolute risk difference [RD] 25.4% versus the reference cohort). Although neither multiplicative nor additive interaction tests achieved statistical significance (multiplicative interaction p = 0.856; RERI 0.432, 95% CI spanning zero), this association suggests a potential trend towards compounded high-risk exposure. The bootstrap-LASSO joint predictive model demonstrated excellent calibration (slope 0.805) and clinical net benefit. E -value analysis (2.16) and probabilistic simulations corroborated that these findings were highly robust against unmeasured confounders, such as baseline neurological deficits (e.g., missing National Institutes of Health Stroke Scale scores). Conclusion The concurrence of systemic inflammatory activation and relatively diminished albumin concentrations constitutes a clinical phenotype robustly associated with long-term adverse outcomes in AIS. By uncoupling collinear biomarkers, this study unmasks complex, non-linear prognostic trajectories that are typically obscured under conventional single-marker or linear assumptions. Furthermore, the compounded high-risk profile of elevated inflammation and diminished albumin provides preliminary epidemiological rationale to inform future investigations into acute-phase anti-inflammatory therapies combined with metabolic homeostasis support.