Approval for the study was granted by the ethics review board of Shenzhen Second People’s Hospital (Ethics Approval Number: 2023-305-01PJ), and the research was carried out following the ethical principles established in the Declaration of Helsinki of 1964 and its later modifications, along with other relevant ethical guidelines. Written informed consent was secured from all participants included in this study.
The RPR value is calculated using the following formula: RPR = RDW/platelet count. In this formula, RDW refers to the standard deviation of red blood cell volume, measured in femtoliters (fL), and the platelet count is expressed in 10⁹/L. Both RDW and platelet count are measured within 24 h after hospital admission.
At 90 days following the onset of AIS, trained follow-up personnel, who have a thorough understanding of the mRS scoring criteria, assessed patients through face-to-face interviews or telephone interviews to collect data on patients’ functional status. During these assessments, we employed the mRS to evaluate functional outcomes, which is a commonly used tool for evaluating functional recovery and independence in daily living among patients with stroke or other neurological diseases. The mRS score ranges from 0 to 6, reflecting various functional states from no symptoms to death. The primary endpoint of this study was the neurological functional outcome at 90 days, classified into two categories: unfavorable (mRS ≥ 3) and favorable (mRS < 3) outcomes. To ensure the accuracy of follow-up, we established a patient information contact database and conducted multiple phone calls to remind patients to participate in the follow-up, thus reducing the occurrence of follow-up loss.
Covariates were chosen based on our clinical expertise and prior studies. The variables identified for inclusion as covariates included: (i) categorical variables including coronary heart disease (CHD), sex, history of prior stroke or transient ischemic attack (TIA), pneumonia, hypertension, atrial fibrillation (AF), DM, smoking status, and stroke etiology; and (ii) continuous variables such as neutrophil count (NEU), age, lymphocyte count (LYC), total cholesterol (TC), the initial score on the National Institutes of Health Stroke Scale (NIHSS) upon admission, triglycerides (TG), fibrinogen (FIB), fasting plasma glucose (FPG), body mass index (BMI), homocysteine (HCY), serum albumin (ALB), low-density lipoprotein cholesterol (LDL-C), hemoglobin concentration (HGB), high-density lipoprotein cholesterol (HDL-C), and D-dimer levels.
At the time of admission to the hospital, specialized research coordinators systematically gathered baseline data pertaining to patients' demographic information and medical histories. This data included information on prior strokes, smoking status, AF, CHD, hypertension, and DM. Neurologists assessed stroke severity at admission using the NIHSS. Stroke subtypes were categorized based on the criteria established by the Trial of Org 10,172 in Acute Stroke Treatment (TOAST). Blood samples were gathered within 24 h following patient admission and were later examined at the laboratory of Shenzhen Second People's Hospital. Qualified technicians adhered to strict quality control protocols during laboratory evaluations, ensuring the confidentiality of patients' baseline data.
In this study, certain covariates exhibited missing data, with the corresponding counts and percentages of absent entries listed as follows: FIB (5, 0.30%), FPG (21, 1.25%), TG (40, 2.38%), HDL-c(40, 2.38%), LDL-c (40, 2.38%), TC (41, 2.44%), HCY (119, 7.07%), and NIHSS score (207, 12.30%). Missing data can undermine the statistical validity of the sample analyzed during the modeling phase. To minimize the bias resulting from these missing variables, we utilized multiple imputation techniques to address the unavailable data. The covariates utilized in the imputation model included age, NEU, LYC, NIHSS score at the time of admission, HGB, RDW, platelet count, BMI, FIB, D-dimer, HCY, TG, HDL-c, ALB, FPG, TC, LDL-c, sex, history of previous stroke or TIA, DM, hypertension, AF, pneumonia, CHD, smoking status, and stroke etiology. The imputation procedure was executed using a linear regression method across ten iterations. The analysis of the missing data was based on the assumption of missing at random (MAR).
Baseline variables were categorized by the quartiles of RPR, facilitating the comparison of characteristics across the groups. Continuous variables with a Gaussian distribution were summarized as means and standard deviations, while non-normally distributed variables were represented using medians and interquartile ranges. Categorical data were expressed as frequencies and percentages. Both analyses of variance (ANOVA) and the Kruskal-Wallis H test were applied to continuous variables, whereas the chi-square (χ²) test was used to evaluate differences among the RPR groups for categorical variables.
This study utilized univariate and multivariate binary logistic regression analyses to develop three distinct models examining the relationship between RPR and the incidence of unfavorable outcomes 90 days after AIS. The models consisted of: (i) Model I: no covariate adjustments; (ii) Model II: adjusted for sex, age, and BMI; and (iii) Model III: adjusted for age, smoking, BMI, CHD, sex, TG, HGB, LDL-c, D-dimer, FPG, stroke etiology, hypertension, HDL-c, DM, and initial NIHSS score.
In order to enhance the reliability of the results, we carried out multiple sensitivity analyses. Initially, RPR was transformed into a categorical variable according to its quartiles, and the trend P-value was computed to evaluate the results of RPR as a continuous variable while investigating the possibility of non-linearity. Second, to address the influence of obesity, hypertension, and DM on the prognosis of AIS patients, we conducted a sensitivity analysis by excluding individuals with a BMI greater than or equal to 28 kg/m², as well as those with hypertension and DM. In addition, we calculated the E-value to evaluate the potential impact of unmeasured confounders on the link between RPR and 90-day unfavorable outcomes.
A logistic regression model using restricted cubic spline functions was employed to explore the potential non-linear association between RPR and 90-day unfavorable outcomes in AIS patients. A recursive approach was utilized to identify the inflection point if a non-linear association was detected. After identifying the inflection point, separate binary logistic regression models were developed for each side of this threshold. The likelihood ratio test was used to select the model that best represented the relationship.
Stratified binary logistic regression models were applied to conduct subgroup analyses across several categories, such as pneumonia, TG, sex, age, history of previous stroke or TIA, AF, FIB, ALB, hypertension, BMI, Smoking, and CHD. In this analysis, continuous variables like TG FIB, ALB, BMI, and age were categorized based on clinically relevant cutoffs. Specifically, TG was categorized using a threshold of 1.7 mmol/L, BMI by 28 kg/m², FIB by 4 g/L, and ALB by 35 g/L. Adjustments were made for sex, LDL-c, CHD, TG, stroke etiology, age, hypertension, Scr, HDL-c, platelet count, FPG, smoking status, DM, and initial NIHSS score, while excluding the stratification variables. Likelihood ratio tests were conducted to evaluate the existence of interaction terms by comparing models that included these terms to those that did not. Finally, a receiver operating characteristic (ROC) curve was constructed to assess the predictive capability of RPR, RDW, and platelet count for unfavorable outcomes in AIS patients.
All findings were written in line with the STROBE statement. Statistical analyses were executed using Empower software (version 4.2) and R (version 3.4.3). A two-tailed p-value of less than 0.05 was considered statistically significant.
