The data utilized in this study were obtained from the National Health and Nutrition Examination Survey (NHANES, https://www.cdc.gov/nchs/nhanes/). NHANES database is a publicly accessible resource under the guidance of the National Center for Health Statistics (NCHS), which is responsible for data collection and quality control. Data extraction from the NHANES database was carried out using a rigorous stratified, multistage, and complex sampling methodology. This robust approach facilitated the acquisition of essential demographic particulars, physical assessments, biochemical blood analysis, as well as comprehensive responses to questionnaires about dietary patterns and medical histories. These data collectively provided an extensive portrayal of the health status and lifestyle factors prevalent within the U.S. population. This survey adheres to ethical guidelines, with all participants having provided informed consent before their inclusion.

Our study collected data from 2001 to 2018, encompassing a total of 91,351 participants. Initially, individuals under 18 years of age (n=37,595) and those without follow-up data (n=88) were excluded. Subsequently, a total of 5,152 individuals were excluded from the analysis due to insufficient information on complete blood cell counts, preventing the calculation of SIRI. Next, those with incomplete covariate information (n=30,128) were removed. Finally, we banned 15,126 participants who were not in CKD status. Eventually, 3262 participants were included in this study ( Figure 1 ).

Neutrophil, lymphocyte and monocyte counts were measured using automated hematology analyzing devices. SIRI is calculated by the following formula (39, 40):

The demographic data encompassed several variables, including age(categorized into 18-39, 40–59, ≥ 60 years old, referred to literature (41)), gender, race, education level, the ratio of family income to poverty (PIR), smoking and drinking habits, dialysis, systolic blood pressure (SBP), and diastolic blood pressure (DBP). Serum albumin, serum creatinine, alanine transaminase (ALT), aspartate transaminase(AST), triglycerides (TG), glycohemoglobin(GHB) and fasting blood glucose were performed under a unified standard. The laboratory utilizes the Jaffe rate method (kinetic alkaline picrate) to determine the concentration of creatinine in serum and plasma. A solid-phase fluorescent immunoassay (FIA) was employed for the quantification of human urinary albumin, while the enzymatic method was utilized to measure urinary creatinine levels. Urinary albumin and urinary creatinine are measured in random urine collected. To calculate the urinary albumin to creatinine ratio (ACR) and estimated glomerular filtration rate (eGFR), we employed the following formulas (41, 42):

It is important to note that Scr denotes serum creatinine concentration (mg/dL), and κ is 0.9 for males and 0.7 for females, and α is −0.411 for males and −0.329 for females.

We defined hypertension as meeting any of the following criteria: (a) self-report of a doctor’s diagnosis of hypertension; (b) current prescription for hypertension medication; (c) three consecutive measurements of systolic blood pressure (SBP) > 140 mmHg and/or diastolic blood pressure (DBP) > 90 mmHg on different days. For diabetes, the definition included meeting any of the following conditions (43): (a) a prior diagnosis of diabetes by a healthcare professional or current use of diabetes medications; (b) fasting blood glucose (FBG) ≥ 7.0 mmol/L; (c) glycated hemoglobin (HbA1C) ≥ 6.5%. According to a comprehensive review of the literature, the assessment of CKD status is based on the following criteria (5, 41, 44–46): (a) eGFR< 60 ml/min/1.73 m², (b) ACR > 30 mg/g. The CKD stage is based on the calculated results of eGFR (ml/min/1.73m²). The classification of CKD stage and the albuminuria grade are by international guidelines (44, 47) (CKD Stage I: eGFR≥90, Stage II: 60≤eGFR<90, Stage III:30≤eGFR<60, Stage IV: 15≤eGFR<30, Stage V: eGFR<15. Albuminuria grade: Mildly:<30mg/g, Moderately:30-300mg/g, Severely:>300mg/g).

The vital status of participants was verified by cross-referencing with the National Death Index (NDI) (https://ftp.cdc.gov/pub/Health_Statistics/NCHS/datalinkage/linked_mortality/).

All-cause mortality encompassed deaths from any cause. Cardiovascular mortality was defined as I00–I09, I11, I13, I20–I51, or I60–I69, as recorded in the National Death Index (48). This definition encompassed both cardiovascular disease (CVD) and cerebrovascular disease. For brevity, we may refer to cardiovascular mortality as CVD mortality. In the codebook for the National Center for Health Statistics (NCHS) 2019 public-use linked mortality files ( Supplementary File 1 ), the ucode_leading for CVD mortality is represented as 001 and 005.

We performed data analysis using the R project (version 4.1.1). Given the complex, multi-stage, and multi-stratified sampling approach used in the NHANES database, we incorporated weighting variables as guided by the data provider’s usage instructions. The “gtsummary” package in the R project was employed for weighted statistical analysis. We initiated our analysis of continuous variables with a normal distribution test, which indicated a departure from a normal distribution. Consequently, we represented these non-normally distributed data by reporting the median, along with the 25% and 75% percentiles, and applied the Wilcoxon test for statistical analysis. As for categorical variables, we illustrated them through numerical values and proportions and conducted statistical analysis employing the chi-square test.

In order to mitigate potential statistical biases, we adopted a weighted Cox regression model rather than the traditional regression model for our survival analysis of the study population. We employed Hazard Ratios (HR) accompanied by 95% Confidence Intervals (CI) to evaluate the impact of SIRI on survival. Next, we conducted a proportional hazards assumption test for the Cox regression models, affirming that our models adhered to the assumption of proportional hazards. Subsequently, we established three Cox regression models with variable adjustments to assess the association between SIRI and survival. Following this, subgroup analyses were conducted. We performed weighted Cox regression analyses on the association between SIRI and the risk of mortality across seven subgroups, including age, gender, the presence of hypertension or diabetes, dialysis status, distinct CKD stages, and diverse albuminuria grades. After that, we generated Kaplan-Meier (KM) survival curves to compare the survival status among different SIRI groups. To elucidate the potential interactions between these seven subgroups and the risk of mortality, we conducted interaction analyses concurrently.

We constructed receiver operating characteristic (ROC) curves to evaluate the specificity and sensitivity of SIRI in predicting mortality. Restricted cubic spline analysis, a scientific and commonly utilized analytical method, is employed to investigate the potential nonlinear relationship between two variables (49, 50). This method enables curve fitting through restricted cubic splines. In the context of our Cox regression model, restricted cubic spline analysis is used to elucidate how the hazard ratio (HR) values vary with changes in SIRI. Statistical significance was considered at a p-value< 0.05 in all hypothesis tests.

This study encompassed a cohort of 3,262 individuals. Among them, 1,535 were male (weighted proportion: 42%), and 2,216 were aged 60 or above (weighted proportion: 59%). According to the CKD staging, these individuals are classified into five groups. The populations in different groups exhibit statistical differences in terms of age, racial distribution, family income situation, drinking habits, BMI, fasting blood glucose, lipid profiles, liver function indicators, and serum creatinine levels. However, no statistically significant differences were found in terms of gender distribution, smoking habits, and educational levels ( Table 1 ).

Advanced stages of CKD appear to be more prevalent in older individuals, those from economically disadvantaged households, individuals with comorbid hypertension, those undergoing dialysis, and those with elevated levels of serum creatinine, ACR, and SIRI. Participants with lower SIRI scores tend to exhibit characteristics similar to individuals with higher glomerular filtration rates (eGFR) and hemoglobin levels.

Three models were developed to assess the independent functionality of SIRI and its correlation with mortality. The investigation entailed the utilization of multivariate weighted Cox regression to examine the interconnection between SIRI and mortality, with meticulous adjustments made to account for potential confounding variables. The findings, inclusive of Hazard Ratios (HRs) and their corresponding 95% CIs, have been presented for comprehensive analysis ( Table 2 ). It is important to note that in the weighted Cox regression analysis, all participants will be reclassified into four groups based on the quartiles of SIRI (Q1:<0.661, 0.661≤Q2<0.978, 0.978≤Q3<1.460, Q4:≥1.460).

In Model 1, no covariates were adjusted. In Model 2, we made adjustments for age, gender, race, education level, poverty status (PIR), and dialysis status. In Model 3, we further adjusted for BMI, hypertension, diabetes, alcohol consumption, smoking status, serum creatinine, fasting blood glucose, glycated hemoglobin, lipid levels, ACR, and transaminase levels. The Cox regression results indicate that higher SIRI levels are associated with an increased risk of cardiovascular mortality (Q4 vs. Q1). Compared to participants in the first quartile of SIRI, those in the fourth quartile of SIRI experienced a 2.5-fold increase in cardiovascular mortality risk and a 3.02-fold increase in all-cause mortality risk.

The Kaplan–Meier curve ( Figure 2 ) corroborates the aforementioned conclusion. It demonstrates that individuals with higher SIRI scores experience a more significant decline in survival over a 20-year follow-up when compared to those with lower SIRI levels in CKD populations (log-rank P for trend<0.0001).

To determine the consistency of the relationship between SIRI and mortality across various subgroups of the population, subgroup analyses were conducted within specific categories, including age, gender, the presence of hypertension and diabetes, dialysis status, CKD stage, and albuminuria grade.

The results from the forest plot ( Figure 3 ) reveal that elevated SIRI remains an independent risk factor for all-cause mortality in individuals aged over 40, not undergoing dialysis, and with CKD stages I-IV. This association holds regardless of the presence of comorbid hypertension and diabetes or the severity of urinary albumin levels. However, the conclusion that elevated SIRI increases the risk of cardiovascular mortality appears to be established only in male participants aged over 60, with comorbid hypertension but not undergoing dialysis. Furthermore, among individuals in CKD stages, I-III, or those with mild to moderate albuminuria, elevated SIRI is associated with an increased risk of cardiovascular mortality. Interaction analysis indicates the absence of interaction between SIRI and these seven factors in jointly influencing cardiovascular or all-cause mortality. This implies that SIRI independently serves as a factor, contributing to an increased risk for both types of mortality.

The establishment of the Receiver Operating Characteristic curve (ROC, Figure 4 ) is integral for evaluating the sensitivity and specificity of SIRI as a prognostic diagnostic tool. SIRI’s predictive capacity for all-cause mortality is exemplified by an Area Under the Curve (AUC) of 0.624 (95% CI: 0.604, 0.644). Likewise, the predictive performance of SIRI for cardiovascular mortality yields an AUC of 0.593 (95% CI: 0.564, 0.622).

In order to elucidate the specific relationship between SIRI and HR (Hazard Ratio), the restricted cubic spline regression analysis was conducted ( Figure 5 ). In the examination of the relationship between SIRI and the Hazard Ratio (HR) for cardiovascular mortality, we discovered a non-linear correlation (Nonlinear p-value = 0.0389, which is<0.05, indicating a significant nonlinear relationship between SIRI and HR for CVD). Therefore, we utilized a non-linear model to elucidate their association. The SIRI value corresponding to an HR of 1 is 1.2. When SIRI is<1.2, there is a decrease in cardiovascular mortality risk within a certain range as SIRI increases. However, when SIRI > 1.2, there is an increase in cardiovascular mortality risk, and as SIRI continues to increase, the HR for cardiovascular mortality changes gradually and levels off.

Interestingly, in the context of all-cause mortality, SIRI and HR exhibit a linear association (non-linear p = 0.4462). As SIRI increases, HR follows suit, and this trend persists beyond the threshold of 1.39, at which point HR surpasses 1.0.