The National Health and Nutrition Examination Survey (NHANES) is a study conducted nationwide in the United States, with the goal of collecting data on the nutrition and health of noninstitutionalized civilians. To obtain baseline data, various information such as demographic details, physical examinations, and laboratory tests were gathered at both the participants’ homes and a mobile center. The protocol used for this study was approved by the Institutional Review Board of the National Center for Health Statistics, and all participants were required to provide informed consent when enrolling in the survey.

For this study, we utilized data from ten NHANES cycles conducted between 1999 and 2018, encompassing a total of 101,316 individuals. We excluded subjects who lacked survival status (n = 42,252), pregnant individuals, and those without a diagnosis of type 2 diabetes (n = 49,665). Additionally, individuals below 18 years old (n = 31) and those with missing SII data (n = 700) were also excluded. Following the application of these exclusion criteria, a total of 8,668 subjects with type 2 diabetes were included in this study. Type 2 diabetes was defined based on one of the following criteria: (1) self-reported doctor-diagnosed diabetes; (2) fasting blood glucose ≥ 7.0 mmol/L; (3) two-hour oral glucose tolerance test blood glucose ≥ 11.1 mmol/L; (4) glycated hemoglobin A1c (HbA1c) ≥ 6.5%; or (5) use of hypoglycemic medication. Figure 1 provides a visual representation of the detailed participant selection process.

Flow diagram of the selection of eligible individuals

The calculation of the SII in this study involved multiplying the neutrophil-to-lymphocyte ratio by the platelet count (10^9 cells/µL), as described in a previous investigation [11]. As the distribution of SII exhibited a right-skewed pattern, the variable was assessed in its continuous form after applying a natural log transformation (lnSII). Subsequently, the lnSII variable was divided into four equal subgroups.

To determine all-cause, CVD, and cancer mortality rates, the study participants were linked to the National Death Index until December 31, 2019. Cause-specific deaths were identified using the International Classification of Diseases, Tenth Revision (ICD-10) codes. CVD mortality was defined by ICD-10 codes I00-I09, I11, I13, I20-I51, or I60-I69, while cancer mortality was defined by ICD-10 codes C00-C97.

Age, gender, ethnicity, education, family income, smoking and drinking habits, comorbidity disease status, and diabetes medication use were assessed through structured interviews conducted at the participants’ homes. At the mobile center, measurements of body mass index (BMI) and collection of blood samples were performed. Ethnicity was classified into four categories: Mexican American, non-Hispanic White, non-Hispanic Black, or other. Education level was categorized as less than high school, high school or equivalent, or college or higher. The ratio of family income to poverty was classified into three groups: 0–1.0, 1.0–3.0, or > 3.0. Smoking status was defined as either a never smoker (smoked < 100 cigarettes in their lifetime), current smoker (smoked ≥ 100 cigarettes in their lifetime and currently smokes some days or every day), or former smoker (smoked ≥ 100 cigarettes in their lifetime and currently does not smoke). Drinking status was grouped as nondrinker (< 12 drinks in their lifetime), low-to-moderate drinker (≤ 1 drink per day for females, ≤ 2 drinks per day for males, or binge drinking on < 2 days per month), heavy drinker (> 1 drink per day for females, > 2 drinks per day for males, or binge drinking on ≥ 2 days per month), or former drinker (≥ 12 drinks in their lifetime and did not drink last year).

Hypertension was defined based on meeting one of the following criteria: (1) self-reported doctor-diagnosed hypertension; (2) mean systolic blood pressure ≥ 140 mmHg or mean diastolic blood pressure ≥ 90 mmHg; or (3) use of antihypertensive medication. Hyperlipidemia was defined based on meeting one of the following criteria: (1) self-reported doctor-diagnosed hyperlipidemia; (2) triglyceride (TG) levels ≥ 150 mg/dL, total cholesterol (TC) levels ≥ 200 mg/dL, high-density lipoprotein (HDL) levels < 40 mg/dL, low-density lipoprotein (LDL) levels ≥ 130 mg/dL; or (3) use of antihyperlipidemic medication. Atherosclerotic cardiovascular disease (ASCVD) was defined as the presence of coronary heart disease, heart attack, angina, or stroke.

At the time of recruitment, various laboratory measurements were conducted, including complete blood count, plasma glucose, HbA1c, TG, TC, HDL, and LDL. To assess insulin resistance, the homeostatic model assessment of insulin resistance (HOMA-IR) was calculated using a previously established method [16]. Further information can be found on the official website (http://www.cdc.gov/nchs/nhanes/).

To account for the complex sampling design of NHANES, all analyses incorporated sample weights, clustering, and stratification. Person-time was calculated from the recruitment date until either the date of death or the end of follow-up (December 31, 2019), whichever occurred first. Weighted means ± standard error (SE) was used for continuous variables, while frequency and weighted percentages were used for categorical variables.

The associations between lnSII and all-cause and cause-specific mortality were estimated using multivariate Cox regression models. Model 1 adjusted for age, sex, and ethnicity. Model 2 further adjusted for BMI, education level, family income-poverty ratio, smoking status, and drinking status. Model 3 additionally adjusted for duration of diabetes, diabetes medication use, HbA1c, and the presence of hypertension, hyperlipidemia, ASCVD, and chronic kidney disease (CKD). The linear trend was assessed by treating the median value of each category as a continuous variable. Multiple imputation was employed to handle missing values for variables.

To assess the relationship between lnSII and all-cause and cause-specific mortality, restricted cubic spline regression with four knots was employed. This analysis incorporated the multivariate adjustment. In order to evaluate potential nonlinearity, a likelihood ratio test was performed. If nonlinearity was detected, a two-piece Cox proportional hazards regression model was utilized, which allows for nonlinearity at the inflection point. The inflection point represents the point at which the relationship between the predictor and outcome variables undergoes a change and was used to determine the two separate components of the model.

Stratified analyses were conducted, considering factors such as age, gender, ethnicity, BMI, smoking status, HbA1c, duration of diabetes, hypertension, and hyperlipidemia. The significance of interactions between lnSII and stratification variables was evaluated using the P value for the product terms.

Sensitivity analyses were performed to test the robustness of the results. Exclusion criteria included individuals with a history of ASCVD or cancer, as well as those who died within 2 years of follow-up to minimize reverse causation bias. Additionally, the multivariate model was further adjusted to include the healthy eating index (HEI) for dietary factors, as well as HOMA-IR, TG, HDL, and LDL. Statistical significance was defined as a two-sided P value < 0.05. All analyses were conducted using R version 4.2.3.