NHANES (https://www.cdc.gov/nchs/nhanes/index.htm), conducted by the Centers for Disease Control and Prevention (CDC), is a survey research program based on the population and cross-sectional, which was used to evaluate the nutritional conditions and the overall health of the entire population in American through a complex survey method. The database includes a wide range of data, including demographic, dietary, examination, laboratory, and questionnaire information. The study procedures were authorized by the Ethical Review Committee of the National Health Statistical Research Center and all individuals agreed to join the study. Data collection was carried out over eight periods between 2001 and 2016. Additionally, the NHANES Public-Use Linked Mortality Files were linked with the NHANES database from 2001 to 2016, enabling researchers to identify participant deaths up until December 31, 2019.

In this cohort study, a total of 91,352 subjects were initially included. After including participants aged 60 years or older, we excluded participants with missing 25(OH)D information (n=2,166), participants with undiagnosed hyperlipidemia (n=2,664), and participants with missing mortality data (n=1,160). Subsequently, we further excluded participants with missing information on covariates, including those with missing information on poverty income rate (PIR), education information, body mass index (BMI), and information on drinking and smoking. A final analysis cohort of 9271 individuals was obtained (shown in Figure 1 ).

In the NHANES 2001-2006 study, 25(OH)D serum levels were measured by the CDC using radioimmunoassay (DiaSorin Assay Kit). Serum samples from 2007 to 2016 were analyzed using a standardized liquid chromatography-tandem mass spectrometry (LC-MS/MS) to estimate the 25(OH)D serum levels. The 25(OH)D serum levels of participants who took part in NHANES 2001-2006 were transformed into equivalent values derived from the standardized LC-MS/MS technique using regression. This transformation was performed to facilitate the utilization and evaluation of 25(OH)D serum levels by researchers. The unit of measurement for 25(OH)D serum levels is nmol/L.

The study incorporated data from the NHANES database, which encompasses details on demographics, examination information, laboratory information, and questionnaire information. The following categories composed the demographic data: age, gender, ethnicity (Mexican American, Other Hispanic, Non-Hispanic Black, Non-Hispanic White, or Other Race), level of education (less than high school, high school or equivalent, and college or above), the status of marital (never married, married, and separated), and PIR. Age and PIR were treated as continuous variables, while race, education, and marital status were considered categorical variables. Examination information included diastolic blood pressure (DBP), systolic blood pressure (SBP), weight, and height information. BMI was calculated by dividing the participant’s weight in kilograms by the square of their height in meters. Laboratory information contained measurements for fasting plasma glucose, hemoglobin A1c (HbA1c), total cholesterol, triglycerides, HDL, and LDL. The questionnaire section contained information about drug use, alcohol consumption, and smoking habits.

In this study, hypertension was identified as a self-reported high blood pressure diagnosis, currently receiving treatment for hypertension, with SBP (average) ≥140 mmHg or DBP (average) ≥90 mmHg. Diabetes was characterized by either self-reported diagnosis, self-reported utilization of medication or insulin to reduce glucose, HbA1c levels of 6.5%, or glucose levels of fasting plasma ≥ 7.0 mmol/L. Hyperlipidemia was characterized by triglycerides at or above 150 mg/dl, total cholesterol at or above 200 mg/dl, LDL at or above 130 mg/dl without using cholesterol-lowering drugs, and HDL lower than 40 mg/dl in men or 50 mg/dl in women.

According to the results of the survey, participants were divided into two groups: nonsmokers and smokers. Smoker status refers to individuals who reported smoking more than 100 cigarettes in their lifetime. The drinking status (yes or no) was determined based on whether alcohol consumption exceeded 12 drinks in the past year.

Considering the potential linear or nonlinear association between 25(OH)D serum levels and various mortality outcomes (including all-cause mortality, cardiovascular disease fatality, malignant neoplasm mortality, and other-cause mortality), subjects were categorized into four groups based on quartiles of 25(OH)D serum levels: Q1 (<49.50 nmol/L), Q2 (49.5-67.50 nmol/L), Q3 (67.5-86.7 nmol/L), and Q4 (>86.7 nmol/L).

Continuous variables are presented as mean (standard error, SE), while categorical data is expressed as frequencies (percentages). Kaplan-Meier curves were used to illustrate survival rates based on 25(OH)D serum levels. To visualize the nonlinear correlation between 25(OH)D serum levels and mortality, restricted cubic splines were employed. Cox proportional hazard models were utilized to calculate adjusted hazard ratios (HRs) and their corresponding 95% confidence intervals for different mortality rates associated with 25(OH)D serum levels. Model 1 represents unadjusted results, while Model 2 adjusts for age, sex, race, and PIR. Model 3, which includes all covariates from Model 2, additionally adjusts for education, marital status, BMI, drinking, smoking, diabetes, and hypertension. Finally, a subgroup analysis was utilized to investigate the robust relationship between 25(OH)D serum levels and fatality rate across different populations.

Every result produced from the data was calculated utilizing a weighted data analysis that was founded on intricate sampling. All statistical analyses were performed using R version 4.0.4 (R-project.org), incorporating weighted data analysis based on complex sampling. Statistical significance was indicated by a P-value below 0.05 in all studies.

Based on the final cohort study, a total of 9,271 participants were included, with males representing 43.28% of the cohort. The average age of participants was 69.58 years, and the mean follow-up period was 88.37 months. Table 1 provides detailed information on survival and mortality rates among the subjects in this study. Age, gender, race, PIR, education level, marital status, BMI, smoking, diabetes, hypertension, and 25(OH)D serum levels were found to have statistically significant associations with mortality rates (P < 0.05). However, no correlation was observed between drinking status and mortality rates. Table 2 examines the relationship between the aforementioned variables and the independent variable, 25(OH)D serum levels. Higher 25(OH)D serum levels were frequently related to increased survival and reduced mortality rates.

The cumulative Kaplan-Meier curve of the quartiles of 25(OH)D levels is presented in Figure 2 . Regarding the overall mortality rate, the cumulative survival probability was significantly lower in the lowest 25(OH)D levels group (P < 0.001). This pattern was consistent for CVD mortality, malignant neoplasm mortality, and other causes of mortality (P < 0.001).

To further investigate the impact of serum 25(OH)D levels on mortality, three Cox regression models were constructed. These models accounted for various factors, including age, sex, ethnicity, education attainment, poverty income rate (PIR), marital status, BMI, smoking status, drinking habits, and history of diabetes or hypertension (Model 3). The adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated as follows: 0.996 (0.994, 0.998), 0.995 (0.991, 0.999), 0.997 (0.993, 1.002), and 0.994 (0.989, 0.999) for all-cause mortality, CVD mortality, malignant neoplasm mortality, and other-cause mortality, respectively.

The HRs and 95% CIs in Table 3 calculated for the quartiles of 25(OH)D serum levels, ranging from the lowest to highest categories, were as follows: 1.000 (reference), 0.776 (0.679, 0.887), 0.784 (0.682, 0.901), and 0.697 (0.590, 0.822) for all-cause mortality (P trend <0.001); 1.00 (reference), 0.796 (0.624, 1.015), 0.776 (0.602, 1.001), and 0.673 (0.508, 0.892) for CVD mortality (P trend = 0.007); 1.00 (reference), 0.699 (0.527, 0.927), 0.714 (0.538, 0.947), and 0.712 (0.520, 0.975) for malignant neoplasm mortality (P trend = 0.028); and 1.00 (reference), 0.788 (0.605, 1.028), 0.825 (0.621, 1.097), and 0.649 (0.456, 0.926) for other-cause mortality (P trend = 0.058).

Figure 3 illustrates the depiction of the association between mortality and 25(OH)D serum levels using a restricted cubic spline. For all-cause mortality, a notable decrease in mortality was observed initially, followed by a stabilization effect as 25(OH)D serum levels gradually increased. The nonlinear relationship demonstrated a saturation point at approximately 67.5 nmol/L. This nonlinear pattern was similarly observed for CVD mortality, malignant neoplasm mortality, and other-cause mortality, with the saturation point consistently around 67.5 nmol/L. Table 4 provides further evidence supporting the nonlinear correlation between 25(OH)D serum levels and mortality.

Finally, to investigate the correlation between mortality and 25(OH)D serum levels in greater detail, subgroup analyses were performed on elderly individuals with hyperlipidemia, considering various characteristics. Notably, the robustness of the negative correlation between all-cause mortality and 25(OH)D serum levels was evident across different subgroups, including age, sex, alcohol consumption, hypertension, and diabetes. Interestingly, among non-obese (BMI < 30 kg/m²) and smoking elderly individuals with hyperlipidemia, this negative correlation was particularly significant ( Table 5 ). Furthermore, robust associations were observed between 25(OH)D serum levels and mortality due to CVD or malignant neoplasms across different subgroups, including age, gender, BMI, tobacco use, alcohol consumption, high blood pressure, and diabetes ( Supplementary Tables S1 , S2 ). In the smoking population, a more pronounced negative relationship was found between other-cause mortality and 25(OH)D serum levels ( Supplementary Table S3 ).

Our study was based on NHANES 2001-2016 data covering 9,271 hyperlipidemic elderly participants, which provided us with sufficient statistical power to detect the association between serum 25(OH)D levels and mortality. In addition, the long follow-up period of our study (mean 88.37 months) allowed us to observe more stable long-term results, increasing the reliability of our findings. Although the reported findings yield intriguing conclusions, it is important to acknowledge certain limitations within this research. Firstly, as an observational study, establishing a causal relationship is not feasible. Secondly, 25(OH)D serum levels were assessed based on a single measurement, failing to capture participants’ status consistently throughout the follow-up period. Thirdly, despite meticulous adjustment for numerous confounding factors, the presence of unmeasured or unknown residual confounders cannot be entirely ruled out. Fourthly, the investigation did not delve into the specific effects of distinct vitamin D supplements on the association between 25(OH)D levels and mortality rates. Therefore, further research is imperative to elucidate the potential role of vitamin D supplementation in this domain. Finally, it is noteworthy that African-American and Hispanic adults are more susceptible to 25(OH)D insufficiency (37), but the subgroup analysis did not account for the potential impact of race on the observed association.