Association of serum 25-hydroxy-vitamin D concentration and risk of mortality in cancer survivors in the United States

This study utilized data from the NHANES, an ongoing national survey orchestrated by the National Center for Health Statistics to appraise the health and nutritional status of Americans across all ages [17]. NHANES comprehensively gathers data relevant to demographics, socioeconomic status, diet, health examinations, medical evaluations, laboratory analyses, and participant questionnaires. Participation was contingent upon obtaining written informed consent from all subjects.

The study cohort consisted of individuals aged between 20 and 80 who had completed the NHANES interview, examination, and the dual 24-h dietary recall process across the 2003–2016 cycles. Mortality status was determined using the Public-Use Linked Mortality File, updated until December 31, 2019. This database connects NHANES participants to the National Death Index, with individual records linked via the Sequence Number (SEQN) provided by the National Center for Health Statistics. Mortality was classified based on the International Classification of Diseases, Tenth Revision (ICD-10); with cardiovascular mortality encompassing heart diseases (ICD-10 codes I00-I09, I11, I13, I20-I51) and cerebrovascular diseases (ICD-10 codes I60-I69), and cancer mortality including deaths from malignant neoplasms (ICD-10 codes C00-C97). Follow-up commenced upon completion of participation at the Mobile Examination Center (MEC) and continued until the cutoff date (December 31, 2019), and person-time calculated from serum 25(OH)D examination to either date of death or end of follow-up. The medical conditions section of the questionnaire, modeled after the U.S. National Health Interview Survey, collected cancer history, detailing type of cancer, age at diagnosis, and occurrences of any secondary cancer.

There were 36165 participants completed NHANES interview and MEC examination. After excluding participants lacking serum 25(OH)D concentrations (n = 1900), those without survival data (n = 39), and individuals who were uncertain, declined to disclose, or missing the information of their cancer diagnoses (n = 697), the study cohort comprised 26,162 participants with no history of cancer and 2,817 who had been diagnosed with the disease. Subsequently, given that various cancer stages and treatment approaches can significantly affect the life expectancy of individuals with cancer, our analysis excluded participants who succumbed to malignancy within 36 months of diagnosis (n = 17), as these cases may represent advanced-stage type or ineffective treatment that could distort overall survival outcomes. Moreover, participants who have diagnosed with a second cancer were also excluded (n = 305), as they might have worse survival expectancy that could generate bias in the analysis. The flow chat of participants enrollment is showed in Fig. 1.

In this study, serum 25(OH)D encompasses 25(OH)D2, 25(OH)D3, and epi-25(OH)D3. The quantification was performed using ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS) [18], and the units are expressed as nanomoles per liter (nmol/L). Detailed methodological procedures are available in the NHANES data documentation (https://​wwwn.​cdc.​gov/​nchs/​data/​nhanes/​2017-2018/​labmethods/​VID-J-MET-508.​pdf) and analytical note (https://​wwwn.​cdc.​gov/​nchs/​nhanes/​vitamind/​analyticalnote.​aspx).

The present study utilized various covariates obtained from enrolled participants, which included: (1) Demographic information such as age, ethnicity (classified into Non-Hispanic White, Non-Hispanic Black, Mexican American, and Other), gender (male or female), and marital status (married/living with partner or never married/separated/widowed/divorced); (2) Physical examination data: Body Mass Index (BMI), calculated as weight in kilograms over height in meters squared and stratified into normal weight (< 25), overweight (25 – 29.9), or obese (≥ 30.0) [19]; and blood collection timing, differentiated by “November 1 through April 30” and “May 1 through October 31”; (3) Lifestyle factors: smoking status (current, never, or former smoker), alcohol consumption (nondrinkers, moderate drinkers with < 2 drinks/day for males and < 1 drink/day for females, and heavy drinkers with ≥ 2 drinks/day for males and ≥ 1 drink/day for females, and not recorded) [20], dietary quality assessed by the Healthy Eating Index-2015 and classified into poor (< 50.0), needs improvement (50.0 – 79.9), or healthy (≥ 80.0) diet quality [21]; and physical activity levels expressed in metabolic equivalent (MET) minutes per week, categorized by the International Physical Activity Questionnaire [22] and the World Health Organization recommendations as low (no activity), moderate (< 600 MET-min/week), or high (≥ 600 MET-min/week) [23]; (4) Health-related information: comorbidity levels determined using self-reported questionnaires and examinations and graded using the Charlson comorbidity index (CCI) [24] (scores of 0–2, 3, or > 3); medication usage within the last 30 days (yes or no); (5) Cancer types, primarily breast, prostate, non-melanoma skin, and other, predicated on their prevalence among the cohort; and time since cancer diagnosis, categorized into ≤ 5 years, > 5 years, or not recorded.

Sample weights, strata, and primary sampling units were used to account for the complex survey design according to the NHANES analytic guidelines [25]. Participants were categorized into tertiles according to their serum 25(OH)D concentrations. Continuous variables underwent normality testing, and the characteristics of the study population were depicted as means (± SE) for continuous variables and as percentages for categorical variables. We assessed the association of continuous variables using analysis of variance (ANOVA) and compared categorical variables via the Chi-square test. Spearman’s correlation coefficients were determined for all variables, and multicollinearity was checked using the variance inflation factor (VIF), with values below 10 denoting the absence of multicollinearity. Detailed results of Spearman’s correlations and VIF for each variable can be found in the Supplementary materials.

We utilized weighted Cox proportional hazards (COXPH) models for survival analysis to explore the relationship between 25(OH)D concentrations and all-cause mortality, confirming the proportional hazards assumption with Schoenfeld residuals (Supplementary Materials). Weighted competing risks regression (CRR) models were used to estimate the association of serum 25(OH)D concentrations with cardiovascular and cancer-specific mortality. The multivariate models included adjustments for age, gender, ethnicity, marital status, and the time of blood draw in the initial model, followed by additional adjustments for body mass index (BMI), Charlson Comorbidity Index (CCI), Healthy Eating Index (HEI), medication intake, smoking status, alcohol consumption, physical activity levels, cancer types, and the time since cancer diagnosis. A test for a linear trend across the tertiles of serum 25(OH)D was conducted by designating median values to each tertile and considering it a continuous variable. Interaction and stratified analyses were conducted on the basis of gender (male and female) and ethnicity (non-Hispanic White and other ethnicities). Additionally, as part of a sensitivity analysis, serum 25(OH)D concentrations were categorized into deficiency (< 50 nmol/L), insufficiency (50—75 nmol/L), and sufficiency (> 75 nmol/L) status [26] to perform survival analyses using the aforementioned procedure.

As Fig. 1 shown, there were 2495 participants meeting the defined criteria that included in our study. We divided participants into three groups at 33% and 67% percentiles of serum 25(OH)D concentration distribution, with the values of 58.1 and 81.6 nmol/L. The characteristics of the included participants were summarized in Table 1. The mean age of participants was 62.10 ± 0.38 years. A majority of 57.4% were female, and 87.8% identified as Non-Hispanic White, followed by 5.0% as Non-Hispanic Black, 2.1% as Mexican American, and 5.1% as Other Races. Regarding 25(OH)D concentrations, 20.8% (n = 518) of the participants were classified as vitamin D deficient (< 50 nmol/L), and 36.5% (n = 910) as insufficient (50—75 nmol/L) [27]. Analysis of participant characteristics by tertiles of 25(OH)D concentrations revealed that those in the highest tertile had the oldest average age (63.81 ± 0.63 years), the highest percentages of females (61.4%) and Non-Hispanic Whites (93.7%), and the greatest prevalence of a CCI score of 3 (36.6%). Conversely, these individuals had the lowest prevalence of obesity (26.6%), poor dietary quality (26.1%) as per the HEI-2015 index, in current smoking status (13.1%), and physical inactivity (17.7%). This group also exhibited the highest proportion of non-melanoma skin cancer (27.5%) and breast cancer (14.8%), yet the lowest proportion of prostate cancer (8.0%) and other types of cancer (49.7%). Participants in the middle tertile showed the highest proportion of males (47.1%) and those married or living with a partner (66.9%), but the lowest percentage with a CCI score greater than 3 (26.1%). This group also exhibited the highest proportion of prostate cancer (10.2%). Among those studied, 78.0% (n = 1,947) reported their age at cancer diagnosis, with an average interval between diagnosis and the end of follow-up of 230.08 ± 3.94 months. Participants who did not report their age at diagnosis had an average post-survey follow-up of 92.3 ± 2.10 months. Over 244,816 person-years of follow-up, a survival rate of 77.1% was observed, with 788 deaths comprising 348 (9.9%) from cardiovascular causes, 216 (6.3%) from malignancies, and 224 (6.7%) from other causes.

As Table 2 shown, a significant correlation was found between serum 25(OH)D levels and the risk of all-cause, cardiovascular, and cancer-specific mortality among participants with a cancer diagnosis. This association persisted after adjusting for multiple covariates. Specifically, for all-cause mortality, the hazard ratios (HRs) and 95% confidence intervals (CIs) for intermediate (58.1 – 81.6 nmol/L) and high (> 81.6 nmol/L) 25(OH)D concentrations were 0.85 (95% CI: 0.68 – 1.05) and 0.70 (95% CI: 0.56—0.87), respectively, as compared to low (< 58.1 nmol/L) concentrations, according to COXPH models. For cardiovascular mortality, the HRs for intermediate and high levels were 0.83 (0.54- 1.30) and 0.53 (0.32—0.86) respectively. In the context of cancer-specific mortality, the HRs for intermediate and high levels were 0.67 (95% CI: 0.40- 1.14) and 0.66 (95% CI: 0.45—0.99) respectively, as analyzed by CRR models. All models, except for those predicting cancer-specific mortality, demonstrated a linear trend between tertiles of serum 25(OH)D concentrations and HRs.

Additionally, when serum 25(OH)D concentration was categorized into deficiency, insufficiency, and sufficiency, it was not found an obviously change on the result that highest risk was still confirmed in the lowest serum 25(OH)D concentration category. As presented in Table 3, deficient serum 25(OH)D concentrations were associated with increased HRs of 1.71 (95% CI: 1.32—2.21), 1.83 (95% CI: 1.14—2.95), and 2.02 (95% CI: 1.26—3.22) for all-cause, cardiovascular and cancer-specific mortality, respectively, in comparison to sufficiency levels.

In the subgroup analysis, HRs for cause-specific mortality still remained similar trend except some variations. According to Table 4, higher serum 25(OH)D concentrations were notably more protective against cause-specific mortality in males than in females. This distinction was particularly marked in the context of cancer-specific mortality, with a discernible interaction effect between serum 25(OH)D concentrations and gender. Male participants with serum 25(OH)D concentrations ranging from 58.1 to 81.6 nmol/L experienced a hazard ratio of 0.28 (95% CI: 0.14–0.55), whereas female participants did not show a significant difference at these levels in comparison to those with serum 25(OH)D concentrations below 58.1 nmol/L. Regarding the ethnic subgroup, higher serum 25(OH)D concentrations corresponded with a greater reduction in cardiovascular mortality among non-Hispanic Whites relative to other ethnicities, characterized by a significant decrease in HRs in accordance with increasing serum 25(OH)D concentrations, and this trend was not observed in other ethnicities subgroup. While other outcomes did not show significant different in HR among ethnicity subgroups.

The RCS models were employed to elucidate the dose–response relationship between serum 25(OH)D concentrations and cause-specific mortality. As illustrated in Fig. 2, significant associations were observed between serum 25(OH)D concentrations and all-cause, cardiovascular, and cancer-specific mortality across all models, HRs and 95% CIs for all-causes, cardiovascular, and cancer-specific mortalities decreased progressively with increasing serum 25(OH)D concentrations up to 87.9 nmol/L, 88.7 nmol/L, and 84.6 nmol/L, respectively, beyond which the HRs plateaued. The RCS curves demonstrated an “L-shaped” configuration for all-cause and cancer-specific mortalities, in contrast, cardiovascular mortality did not exhibit a non-linear relationship with serum 25(OH)D concentrations.