National Health and Nutrition Examination Survey (NHANES) is a nationally representative cross-sectional survey that utilizes a stratified multistage sampling design to evaluate the health and nutritional status of the non-institutionalized population in the United States (20). The research protocol was approved by the Institutional Review Board of the National Center for Health Statistics (NCHS), and written informed consent was obtained from all participants. Since the de-identified data analyzed during the current study were publicly available from NHANES, the study did not require any review board approval again.

This study included a total of 50,201 participants aged ≥20 years from 2001 to 2018. We excluded non-long-term prescription opioid users (n = 47,426), individuals receiving medications for opioid dependence or withdrawal (including buprenorphine; naloxone, buprenorphine; methadone, n = 79), those with missing vitamin D data (n = 331), those with missing follow-up data (n = 1), and those with missing covariate data (n = 508). Ultimately, the analysis included 1856 participants. The participant inclusion process was detailed in Supplementary Figure S1.

Before 2005–2006, serum 25(OH)D concentrations were measured using the DiaSorin RIA kit (Stillwater, MN, USA). Since 2005–2006, a standardized liquid chromatography–tandem mass spectrometry (LC–MS/MS) method had been used to measure serum 25(OH)D concentrations. Following the analytical guidelines of Centers for Disease Control and Prevention (CDC), serum 25(OH)D concentrations from 2005 to 2006 and earlier were converted to equivalent concentrations from LC–MS/MS using regression, and LC–MS/MS equivalent data were utilized in all analyses.

During the household interview survey, participants were asked if they had taken a medication in the past month for which they needed a prescription. Participants who answered “yes” were asked to report the names of the medications through medication containers, verbal reports, or pharmacy receipts. Additionally, participants were also asked how long they had been taking the medication. All recorded medications were classified using the Multum Lexicon therapeutic classification system. Participants who reported the use of narcotic analgesics or narcotic analgesic combinations were considered as prescription opioid users, and the duration of prescription opioid use ≥90 days was considered as long-term use (21). Methadone, naloxone, and buprenorphine were often used to treat opioid dependence or withdrawal, we exclude participants taking these drugs from prescription opioid users.

Mortality status and follow-up time were determined by matching with the National Death Index (NDI) records available until December 31, 2019. According to the International Classification of Diseases, 10th Revision (ICD-10), cardiovascular disease-specific mortality was defined as deaths due to heart disease (I00-I09, I11, I13, I20-I51) or cerebrovascular disease (I60-I6), while cancer-specific mortality was defined as deaths due to malignant neoplasms (C00-C97).

Sociodemographic information [age, sex, race/ethnicity, education, poverty-income ratio (PIR)], living habits (physical activity, cotinine, alcohol consumption), comorbidity (history of hypertension, cardiovascular disease (CVD), diabetes, and cancer), as well as body mass index (BMI), were obtained from NHANES. Race/ethnicity was categorized as Non-Hispanic White, Non-Hispanic Black, Mexican American, and Other. Education was divided into less than high school, high school, and more than high school. PIR was classified as ≤130, 130–300%, and > 300% (22). Physical activity was calculated as weekly minutes of moderate and vigorous activity multiplied by metabolic equivalent (MET) levels. Alcohol consumption was categorized as never (<12 drinks in a lifetime), former (≥12 drinks in a year or ≥ 12 drinks in a lifetime but none in the past year), mild (female: <2 drinks/d, male: <3 drinks/d), moderate (female: 2 to <3 drinks/d, male: 3 to <4 drinks/d, or 2–4 binges/month), and heavy (female: ≥3 drinks/d, male: ≥4 drinks/d, or ≥ 5 binges/month) (23). BMI was calculated as weight (kilograms) divided by the square of height (meters) and classified as <25, 25–30, and ≥ 30 (24). Hypertension was assessed by systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 80 mmHg, or taking antihypertensive drugs (25). Diabetes was determined by self-reported diagnosis by doctor or glycosylated hemoglobin (HbA1c) level ≥ 6.5% or taking anti-diabetic drugs. CVD was confirmed through self-reported diseases, including coronary heart disease, congestive heart failure, heart attack, stroke, and angina.

All analyses were conducted using appropriate sample weights to account for the complex sampling design of NHANES. Serum 25(OH)D concentrations were classified into four groups: severe deficiency (<25.00 nmoL/L), deficiency (25.00 to <50.00 nmoL/L), insufficiency (50.00 to <75.00 nmoL/L), and sufficiency (≥75.00 nmoL/L) (26). The reference group was defined as those with serum 25(OH)D concentrations <25.00 nmoL/L. Continuous variables and categorical variables were presented as means (standard errors, SE) and frequencies (weighted percentages), respectively. Differences between the four 25(OH)D groups were assessed using ANOVA for continuous variables and chi-square test for categorical variables. Multivariable Cox proportional hazards regression models were created to investigate the association between serum 25(OH)D concentrations and risks of all-cause or cause-specific mortality. Model 1 made no adjustments; model 2 was adjusted for age, sex, race/ethnicity; model 3 was further adjusted for education, alcohol consumption, cotinine, physical activity, BMI, PIR, hypertension, diabetes, cancer, and CVD. Restricted cubic spline analysis (RCS) with 5 knots (5th, 28th, 50th, 73th, and 95th) was applied to examine the non-linear relationship between serum 25(OH)D concentrations and all-cause mortality. Stratified analyses were further performed by age (<60 or ≥ 60 years), sex (male or female), race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Mexican American, and Other), BMI (< 25, 25–30, and ≥ 30), and cancer (yes or no). We also conducted several sensitivity analyses: (1) Participants who died within 2 years of follow-up were excluded to minimize reverse causality; (2) Participants with a history of cardiovascular disease or cancer were further excluded; (3) Additionally, adjustments were made for the blood drawing season to account for the influence of seasonal variations on vitamin D status.

All analyses were conducted using R 4.2.3 (Boston, MA, USA), and a two-tailed p-value <0.05 was considered statistically significant.

Of the 1856 participants with long-term prescription opioid use (mean [SE] age, 54.12 (0.5) years; 808 male [weighted, 40.7%]; 1,060 [weighted, 78.5%] non-Hispanic White; mean [SE] serum 25(OH)D concentrations, 70.19 (1.2) nmol/l), 32.4% were with deficient vitamin D (<50 nmoL/L), and 67.2% were with insufficient vitamin D (<75.00 nmoL/L). Table 1 shows the baseline characteristics of the study participants stratified by serum 25(OH)D concentrations. Compared to sufficient vitamin D (≥75.00 nmoL/L), the age of participants with severe deficient vitamin D (<25.00 nmoL/L) was younger. Severe deficient vitamin D (<25.00 nmoL/L) was more likely to occur in participants with of female, Non-Hispanic Black and Mexican American, with less physical activity, with obesity or history of CVD and diabetes. In addition, participants with severe deficient vitamin D were less likely to have higher PIR (>300%), have history of cancer, and have mild alcohol consumption.

During a median follow-up of 7.75 years, we identified 443 all-cause deaths (weighted, 18.7%), including 135 CVD deaths (weighted, 4.9%), and 94 cancer deaths (weighted, 3.8%).

Table 2 displays the results of Cox proportional hazards regression analyses between serum 25(OH)D concentrations and mortality. After multivariable adjustment (model3), compared to participants with serum 25(OH)D concentrations <25.00 nmol/L, the hazard ratio (HR) for all-cause mortality for those with serum 25(OH)D concentrations ranging from 50.00 to <75.00 nmol/L was 0.50 (95% confidence interval [CI] 0.29, 0.86), for those with serum 25(OH)D concentrations ≥75 nmol/L was 0.54 (95% CI 0.32, 0.90). Nevertheless, no significant inverse associations were observed for serum 25(OH)D concentrations with CVD-specific mortality and cancer-specific mortality after multivariable adjustment. It is important to note that there was no linear association between serum 25(OH)D concentrations and all-cause mortality (p for trend = 0.19).

We used the RCS curve to estimate the dose–response relationship between serum 25(OH)D concentrations and all-cause mortality, and a non-linear association was observed after adjusting all confounders (p for non-linear = 0.01) (Figure 1). The inflection points for serum 25(OH)D concentrations were 62.17 nmoL/L and 94.52 nmol/L. Subsequently, piecewise COX proportional hazards regression was performed to analyze a threshold effect of serum 25(OH)D concentrations on all-cause mortality (Table 3). When serum 25(OH)D concentrations were < 62.17 nmoL/L, per 1-unit increment in serum 25(OH)D concentrations, there was a 2% lower risk of all-cause mortality (HR 0.98, 95%CI [0.97, 1.00]). However, there was no benefit effect for lower risk of all-cause mortality when elevate serum 25(OH)D concentrations for participants with serum 25(OH)D concentrations were 62.17- <94.52 nmol/L (HR 1.01, 95%CI [0.98, 1.03]) or ≥ 94.52 nmol/L (HR 0.97, 95%CI [0.93, 1.02]).

In stratified analyses, the benefit of high serum 25(OH)D concentrations (≥ 62.17 nmol/L) and low serum 25(OH)D concentrations (< 62.17 nmol/L) for the survival rate of participants with long-term prescription opioid use stratified by age, sex, race/ethnicity, BMI, and cancer was similar. No modification factors were found between serum 25(OH)D concentrations and these stratified variables (all p for interaction >0.05) (Figure 2). In sensitivity analyses, the results were also typically robust when excluding deaths that occurred within 2 years (Supplementary Table S1); excluding participants with history of CVD or cancer (Supplementary Table S2); further adjustment of blood drawing season (Supplementary Table S3).