Source of data
National Health and Nutrition Examination Surveys (NHANES) are national cross-sectional surveys providing health and nutrition information on a representative sample of non-institutionalized U.S. residents of all ages that is generalizable to the entire U.S. population [29] . For the current study, we used data collected from NHANES 2005-2006 because it provided data relevant not only to arthritis and CVD, but also to serum measurement of 25(OH)D. Participants signed written informed consent. Data were collected with health questionnaires, physical examination and blood tests [29]. The questionnaires were standardized and filled out during home interviews, while the physical examination and the blood collection were done in specifically equipped mobile centers travelling when needed to the participant’s home [29].
Sample selection
Participants included were aged between 20 and 69 years. Exclusion criteria were mainly linked to the blood samples: 1) hemophilia, 2) received chemotherapy in the last 4 weeks, 3) presence of the following on two arms: rashes, gauze dressings, casts, edema, paralysis, open wounds, missing limbs, sclerotic veins, allergies to cleansing agents, burns or scar tissues. Besides the NHANES exclusion criteria, we excluded pregnant women and participants with incomplete physical exam or missing data related to vitamin D status.
Figure 1 shows the detailed participants’ selection: 10,348 individuals participated to NHANES 2005-2006, of whom 382 were pregnant women, 5323 participants were younger than 20 years, 882 were older than 70 years, 140 (1.4%) had not completed the physical examination and 215 (2.1%) had vitamin D data missing. The final sample size available for the current study was 3406 (90.6% of those eligible).
Main independent and dependent variables
Participants reported their health conditions through the NHANES 2005-2006 “Medical Conditions Questionnaire”. The main independent variable was the presence of arthritis; participants answered if they had been diagnosed as having arthritis, which included several diagnoses such as osteoarthritis, rheumatoid arthritis, spondylarthritis, psoriatic arthritis, etc. If the answer was “Yes” to the question: “Has a doctor or other health professional ever told you that you had arthritis?”, then we considered them as having arthritis.
The dependent variable was the diagnosis of CVD; the questions were: “Has a doctor or other health professional ever told you that you had -congestive heart failure?”; “-coronary heart disease?”; “-angina pectoris?”; “-a heart attack?”; “-a stroke?” A positive answer to any of these questions was considered as a positive CVD diagnosis. This is the approach that was used in NHANES to estimate the prevalence of CVD [30].
Other independent variables
Several variables described participants’ sociodemographic, lifestyle and physical characteristics (Table 1). Some were considered as potential confounders of the association between arthritis and CVD, based on biological plausibility and the literature [8, 16, 31-34]. Covariates included in analyses were age, sex, ethnicity, body mass index (BMI), total blood cholesterol concentration, 25(OH) D serum concentration and monthly dose of vitamin D supplementation. In addition, we included the administration of glucocorticoids and anticonvulsants as potential confounders. In fact, some drugs (anti-convulsants, corticosteroids) adversely affect the vitamin D absorption [35,36]. BMI was calculated as weight in kilograms divided by the square of height in meters, and divided into 4 categories: underweight (BMI<18.5), normal weight (18.5<=BMI<24.9), overweight (24.9<=BMI<30), and obese (BMI >=30). Total serum cholesterol (mg/dL) was considered as a continuous variable in analyses.
Details concerning the administration of glucocorticoids and anticonvulsants were extracted from NHANES 2005-2006 by matching two drug files (“Dietary Supplements” and “Prescription Medications”) into a computer where an automatic match to a prescription drug database was performed [37].
Laboratory data provided measurements of serum concentrations of 25(OH)D for participants. The Diasorin 25-Hydroxyvitamin D (25(OH)D) assay, an accurate kit measuring serum concentrations of 25(OH)D [38], was performed to first extract 25(OH)D metabolites from serum, and then separate and measure them [39]. NHANES uses different procedures to control the quality of the analyses performed by the laboratories [40]. Serum concentrations of 25(OH)D were dichotomized at >20 ng/ml [17-20]. Finally, information on the participants’ vitamin D supplementation and its monthly dose were taken from several dietary data files available in the NHANES 2005-2006 database, that were merged.
Statistical analyses
SAS software (version 9.3) was used to perform statistical analyses. Sample weights were used to produce an unbiased national estimate of the civilian noninstitutionalized U.S. population[41]. Comparisons between arthritis and non-arthritis groups were conducted using chi-squared tests for categorical variables and Student’s t test for continuous variables. To visualize bivariate associations of CVD with each of the independent variables, chi-squared tests were performed and the prevalence of CVD was calculated for each variable category. Multiple logistic regression was used to assess the association between arthritis and CVD. Age, total cholesterol and monthly dose of vitamin D supplementation were treated as continuous variables, while other variables were left categorical. Prevalence odds ratios (OR) with 95% confidence intervals (95% CI) were calculated. Age, sex, ethnicity, BMI, total cholesterol, the administration of glucocorticoids and anticonvulsants, 25(OH)D serum concentration of vitamin D (<20 ng/ml vs >20 ng/ml), and vitamin D supplementation use were considered as potential confounding factors. Confounding was defined as a change in the OR ≥ 10% brought about by taking off one variable at a time compared to the full model. Effect modification was tested on a full model that included three double interaction terms (sex*arthritis, vitamin D*arthritis and sex*vitamin D) and a triple interaction term (arthritis*sex*vitamin D). This is because the main interest in this study were possible modifying effects of sex and vitamin D serum concentrations on the association between arthritis and CVD. To measure such effects, ratios of ORs (ROR) were calculated with their 95% confidence intervals; here, a ROR of 1.00 means no effect modification.