Methods

We used publically available data from the continuous National Health and Nutrition Examination Survey (NHANES), an ongoing, multistage, probability sample survey designed to assess the health and nutritional status of the civilian, noninstitutionalized population of the United States. Detailed interviews, physical examinations, and serum samples were obtained from >17,000 subjects in the survey conducted from 2001 to 2006. Details of sampling procedures and data collection techniques have been described previously and are available online (available at: http://www.cdc.gov/nchs/nhanes/about_nhanes.htm , accessed March 10, 2011). To create a large nationally representative sample, the data from three 2-year cycles of continuous NHANES survey were combined from 2001 to 2006. Sample weights were constructed, with rescaling of weights such that sum of weights matched the survey population at the midpoint of each survey period.

The demographic information was ascertained from self-reported responses to the questionnaire administered by trained interviewers. The body mass index was calculated by dividing the body weight in kilograms by the height in meters squared. Obesity was defined as a body mass index of ≥30 kg/m ² and overweight if the body mass index was ≥25 kg/m ² but <30 kg/m ² . Blood pressure was recorded for ≤4 readings. Hypertension was defined as a mean systolic blood pressure of ≥140 mm Hg, mean diastolic blood pressure of ≥90 mm Hg, a diagnosis of hypertension, or the current use of antihypertensive medications. The glomerular filtration rate was calculated using the Modification of Diet in Renal Disease formula. The serum glucose level was measured using the Beckman Synchron LX20 test (Beckman Coulter, Brea, CA) on refrigerated specimens, and the cholesterol level was measured enzymatically in a series of coupled reactions that hydrolyze cholesteryl esters and oxidize the 3-OH group of cholesterol. The participants were categorized as smokers if they were currently smoking or had ever smoked >100 cigarettes. The Beckman Coulter method was used to determine the white blood cell count combined with automatic diluting and mixing device for sample processing. The DiaSorin (formerly Incstar) radioimmunoassay (Nutritional Biochemistry Branch, Division of Laboratory Sciences, National Center for Environmental Health, Atlanta, Georgia), a 2-step procedure, was used to assay 25(OH)D. The first procedure involved extraction of 25(OH)D and other hydroxylated metabolites from serum with acetonitrile. Next, the treated sample was assayed using an equilibrium radioimmunoassay procedure. Updated and adjusted data files were used for 25(OH)D to address assay drift. CRP was quantified using latex-enhanced nephelometry and a Behring Nephelometer II Analyzer (Immunology Division, Department of Laboratory Medicine, University of Washington Medical Center, Seattle, Washington). CRP was log-transformed to meet the assumptions of residual normality.

All analyses were performed with adjustments for the complex survey sampling method of the NHANES data. In continuous NHANES, primary sampling units represent variance (sampling units used to estimate sampling error) units. These sampling weights were assigned to each subject, reflecting adjustment for the unequal probability of selection and nonresponse and adjustments to independent population controls. The participants were oversampled for certain population subgroups, especially the subgroups of population such as black and Mexican Americans, to ensure reliability and precision of the health estimates indicators in these subgroups. Masked variance units were constructed to protect the confidentiality of data obtained from sample participants. Masked variance units, used to define the strata, were created for each 2-year cycle of the continuous NHANES data, allowing any combination of data cycles without recoding by the users. The sample weights were constructed for 6 years of combined survey data.

To address the hypothesis that the favorable association between vitamin D and serum CRP is present only at lower serum vitamin D level, we introduced a spline in the regression models, with single knot at the population median of 25(OH)D (i.e., 21 ng/ml), hypothesizing that this association exists up to the population median but not above it. Analyses were performed using Stata/IC, version 10.1 (StataCorp, College Station, Texas) using survey-specific commands. Adjusted coefficients and their 95% confidence intervals (CIs) were estimated using univariate and multivariate linear regression models. The analysis was limited to subjects >18 years old and adjusted for demographic variables, obesity, hypertension, serum glucose, cholesterol, smoking, white blood cell count, and renal function. A p value of <0.05 was considered statistically significant.