Data from the 2003–2006 cycles of the National Health and Nutrition Examination Survey (NHANES) were used (http://www.cdc.gov/nchs/nhanes.htm). During NHANES 2003–2006, 4,538 participants aged ≥20 years provided data on the study variables. All study procedures were approved by the National Center for Health Statistics review board; all participants provided informed consent.

RDW was derived from the coefficient of variation of the NHANES RDW histogram and recorded as a percentage.

Accelerometers were initialized to collect data in 1-minute epochs. SB was measured during waking hours and was assessed here among participants with ≥10 hours/day of monitoring on at least 4 days (in a 1-week period) using the ActiGraph 7164 accelerometer (ActiGraph LLC, Pensacola, Florida). An activity level of less than 100 counts per minute was classified as SB (6), and to control for physical activity, counts/minute greater than or equal to 2,020 were classified as moderate-to-vigorous physical activity. Nonwear was defined as a period of at least 60 consecutive minutes of zero activity counts, with the allowance of 1–2 minutes of activity counts between 0 and 100 (5).

Covariates included age, sex, race/ethnicity, poverty-to-income ratio, body fat percentage (as determined by dual-energy x-ray absorptiometry (DEXA)), cotinine level, history of diabetes, history of cancer, history of coronary artery disease, mean arterial blood pressure, estimated glomerular filtration rate, high-density lipoprotein cholesterol level, diet, accelerometer-determined physical activity, accelerometer wear time, serum iron level, and C-reactive protein level.

Total body fat percentage was estimated from whole-body DEXA scans (Hologic, Inc., Bedford, Massachusetts). Serum cotinine concentration was measured by means of isotope dilution–high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometry. Blood pressure was measured up to 4 times, and the average mean arterial pressure ([diastolic blood pressure × 2 + systolic blood pressure]/3) was calculated. Iron concentration was measured from a blood sample using the LX20 Pro Chemistry Analyzer (Beckman Coulter Inc., Brea, California), with high-density lipoprotein cholesterol measured enzymatically directly in serum. Dietary behavior was assessed using the 2005 Healthy Eating Index (6–8). Glomerular filtration rate was assessed from the Chronic Kidney Disease Epidemiology equation (9). High-sensitivity C-reactive protein concentration was quantified using latex-enhanced nephelometry.

Analyses were performed using survey data procedures in Stata (StataCorp LP, College Station, Texas) to adjust for the complex survey design. Multivariable logistic regression was used to examine the association between 5 SB quintiles (independent variable) and elevated RDW (≥75th percentile; ≥13.1%). Four logistic regression models were fitted, including an unadjusted model (model 1), a minimally adjusted model (model 2), a further adjusted model (model 3), and a fully adjusted model (model 4). The minimally adjusted model (model 2) controlled for age, sex, race/ethnicity, poverty level, and DEXA-determined body fat percentage. The further adjusted model (model 3) controlled for the same covariates as model 2 plus cotinine, diabetes, cancer, coronary artery disease, mean arterial blood pressure, estimated glomerular filtration rate, high-density lipoprotein cholesterol, diet (2005 Healthy Eating Index), accelerometer-determined physical activity, and accelerometer wear time. The fully adjusted model (model 4) controlled for the same covariates as model 3 plus serum iron and C-reactive protein. Statistical significance was established as a 2-sided P value less than 0.05.