Nutrient intake differs among people with celiac disease and gluten-related disorders in the United States
The NHANES is a series of nationally representative cross-sectional surveys of the U.S. population with oversampling of non-Hispanic blacks, Hispanics, Asians (2011-2014), low-income whites, and older adults. 80 years or older that are conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention in 2-year cycles13. The survey includes household interviews, physical examinations and laboratory data collected using a complex multistage, stratified and clustered probability sampling design. We analyzed survey data collected from 2009 to 2014, years in which gluten-related disorders were assessed by interview and serology. All interviews and examinations were conducted in accordance with applicable guidelines and regulations.
CD definitions and gluten avoidance without CD
Participants in the 2009-2014 NHANES were asked the following questions: (1) Has a doctor or other healthcare professional ever told you that you have celiac disease, also known as sprue? and (2) Do you follow a gluten-free diet?14 Serum samples were shipped to the Celiac Disease Research Laboratory at Mayo Clinic, Rochester, MN, for serological testing. Serum was tested for tissue transglutaminase immunoglobulin A (tTG IgA) as a screening test (sensitivity, ~98%) with an enzyme immunoassay that uses recombinant human tTG (Inova Diagnostics, San Diego, CA, USA), and results were categorized as positive (>10 U/ml), weak positive (4–10 U/ml) or negative (4.15.
Individuals were identified as having diagnosed CD if they had a clinical diagnosis (provider) and positive (or weakly positive) serology or adherence to GFD. Similarly, people with undiagnosed CD had positive (or weakly positive) serology without a clinical diagnosis. PWAG was defined as adherence to GFD without diagnosis of CD and with negative serology. Individuals without a CD and without a GFD are referred to herein as “controls”. These case definitions are shown in Fig. 1 for reference. Contrary to previous reports, we included all 2009-2014 NHANES participants with positive IgA tTG alone, rather than serially positive IgA tTG and endomysial IgA antibodies, to increase the number of cases due to low prevalence of CD in the general population2.3.
Nutrient intake estimates
Dietary intakes were obtained from two 24-hour dietary recall interviews in which respondents reported all foods and beverages consumed during the previous 24 hours (midnight to midnight). The first dietary recall interview was collected in person at the mobile examination center visit and the second interview was collected by telephone 3 to 10 days later. Detailed descriptions of food maintenance methods and protocol are fully documented16. Daily nutrient and food energy intakes were estimated using the respective Foods and Nutrients Database for Dietary Studies for participants in each NHANES cycle17.
Information on dietary supplement use was collected by both a 30-day frequency questionnaire and two 24-hour dietary recall interviews. For this analysis, we used additional data from 24-h recalls. Participants were asked about the 24-hour period preceding the interview (from midnight to midnight). People who reported using prescription or over-the-counter dietary supplements as well as over-the-counter antacids containing calcium and magnesium were asked to show the interviewer the bottles of each supplement product taken. The amount and duration of use were then queried during the interview. Daily nutrient intakes from dietary supplements and antacids were estimated using the NHANES dietary supplement database. More information is provided in the NHANES documentation18.
Data were collected on demographic and clinical characteristics and examined in relation to gluten-related disorders19.20: age (years), sex, race-ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, other), education, income and BMI. The highest year of school completed was reported and categorized as less than a high school diploma, high school diploma, or education beyond high school. Income was measured by the poverty income ratio (ratio of family income to the poverty line) and classified into tertiles (2 normal weight (3. Latitude (° North) was geocoded by the NCHS from the participant’s residential address and categorized as 21. Geographic data was used by the NCHS Research Data Center22. Serum was tested for hemoglobin, standard biochemistry, lipid profile, and selected vitamins and minerals as previously described23.
Of 25,426 2009-2014 NHANES participants aged 20 years and older, 16,966 (67%) attended a study visit to a mobile examination center and were included. A total of 15,610 (92%) participants completed at least one of two 24-h dietary recall interviews, and 13,825 participants (81% of those examined) provided two reliable dietary recalls. We included in the analyzes all people who performed at least one reliable 24-h dietary recall (Fig. 1). People with missing data on CD and PWAG status were excluded.
Characteristics of participants according to gluten-related disorders were compared using a chi-square (χ2) test and linear regression. Unadjusted means and standard errors of serum biochemical and nutritional markers were calculated as a function of CD and PWAG status using linear regression.
We estimated the distribution of usual nutrient intakes using methods developed by the National Cancer Institute (NCI)24. The NCI method is useful for estimating within- and between-individual variances and correcting for the large intra-individual daily variation commonly seen in 24-h recalls. A two-step process is involved in the NCI method. The first step uses the MIXTRAN macro to fit a nonlinear mixed effects model using the SAS procedure NLMIXED to obtain parameter estimates of average usual intake. Quantity data were transformed to approximate normality using the Box-Cox transformation. In the second step of the NCI method, the macro DISTRIB uses parameter estimates from MIXTRAN and a Monte Carlo method to estimate the usual intake distribution of a nutrient24. To account for the complex sampling design of the survey, variance estimation was performed by the balanced repeated replicate technique with Fay’s modification25. Since micronutrient intake can come from both diet and supplements, these two amounts were first summed for each participant’s recall day before applying the NCI method. Mean nutrient intakes were estimated unadjusted, adjusting for age, gender, and ethnicity, as well as education, income, and BMI. First-day dietary recall weights provided by NHANES were used in all analyzes of nutrient intake.
The relationship between nutrients and gluten-related disorders was further examined by latitude using linear regression analysis (SUDAAN PROC REGRESS) to calculate mean (least squares) adjusted estimates for l age, gender and ethnicity. For these analyses, the intra-individual mean (2-day mean) of both 24-h food and supplement recalls was used because it was not possible to use NCI methods through the NCHS Research Data Center.
Multivariate adjusted analyzes excluded people with missing values for any factor included in the model. P-values were two-sided, and one P-a value 26. SAS 9.4 (SAS Institute, Cary, NC, USA) and SUDAAN 11 (RTI, Research Triangle Park, NC, USA) were used for all analyses.
Ethical approval and consent to participate
The NCHS Research Ethics Board approved the survey and all participants provided written informed consent.