Consumption of whole grains and cereal fiber and total and cause-specific mortality: prospective analysis of 367,442 individuals

The NIH-AARP Diet and Health Study included 566,399 AARP members aged 50 to 71 from six US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and two metropolitan areas (Atlanta, Georgia, and Detroit, Michigan) [10]. Participants responded to a questionnaire mailed in October 1995 and December 1997. Details of the NIH-AARP Study have been previously described [11]. Among participants who returned the questionnaires with satisfactory dietary data, we excluded individuals who indicated that they were proxies for the intended respondent (n = 15,760) as well as those who had cancer (n = 50,591), heart disease (n = 80,254), stroke (n = 12,812), diabetes (n = 52,647), or self-reported end-stage renal disease at baseline (n = 1,299). We also excluded those who reported extreme consumption (>2 times the interquartile ranges of Box-Cox transformed intake) of total energy (n = 3,771) and dietary fiber (n = 3,324). Exclusion of individuals reporting extreme energy intake is widely used in nutritional epidemiology studies since these participants are more likely to over- or under-report their intake [12]. After exclusions (n = 198,957), the analytic cohort included 367,442 individuals. The NIH-AARP Diet and Health study was approved by the Special Studies Institutional Review Board of the US National Cancer Institute. All participants provided written informed consent.

At baseline, dietary intake was assessed with a self-administered 124-item food frequency questionnaire (FFQ), which was an early version of the Diet History Questionnaire developed at the National Cancer Institute [13,14]. Participants were asked to report their usual frequency of intake and portion size over the past 12 months using 10 predefined frequency categories ranging from never to 6+ times per day for beverages and from never to 2+ times per day for solid foods with three portion size categories. The food items, portion sizes, and nutrient database were constructed using the US Department of Agriculture’s 1994–1996 Continuing Survey of Food Intakes by Individuals. The FFQ used in the study was calibrated using two non-consecutive 24-hour dietary recalls in 1953 NIH-AARP study participants. The nutrient database for dietary fiber was based on AOAC methods.

The whole grains were defined as the whole grain part of each product. The US Department of Agriculutre’s Pyramid Servings Database enabled us to accurately estimate whole-grain intake from all foods in the FFQ. The sources of whole-grain intake in the FFQ used in our study were ready-to-eat cereals, high-fiber cereals, other fiber cereals, whole-grain breads or dinner rolls, cooked cereal, popcorn, pancakes, waffles, French toast or crepes, rice or other cooked grains, bagels, English muffins, tortillas, pasta, crackers, chips, cookies or brownies, sweet pastries, and pies. In this Continuing Survey of Food Intakes by Individuals dataset., whole grain foods were defined as those containing at least 25% whole grains and/or bran. Main fibers are from fruit, grains, vegetables, and beans in the present study. Cereal fiber was defined as fiber from all cereals (e.g., ready-to-eat cereals, high-fiber cereals, cooked cereal, and other fiber cereals) and grain-based products.

In specifying numbers of deaths by intake quintile, the number of deaths is determined by the energy-adjusted intake quintile for the entire population; when deaths are specific to a sex, we used the quintiles within the sex. We also collected demographic, anthropometric, and lifestyle information, including history of smoking (the number of cigarettes smoked per day), time of smoking cessation (<1 years, 1 to 5 years, 5 to 10 years, or ≥10 years before baseline), physical activity (never, rare, 1 to 2, 3 to 4, ≥5 hours/week), alcohol intake (g/day) family history of cancers, menopausal hormone therapy use in women, and some medical conditions at baseline.

The AARP dataset denotes date of death and cause of death. There are 22 broad categories for cause of death. The modeling analysis for specific cause of death is performed with the study end date of 2008. For total mortality, models for study end dates in 2008 and 2009 were designed. Subjects with date of death after the study’s end date are treated as alive at the end of the study, with no death or cause of death in the model. When the study end date was 2008, and there was a date of death but no cause of death for 2008 or earlier, the subject was not included in the modeling for cause of death, but only total mortality. Thus, when specifying numbers of deaths it depends on both the study end date and whether the cause of death field is missing. Vital status was determined through a periodic linkage of the cohort to the Social Security Administration Death Master File and follow-up searches of the National Death Index Plus for participants who matched the Social Security Administration Death Master File, cancer registry linkage, questionnaire responses, and responses to other mailings. The International Classification of Diseases, Ninth Revision [15] and the International Statistical Classification of Diseases, 10th Revision [16] were used to define death as follows: cancer (ICD-9, 140–239; ICD-10, C00–C97 and D00–D48), CVD (ICD-9, 390–398, 401–404, 410–429, and 440–448; ICD-10, I00–I13, I20–I51, and I70–I78), diabetes (ICD-9, 250; ICD-10, E10–E14), respiratory disease (ICD-9, 480–487 and 490–496; ICD-10, J10–J18 and J40–J47), infections (ICD-9, 001–139; ICD-10, A00–B99), and all other/unknown causes.

We used the Cox proportional hazards model to estimate hazard ratios (HRs) and two-sided 95% confidence intervals (CIs) using the SAS PROC PHREG procedure (Version 9.1; SAS Institute Inc., Cary, NC, USA). Person-years of follow-up were calculated from the date of the baseline questionnaire until the date of death or the end of follow-up (December 31, 2009), whichever occurred first. Intake of whole grains and cereal fiber were adjusted for total energy intake using the residual method [17], and were categorized into quintiles.

We presented the results from four analysis models. Model 1, adjusted for age and gender; Model 2, adjusted for age, gender, the number of cigarettes smoked per day, and time of smoking cessation (<1 years, 1 to 5 years, 5 to 10 years, or ≥10 years before baseline); Model 3, adjusted for age, gender, the number of cigarettes smoked per day, time of smoking cessation (<1 years, 1 to 5 years, 5 to 10 years, or ≥10 years before baseline), race or ethnicity group, alcohol intake, education level, marital status (yes, no), health status (poor, fair, good, very good), obesity (underweight, overweight, obesity), physical activity, consumption of red meat (processed and fresh meat), total fruit and total vegetables, total energy intake, and hormone usage; and Model 4, based on Model 3 further adjusted for cereal fiber (whole grains analysis).

For missing data in each covariate, we created indicator variables. Overall, missing data was less than 5%. The model results summary includes the results of statistical tests for trend in the response for the risk variable. Quintiles Trend P denotes the P value when the median value within the risk variable quintile is included in the hazard model as linear.