Dietary Fiber Intake and Mortality in the NIH-AARP Diet and Health Study

Dietary fiber is defined as the edible parts of plants or analogous carbohydrates that are resistant to digestion and absorption in the human small intestine, with complete or partial fermentation in the large intestine.¹ It has been hypothesized to lower the risk of coronary heart disease, diabetes, some cancers, obesity, and premature death because it is known to (1) improve laxation by increasing bulk and reducing transit time of feces through the bowel; (2) increase excretion of bile acid, estrogen, and fecal procarcinogens and carcinogens by binding to them; (3) lower serum cholesterol levels; (4) slow glucose absorption and improve insulin sensitivity; (5) lower blood pressure; (6) promote weight loss; (7) inhibit lipid peroxidation; and (8) have anti-inflammatory properties.^2,3

A limited number of observational studies have examined the effect of dietary fiber on mortality and reported inconsistent results. The Scottish Heart Health study (N = 11 629; total No. of deaths, 591)⁴ found that dietary fiber intake was inversely related to total mortality in men but not in women. The Zutphen Study in the Netherlands,⁵ which followed up 1373 men for 40 years, found a 9% lowered risk of total death per 10 g/d of dietary fiber intake. Another study, conducted in an Israeli population,⁶ also observed a 43% lowered risk of total death in persons consuming 25 g/d or more of dietary fiber compared with those with less than 25 g/d of dietary fiber intake. On the other hand, the National Health and Nutrition Examination Survey I Epidemiologic Follow-up Study (N = 9776; total No. of deaths, 2632)⁷ found no association between dietary fiber intake and total mortality. However, it assessed dietary fiber intake using a single 24-hour dietary recall, which is unlikely to reflect usual dietary intake.

Studies of dietary fiber in relation to cause-specific death, except death from cardiovascular disease (CVD), are sparse. Furthermore, previous studies examining the association between dietary fiber and mortality were limited by small sample sizes, narrow ranges of dietary fiber intakes, and inadequate control for confounding, leading to decreased power, intakes without the necessary ranges to observe associations, and residual confounding. Therefore, we investigated dietary fiber intake in relation to total and cause-specific mortality in a large prospective cohort of men and women in which more than 30 000 deaths occurred during an average of 9 years of follow-up and a wide range of dietary intakes.

The NIH (National Institutes of Health)-AARP Diet and Health Study was initiated when 567 169 AARP members aged 50 to 71 years from 6 US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and 2 metropolitan areas (Atlanta, Georgia, and Detroit, Michigan) responded to a mailed questionnaire in 1995 and 1996. Details of the NIH-AARP Study have been described previously.⁸ Among participants who returned the questionnaires with satisfactory dietary data, we excluded individuals who indicated that they were proxies for the intended respondent (n = 1265) as well as those who had any prevalent cancer except nonmelanoma skin cancer (n = 23 954), heart disease (n = 68 664), stroke (n = 6434), diabetes (n = 30 881), or self-reported end-stage renal disease at baseline (n = 371). Furthermore, we excluded individuals who reported extreme intakes (>2 times the interquartile ranges of Box-Cox log transformed intake) of total energy (n = 1609) and dietary fiber (n = 891). After exclusions, the analytic cohort comprised 219 123 men and 168 999 women. The NIH-AARP Diet and Health study was approved by the Special Studies Institutional Review Board of the US National Cancer Institute.

At baseline, dietary intakes were 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.⁹ 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 as well as 3 categories of portion size. 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.¹⁰ The nutrient database for dietary fiber was informed by the Association of Official Analytical Chemist method.¹¹

The FFQ used in the study was calibrated using 2 nonconsecutive 24-hour dietary recalls in 1953 NIH-AARP study participants.¹² The energy-adjusted correlation coefficients of dietary fiber intake between an FFQ and 24-hour recalls were 0.72 in men and 0.66 in women. We also collected demographic, anthropometric, and lifestyle information, including history of smoking, physical activity, family history of cancers, menopausal hormone therapy use in women, and some medical conditions at baseline.

We ascertained vital status 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 to the Social Security Administration Death Master File, cancer registry linkage, questionnaire responses, and responses to other mailings. We used the International Classification of Diseases, Ninth Revision and the International Statistical Classification of Diseases, 10th Revision to define death due to CVDs (I00-I78); cancer (C00-C97); infectious diseases (eg, tuberculosis, septicemia, and other infectious and parasitic diseases [A00-B99, excluding B20-B24]); respiratory diseases (eg, pneumonia, influenza, chronic obstructive pulmonary diseases, and allied conditions [J10-J18, J40-J47]); accidents (eg, accident, suicide, and homicide [V01-X59]); and other causes (all other causes except those previously mentioned).

We used the Cox proportional hazards model¹³ to estimate relative risks (RRs) and 2-sided 95% confidence intervals (CIs) using the SAS PROC PHREG procedure (Version 9.1; SAS Institute Inc, Cary, North Carolina). 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, 2005), whichever occurred first. We evaluated the proportional hazards assumption and confirmed it by modeling interaction terms consisting of the cross product of time and dietary fiber intake.

Although there was no statistically significant interaction by sex (P = .17), we performed sex-specific analysis and reported the results by sex. Intakes of dietary fiber and fiber from food sources were adjusted for total energy intake using the residual method¹⁴ and were categorized into quintiles. We estimated the RRs for quintiles of fiber intake as well as continuous intake. To test linear trends across quintiles of fiber intake, we created a continuous variable based on the median value in each quintile and regressed the risk of death on this variable.

We presented an age-adjusted model and 2 multivariate models. Multivariate model 1 was adjusted for age, smoking status, smoking dose, and time since quitting smoking because smoking was a strong confounder. In multivariate model 2, we also adjusted for race/ethnicity, education, marital status, self-rated health status, body mass index (BMI), physical activity, menopausal hormone therapy use in women, and intakes of alcohol, red meat, fruits, vegetables, and total energy. Furthermore, we adjusted for aspirin use, high cholesterol level, and high blood pressure and found that the results did not change. For missing data in each covariate, we created an indicator variable reflecting missing data. Generally, missingness was less than 5%. We tested whether the association between dietary fiber intake and mortality was modified by smoking and BMI. The test for interaction was performed using the likelihood ratio test entering a cross product term of both dietary fiber intake and BMI as continuous variables and smoking as an ordinal variable. In analyses of fiber from food sources, fiber intake from grains, fruits, vegetables, and beans was mutually adjusted.

Also, we carried out measurement error corrections using a regression calibration method^15,16 and a SAS macro¹⁷ for an age-adjusted model and a parsimonious multivariate model that adjusted for age, smoking, and total energy intake. The RRs for dietary fiber intake were corrected for measurement error by regressing intake from the reference dietary assessment method—2 nonconsecutive 24-hour recalls—on intake from the FFQ.

During an average of 9 years of follow-up, we identified 20 126 deaths in men and 11 330 deaths in women. There were 5248 CVD deaths and 8244 cancer deaths in men and 2417 CVD deaths and 4927 cancer deaths in women. Age-adjusted total, CVD, and cancer mortality rates per 100 000 person-years were 991, 258, and 406, respectively, in men and 716, 153, and 311, respectively, in women. Energy-adjusted dietary fiber intake ranged from 13 (10th percentile) to 29 (90th percentile) g/d in men and from 11 to 26 g/d in women. The ranges (10th-90th percentiles) of energy-adjusted fiber intake from grains, fruits, vegetables, and beans were 3.3 to 11.1, 0.9 to 8.0, 3.1 to 10.6, and 0.6 to 4.8 g/d, respectively, in men and 2.5 to 8.8, 1.0 to 8.1, 2.9 to 10.3, and 0.3 to 3.4 g/d, respectively, in women. Compared with individuals in the lowest quintile of dietary fiber intake, persons in the highest quintile were more likely to have higher education, to have self-rated their health as being very good/excellent, to have a lower BMI, to be physically active, and to use menopausal hormone therapy in women but were less likely to smoke, to drink alcohol, and to consume red meat (Table 1).

We found that dietary fiber intake was significantly inversely associated with risk of total death in both men and women (Table 2). Comparing the lowest quintile of dietary fiber intake, both men and women in the highest quintile had a 22% lower risk of total death (multivariate RR_{Q5 vs Q1}, 0.78 [95% CI, 0.73-0.82] in men and 0.78 [95% CI, 0.73-0.85] in women). For an increment of 10 g/d of dietary fiber intake, the multivariate RR for total death was 0.88 (95% CI, 0.86-0.91) in men and 0.85 (95% CI, 0.82-0.89) in women. We also performed the propensity score analysis to better control for confounding and found that the results did not materially change. Comparing the highest quintile of dietary fiber intake with the lowest, the RR for total death was 0.82 (95% CI, 0.80-0.84) in men and 0.84 (95% CI, 0.81-0.87) in women. Because smoking was a strong confounder in our analysis, we performed analyses stratified by smoking status. The inverse association between dietary fiber intake and total death remained significant in never smokers in both men (multivariate RR_{Q5 vs Q1}, 0.81 [95% CI, 0.71-0.93; P for trend, <.001]) and women (multivariate RR_{Q5 vs Q1}, 0.83 [95% CI, 0.72-0.97; P for trend, .05]). Dietary fiber intake was also significantly inversely related to total death in former and current smokers in both men and women. A significant association with dietary fiber intake was also observed across categories of BMI. Furthermore, we examined the association stratified by age at baseline (<60 and ≥60 years), self-rated health condition (excellent/good and fair/poor), and menopausal hormone therapy use in women (never and ever) and observed a consistently inverse association between dietary fiber intake and total death across all categories examined (data not shown).

Correction for measurement error in the assessments of dietary fiber intake strengthened the association with total mortality. For an increment of 10 g/d of dietary fiber intake, the RR of total death was 0.86 (95% CI, 0.84-0.88) in men and 0.83 (95% CI, 0.80-0.86) in women after adjustment for age, smoking, and total energy intake but before measurement error correction. After measurement error was corrected for, the RR was 0.76 (95% CI, 0.72-0.80) in men and 0.71 (95% CI, 0.66-0.76) in women.

Dietary fiber intake was also inversely related to risk of death from CVD, cancer, and infectious and respiratory diseases in men (Table 3). Comparing the highest with the lowest quintile of dietary fiber intake, men had a 24% to 56% lower risk of death from CVD, cancer, and infectious and respiratory diseases. Comparable findings in women were a 34% to 59% lower risk of death from CVD and infectious and respiratory diseases and no association with cancer death (Table 4). For every 10-g/d increase in dietary fiber intake, the multivariate RRs for death from CVD, cancer, and infectious and respiratory diseases were 0.88 (95% CI, 0.86-0.91), 0.92 (95% CI, 0.88-0.96), 0.66 (95% CI, 0.52-0.84), and 0.82 (95% CI, 0.74-0.93), respectively, in men and 0.76 (95% CI, 0.69-0.84), 0.97 (95% CI, 0.91-1.04), 0.61 (95% CI, 0.44-0.85), and 0.66 (95% CI, 0.56-0.78), respectively, in women.

We further adjusted for aspirin use, high blood pressure, and cholesterolemia and found no appreciable differences. We also analyzed the data by excluding death that occurred during the first 2 years of follow-up and the first 4 years of follow-up and found no change in results. When we restricted analyses to never smokers, the multivariate RRs_{Q5 vs Q1} of death from CVD, cancer, and infectious and respiratory diseases in men were 0.95 (95% CI, 0.74-1.21; P for trend, .81 [1134 deaths]), 0.82 (95% CI, 0.65-1.03; P for trend, .009 [1405 deaths]), 0.22 (95% CI, 0.08-0.61; P for trend, .002 [70 deaths]), and 0.60 (95% CI, 0.25-1.44; P for trend, .20 [90 deaths]), respectively. In female never smokers, the multivariate RRs_{Q5 vs Q1} were 0.69 (95% CI, 0.50-0.95; P for trend, .02 [729 deaths]) for CVD death, 1.14 (95% CI, 0.90-1.44; P for trend, .11 [1342 deaths]) for cancer death, 0.28 (95% CI, 0.10-0.80; P for trend, .02 [70 deaths]) for infectious disease death, and 0.50 (95% CI, 0.22-1.14; P for trend, .07 [103 deaths]) for respiratory disease death.

Fiber intakes from food sources such as grains, fruits, vegetables, and beans were also examined (Figure). We found that dietary fiber from grains was significantly inversely related to the risk of total, CVD, cancer, and respiratory disease deaths in both men and women. Comparing the highest with the lowest intake of fiber from grains, men had a 23% lower risk of total death (multivariate RR, 0.77 [95% CI, 0.73-0.81]) and women had a 19% lower risk of total death (multivariate RR, 0.81 [95% CI, 0.76-0.86]). Fiber from vegetables and beans was also weakly associated with a lower risk of total death in both men and women. However, fiber from fruits was not related to total and cause-specific deaths in men and women.

In this large prospective cohort study, we found that dietary fiber intake was significantly inversely associated with the risk of total death and death from CVD, infectious diseases, and respiratory diseases in both men and women. Dietary fiber intake was also related to a lower risk of death from cancer in men but not in women. Among specific sources of dietary fiber, fiber from grains showed the most consistent inverse association with risk of total and cause-specific deaths.

Three of 4 previous studies that examined dietary fiber intake in relation to total mortality reported a 9% to 43% lowered risk of total death among persons with a higher consumption of dietary fiber in various populations.^4-6 Consistent with most of these studies, our study found a 22% lower risk of total death comparing the highest with the lowest quintile of intake in both men and women. Furthermore, the association was not modified by smoking status or BMI. The findings remained robust when we corrected for dietary intake measurement error using calibration study data; in fact, the association was even stronger with measurement error correction.

Dietary fiber intake in relation to CVD death has been studied in several studies,^5,18-21 consistently showing protection for CVD death, in agreement with our study. A pooled analysis of 10 prospective cohort studies estimated that the risk of CVD death decreased by about 19% (multivariate RR, 0.81 [95% CI, 0.73-0.91]) per 10-g/d increment of dietary fiber intake.²² This pooled study also found that fiber intake from both grains and fruits was associated with a lower risk of CVD death. Our findings were consistent for fiber intake from grains but not from fruits. Several plausible mechanisms, such as improving serum lipid profiles,²³ postprandial absorption and insulin resistance,²⁴ and lowering blood pressure,²⁵ have been suggested to underlie the beneficial effects of dietary fiber on CVD.

Studies of dietary fiber intake in relation to cause-specific death other than CVD are limited. The role of dietary fiber in the pathogenesis of cancer has long been debated. Despite plausible physiologic mechanisms, the association between dietary fiber and cancer has been inconsistent. The consensus report from the World Cancer Research Fund and the American Institute for Cancer Research concluded that dietary fiber intake probably lowers the incidence of colorectal and esophageal cancer,²⁶ but little is known about dietary fiber intake in relation to cancer death. In our study, we observed an inverse association between dietary fiber intake and cancer death in men but not in women. The observed sex difference may, in part, be explained by the differences in leading organ sites for cancer death between men and women. Men have higher mortality rates of cancers of the head and neck, esophagus, liver, urinary bladder, and kidney than women. Considering data indicating that the risks of these cancers are influenced by several dietary factors, including grains, fruits, and vegetables,²⁷ our findings of the protective effect of dietary fiber on total cancer death may have been the result, in part, of the lowering mortality of these specific cancers in men. Nonetheless, we cannot rule out the possibility that our findings of different associations between dietary fiber intake and cancer death between men and women are due to chance. Future research might elucidate this finding.

Interestingly, our study found that dietary fiber intake, especially from grains, was inversely associated with the risk of death from infectious and respiratory diseases. Inflammation, a predominant pathphysiologic response in many infectious and respiratory diseases, has been suggested to contribute the progression of these diseases.²⁸ Studies have shown that dietary fiber has anti-inflammatory properties: dietary fiber intake was associated with lower levels of inflammation markers such as C-reactive protein, interleukin 6, and tumor necrosis factor α receptor 2, which play roles in chronic inflammatory conditions.^29-32 The anti-inflammatory properties of dietary fiber could explain, in part, significant inverse associations of dietary fiber intake with infectious and respiratory diseases as well as with CVD death. The Iowa Women's Health Study³³ found that women consuming a large amount of whole grains—rich sources of fiber, mineral, and other phytochemicals—had a 34% lower risk of death from noncardiovascular, noncancer inflammatory diseases (RR_{Q5 vs Q1}, 0.66 [95% CI,: 0.54-0.81; P for trend, .008]) and a 40% lower risk of death from respiratory system diseases (RR_{Q5 vs Q1}, 0.60 [95% CI, 0.46-0.80; P for trend, .006]). A few other studies have also suggested that dietary fiber intake lowers the risk of inflammatory diseases such as duodenal ulcer and diverticular disease.^34,35

Our study has several strengths. First, it is a large prospective cohort study in which diet was measured at baseline, thus decreasing the likelihood of recall bias. Also, our study has wide ranges of dietary fiber intake and a large number of deaths, providing good statistical power and sample size to examine the associations in never smokers and relationships with specific causes of death, such as infectious and respiratory diseases. Our findings of the beneficial effect of dietary fiber intake for lowering the risk of death from infectious and respiratory diseases are interesting but warrant further investigation. Second, we excluded persons who reported chronic diseases such as heart disease, stroke, diabetes, and end-stage renal disease at baseline. This may have minimized the attenuation of associations. Given the perception that dietary fiber is a healthy dietary constituent, persons who presumably are already at high risk of death may have been prone to biased reporting of dietary fiber intake, thus attenuating associations. Third, we extensively controlled for smoking and many other risk factors of mortality. Finally, we performed analyses in never smokers only to minimize residual confounding by smoking, the strongest confounder, and still observed consistently significant inverse associations.

Several potential limitations also need to be considered. We cannot rule out the possibility that dietary fiber intake is a marker of a healthy diet that is high in grains, fruits, and vegetables—foods that are rich in vitamins, minerals, antioxidants, and phytochemicals—or that dietary fiber intake is a marker of a healthy lifestyle. Persons who consumed a large amount of dietary fiber may engage in other healthy behaviors that were incompletely assessed in the study. Therefore, our findings may in part due to residual confounding by healthy lifestyle. However, we found significant inverse associations even after controlling for all these factors in multivariate models. Also, inverse associations between dietary fiber intake and total and cause-specific deaths were observed in never smokers and in persons with a BMI of less than 25 (calculated as weight in kilograms divided by height in meters squared). Measurement error is always an inherent limitation in self-reported dietary assessment. Nevertheless, our measurement error correction analyses strengthened our associations. This suggests that measurement errors in dietary fiber intake may have attenuated the true RRs, but the measurement errors were not substantial; therefore, we were able to detect an association in the study. We recognize, however, that the reference method—24-hour recalls—used in our measurement error correction may underestimate or overestimate true intake. Therefore, it is possible that substantial changes in risk estimates could occur if an objective marker of intake, which may be close to the true intake, is used in measurement error correction.

In conclusion, our study shows that dietary fiber may reduce the risk of premature death from all causes, especially from CVD and infectious and respiratory diseases. The current Dietary Guidelines for Americans recommend choosing fiber-rich fruits, vegetables, and whole grains frequently and consuming 14 g/1000 calories of dietary fiber. A diet rich in dietary fiber from whole plant foods may provide significant health benefits.

Correspondence: Yikyung Park, ScD, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 6120 Executive Blvd, Rockville, MD 20852 (parkyik@mail.nih.gov).

Accepted for Publication: November 29, 2011.

Published Online: February 14, 2011. doi:10.1001/archinternmed.2011.18

Author Contributions:Study concept and design: Park, Hollenbeck, and Schatzkin. Acquisition of data: Subar, Hollenbeck, and Schatzkin. Analysis and interpretation of data: Park and Subar. Drafting of the manuscript: Park and Subar. Critical revision of the manuscript for important intellectual content: Park, Subar, Hollenbeck, and Schatzkin. Statistical analysis: Park. Obtained funding: Schatzkin. Administrative, technical, and material support: Park and Hollenbeck. Study supervision: Park and Schatzkin.

Financial Disclosure: None reported.

Funding/Support: The study was supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health.

Additional Contributions: We are indebted to the participants in the NIH-AARP Diet and Health Study for their outstanding cooperation. We also thank Sigurd Hermansen and Kerry Grace Morrissey from Westat for study outcomes ascertainment and management and Leslie Carroll at Information Management Services for data support and analysis.