Courtesy of http://topnews.in/
That’s right folks, it is time to fiber up! Take up those whole grain cereals and those bran muffins, and be proud to be regular!
In all seriousness, dietary fiber, or edible parts of plants or analogous carbohydrates that are resistant to digestion and absorption in the human small intestine (Park, Subar, Hollenbeck, & Schatzkin, 2011, p. 1), has been proven to provide countless benefits to the human body. According to the Mayo Clinic (2009), dietary fiber has the following benefits:
- Normalizes bowel movement (in the words of my father: “Keeps you regular!”)
- Maintains bowel integrity and health
- Lowers blood cholesterol levels
- Helps control blood sugar levels
- Assists in weight loss
- Lowers the chance of colorectal cancer (studies on this are mixed however)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
A recent study was released (2/14/2011) by the Archives of Internal Medicine titled Dietary Fiber Intake and Mortality in the NIH-AARP Diet and Health Study, as to the health benefits of dietary fiber intake (Park, Subar, Hollenbeck, & Schatzkin, 2011). More precisely, the article demonstrates an inverse association between dietary fiber intake and mortality, especially with respect to cardiovascular, infectious, and respiratory diseases. The New York Times then made the article public (2/21/2011) by summarizing the article in the Health section of their newspaper. The New York Times article was titled Diet: High Fiber to Combat Death and Disease (Rabin, 2011).
The study conducted was a prospective cohort study. We know this because the researchers started with individuals that were categorized as being either exposed (consumes a high amount of dietary fiber) or non-exposed (consumes a low amount of dietary fiber), and thus our independent variables, in order to determine the dependent variable, or the disease, which in this case is mortality.
(Image Removed)
To the left is a diagram outlining the logic behind a prospective cohort study design. As mentioned before, exposed in this study indicates high dietary fiber, non-exposed indicates low dietary fiber, disease indicates death, and no disease indicates vitality.
This specific cohort study was set up by sending mailed questionnaires to 567,169 AARP members aged 50 to 71 years of age between 1995 and 1996. The mailed questionnaires were self-administered 124-item food-frequency questionnaires, or what are known as FFQs. Participants were each asked to record their usual consumption and portion size over the past 12 months. After numerous participants were excluded from the study (mostly due to the presence of chronic diseases, which were thought to be potential risk factors), the cohort study contained a total of 219,123 men and 168,999 women. During a total of nine years of follow-up (until 12/31/2005), there were 20,126 deaths in men and 11,330 deaths in women. Person-years were used to quantify the total follow-up time, which was calculated from the baseline questionnaire until death or the end of that specific individual’s follow-up time, for a combined 100,000 person-years for the entire study.
With this design and set-up, there are already a number of issues that need to be addressed. Usually with a cohort study, there are a number of subjects who drop out of the study, whether it is simply quitting the study or dying. We do know the number of individuals that died, as it is directly a part of the study itself, we however are not given the exact number of subjects who quit the study and were not included in the entire nine years of follow-up. Therefore, we are unable to determine the percentage of attrition. Next, questionnaires of this nature can be potentially troublesome. Depending on how each question is worded, each question could be interpreted differently from person to person. Additionally, there are issues with reporting biases. A bias by definition is “any systematic error in the design, conduct, or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of a disease” (Gordis, 2009, p. 247). In this case, a reporting bias is when “a subject may be reluctant to report an exposure (or non-exposure in this study) he is aware of because of attitudes, beliefs, and perceptions” (Gordis, 2009, p. 250). Also, the fact that subjects were required to recall their average fiber intake over the last 12 months may be troublesome, as some people become forgetful of such matters, which only gets worse with age.
There is also a form of information bias that is present here as a result of the long questionnaire (124 items) that may result in survey fatigue. Information bias is “when the means for obtaining information about the subjects in the study are inadequate (too long) so that as a result some of the information gathered regarding exposures is incorrect” (Gordis, 2009, p. 249). Finally, only AARP members were used in the study, which in itself may provide a source of bias. AARP is a nonpartisan, nonprofit, organization that lobbies on behalf of the interests of its members, all of whom are age 50 and over. Thus, members of AARP may on average be more educated, and may be more aware of current events and research surrounding individuals of their age group, including the many benefits of a healthy lifestyle (all relative to the general population).
An additional bias that is common with cohort studies is bias in assessment of the outcomes (Gordis, 2009, p. 174). This form of bias, however, is not a problem in this study, because since researchers have preconceptions as to what they believe the results to be, based on their hypothesis, they may be biased in their interpretation of the diagnosis of a disease. But, in this study, the outcome is death, and dead is dead….not much argument there…we hope. On the other hand, since researchers do indeed have preconceptions about the results, there will always be some degree of analytic bias, which results in the “unintentional introduction of their biases into their data analyses and into their interpretation” (Gordis, 2009, p. 174).
At this point, you may be thinking to yourself: “Wow, this guy is really picky! How will any study measure up to his expectations!” Yes, I am being critical here, but the truth is no study is perfect, and every study is bound to have flaws. This doesn’t mean that we have to throw the study out, but we do need to be aware of the biases that are possible and present, as they help to determine the strength of the association being hypothesized in the study.
(Image Removed)
With a study such as this, involving dietary fiber and its association with mortality rates, there are many potential confounders that can affect the results. A confounder is an additional variable that may be different between the study groups thereby potentially resulting in a biased result. Put another way, it is possible that the effect of the exposure on the disease, which in our example would be the effect of dietary fiber on specific forms of mortality, is actually due to the effect of a confounder. A confounder could be age, sex, occupation, urbanization, etc. In order to control confounding, researchers will either stratify the results or adjust for specific confounder variables in a multivariate manner.
(Image Removed)
Above is table 1 from the original article. This table displays various selected characteristics of all the subjects used in the study that were later adjusted for to avoid the influence of them as confounders. First, men and women were stratified into two groups to avoid any gender biases in the data. Then, the amount of dietary fiber intake was stratified into quintiles based on the amount consumed. This “dose-response relationship” helps to establish an inverse association between dietary fiber intake and specific mortality rates. The median amount of dietary fiber consumption for the lowest quintile was 12.6 g/d (grams/day) for men and 10.8 g/d for women, while for the highest quintile was 29.4 g/d for men and 25.8 g/d for women. Things to notice in table 1 is that college and post-college, health status, physical status, alcohol consumption, smoking status and red meat intake all seem to correspond with the specific fiber quintiles, with better results moving hand-in-hand with increased fiber intake. This is why these variables are considered confounders. People who intake a lot of fiber are probably very health conscious, and thus probably smoke less, exercise more, etc.
(Image Removed)
In table 2, the authors show all five quintiles, for both men and women. Most of the information provided in this table is in the form of relative risks. Relative risk is the risk of developing the disease among exposed individuals relative to individuals not exposed. In our case, it would be the risk of death among those individuals with a relatively high intake of fiber relative to those with a relatively low intake of fiber. Relative risk is a powerful tool because it estimates the strength of an association between the disease and exposure. Next to all the relative risk calculations is an interval, this is the confidence interval, which at 95% tells us you can be 95% confident that the true value is within this range. In our case, if the lowest quintile is our “reference” or “base” to measure all the other risks with, the relative risk of 0.78 for the highest quintile would tell us those individuals had a 22% lower risk of total death relative to the lowest quintile individuals.
In fact, in general, if relative risk calculations are over one, this indicates that the risk of disease in greater in exposed individuals than in non-exposed individuals. If the relative risk is exactly one, this indicates that risk is equal between exposed and non-exposed individuals, and thus no association exists between the exposure and disease. In our case, the relative risk amounts are mostly under one, which indicates the risk of disease (death in our case) is lower in exposed individuals (high fiber) relative to non-exposed (low fiber), thereby indicating a possible protective quality of the exposure (high fiber). With this in mind, we can determine if the relative risk value is significant or not via studying the confidence interval. If the confidence interval contains one, then there is a possibility that the true relative risk is really one, and thus nullifying the association between exposure and disease.
Getting back to confounders, in order to control for age, the authors decided to first analyze relative risk calculations with the results age-adjusted. To adjust for age means to nullify age as a factor that differs between different quintiles. Next, the authors decided to adjust for multiple factors at once, via two different multivariate analyses. The first multivariate analysis, controls for numerous stratifications of smoking. The second multivariate analysis, controls for all the characteristics listed in table one. Then, the authors took the results from the second multivariate analysis and stratified it via smoking status and via BMI figures. Since the overall study is so large participant-wise, the stratifications didn’t hurt the power of the study. Power in a statistical sense means “the probability of obtaining a statistically significant p-value when the null hypothesis is truly false” (Machin, Campbell, & Walters, 2007, p. 109).
(Image Removed)
From table 2, we can see from the stratified analysis that the association was not changed and thus is equally as strong as the non-stratified results, thus indicating no significant interaction from smoking and BMI levels on the death rates relative to the already adjusted multivariate analyses.
(Image Removed)
(Image Removed)
From tables 3 and 4, we can see that for both men and women, the risk of death from cardiovascular, infectious, and respiratory diseases was lowered by greater fiber intake. Furthermore, we observe a similar association between fiber and cancer, but in this circumstance, the results for cancer deaths were only significant for men. On this note, the authors stated that the difference may be due to chance, and that further research would be necessary. For “deaths from accidents,” we would not expect fiber intake to reduce the risk of accidental death, and that is exactly what we see, as the p-values and confidence intervals indicate the relative risk calculations are insignificant statistically. Due to convention, p-values above 0.05 are considered insignificant. The p-value is the probability of obtaining the observed statistic, or one more extreme, given that the null hypothesis is true.
(Image Removed)
Finally, in the above figure, the authors compared the different types of fibers and their respective relative risk factors. The fibers compared were grains, fruits, vegetables, and beans. Grains appeared to have the greatest association with decreased risk of death for both men and women, while fruits and vegetables did not. Beans on the other hand, appeared to have an overall significant association with reduced risk of death for women, but not for men.
In the end, the authors were able to conclude that dietary fiber has the potential to reduce the risk of death from all causes, but especially from cardiovascular, infectious and respiratory diseases. With respect to the lower cardiovascular death risk, the authors credited fiber’s ability to “improve serum lipid levels, postprandial absorption and insulin resistance, and lower blood pressure” (Park, Subar, Hollenbeck, & Schatzkin, 2011, p. 5). With respect to the lower infectious and respiratory death risk, the authors credited fiber’s anti-inflammatory properties.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
After a thorough analysis of this study, I think it was a very well-designed study. The sample size is in the hundreds of thousands, and the authors attempted to adjust for almost every important potential confounder. My only reservation with this, is the general confounder of a healthy lifestyle overall. People with a healthy lifestyle are more likely to consume large amounts of dietary fiber, and are more likely to live longer lives. Nonetheless, I think the study design was excellent and well thought out. The New York Times article, on the other hand, was not responsibly reported in my opinion. While the NYT article did summarize the main points accurately, the author failed to explain the details of why the results occurred. While the general public would probably not be able to appreciate the full scientific article without a statistical and epidemiological background, I still feel the NYT article is an insult to the real article and all the hard work that was put into it, in both money and time. After reading just the NYT article, I was not convinced at all, as it just read like any other random health article. However, after reading the scientific article, I am fully convinced of the benefits of dietary fiber.
I chose this article because I have a fondness for nutrition, especially fiber. After being educated thoroughly by my colorectal specialist as to the importance of fiber (and water) I have taken my fiber intake seriously. Also, in the last year and a half, I have lost over 40 pounds of weight due to a changed and improved lifestyle, but not a diet. Diets are temporary, lifestyle changes are permanent. I take this subject both seriously and personally. I had one grandfather die from diabetes, and another die from heart disease, so I take my personal lifestyle seriously now, as everyone should. Both nutrition and exercise are important to me, especially my dietary fiber intake. I only eat 100% whole wheat bread, I love beans, vegetables high in fiber like corn, peas and carrots, raisin bran cereal, and dried fruit. Thus, I think this article is really important, as fiber has multiple benefits as outlined in this blog post, and we all need to stop and think about what we are and are not putting into our bodies. From reading this article, I feel like I have grown in my love and interest for nutritional health, and its potential implications on future public policy. Thank you.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Gordis, L. (2009). Epidemiology (4th Edition ed.). Philadelphia, PA, USA: Saunders Elsevier.
Machin, D., Campbell, M. J., & Walters, S. J. (2007). Medical Statistics: A Textbook for the Health Sciences (4th Edition ed.). England, UK: John Wiley & Sons.
Mayo Foundation for Medical Education and Research. (2009, November 19). Dietary fiber: Essential for a healthy diet. Retrieved February 28, 2011, from Mayo Clinic: http://www.mayoclinic.com/health/fiber/NU00033
Park, Y., Subar, A. F., Hollenbeck, A., & Schatzkin, A. (2011, February 14). Dietary Fiber Intake and Mortality in the NIH-AARP Diet and Health Study. Archives of Internal Medicine , 1-8.
Rabin, R. C. (2011, February 21). Diet: High Fiber to Combat Death and Disease. The New York Times , p. D6.
Epidemiology...it's everywhere! 
