The Application of Evidence-Based Methods to Nutrition Research and Challenges Encountered: Fiber and Cardiovascular Outcomes.
Yu, Winifred.
2011
-
Abstract: The use
of systematic review methodology to summarize all available data to answer a research
question is increasing in the nutrition field. However, methodological issues unique to
nutrition have not yet been adequately adjudicated. Research that focuses on factors
influencing the validity and quality in conducting evidence-based review in nutrition is
needed. The overall aim of ... read morethis thesis research was to identify the challenges and to
explore methods to facilitate the application of the evidence-based approach to
nutrition research. Using the relationship of dietary fiber and cardiovascular clinical
and risk factor outcomes as the test case, we qualitatively and quantitatively described
the published literature by first developing an evidence map. An evidence map is a
description of the type and amount of evidence in the literature for a particular
research area. According to the evidence map, we found that the available literature was
diverse. Two areas were identified where the available literature was judged adequate
for conducing systematic reviews - (i) the association between dietary fiber types
(total, cereal, vegetable, fruit, soluble, insoluble) and fiber-rich foods (whole
grains, legumes), and the incidence and mortality of coronary heart disease (CHD) among
observational studies, and (ii) the effect of oat, barley or β-glucans supplement,
and serum lipid concentrations or blood pressure among interventional studies. We found
that evidence among observational studies for the association between dietary fiber and
incidence and mortality of CHD was inconclusive. Among interventional studies, we found
modest protective effects of oat, barley or β-glucans supplementation on total and
low-density lipoprotein cholesterol concentrations, but not for high-density lipoprotein
cholesterol and triglyceride concentrations, and blood pressure. While conducting the
above systematic reviews, we identified challenges when synthesizing dose data from
epidemiological studies because of inconsistent reporting. This led to an exploration of
different statistical techniques for conducting meta-analyses of epidemiological
dose-response data, and comparison of statistical methods between a) meta-analyses using
only data from extreme exposure categories versus meta-analyses using data from all
exposure categories, b) meta-analyses using observed (crude, unadjusted) count data
versus estimates obtained from adjusted data, and c) meta-analyses using the 2-step
approach, where study-specific effect sizes were first calculated and then combined
across studies, versus the 1-step approach, where data from all primary studies are
analyzed in one regression model. This comparison was done among 26 identified nutrition
and health topics of meta-analyses. Compared to methods that uses all exposure
categories, meta-analysis including extreme exposure categories only resulted in
consistently greater magnitude of effects and wider confidence intervals. There was
little difference in the results between meta-analyses using the 1-step approach versus
the 2-step approach. Comparisons between meta-analysis using unadjusted data and
adjusted data suggested no consistent pattern when discordant results were identified in
the direction, statistical significance, or magnitude of effect. In summary, this thesis
research identified the challenges encountered when applying the evidence-based approach
to nutrition research and explored statistical methods for synthesizing dose-response
data, using the relationship between dietary fiber and cardiovascular clinical and risk
factor outcomes as a test case. Critical components of successfully conducting a
systematic review for a nutrition topic include thoughtful research question refinement,
rigorous protocol development, meticulous evaluation of the adequacy and types of
available literature, and appropriate processing of the available data. When
statistically synthesizing data from epidemiological studies, meta-analytic methods that
consider the unique characteristics of dose-response data should be
applied.
Thesis (Ph.D.)--Tufts University, 2011.
Submitted to the Dept. of Nutritional Epidemiology.
Advisor: Alice Lichtenstein.
Committee: Joseph Lau, Thomas Trikalinos, and Christopher Schmid.
Keywords: Epidemiology, Nutrition, and Medicine.read less - ID:
- bz60d779k
- Component ID:
- tufts:20634
- To Cite:
- TARC Citation Guide EndNote
