Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
1 \` - ~ Background Historically, the role of diet in promoting good health and preventing dis- ease has been of great interest to many segments of society. Scientists are faced with the challenge of identifying dietary factors that influence specific diseases and defining their pathophysiological mechanisms. Government agencies, public health policymakers, the food industry, consumer groups, and others seek to interpret and translate that information as they develop regulations, policies, and products and provide advice to the public. The science relative to potential bene- fits or harm of specific nutrients or dietary supplements for health maintenance and disease risk reduction undergoes an evolutionary process, which frequently makes it difficult for the public to comprehend the current state of the science. The evolutionary process can at times include contradictory evidence from ap- parently carefully controlled studies. This may be due in part to differences in the dose of the nutrient, size or duration of the study, or characteristics of the population studied. In addition, scientists do not always agree on what consti- tutes sufficient scientific evidence to warrant changing recommendations to the public. It is not uncommon for a debate to ensue about how much and what kind of evidence justifies giving dietary advice to the public. The increased availabil- ity and potentially harmful use of dietary supplements has focused particular attention on nutrient-disease relationships and the role of increased nutrient in- take in health promotion and disease prevention. As in many scientific fields, early results in nutrition often receive wide public circulation and are applied or adopted without a proper evaluation of the scientific merit of the evidence, thus potentially leading to confusion and rec- ommendations that may not be beneficial and could even be harmful. Advances 9
10 EVOLUTION OF EVIDENCE in multimedia technology and new communication channels such as the Internet have facilitated access to such information and the potential for benefit and harm (SciPICH, 1999~. The resulting confusion can dilute important and substantiated public health messages and lead to a lack of confidence in many dietary recom- mendations. Given the exposure to and interest in an early application of evolv- ing information about diet and health relationships, it is important to determine if it is possible to streamline the scientific method by identifying patterns in types of evidence so that one can potentially predict whether a preliminary rela- tionship will be substantiated by further scientific study. The Committee on Ex- amination of the Evolving Science for Dietary Supplements undertook this task with regard to nutrients and disease at the request of the Food and Drug Admini- stration (FDA) of the U.S. Department of Health and Human Services. THE COMMITTEE AND ITS CHARGE In accordance with the Institute of Medicine committee process, a seven- member expert committee was appointed upon recommendation of the Food and Nutrition Board. The committee's task was to retrospectively review selected case studies of diet and health relationships identified as important in the 1989 report, Diet and Health: Implications for Reducing Chronic Disease Risk (D&H) (NRC, 1989), with relevance to dietary supplements. The committee was to determine the extent to which subsequent scientific evidence from the peer- reviewed literature used in the published Dietary Reference Intakes (DRI) re- ports agreed with the preliminary evidence used to support the relationship iden- tified originally in the 1989 review, or modified significantly the original hy- pothesis and preliminary conclusions. Based on the review of these studies, the committee was to provide an analysis that included characteristics of research that appear to have a high probability of predicting future confirmation by new science in support of a diet and health relationship over time. The analysis was also to include characteristics of information useful to consumers regarding the health effects of such food components or nutrients that allow them to make scientifically informed judgments regarding the role that a specific food compo- nent or nutrient plays in health. The committee had expertise in nutritional epidemiology, evidence-based medicine, research design methodology, clinical medicine, dietetics, and nutri- tional sciences, including micronutrients, B vitamins, antioxidants, and calcium and related nutrients. Five members had participated in the DRI project, and one on the D&H study. Biographical sketches are included in Appendix B. The committee met three times to consider its scope of work, review the relevant scientific evidence, and develop its findings. Once the committee com- pleted its initial draft report, a set of reviewers familiar with the issues under discussion and approved by the National Research Council Report Review Committee individually reviewed and commented on the draft report. These .
BACKGROUND 11 reviewers remained anonymous until the report was finalized. The review proc- ess is intended to ensure that the report addresses the committee's charge, that the conclusions and recommendations are based on scientific evidence, and that the report is presented in an effective and impartial manner. THE REPORT This report looks at the evolution of evidence for selected nutrient-disease relationships by comparing earlier reports prepared using similar mechanisms at different times. For selected case studies, it reviews the types of evidence cited in the D&H report and compares them with types of evidence used to draw con- clusions in the DRI series (IOM, 1997, 1998, 2000a, 2001~. The report is divided into three parts. The remainder of this chapter de- scribes how the committee approached its task, identified the case studies, and assessed the evidence. Chapter 2 describes the evidence in the D&H and DRI reports and the evolution of the evidence for each case study. Chapter 3 presents the committee's findings about patterns of evidence. THE COMMITTEE'S APPROACH As instructed in its charge, the committee confined its analysis to the nutri- ent-disease relationships and related scientific evidence specifically referenced in D&H or in one of the four reports from the DRI series published by the time it began its task. The DRI reports examined were Dietary Reference Intakes for Calcium, Phosphorus, Magnesium, Vitamin D, and Fluoride (IOM, 1997~; Die- tary Reference Intakes for Thiamin, Riboflavin, Niacin' Vitamin B6, Folate, Vi- tamin By, Pantothenic Acid, Biotin, and Choline (IOM, 1998~; Dietary Refer- ence Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids (IOM, 2000a); and Dietary Reference Intakes for Vitamin A, Vitamin K, Arsenic, Bo- ron, Chromium, Copper, Iodine, Iron, Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc (IOM, 2001~. A report on macronutrients was under devel- opment and was not included in the committee's review. The committee recognized that although the D&H and DRI reports both in- cluded assessments of scientific evidence, different purposes of each made pre- cise comparison of types of evidence difficult. The D&H report was an "in- depth analysis of the overall relationship between diet and the full spectrum of major chronic diseases" (p. 4~. It was intended to move beyond assessment of dietary risk factors for single chronic diseases and address the complex task of determining how these risk factors influence a number of chronic diseases. The intended outcome of the assessment was recommendations for dietary patterns that would reduce risk for chronic disease. The D&H report examined data on the association among diet, health, and chronic disease. Whenever possible, primary sources of data in the literature were used. Works of other evaluative
12 EVOLUTION OF EVIDENCE bodies were secondary sources. Strengths and weaknesses of each type of study were assessed and the total evidence was evaluated against six criteria: strength of association, dose-response relationship, temporally correct association, con- sistency of association, specificity of association, and biological plausibility. The strength, consistency, and preponderance of the data and the degree of con- cordance in epidemiological, clinical, and laboratory experiments determined the strength of a conclusion in the report (pp. 38, 653~. The DRI reports present "reference values that are quantitative estimates of nutrient intake to be used for planning and assessing diets for apparently healthy people" (IOM, 2000a, p. 2~. The recommended intakes in the DRI re- ports were based on reviews of scientific data from observational and experi- mental studies published in peer-reviewed journals. When possible, the relation- ships between nutrient intake and prevention of disease were reviewed primarily using randomized trials, but observational studies were reviewed as well. Many of the questions raised about requirements for and recommended intakes of nu- trients could not be answered fully because of inadequacies in the database. Thus, with few exceptions, following review of evidence that included examina- tion of the extent of congruent findings, scientific judgment determined the basis for establishing recommendations for intakes of nutrients (IOM, 2000a). In spite of the difference in purpose, the D&H and DRI reports were delib- erative reviews of the scientific literature spanning a decade or more. Thus, it was felt that these reports were valuable sources for an analysis of the evaluation of scientific discovery related to the role of specific nutrients in disease preven- tion, and could be used in a qualitative description of selected case studies and an assessment of possible patterns in relationships. Given its charge to look at patterns in evidence for potential relationships, not to conduct a new evidence-based review, the committee accepted as given the conclusions in the D&H and DRI reports about specific nutrient-disease relationships. The committee was aware of, but did not include, evidence pub- lished since release of the relevant DRI report because that was not part of its charge. The committee acknowledges that publication bias (i.e., the greater chance of publishing positive rather than negative findings) exists and that evaluation of nutrient-disease associations could be biased in favor of a positive association. Both reports appear to have evaluated the quality of the evidence in generally similar ways and based conclusions on the totality of the evidence available to them. The D&H report, for example, indicates that "Overall, the strength, consistency, and preponderance of data and the degree of concordance in epidemiologic, clinical, and laboratory evidence determined the strength of the conclusions in the report" (p. 39~. The committee relied on the summary statements of the D&H and DRI reports; it did not reevaluate the analyses in those reports. Given that both reports were prepared with similar analyses of the quality of evidence, one would expect that any publication bias would be similar in both reports and would not be expected to affect this committee's findings. .
BACKGROUND 13 SELECTION OF CASE STUDIES The case studies selected were to be identified as important in the D&H re- port and of relevance to diet-disease relationships. For this study, the committee assumed that nutrient-disease dyads listed in the D&H report for which there was some indication of a possible relationship in a summary statement met these two conditions and were potential case studies. The committee sought to be sys- tematic and consistent in its identification of a manageable number of case stud- ies by using the approach described here, and recognizes that judgment was also involved in applying its classification scheme. The general guidelines for selec- tion were: (1) either a summary statement about a relationship in the D&H re- port or an inconclusive statement in the D&H report and a possible or positive summary statement in the DRI report, in order to discuss evolution, (2) a dyed expressing a potential beneficial effect, and (3) high likelihood that it was an active research area, to increase the opportunity to observe any patterns that existed. Using a staff-generated list of 85 potential nutrient-disease combinations in the D&H report, the committee members examined statements about each dyed to determine whether a conclusion was drawn about a nutrient-disease relation- ship. The initial list was limited to those nutrients included in the published DRI reports. If there was a summary statement in the D&H report about a possible positive, beneficial relationship, the dyed was selected for review as a case study. If the relationship was inconclusive, it was kept as a candidate for a case study unless the DRI report also indicated an inconclusive relationship. Most dyads were found to be inconclusive in both the D&H and DRI reports and were not included in the case studies (e.g., selenium and cancer). Also not included were dyads related to adverse effects (e.g., vitamin B6 and neuropathy). There were several exceptions. During its review the committee identified several promising or accepted dyads in the relevant DRI report that were not mentioned in the D&H report. These dyads are folate and neural tube defects (NTDs), folate and colorectal cancer, and vitamin E and prostate cancer. Be- cause these represented an evolution of evidence since the publication of the D&H report, they were included in the committee's analysis, using the evidence in the DRI report to classify the types of evidence. Dyads discussed only in a DRI report that were inconclusive in that report (e.g., folate and homocysteine) were not selected. The dyads selected for review (case studies) are shown in Box 1-1. Once the dyads were selected, committee members reviewed the papers cited in the reports to classify types of evidence, drafted and discussed the case studies, and looked for patterns in evidence.
14 EVOLUTION OF EVIDENCE EXAMINATION OF EVIDENCE Classification of Types of Evidence The committee used a categorization of evidence that was adapted from other classifications (IOM, 2000b; NIH, 1998; USPSTF, 1996), as shown in Box 1-2. The committee added the two categories of"animal studies" and "mecha- nistic studies." Thus, the studies reviewed from the examined reports were clas- sified as either animal studies, mechanistic studies, observational studies (case control or cohort), or clinical trials (nonrandomized, small randomized t< 1,000 subjects), and large randomized [> 1,000 subjects]) (see Box 1-2~. Animal studies are experiments that test the effects of nutrient deficiency or augmentation (through either supplements or foods) on physiological variables and disease outcomes. These studies may use randomized or nonrandomized designs. Mechanistic studies are studies that assess the effects of nutrient intake on intermediate physiological variables (e.g., the effects of vitamin E on platelet adhesion) or biochemical actions (e.g., the effects of vitamin E on protein kinase C) rather than well-accepted preclinical or clinical outcomes. Observational case-control studies are retrospective, observational studies that compare nutri- ent status, as assessed by dietary intake or biomarkers, in persons who have a
BACKGROUND 15 disease (cases) and persons without the disease (controls). Observational cohort studies are observational epidemiological studies that typically assess in a pro- spective fashion the risk of developing a disease according to baseline nutrient status in persons who are free of apparent disease at enrollment. Nonrandomized clinical trials are a heterogeneous group of studies that test the effects of nutrient intake in the setting of a nonrandomized experimental design, for example, depletion-repletion studies. The hallmark of randomized trials is random allocation of individuals to different groups, typically a control group or a condition. The committee further subdivided randomized trials by the
16 EVOLUTION OF EVIDENCE number of participants as less than 1,000 (small) or greater than 1,000 (large). Although the cutpoint of 1,000 is somewhat arbitrary, it served to consistently distinguish between smaller and larger studies. The committee acknowledges limitations of each type of study and the im- perfections of such classifications, specifically the recognition that the strongest evidence might come from studies classified lower on this scale. For instance, a well-done study applying an appropriate animal model might be more persua- sive than observational studies or small clinical trials in humans. Also, if one applies a "best evidence" approach, one large, well-done clinical trial is more persuasive than a corresponding meta-analysis of multiple trials of small size or variable quality, which may also have the inherent drawback of excluding un- published negative studies. Limitations of studies of diets or nutrients and dis- ease outcomes are described in the D&H and DRI reports. Classification of Relationships by Level of Confidence For each case study, the committee categorized the dyed by the level of con- fidence for a positive relationship as expressed in the D&H and then in a DRI report. The level of confidence that a relationship was real was described as ac- cepted, promising, uncertain, or no relationship. For this analysis, a relationship was considered accepted (A) if the evidence was sufficiently strong that the re- port specifically recommended increased or decreased nutrient consumption as a means to prevent disease and possibly develop public policy. That is, the report considered the totality of the evidence strong enough to initiate or consider a public policy recommendation. A relationship was considered promising ('P) if the evidence was sufficiently strong to initiate a large clinical trial or to advocate additional research but not strong enough for public policy recommendations. A relationship was considered uncertain (U) if the evidence was sparse or incon- sistent and the directives for future research not prominent. A fourth category, no relationship (N9, indicates that the report indicated either that no relationship existed or made no concluding comment about a relationship. The committee used the four categories to classify the evidence for selected dyads reported in the D&H and DRI reports. It did not reevaluate the previous conclusions. The system should not be interpreted as a recommendation for or against increased consumption of a given nutrient. Patterns in Evidence The committee used a qualitative approach to examine differences in level of confidence for a positive relationship in the two reports and to search for pat- terns. For this analysis, an increased level of confidence meant that there was some progression in the level of evidence (e.g., U to A), or that the indications for augmented intake applied to a broader population (e.g., the role of fluoride as
BACKGROUND 17 a means to prevent caries expanded from just children to adults). A decreased level of confidence meant that the evidence on balance was less compelling in the DRI report than previous evidence in the D&H report. An unchanged level of confidence meant that there was no alteration in the apparent strength of the relationship, potentially as a result of new but inconsistent data or a lack of sub- stantive new data. Relationships not mentioned in the D&H report that subse- quently became accepted (folate and NTDs) or promising (folate and colorectal cancer, vitamin E and prostate cancer) were also considered changed and placed in a separate category, not in D&H. This category indicates that the D&H report did not state a potential relationship between the nutrient and disease, either be- cause there was no hypothesized effect (e.g., vitamin E and prostate cancer) or because the D&H report focused on the impact of nutrients on chronic diseases rather than adverse outcomes during pregnancy (e.g., neural tube defects). The committee then prepared a tabular summary of types of evidence in the D&H and DRI reports for each dyed, grouping the dyads as increased, de- creased, unchanged, or not in D&H. The committee's findings about patterns were based in large part on its tabular summary. Qualitative evidence was used in lieu of factorial or regression analysis due to the small number of case studies available for review. The committee's findings are discussed in Chapter 3. 1
