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.
J I,, ~x aid,, ~ Findings FINDINGS Using a qualitative case study approach, the committee assessed changes in the levels of confidence in 14 nutrient-disease relationships (dyads). Eleven were identified in both the D&H and DRI reports; three, for which evidence had evolved, were discussed only in a DRI report. Applying its classifications of accepted, promising, uncertain, and no rela- tionship to the selected dyads in the D&H report, the committee found that only fluoride and dental caries fell in the accepted category. Two were promising (~- carotene and lung cancer, vitamin C and gastric cancer), and eight were uncer- tain (calcium and bone status, vitamin D and bone status, vitamin E and cancer, vitamin E and coronary heart disease [CHD], vitamin C and colds, folate and cervical dysplasia, phosphorus and bone status, chromium and diabetes). By publication of the DRI reports between 1997 and 2001, the level of confidence had changed for six dyads, in some cases in surprising ways. Of the three dyads reviewed that were discussed only in the DRI reports, one was accepted (folate and neural tube defects tNTD]) and two were promising (folate and colorectal cancer, vitamin E and prostate cancer). The dyads are shown in Table 3-1. Table 3-2, which displays the types of evidence and changes in confidence, is the basis for the committee's findings, described in the following paragraphs. Confidence in nutrient-disease relationships can change, often in unex- pected directions. An important finding is that preliminary evidence in support of a nutrient- disease relationship is often not confirmed. Neither promising relationship 51
52 EVOLUTION OF EVIDENCE TABLE 3-1 Description of Relationships . Change in Confidence Level of About Existence of Confidence Positive Relationship (D&H ~ DRI) a Dyad Increased A ~ A+ Fluoride ar~ddentalcaries U ~ A Calcium arid bone status U ~ A Vitamin D end bone statue Decreased P ~ No ,B-carotene end lung cancer P ~ U Vitamin C arid gastric cancer U ~ N Vitamin E and cancer (except prostate) Unchanged U ~ P ~ U Vitamin E and coronary heart disease U ~ U Vitamin C end colds U ~ U Folatear~dcervicaldysplasia U ~ U Phosphorus and bone status U ~ U Chromium and diabetes Not in Diet and Health ~ A Folate and neural tube defects P Folate and colorectal carder ~ P Vitamin E and prostate cancer a D&H = Diet and Health: Implications for Reducing Chronic Disease Risk (NRC, 1989), DRI = Dietary Reference Intake reports (IOM, 1997, 1998, 2000a, 2001), A = accepted, U = uncertain, P = promising, N = no relationship. ~ P to U for dietary p-carotene and U to N for supplementary ,B-carotene. from the D&H report (~-carotene and lung cancer, vitamin C and gastric cancer) was subsequently accepted in a DRI report. Of the eight uncertain dyads from the D&H report, two were subsequently found to be accepted (calcium and bone status, vitamin D and bone status), one was found not to be a relationship (vita- min E and cancer "excluding prostate cancer]), and five remained uncertain (vi- tamin E and CHD, vitamin C and colds, folate and cervical dysplasia, phospho- rus and bone status, chromium and diabetes). High-dose p-carotene and lung cancer is illustrative. An impressive body of evidence, including numerous ob- servational studies, suggested that an increased intake of foods rich in ,8- carotene might reduce the risk of developing lung cancer. This appealing hy- pothesis was evaluated by testing high-dose Q-carotene administration in three large-scale, long-term trials, two of which focused on populations at high risk for lung cancer. In contrast to expectations, supplementation with p-carotene significantly increased the risk of lung cancer in the two studies that enrolled persons from high-risk populations. In the third trial, involving male physicians, p-carotene supplementation had no significant effect. Hence, not only was con- fidence in the putative benefit of p-carotene reduced, but also the direction of the relationship changed because the available evidence suggested that ,B- carotene supplementation may increase the risk of lung cancer in high-risk populations.
FINDINGS 53 No pattern of evidence clearly predicts change in the confidence of relation- ships, particularly those initially deemed uncertain or promising. The evidence cited for each dyed was quite heterogeneous. The committee could not identify any pattern of evidence that consistently predicted a change in the level of confidence that a positive relationship existed. The committee ob- served three instances in which confidence in a relationship decreased from the D&H report to the DRI report (~-carotene and lung cancer, vitamin C and gas- tric cancer, vitamin E and cancer [except prostate cancers. In each instance, a common characteristic was an absence of trial citations (even a small trial of less than 1,000 participants) in the D&H report and the presence of trial citations in a DRI report. The case studies suggest that there is a tendency for large trials to be developed when smaller trials are promising, but the outcome of larger trials remains unpredictable. Even the citation of small clinical trials in the D&H re- port did not predict the nature of the relationship of the corresponding dyads in the DRI reports. The committee also observed that few meta-analyses and sys- tematic reviews were referenced in the D&H and DRI reports. This may have occurred because those techniques were not commonly used prior to publication of the D&H report and because the DRI reports drew on original research stud- ies to set nutrient requirements. For the dyads discussed in both the D&H and DRI reports, the committee also observed references to animal studies more often in the D&H report (nine dyads) than in the DRI reports (two dyads). The differential citation of animal studies likely reflects the different purposes of the reports and the reliance on human studies to set nutrient requirements. Large randomized trials have the greatest impact in changing the level of confidence in a nutrient~isease relationship. Not surprisingly, large clinical trials were cited for only two dyads in the D&H report (fluoride and dental caries, vitamin C and colds) and nine dyads in the DRI reports (fluoride and dental caries, calcium and bone status, vitamin D and bone status, Q-carotene and lung cancer, vitamin C and gastric cancer, vita- min E and cancer [except prostate cancer], vitamin E and CHD, folate and NTDs, and vitamin E and prostate cancer). The latter two dyads were not men- tioned in the D&H report (see Table 3-2~. For a dyed considered accepted in either the D&H or DRI reports, a large clinical trial was cited. Also, in those instances with an increase or decrease in the level of confidence in a positive relationship, a large clinical trial typically was cited in the DRI report. For vita- min E and CHD, considerable interest developed as a result of prospective ob- servational studies published after the D&H report that suggested the relation- ship to be promising. However, large clinical trials published prior to issuance of the corresponding DRI report failed to demonstrate a beneficial effect of vitamin E on CHD.
54 EVOL UTION OF EVIDENCE TABLE 3-2 Change in Confidence In a Positive Relationship by Type of Evidence for Nu~ient-Disease Dyads Types of Evidence in D&H and DRI Reportsa Animal Mechanistic Change in Confidences Dyad Increased A ~ A+c Fluoride and dental caries O ~ O U ~ A Calcium and bone status O O U ~ A Vitamin D end bone statue O O Decreased P ~ N`i Q-carotene and lung cancer O O P ~ U Vitamin C end gastric cancer O O U ~ N Vitamin E and cancers O Unchanged U ~ P ~ U Vitamin E and CHDf O U ~ U Vitamin C and colds U ~ U Folate and cervical dysplasia U ~ U Phosphorus end bone status" O O U ~ U Chromium and diabetes O Not in D&H A Folate and neural tube defects P Folate and colorectal cancer ~ P Vitamin E and prostate cancer a 0 = Diet and Health: Implications for Reducing Chronic Disease Risk (D&H) report (NRC, 1989), ~ = Dietary Reference Intake (DRI) reports (IOM, 1997, 1998, 2000a, 2001~. See text for description of types of studies. h A = accepted, P = promising, U = uncertain, N = no relationship. Some animal and mechanistic studies may have been cited in review articles in the DRI reports. c Indicates extension to include adults, not just children. Confidence changes from promising to uncertain for diet and Tom uncertain to no relationship for dietary supplements.
I FINDINGS 55 Observational Clinical Trials Small Large Case Control Cohort Randomized Randomized Retrospective Prospective Nonrandomized (< 1,000) (> 1,000) O O O O o O O lo O o O o O O o O O O O O lo O O O lo lo e Except prostate cancer. f CHD = coronary heart disease. g For phosphorus, the D&H and DRI reports did not include any studies that directly assessed the effect of dietary intake of phosphorus on bone, but only on phosphorus absorption or serum phosphorus. In the DRI report, for young children only, data on measures of whole body bone mineral content were used to estimate accretion of phosphorus in the body during growth
56 EVOLUTION OF EVIDENCE In the case of folate and NTDs, large clinical trials substantiated the role of folate in NTD risk reduction. The large trials reported in 1991 and 1992 were supported by the collective findings of observational studies published primarily between 1988 and 1995. None of the observational studies were able to identify folate specifically as the responsible vitamin. The definitive data from the clini- cal trials led scientists and public health policymakers to accept as conclusive the positive association between supplemental periconceptional folate use and NTD risk reduction. Large, randomized controlled trials are generally given greater weight in evaluating nutrient-disease relationships. In such trials, participants are assigned randomly to a study group that receives the treatment or to a control group. This assignment is intended to evenly distribute known and unknown confounding factors so that the randomized groups are comparable. Even under these circum- stances, large-scale trials have limitations (IOM, 1997, 1998, 2000a, 2001; NRC, 1989) that may lead to failure to find an effect, either positive or negative. These include short duration of trial, level of nutrient supplement given, con- founding by other nutrients administered, and incomplete compliance of the study group. Trends in nutrient intake or disease risk factors can also alter a rela- tionship over time. Perhaps most importantly, the case studies of vitamin C and gastric cancer and of vitamin E and CHD highlight the difficulty in conducting large-scale trials to investigate potential beneficial effects of single nutrients in reducing risk for a chronic disease like cancer or CHD, especially when compared with conditions that develop over a relatively short time, like NTDs. Chronic diseases develop over a long period of time, typically over decades, and may be affected by various factors at different times in the disease process. Nutrient trials have been more successful in establishing causality for conditions that develop over a much shorter time, such as was the case for trials aimed at preventing birth de- fects and caries. CONCLUDING REMARKS Conclusive evidence about a relationship between specific nutrients and a disease or diseases remains typically elusive for a number of reasons. First, sci- entific advances in understanding relationships between specific nutrients and diseases do not necessarily emerge within a short time, and progress is often erratic. Some gaps are filled while others are created. Initial findings based on preliminary evidence are sometimes confirmed and other times not supported. Not surprisingly, the process is time consuming and costly. Second, while pre- liminary evidence, typically from mechanistic studies, animal studies, and ob- servational studies in humans, can generate exciting new hypotheses about nu- trient-disease relationships, evidence from these studies has limitations. For instance, even in well-designed, large-scale observational studies, it is difficult
FINDINGS 57 to isolate the effects of a single nutrient under investigation from the confound- ing effects of other nutrients and from non-nutritional factors, or a mechanistic study may not consider multiple actions of a nutrient. Third, the etiology of dis- ease, especially chronic disease, is commonly multifactorial. Even if diet has a prominent role, it is extremely unlikely that a single nutrient is directly respon- sible for a chronic disease, and conversely that addition of a single nutrient will eliminate disease risk. Fourth, clinical trials, which are generally considered to provide the strongest evidence about the effects of nutrient intake on subsequent disease, are complex, expensive, and time consuming, especially for chronic diseases, which develop over decades and are influenced by a host of genetic, physiological, and environmental factors that may also affect risk. Fifth, it is possible that a focus on specific nutrients as risk factors for dis- eases in relatively homogenous populations has led to a number of spurious as- sociations that are not subsequently supported by clinical trials. Historically, the substantial success of understanding the relationships between nutrients and classic nutrient deficiency diseases in humans has been driven by clinical obser- vations where the risk of disease was much higher in a subset of the population. Additionally, the etiology was specific to one disease, and short-term interven- tion trials could prevent the disease. Although the multifactorial nature of chronic diseases in general increases the complexity of nutrient~isease rela- tionships, greater emphasis on understanding the interrelationships between ge- netic factors and specific nutrients might contribute to more developed hypothe- ses for further clinical testing, as might development of biomarkers and surrogate endpoints. Unfortunately, society's expectation of a rapid translation of scientific dis- covery into application lacks an appreciation of the constraints on defining nu- trient~isease relationships as described above. Advances are commonplace in other fields, and advancement in medical research is likewise expected to occur quickly. It is understandable that persons at high risk for a disease or afflicted with a disease sometimes view the scientific method as tedious and unrespon- sive to their needs. Likewise, entrepreneurs and businesses that have invested in research and development want their products to reach and succeed in the mar- ketplace as quickly as possible. In its qualitative assessment of 14 case studies of nutrient-disease relation- ships, the committee was unable to find a way to predict from preliminary evi- dence whether there is a real beneficial effect of a single nutrient in reducing risk for a given disease or condition. Indeed, while the committee recognizes the urgency of understanding the potential benefits of nutrients in disease preven- tion, the findings suggest a cautious approach. The dearth of confirmatory evi- dence on aspects of vitamin E in preventing CHD, and the discovery of possible harmful effects from trials of high-dose p-carotene supplements, illustrate the committee's concerns.
58 EVOL UTION OF EVIDENCE Using a case study approach, the committee looked for patterns that could streamline the scientific process and bring useful recommendations and informa- tion to consumers more rapidly. It did not find a "pattern express train." The committee's review of differences in evidence available in the D&H report (1989) and the DRI reports (1997-2001) suggests a skeptical approach to state- ments about beneficial effects of single nutrients based on animal, mechanistic, or observational studies alone, and argues against premature claims of benefit. For consumers, policymakers, and regulators, the committee's assessment is as follows: · Large, randomized controlled studies play an important role in establish- ing the relationship between nutrient intake and the risk of disease. Ideally, con- sumers should base decisions to change intake of specific nutrients on evidence from trials. Likewise, regulators and policymakers should rely heavily upon such evidence to guide nutrient recommendations. · Caution should be exercised in using preliminary evidence from non- controlled studies as the basis for recommendations for increased intakes of a nutrient. · Claims about nutrient-disease relationships are more easily made than scientifically supported. Because the implications for public health are so impor- tant, caution is urged prior to accepting such claims without supportive evidence from appropriately designed, typically large, clinical trials.
