U. S. Food and Drug Administration
Center for Food Safety and Applied Nutrition
Office of Nutritional Products, Labeling, and Dietary Supplements
May 4, 2001

Letter Regarding Dietary Supplement Health Claim

for Antioxidant Vitamins and Certain Cancers

(Docket No. 91N-0101)

Dear Mr. Emord:

This letter is in reference to the court decision directing the Food and Drug Administration (FDA or the agency) to reconsider the health claim "Consumption of antioxidant vitamins may reduce the risk of certain kinds of cancer" in dietary supplement labeling (Pearson v. Shalala, 164 F.3d 650 (D.C. Cir. 1999)). FDA previously sent to you replies on the three other health claims that the court directed FDA to reconsider, namely, folic acid and neural tube defects, fiber and colorectal cancer, and omega-3 fatty acids and coronary heart disease. We regret the delay in responding to you on this claim.

I. Procedure and Standard for Evaluating the Claim

In reconsidering this claim and the three other health claims that were the subject of Pearson, FDA has proceeded as described in the October 6, 2000, Federal Register notice entitled "Food Labeling; Health Claims and Label Statements for Dietary Supplements; Update to Strategy for Implementation of Pearson Court Decision" (hereinafter "the October 6 notice"). 65 Fed. Reg. 59,855 (2000). As noted below in section IV., FDA first gathered new scientific evidence on the claims by contracting for a literature search and publishing two notices in the Federal Register soliciting comments and data. After reviewing the updated body of evidence on the claims, FDA applied the "significant scientific agreement" standard by which the health claim regulations require the agency to evaluate the scientific validity of claims. Under this standard, FDA may issue a regulation authorizing a health claim only "when it determines, based on the totality of publicly available scientific evidence (including evidence from well-designed studies conducted in a manner which is consistent with generally recognized scientific procedures and principles), that there is significant scientific agreement, among experts qualified by scientific training and experience to evaluate such claims, that the claim is supported by such evidence." 21 C.F.R. § 101.14.

For claims that did not meet the significant scientific agreement standard, FDA next considered whether to exercise enforcement discretion for qualified claims about the substance-disease relationship. Consistent with the Pearson opinion, the agency considered whether consumer health and safety would be threatened by the claim, and, if not, whether the evidence in support of the claim was outweighed by evidence against the claim, either quantitatively or qualitatively. See 164 F.3d at 650, 659 & n.10. If the evidence for the claim outweighed the evidence against the claim and there was no health or safety threat, the agency went on to consider whether a qualified claim could meet the general health claim requirements of 21 C.F.R. § 101.14, other than the requirement to meet the significant scientific agreement standard and the requirement that the claim be made in accordance with an authorizing regulation. These requirements were not challenged in Pearson and therefore still apply.

In the October 6 notice, FDA explained that it would consider exercising enforcement discretion for a dietary supplement health claim that did not meet the significant scientific agreement standard if the scientific evidence for the claim outweighed the scientific evidence against the claim, if the claim included appropriate qualifying language, and if the other criteria listed in the notice were met. In that event, the agency explained, FDA would send a letter to the petitioner outlining the agency's rationale for its determination that the evidence did not meet the significant scientific agreement standard and stating the conditions under which the agency would ordinarily expect to exercise enforcement discretion for the claim. 65 Fed. Reg. at 59,856. The agency also stated that, conversely, if the scientific evidence for the claim did not outweigh the scientific evidence against the claim, or the substance posed a threat to health, or the other criteria for the exercise of enforcement discretion were not met, FDA would issue a letter denying the claim and explaining its reasons for doing so. Id.

Although the deadlines for FDA action in 21 C.F.R. § 101.70(j) apply to health claims that are submitted by petition, they do not apply to the four claims that were the subject of Pearson. FDA is reconsidering those claims under a court order that sets no specific deadlines but clearly contemplates prompt action because of First Amendment concerns and the agency's obligation to comply with court orders as soon as possible. FDA is issuing this decision letter on May 4, 2001.

II. Summary of Review

In the January 6, 1993 final rule concerning a health claim for antioxidant vitamins and cancer for conventional food (hereinafter "the 1993 final rule"), FDA considered the relationship between nutrients identified at that time as antioxidant vitamins (i.e., beta-carotene, vitamin C, and vitamin E) and cancer. 58 Fed. Reg. 2622 (1993). FDA authorized a health claim (codified at 21 CFR § 101.78) relating substances in diets that are low in fat and high in fruits and vegetables (foods that are low in fat and may contain dietary fiber, vitamin A and vitamin C) to a reduced risk of cancer. Id. While FDA did conclude that evidence supported an association of reduced risk of cancer and diets low in fat and high in fruits and vegetables, FDA also concluded that the evidence available at the time did not support an association of antioxidant vitamins, alone or in combination, and reduced risk of cancer. Id. at 2634. The available evidence did not resolve whether the observed protective effects of fruit and vegetable consumption against cancer risk are due to a single or combined effect of the antioxidant vitamins and other nutrients with antioxidant functions (i.e., selenium), to other nutritive components of such foods (e.g., dietary fiber), to unmeasured components of such diets (e.g., carotenoids, indoles or flavonoids), or to displacement of other known risk components (such as fats and calories) within the total diet. Id. Rather, FDA found that vitamins A and C and fiber are characteristic of protective foods and may serve as useful markers for identifying the types of foods which contribute to a dietary pattern that is associated with a reduced cancer risk. Id. at 2634-35.

Therefore, because of the limitations in the evidence, the authorized health claim for fruits and vegetables and cancer in 21 CFR § 101.78 characterizes the association between the reduced risk of cancer and consumption of fruits and vegetables, not the antioxidant vitamin component or some other components of those foods. Id. at 2635. The agency found that the scientific evidence was not sufficient to conclude that antioxidant vitamins are responsible for the protective effect of fruit and vegetable consumption against cancer risk. Id. FDA concluded that the scientific evidence did not provide the basis for significant scientific agreement among qualified experts that there is a relationship between antioxidant vitamins and a reduced risk of cancer and therefore did not authorize a health claim for that relationship. Id. at 2622. As explained in more detail in section IV. below, FDA also did not authorize a health claim for antioxidant vitamins and reduced risk of cancer for dietary supplements.⁽¹⁾

In response to Pearson, FDA has reconsidered the scientific evidence on the putative relationship between antioxidant vitamins and the risk of certain kinds of cancer. Both the agency's original 1991 - 1993 scientific review and its evaluation of the evidence that has become available since that time were conducted consistent with the principles and procedures articulated in FDA's Guidance for Industry: Significant Scientific Agreement in the Review of Health Claims for Conventional Foods and Dietary Supplements (December 1999).

Based on its review of the scientific evidence, including evidence published after January 6, 1993, FDA finds that: 1) The totality of the publicly available scientific evidence demonstrates a lack of significant scientific agreement among qualified experts as to the validity of a relationship between the intake of antioxidant vitamins (i.e., vitamin C and vitamin E)⁽²⁾ and reduced risk of certain kinds of cancer in the general population, and 2) the weight of the scientific evidence against the relationship between vitamin C or vitamin E, alone or in combination, as antioxidants, and reduced risk of certain kinds of cancer is greater than the weight of the scientific evidence for the relationship.⁽³⁾ Thus, the agency is not authorizing a health claim for a relationship between vitamin C or vitamin E, alone or in combination, and the risk of certain kinds of cancer or individual cancers (i.e., cancer of the bladder, breast, cervix, colon and rectum, oral cavity/pharynx/esophagus, lung, prostate, pancreas, skin, stomach). Further, based on this review, FDA is not exercising enforcement discretion for a qualified claim for a relationship between vitamin C or vitamin E, alone or in combination, and the risk of certain kinds of cancer or of individual cancers (i.e., cancer of the bladder, breast, cervix, colon and rectum, oral cavity/pharynx/esophagus, lung, prostate, pancreas, skin, stomach).

III. Safety Review

Under 21 C.F.R. § 101.14(b)(3)(ii), which was not challenged in Pearson and still applies to FDA's review of a proposed dietary supplement health claim, the use of vitamin C and vitamin E at levels to justify a claim must be demonstrated by the proponent of the claim, to FDA's satisfaction, to be safe and lawful under the applicable food safety provisions of the Federal Food, Drug, and Cosmetic Act (the act).⁽⁴⁾

The applicable safety provisions require, for example, that the dietary ingredient not present a significant or unreasonable risk of illness or injury under conditions of use recommended or suggested in the labeling or under ordinary conditions of use. 21 U.S.C. 342(f)(1)(A). Further, a dietary supplement must not contain a poisonous or deleterious substance which may render the supplement injurious to health under the conditions of use recommended or suggested in the labeling. 21 U.S.C. 342(f)(1)(D). Ensuring the safety of a dietary supplement that may bear a qualified claim is also consistent with the Pearson decision, in which the court stated that the agency could be justified in banning certain health claims outright if, for example, consumer health and safety would be threatened. See Pearson, 164 F.3d 650 at 657-60.

In its safety review in this matter, FDA considered its 1991 proposed rule (56 Fed. Reg. 60,624; November 27, 1991) and its 1993 final rule on antioxidant vitamins and cancer, in which FDA addressed the safety of vitamins C and E. In the 1991 proposed rule, FDA noted that the Surgeon General's report stated that amounts of vitamin C in excess of the Recommended Dietary Allowances (RDAs) may cause rare adverse effects, including gastrointestinal disturbances, iron overload in susceptible individuals, altered metabolism of certain drugs, precipitation of calcium oxalate kidney stones, altered absorption of several minerals, and interference with clinical laboratory tests. 56 Fed. Reg. at 60,635. Regarding vitamin E, FDA noted that the National Research Council's report entitled "Diet and Health" cited scientific evidence suggesting that large doses of vitamin E are relatively nontoxic. Id. at 60,637-38. However, as discussed below, vitamin E supplementation may increase the risk of prolonged bleeding time for some individuals.

FDA also considered the April 11, 2000 report of the Food and Nutrition Board, Institute of Medicine (IOM), National Academy of Sciences (NAS) on Dietary Reference Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids (hereinafter the April 2000 DRI Report). The April 2000 DRI Report (at 155) states that IOM's review of the scientific literature indicates that high vitamin C intakes generally are associated with low toxicity. IOM noted that adverse effects associated with very high vitamin C intakes include: diarrhea and other gastrointestinal disturbances, increased oxalate excretion and kidney stone formation, increased uric acid excretion, pro-oxidant effects, systemic conditioning (rebound scurvy), increased iron absorption leading to iron overload, reduced vitamin B₁₂ and copper status, increased oxygen demand, and erosion of dental enamel.

With respect to vitamin E, the IOM reported, in the April 2000 DRI Report (at 253), that some uncontrolled studies have found various adverse effects to be associated with excess intake of vitamin E, including fatigue, emotional disturbances, thrombophlebitis (i.e., inflammation of the veins), breast soreness, creatinuria, altered serum lipid and lipoprotein levels, gastrointestinal disturbances, and thyroid effects. Side effects have been reported with extended intakes of 1,600 to 3,200 milligrams per day. However, the IOM noted that these effects are not severe and subside rapidly upon reducing the dosage or discontinuing use. The April 2000 DRI Report (at 253) notes that hemorrhagic effects have been seen in experimental animals with very high doses of vitamin E and are corrected with supplemental vitamin K. The IOM reported in the April 2000 DRI Report (at 253) that vitamin E supplementation may increase the risk of prolonged bleeding time for individuals routinely ingesting non-steroidal anti-inflammatory drugs, such as aspirin, and anticoagulant drugs, or for individuals who have a vitamin K deficiency. The IOM noted that caution must be exercised in judgments regarding the safety of supplemental doses of vitamin E over multi-year periods, as available human data are based on small studies of relatively short duration.

Another potential concern about the safety of supplemental vitamin E raised by the IOM in its April 2000 DRI Report (at 254) was the apparent increase in mortality from hemorrhagic stroke seen in the Alpha-Tocopherol, Beta-Carotene (ATBC) Cancer Prevention Study. However, the IOM considered the findings in the ATBC study preliminary and provocative but not convincing until the findings are corroborated or refuted in further large-scale clinical trials.

Based on its review, the IOM has established "Tolerable Upper Intake Levels" (ULs) for both vitamin C and vitamin E. A UL is the highest level of daily nutrient intake that is likely to pose no risk of adverse health effects in almost all individuals. The IOM, in its April 2000 DRI Report (at 162), established the UL for adults for vitamin C at 2,000 milligrams per day from both food and supplement sources, based on the adverse effect of osmotic diarrhea. The April 2000 DRI Report (at 257) states that the UL for adults for vitamin E is 1,000 milligrams per day⁽⁵⁾ from sources other than those that occur naturally in foods, based on the potential adverse effect of an increased tendency to hemorrhage.

Finally, FDA notes unexpected increases in the incidence of some cancers in association with consumption of antioxidant vitamins, as reported in two of the studies identified in its current review. An intervention trial in Linxian, China found that there was an increased prevalence of gastric dysplasia and cancer among subjects receiving dietary supplements of 120 mg vitamin C and molybdenum (Wang et al., 1994). Because the supplement combined vitamin C and molybdenum, it is not clear that the increased prevalence of gastric cancer was a vitamin C effect; however, it cannot be ruled out that vitamin C contributed to this adverse finding. The incidence of bladder and stomach cancer, in the Alpha-Tocopherol, Beta-Carotene (ATBC) intervention trial in Finnish male smokers (ATBC Study Group, 1994), among subjects receiving vitamin E supplements, was reported to be above the incidence for subjects not supplemented with vitamin E. Based on a significant body of observational studies, this trial was designed to evaluate the effect of dietary supplement vitamin E on lung cancer; it was not designed to evaluate the effect of vitamin E on any other cancer. Thus, the enrollment protocols were not designed to evaluate and control for risks associated with cancers other than lung cancer, and therefore the occurrence of other cancers is subject to potential bias. Importantly, despite the availability of a significant body of human observational studies prior to this study, the vitamin E and lung cancer relationship was not supported. Moreover, the post-hoc analyses presented very mixed results at other cancer sites, with cancers at two sites appearing to benefit from vitamin E supplementation (i.e., prostate and colorectal) and two cancers appearing to have increased risk associated with vitamin E supplementation (i.e., bladder and stomach). The post-hoc findings are useful in generating hypotheses. The primary and post-hoc findings from this large, well-designed and well-conducted trial raise serious questions about the safety and effectiveness of vitamin E supplementation on cancer risk and underscore the critical need for more research to ensure that any suggestion of benefit or increased risk from vitamin E supplementation is real and that safe conditions of use of vitamin E supplementation can be ascertained.

The agency recognizes that there are potential safety concerns with the use of supplemental vitamins C and E that are currently not well defined. FDA is not currently authorizing a health claim nor exercising enforcement discretion for a qualified health claim for vitamin C or vitamin E, alone or in combination, and their relationship to certain kinds of cancer or to any individual cancers. As a result, FDA does not have to evaluate the safety of vitamin C or vitamin E dietary supplements. Should the scientific evidence change in the future, such that the agency would consider authorizing a health claim or exercising its enforcement discretion for a qualified health claim, FDA would consider these potential safety concerns at that time.

IV. Review of the Scientific Evidence

A. 1991-1993 Scientific Review

Congress enacted the health claims provisions of the Nutrition Labeling and Education Act of 1990 (the NLEA) to help consumers maintain good health through appropriate dietary patterns and to protect consumers from unfounded health claims. The NLEA specifically required the FDA to determine whether claims respecting 10 nutrient/disease relationships met the statutory requirements for health claims. Pub. L. No. 101-535, § 3(b)(1)(A), 104 Stat. 2353, 2361. The relationship between antioxidant vitamins and cancer was one of these 10 claims the agency was required to evaluate.

FDA began its review of these 10 claims by publishing a notice in the March 28, 1991, Federal Register (56 Fed. Reg. 12,932) requesting scientific data and information relevant to the claims. The agency also contracted with the Life Sciences Research Office (LSRO) of the Federation of American Societies of Experimental Biology (FASEB) for an independent scientific review of recent evidence on antioxidant vitamins and cancer. In November 1991, FDA published, in the Federal Register, a proposed rule (the 1991 proposed rule) setting forth its review of available scientific evidence and tentative conclusions with respect to authorization of a health claim for the relationship between antioxidant vitamins and cancer risk. 56 Fed. Reg. 60,624. In the 1991 proposed rule, the agency proposed not to authorize the use on foods, including dietary supplements, health claims relating to the association between antioxidant vitamins and cancer. The agency found that the data on the relationship between vitamin C and cancer risk were not consistent and had mostly been obtained in studies of consumption of foods containing high levels of vitamin C. Id. at 60,635-36. Regarding vitamin E, FDA found that the evidence for an effect of vitamin E on cancer risk was limited and inconclusive. Id. at 60,625. FDA tentatively concluded that there was not significant scientific agreement to support the use of a health claim relating to antioxidant vitamins and cancer. Id. at 60,624 and 60,638. The agency found that strong epidemiologic evidence existed that showed that consumption of fruits and vegetables, which tend to be rich in the carotenoids and vitamin C, were associated with reduced risk of cancers in some sites. Id. at 60,631 and 60,636. However, the agency found that, in most studies, it was not possible to determine from the available data whether a protective effect was due to the presence of vitamin C, beta-carotene, other nutrients, or combined effects of both vitamins and other dietary factors, such as fiber. Id. at 60,635-36.

While the proposed rule was pending, Congress passed the Dietary Supplement Act of 1992 (the DSA). Pub. L. No. 102-571, 106 Stat. 4500. The DSA imposed a moratorium on FDA's implementation of the NLEA with respect to dietary supplements until December 15, 1993. The DSA also directed FDA to repropose implementing regulations for dietary supplements by June 15, 1993, and provided that the proposed regulations would become final by operation of law if final rules were not issued by December 31, 1993.

In the 1993 final rule, FDA concluded that diets rich in fruits and vegetables, which are low in fat and generally are good sources of vitamin A (as beta-carotene), vitamin C, and dietary fiber, are associated with a reduced risk of cancer. 58 Fed. Reg. at 2634. However, the agency found that there was not significant scientific agreement as to whether the observed protective effects of fruit and vegetable consumption against cancer risk are due to a single or combined effect of the antioxidant vitamins and other nutrients with antioxidant functions (i.e., selenium), to other nutritive components of such foods (such as dietary fiber), to unmeasured components of such diets (for example, nonnutritive components such as carotenoids, indoles or flavonoids), or to displacement of other known risk components (such as fats and calories) within the total diet. Id. Regarding vitamin C, FDA found that the data were not sufficient to identify vitamin C, from among other substances in these foods, as being responsible for the observed protective effect against cancer and therefore, the data did not support a relationship between vitamin C and a protective effect against cancer. Id. at 2634-35. With respect to vitamin E, FDA found that the data were not sufficient to associate vitamin E's antioxidant effects with protection against cancer. Id. at 2633. The agency concluded that the scientific evidence does not provide the basis for significant agreement among qualified experts that there is a relationship between antioxidant vitamins (i.e., beta-carotene, vitamin C, or vitamin E) and a reduced risk of cancer. Id. at 2633. Therefore, FDA did not authorize a health claim for a relationship between intake of antioxidant vitamins and a reduced risk of cancer. Id. at 2634-35.

Because of the DSA's moratorium on implementation of the NLEA with respect to dietary supplements, the 1993 final rule applied only to health claims for conventional foods, not for dietary supplements. In response to the DSA's directive to issue proposed regulations specific to dietary supplements, FDA proposed, on October 6, 1993, not to authorize a health claim for antioxidants and cancer in the labeling of dietary supplements. 58 Fed. Reg. 53,296 (1993). The October 1993 proposal relied on the scientific review conducted as part of the antioxidant-cancer health claim rulemaking that concluded in January 1993. Because FDA did not issue a final rule by December 31, 1993, the October 1993 proposal became final by operation of law. 59 Fed. Reg. 436 (1994). Therefore, the only authorized health claim related the substances in diets that are low in fat and high in fruits and vegetables (foods that are low in fat and may contain dietary fiber, vitamin A, and vitamin C) to a reduced risk of cancer.

B. Current Scientific Review

FDA considered the antioxidant vitamins to include vitamins C and E and beta-carotene when the agency published the 1991 proposed rule and 1993 final rule concerning a health claim for antioxidant vitamins and cancer. 56 Fed. Reg. at 60,625 and 58 Fed. Reg. at 2622. Recently, the IOM/NAS evaluated the nutritional requirements for antioxidant-related nutrients. In its April 2000 DRI Report (at 42), the IOM defined a dietary antioxidant as "a substance in foods that significantly decreases the adverse effects of reactive species, such as reactive oxygen and nitrogen species, on normal physiological function in humans." The IOM concluded in its April 2000 DRI Report (at 43-44) that although beta-carotene and other carotenoids display antioxidant activity in vitro, there is inadequate evidence that they have antioxidant activity in vivo when consumed in food by humans and, therefore, do not meet the definition of a dietary antioxidant. The IOM considered only vitamins C and E and the mineral selenium to be dietary antioxidants. FDA concurs with the IOM definition of "dietary antioxidant" and the rationale expressed in the April 2000 DRI Report for why beta-carotene and other carotenoids do not meet that definition. Therefore, FDA does not now believe that it is appropriate to consider beta-carotene as an antioxidant vitamin in its review of the proposed health claim for a relationship between antioxidant vitamins and a reduced risk of certain kinds of cancer. Consequently, FDA is considering only vitamins C and E in this review and will refer to them as the "antioxidant vitamins" throughout the remainder of this letter.⁽⁶⁾

FDA's initial step in reconsidering the health claim for antioxidant vitamins and reduced risk of certain kinds of cancer in response to Pearson was to gather the relevant scientific evidence that had become available since the previous rulemaking on this topic. To update its previous review, the agency reviewed comments⁽⁷⁾ and data submitted in response to two Federal Register notices requesting scientific data and information, as well as data identified in a literature search. See 64 Fed. Reg. 48,841 (1999); 65 Fed. Reg. 4,252 (2000). The literature search covered publications that were issued after 1992.

During its 1991-93 review, FDA considered preclinical studies (studies not performed in humans) because such studies are useful for developing hypotheses or investigating mechanisms of putative relationships between food substances and physiological changes associated with disease risk. The available clinical data at the time of FDA's 1991-1993 review specifically relating to antioxidant vitamins, as opposed to data for foods containing antioxidant vitamins, were limited. However, the usefulness of data from preclinical studies is limited in that such studies cannot fully simulate human disease and physiology. Additionally, such studies cannot accurately estimate appropriate intake levels or the magnitude of effects in humans. Since FDA's 1991-93 review, results from a number of new human studies with antioxidant vitamin data have become available. In the current review, therefore, FDA focused its attention on human studies that quantitatively measured or estimated the intakes of vitamin C and vitamin E (alone or in combination) and that were specifically designed to test the effect of these antioxidant vitamins on cancer risk. The threshold criteria for selection of human studies as part of the evaluation were the same as those used in the 1991-93 FDA review of this health claim topic. See 56 Fed. Reg. at 60,629.

1. Intervention Trials

In an intervention study, the investigator controls whether the subjects receive an exposure (the intervention), whereas in an observational study, the investigator does not have control over exposure. Therefore, intervention studies generally provide the strongest evidence for an effect. Unlike observational studies, which provide evidence of an association between the substance and disease of interest, but not necessarily a cause and effect relationship, intervention studies can provide evidence of causal relationships or the lack thereof. Randomized controlled clinical trials are considered the most persuasive studies. When the results of such studies are available, they will be given the most weight in the evaluation of the totality of the evidence. See Guidance for Industry: Significant Scientific Agreement in the Review of Health Claims for Conventional Foods and Dietary Supplements, at 5.

A number of randomized, controlled, clinical intervention trials of vitamin C and vitamin E alone and in combination have been published since 1992 (Blot et al., 1993; Li et al., 1993; Roncucci et al., 1993; Zaridze et al., 1993; ATBC Study Group, 1994 (and other reports based on this study population); Dawsey et al., 1994; Greenberg et al., 1994; Kaugars et al., 1994; Wang et al., 1994; Hofstad et al., 1998; Liede et al., 1998; Mackerras et al., 1999; and Correa et al., 2000). These trials were most useful when they provided specificity regarding measurement of the substance (i.e., antioxidant vitamin), measurement of the disease or health-related condition, and evidence for evaluating a relationship between the substance and the disease or health-related condition. For example, some of these trials directly addressed the intake of dietary supplements of vitamin C, vitamin E, or a combination of both and a cancer endpoint (e.g., ATBC Study Group, 1994 (lung cancer)). However, some trials included other substances with vitamin C or vitamin E and thus lacked specificity of substance (e.g., Blot et al. (1993) and Wang et al. (1994): beta-carotene and selenium; and Li et al. (1993) and Dawsey et al. (1994): multivitamin and mineral dietary supplements). FDA also considered evidence from post-hoc analysis of intervention trials (e.g., Hartman et al., 1998 and Heinonen et al., 1998). A post-hoc analysis of an intervention trial is an analysis of data on an endpoint other than the primary endpoint tested in the intervention trial. Such a post-hoc analysis must be interpreted cautiously. Because the original intervention trial was not specifically designed to look at post-hoc endpoints, factors that may affect the results may not be controlled in the original intervention trial, thus potentially introducing bias into the results.

A number of clinical intervention cancer trials that investigated the relationship between vitamin C and vitamin E and the risk of colon cancer used a surrogate marker of cancer risk. A surrogate marker is a biological parameter that is associated with a disease, and for which there is evidence that altering the parameter can reduce the risk of the disease. A surrogate marker for cancer must be validated by evidence demonstrating that altering the surrogate marker does, in fact, affect the risk of developing cancer. Several studies used colorectal adenomatous polyp recurrence as a surrogate marker of colorectal cancer risk (e.g., McKeown-Eyssen et al., 1988; DeCosse et al., 1989; Roncucci et al., 1993; Greenberg et al., 1994; and Hofstad et al., 1998). Development of colorectal cancer is a multi-step process beginning with adenomatous polyps. Most colorectal adenomatous polyps remain as small non-malignant tubular polyps, but a small proportion grow into larger, more dysplastic polyps, which in turn evolve into malignant adenocarcinomas. Because all colorectal cancers are believed to develop from adenomatous polyps, polyp appearance is considered to be a surrogate marker for the cancer endpoint (Einspahr et al., 1997). Further, it has been established that removal of adenomatous polyps prevents the development of colorectal cancer (Winawer et al., 1993); that is, colorectal cancer does not develop in the absence of adenomatous polyps. Thus, the link between adenomatous polyps and subsequent colorectal cancer risk in humans has been established.

In a study of cervical cancer, Mackerras et al. (1999) used the rate of progression in cervical intraepithelial neoplasia (CIN) lesions as a surrogate marker for invasive cervical cancer risk. CIN is a precursor of cervical cancer. CIN lesions are pre-malignant tumors typically localized in the cervical epithelium. Some CIN lesions progress to more dysplastic stages and grow through the epithelial layer to become invasive cervical cancer. Development of invasive cervical cancer is a continuum from pre-invasive CIN stages to the invasive cancer stages that have spread through the epithelial wall (Rock et al., 2000). Therefore, the risk of developing invasive cervical cancer is directly related to the rate of progression of existing CIN lesions.

2. Observational Studies

FDA also reviewed observational studies in humans that specifically estimated the intake of antioxidant vitamins from food sources (e.g., fruit and vegetable or dietary supplement consumption), or that measured the level of antioxidant vitamins in the body (e.g., serum levels), and the impact on certain kinds of cancer. Though less persuasive than intervention studies, particularly with regard to the quantitative measure of antioxidant vitamins and to attribution of any relationship to antioxidant vitamins per se, observational studies can provide evidence of an association between the intake of the dietary substance and the disease or health-related condition. However, these studies often do not provide a sufficient basis to determine if this association is causal or coincidental. In general, observational studies (also commonly called "epidemiological" studies) include, in descending order of persuasiveness, cohort studies, nested case-control studies, case-control studies, cross-sectional studies, and population or ecological studies. In the prospective studies (cohort and some nested case-control), investigators recruit subjects and observe them prior to the occurrence of the outcome. In retrospective studies (case-control), investigators review the records of subjects and interview subjects after the outcome has occurred. Retrospective studies are usually considered to be more vulnerable to recall bias (error that occurs when subjects are asked to remember past behaviors) and measurement error (e.g., measurement of the substance using intake data or serum or plasma levels). Temporal association between dietary exposure and disease outcome is also difficult to establish. Accordingly, prospective studies are generally more persuasive than retrospective studies.

In prospective cohort studies disease-free subjects are recruited within a specified group of people (the cohort) and the intakes of the subjects are determined. The study tracks the subjects over an extended period of time to see whether they develop the disease under investigation. At the end of the follow-up period, the intakes of subjects who developed the disease during the follow-up period are compared to those subjects who did not develop the disease to discern intake patterns that are associated with the risk of the disease. FDA generally weighted the prospective cohort studies more heavily than other types of observational studies because prospective studies are generally considered the most persuasive type of observational study. Nested case-control studies are case-control studies that are embedded in prospective cohort studies. Nested case-control studies may be more like prospective or more like retrospective studies, depending on when and how the intake estimates were performed.

In retrospective case-control studies, subjects with existing diagnosed disease are enrolled in the study (the cases) and are matched by identifiable characteristics (e.g., age, race, gender) to disease-free subjects (the controls). The intakes of the two groups are compared to identify differences in intake patterns associated with risk for the disease. In cross-sectional studies, at a single point in time the individuals with a disease who have received a specific exposure are compared to the individuals without the disease who did not receive the exposure. Population (ecological) studies use grouped data to examine the relationship between dietary exposure and health outcome among populations. In these studies, the rate of a disease is compared across different populations and the investigators seek to identify population traits that may cause the disease. In this evaluation, FDA focused its attention on the more persuasive types of observational studies that evaluated the association of vitamin C or vitamin E, alone or in combination, with certain kinds of cancer in individuals. See Guidance for Industry: Significant Scientific Agreement in the Review of Health Claims for Conventional Foods and Dietary Supplements.

One of the inherent limitations of observational studies is the extent to which vitamin C or vitamin E intake of the subjects can be accurately assessed. Also, it often is not possible to isolate effects from vitamin C or E intake from the intake of other dietary components, as we noted in our 1993 review of these issues (see section II. above). These difficulties are encountered whether intake is assessed from dietary data, or from serum or plasma data as a surrogate for dietary data.

Intake of vitamins C and E based on food or supplement recall is difficult to accurately estimate. In general, there is considerable uncertainty in the quantitative measurement of habitual intake over long periods of time. Some studies typically use a retrospective food frequency or supplement questionnaire in which the study subjects are asked to recall their typical intakes (in terms of foods eaten, frequency of eating and serving sizes, and information about supplements used) during prior time periods. Such techniques are subject to recall bias, particularly for dietary factors thought possibly related to the disease. Further, there is uncertainty in the translation of food intake data into antioxidant vitamin intake data by calculation from food composition tables. The natural variability of foods and the effects of processing, storing, and preparation of the food on vitamin content make it impossible to accurately calculate antioxidant vitamin intake from food intake data. Moreover, it is not possible to isolate the effect of the nutrients of interest from the effects of other components in foods or dietary supplements. Problems also are encountered with obtaining data on composition of any supplements used, including data that reflect actual levels of nutrient intake. This makes it difficult to establish whether antioxidant vitamins or some other component of the diet is responsible for any observed benefit. In short, there are significant limitations to assessing dietary antioxidant intake data from observational studies and associating intake with the disease. Because the variable assessed in these studies may include the diet, dietary supplement intake, or a biological marker of dietary patterns and there is uncertainty involved in the estimates of antioxidant vitamin exposures from such data, the usefulness of these types of studies to differentiate effects of the dietary antioxidant vitamin component of the food from effects of other components of the food is limited.

In the case of vitamin E, dietary intake estimates from observational studies are particularly prone to difficulties in obtaining accurate measurements (see, for example, the April 2000 DRI Report at 245 and 247-8). Most nutrient data bases and analytical methods do not distinguish among the alpha-, beta-, gamma-, and delta- forms of tocopherols that occur in food. Only one tocopherol (i.e., alpha-tocopherol) is retained in the body and is the most bioactive form. Additionally, vitamin E intakes are dependent on the types and amounts of oils in the diet; some vegetable oils contain more gamma- than alpha-tocopherol. Accurate information on both amounts and composition of oils used in food processing and preparation is often not asked in interviews nor known by most study respondents.

A serum or plasma level of either vitamin C or E is difficult to interpret as a surrogate marker for intake unless frank deficiency is present. Vitamin C deficiency is rare in developed countries such as the United States, and vitamin E deficiency is so rare in humans generally that medical indications of deficiency cannot be compared with vitamin E intake (April 2000 DRI Report at 101, 202 and 210). In retrospective observational studies (and some prospective cohort studies), the serum or plasma measure is taken in subjects with existing disease. When such a measure is taken, it is not possible to determine whether a low plasma vitamin E or C level in a cancer patient is due to a lower nutritional status or is low as a result of the disease process itself. Hence, predictions based on such findings of higher or lower serum levels of a nutrient in test subjects versus controls are not scientifically credible because the nutrient may either be a contributing cause or a consequence of the disease.

In addition to the generic concerns with the use of plasma or serum values to estimate dietary intake of vitamin E or C, a plasma or serum vitamin E level is not a reliable surrogate measure of dietary intake. The correlation, if any, between dietary vitamin E intake and normal vitamin E plasma concentrations is not strong (April 2000 DRI Report at 210). Consequently, results from observational studies based on blood measurement of vitamin E intakes are not reliable and are particularly difficult to interpret. Predictions based on outcomes of such vitamin E observational studies are problematic and clinical intervention trials, that quantitatively measure actual vitamin E intakes, are needed to meaningfully evaluate a possible relationship between vitamin E intake and a reduced risk of cancer.

Serum or plasma measures of vitamin C also pose interpretive problems when intake is outside of typical dietary ranges. Dose-dependent absorption and renal regulation limit plasma levels when intakes are high and conserve body stores when intakes are low (April 2000 DRI Report at 100). Nonetheless, there is a direct relationship between serum or plasma vitamin C levels and recent vitamin C intake when that intake is within the range typically obtained from the diet. Thus, the scientific credibility of prospective studies of a possible association between vitamin C and cancer risk will depend on study design and subject population.

As a consequence of these and other inherent limitations, observational studies are much less useful than intervention studies in resolving the key issue from the 1993 evaluation; that is, whether the vitamin C or vitamin E, alone or in combination, is responsible for reducing the risk of some kinds of cancer that is observed with diets low in fat and high in fruits and vegetables, and that may contain dietary fiber, vitamin A, and vitamin C. The agency gave greater weight to well-designed and well-conducted intervention trials that directly addressed the intake of vitamin C or vitamin E, alone or in combination, in relation to a cancer endpoint. The agency gave the observational studies relatively low weight.

C. Evaluation of the Scientific Evidence

As with the agency's review of the available data in the 1993 final rule, the more recently available studies that the agency evaluated concerned the putative relationship between antioxidant vitamins and certain kinds of cancer. Because associations between intake patterns⁽⁸⁾ and cancer risk appear to be site related, the data from the 1991-1993 review and the current review are summarized by cancer sites: cancer of the bladder, breast, cervix, colon and rectum, oral cavity/pharynx/esophagus, lung, prostate, pancreas, skin, and stomach. Because of the sheer number of relationships between the antioxidant vitamins and types of cancer to be evaluated, FDA included the data used in its 1991-1993 review in the current scientific review.

1. Bladder Cancer

a. Vitamin C

FDA's review of the available scientific evidence did not identify any intervention trials that evaluated a possible relationship between vitamin C and the reduced risk of bladder cancer. Without relevant data from intervention trials, the agency must evaluate the results of observational studies to determine whether there is a potential relationship between vitamin C and bladder cancer risk. Regarding the relevant observational studies, FDA identified two prospective cohort studies (Shibata et al., 1992 and Enstrom et al., 1992) and two retrospective case-control studies (La Vecchia et al., 1989 and Vena et al., 1992).

No statistically significant association between vitamin C intake (both dietary and supplemental intake) and bladder cancer risk was found in a ten-year follow-up of the First National Health and Nutrition Examination Survey (NHANES I) cohort of 11,348 adults (Enstrom et al., 1992). Similarly, an eight-year follow up of 11,580 elderly subjects in a California retirement community cohort found no statistically significant association of dietary vitamin C intake and bladder cancer incidence among elderly men; too few cancer cases occurred among the females in the cohort to evaluate cancer risk (Shibata et al., 1992). Shibata et al. (1992) did find, however, a statistically significant inverse association between use of vitamin C-containing multivitamin/mineral dietary supplements and bladder cancer incidence in elderly men. No statistically significant association between vitamin C and bladder cancer risk was found in either case-control study (La Vecchia et al., 1989 and Vena et al., 1992).

i. Consideration of Significant Scientific Agreement

Without relevant intervention trials to evaluate whether there is a relationship between vitamin C and bladder cancer risk, the agency considered whether the data from the observational studies are sufficient to establish such a relationship. Of the available evidence, one prospective cohort study (Enstrom et al., 1992) found no statistically significant association of bladder cancer risk with dietary or supplemental vitamin C intake. Another prospective cohort study found no statistically significant association of bladder cancer risk in elderly men with dietary vitamin C intake, but found a statistically significant inverse association with use of vitamin C-containing multivitamin/mineral dietary supplements in such men (Shibata et al., 1992). Neither of the retrospective case-control studies found any association between vitamin C and bladder cancer risk (La Vecchia et al., 1989 and Vena et al., 1992).

The lone finding of an association in a prospective cohort study (Shibata et al., 1992) provides insufficient scientific evidence to support a relationship between vitamin C and reduced bladder cancer risk. A single non-replicated result from an observational study does not provide a sufficient body of scientific evidence to permit a determination of whether a change in the dietary intake of the substance will result in a change in a disease endpoint. (See memorandum to the file in Docket 91N-0101 - "Replication of research findings" April 30, 2001.) Further, Shibata et al. (1992), which found a statistically significant inverse association between the use of vitamin C-containing multivitamin/mineral dietary supplements and bladder cancer incidence in elderly men, could not isolate the effect of vitamin C from other substances in the supplement products as being responsible for a possible association. In addition, the association found in Shibata, et al. (1992) with the use of vitamin C-containing multivitamin/mineral dietary supplements and bladder cancer incidence was not consistent with the findings with dietary vitamin C intake in that same study. Moreover, an association between dietary or supplemental vitamin C intake was not found in either the prospective cohort study (Enstrom et al., 1992) or the retrospective case-control studies (La Vecchia et al., 1989 and Vena et al., 1992). Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin C intake and a reduced risk of bladder cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin C intake and reduced risk of bladder cancer.

ii. Weight of the Evidence

The agency noted that the available evidence consisted of only four observational studies; two prospective and two retrospective studies. One prospective cohort study (Enstrom et al., 1992) found there was not a statistically significant association of bladder cancer risk with dietary or supplemental vitamin C intake. Another prospective cohort study found no statistically significant association between dietary vitamin C intake but a statistically significant inverse association between vitamin C-containing multivitamin/mineral dietary supplement use and reduced bladder cancer risk in elderly men (Shibata et al., 1992). Neither of the retrospective case-control studies found any association between vitamin C and bladder cancer risk (La Vecchia et al., 1989 and Vena et al., 1992).

The available evidence is limited and of low persuasiveness. This evidence includes one non-replicated observational study that suggests a relationship between vitamin C and reduced bladder cancer risk, with another similar type of observational study and a few other less persuasive observational studies that show no such relationship. The single finding of a suggested benefit is both unconfirmed and inconsistent with the results of the other available studies. The agency finds that there is an insufficient body of sound, relevant scientific evidence to support even a qualified claim⁽⁹⁾ about a relationship between supplemental vitamin C and reduced risk of bladder cancer in the general population. In order to make suggestions about any benefit of ingesting a substance to reduce the risk of cancer, without being false or misleading, there must be a credible scientific basis to do so. Thus, a certain threshold level of scientific evidence supporting the purported substance-disease relationship must be met to make a claim about such a relationship, even with a disclaimer that the available evidence is inconclusive or suggestive.⁽¹⁰⁾ Below this threshold, the agency would deem any qualified claim about such a relationship to be inherently misleading because there would be an insufficient scientific basis for the claim.

Thus, the agency concludes that the available observational data do not provide a sufficient body of sound, relevant scientific evidence to support the use of a qualified claim for a relationship between vitamin C and bladder cancer risk. Therefore, the agency is not providing for the use of a qualified claim about the use of vitamin C and a reduced risk of bladder cancer.

b. Vitamin E

FDA's review of the available scientific evidence identified a single intervention trial that evaluated, on a post-hoc basis, a possible relationship between vitamin E and bladder cancer risk (ATBC Study Group, 1994). The agency also identified three relevant prospective cohort studies (Shibata et al., 1992; Comstock et al., 1991; and Wald et al., 1987) and two retrospective case-control studies (Vena et al., 1992 and Riboli et al., 1991).

The Alpha-Tocopherol, Beta-Carotene Cancer Prevention intervention trial (ATBC Study Group, 1994) was designed to investigate the effects of beta-carotene and 50 milligrams of vitamin E daily on lung cancer risk among male Finnish smokers; incidences of cancers at other sites were also recorded. This 1994 ATBC study report states that there was a higher incidence of cancers of the bladder (9.6 versus 8.7 cases per 10,000 person-years) in the participants who received vitamin E supplements than in participants who received a placebo. Because this trial was designed to evaluate the effect of vitamin E on lung cancer, the enrollment protocols were not designed to evaluate and control for risks associated with other cancers, nor systematically to screen for and diagnose other cancers. Thus, the results with respect to bladder cancer risk must be interpreted with caution. With this caution in mind, FDA notes that the observation of a higher cancer incidence at two sites other than the lung (i.e., bladder and stomach), despite observation of a lower cancer incidence at two other sites (i.e., prostate and colorectal), suggests that there may be potential safety concerns. The results from the ATBC lung cancer prevention trial raise concerns about the safety of vitamin E supplementation and the ability of observational studies to predict benefit. These results underscore the critical need for more clinical research to ensure that any suggestion of benefit or increased risk from vitamin E supplementation is real, and that safe conditions of use for vitamin E supplementation can be ascertained.

No statistically significant association between vitamin E and reduced risk of bladder cancer was found in any of the three prospective cohort studies (Shibata et al., 1992; Comstock et al., 1991; and Wald et al., 1987). One retrospective case-control study (Vena et al. 1992) found no statistically significant association between vitamin E and bladder cancer risk. Conversely, the other retrospective case-control study (Riboli et al. 1991) reported a marginally significant reduction of bladder cancer risk associated with vitamin E intake. However, the Riboli et al. study may have introduced bias by including prevalent cancer cases (approximately 40 percent of the cancer cases). Prevalent cases include both patients who have survived the disease for a period of time and newly diagnosed patients. Case-control studies typically rely upon incident cases (newly diagnosed) rather than prevalent cases because the characteristics that contribute to survival of the prevalent cases may modify potential risk factors for the disease. Therefore, although the results of the two relevant case-control studies were mixed, the study finding an association (Riboli, et al., 1991) had design limitations which produced questionable results.

i. Consideration of Significant Scientific Agreement

The agency considered whether the available scientific evidence was sufficient to establish a relationship between vitamin E and reduced risk of bladder cancer. Because the ATBC trial was designed as a lung cancer prevention trial, the results cannot be relied upon to support any effect of vitamin E supplements other than those on lung cancer incidence. However, as already noted, the results from the ATBC trial do raise safety concerns about vitamin E supplementation and the ability to predict effectiveness of vitamin E supplementation on cancer risk. Thus, more research is needed to ascertain conditions of safe use and whether such use is associated with benefit or risk for certain cancers.

None of the three prospective cohort studies reported a statistically significant association between vitamin E and bladder cancer risk (Shibata et al., 1992; Comstock et al., 1991; and Wald et al., 1987). One retrospective case-control study also found no statistically significant association between vitamin E and bladder cancer risk (Vena et al., 1992). The single case control study (Riboli et al., 1991) that suggested an association between vitamin E intake and reduced bladder cancer risk was the least persuasive evidence available, and also had design limitations resulting in questionable results. Moreover, Riboli et al. (1991) could not isolate vitamin E from other substances in the diet as being responsible for a possible association.

A single non-replicated result from an observational study does not provide a sufficient body of scientific evidence to permit a determination of whether a change in the dietary intake of the substance will result in a change in a disease endpoint. The results from the ATBC intervention trial underscore the difficulty of predicting the safety or effectiveness of vitamin E supplementation on cancer risk. This study not only failed to support the hypothesized effect based on a body of observational studies, but also suggested that vitamin E supplementation is associated both with reduced cancer incidence at some sites and increased cancer incidence at other sites, including bladder cancer. This low predictability and confusion about the role of vitamin E dietary supplements in modifying cancer risk can be resolved only by further clinical intervention research to ensure that any suggestion of benefit or increased risk from vitamin E supplementation is real and that safe conditions of use from such supplementation can be ascertained. Thus, there is no strong, relevant, consistent body of observational evidence to support a causal relationship between vitamin E and bladder cancer. Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin E intake and a reduced risk of bladder cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin E intake and reduced risk of bladder cancer.

ii. Weight of the Evidence

The agency first considered the only available intervention trial evidence (ATBC Study Group, 1994). The ATBC Study Group (1994) results raise concerns about the safety of vitamin E supplementation, including bladder cancer, and the ability of observational studies to predict benefit. These results underscore the critical need for more research to ensure both that any suggestion of benefit or increased risk from vitamin E supplementation is real and that safe conditions of use for vitamin E supplementation can be ascertained.

In evaluating the observational evidence, the agency noted that the results from all three of the prospective cohort studies are consistent in finding no association between vitamin E and bladder cancer risk (Shibata et al., 1992; Comstock et al., 1991; and Wald et al., 1987). Concerning the less persuasive observational data, the results of the retrospective case-control studies were mixed. One retrospective case-control study found no statistically significant association (Vena et al., 1992), while another retrospective case-control study reported a marginally significant association between vitamin E and reduced bladder cancer risk (Riboli et al., 1991). The agency placed less weight on Riboli et al. (1991), compared to Vena et al. (1992), because of a limitation in the Riboli et al. (1991) study design. Therefore, the only evidence suggesting an association is a single retrospective case-control study with design limitations (Riboli et al., 1991), the results of which were marginally significant and questionable at best.

The results from the ATBC intervention trial, which suggest that vitamin E supplementation might be associated with both reduced cancer incidence or increased cancer incidence depending on the cancer site, raises serious questions and cause confusion about the role, if any, of vitamin E dietary supplements in modifying cancer risk, such that no disclaimer could render a claim for a relationship of vitamin E and reduced risk of cancer non-misleading. Further, FDA explained earlier in section IV.B.2. the difficulties in interpreting the results of observational studies of vitamin E and cancer. After reviewing the available data, including the post-hoc results from the ATBC intervention trial and the limitations associated with observational data on vitamin E, the agency concludes that the quality and quantity of the available scientific evidence do not support the use of a qualified claim for a relationship between vitamin E and reduced bladder cancer risk. Therefore, the agency is not providing for the use of a qualified claim about the use of vitamin E and reduced risk of bladder cancer.

2. Breast Cancer

a. Vitamin C

FDA's review of the available scientific evidence did not identify any intervention trials that evaluated a possible relationship between vitamin C and the reduced risk of breast cancer. Without relevant data from intervention trials, the agency must evaluate the results of observational studies to determine whether there is a potential relationship between vitamin C and breast cancer risk. Regarding the relevant observational studies, FDA identified seven prospective cohort studies (Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Verhoeven et al., 1997; Shibata et al., 1992; Zhang et al., 1999; and Enstrom et al., 1992), one prospective nested case-control study (Rohan et al., 1993), thirteen retrospective case-control studies (Landa et al., 1994; Ronco et al., 1999; Freudenheim et al., 1996; Rosenblatt et al., 1999; Bohlke et al., 1999; Ramaswamy et al., 1996; Yuan et al., 1995; Gerber et al., 1991; Graham et al., 1991; Katsouyanni et al., 1988; Toniolo et al., 1989; Zaridze et al., 1991; and Mannisto et al., 1999), and one meta-analysis (Howe et al. 1990).⁽¹¹⁾

Six of the seven prospective cohort studies found no statistically significant association between vitamin C and breast cancer risk (Enstrom et al, 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Shibata et al., 1992 and Verhoeven et al., 1997). The single prospective cohort study (Zhang et al., 1999) that found a statistically significant association between total dietary vitamin C and reduced breast cancer risk also found no such statistically significant association with use of vitamin C-containing multivitamin/mineral dietary supplements. These findings from Zhang et al. (1999), that vitamin C in the diet but not vitamin C in supplements was associated with breast cancer risk, suggest that dietary components of the types of foods that are high in vitamin C, but not vitamin C itself, affected the breast cancer risk, and that dietary vitamin C may have been a marker for those other dietary components. Similar to the findings of the majority of the prospective cohort studies, the results of the prospective nested case-control study also found no association between vitamin C and breast cancer risk (Rohan et al., 1993).

Most of the retrospective case-control studies also reported no statistically significant association between vitamin C and breast cancer risk (Katsouyanni et al., 1988; Toniolo et al., 1989; Graham et al., 1991; Gerber et al., 1991; Freudenheim et al., 1996; Rosenblatt et al., 1999; Ramaswamy et al., 1996; Bohlke et al., 1999; and Mannisto et al., 1999). The remaining studies reported a statistically significant decreased risk of breast cancer associated with vitamin C intake (Zaridze et al., 1991; Landa et al., 1994; Yuan et al., 1995; and Ronco et al., 1999). A 1990 meta-analysis of 12 retrospective case-control studies was identified in FDA's 1991 proposal (56 FR 60,624 at 60,633-34) as having found an association between estimated vitamin C intakes and breast cancer risk (Howe et al., 1990). However, meta-analyses must be reviewed with caution because such analyses are potentially subject to publication biases⁽¹²⁾ and can also magnify biases that are present in individual studies. Moreover, the results of this meta-analysis of pre-1990 case-control studies are not consistent with more recent evidence, including many prospective studies, showing no association between vitamin C and breast cancer risk.

i. Consideration of Significant Scientific Agreement

Without relevant intervention trials to evaluate whether there is a relationship between vitamin C and reduced breast cancer risk, the agency considered whether the data from the observational studies are sufficient to establish such a relationship. The evidence from the prospective studies consistently showed no statistically significant association between vitamin C and breast cancer risk (Enstrom, et al, 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Shibata et al., 1992; Verhoeven et al., 1997; Zhang et al., 1999; and Rohan et al., 1993). The single prospective study (Zhang et al., 1999) that showed a possible association between vitamin C and breast cancer risk found the association with vitamin C-containing fruit and vegetable intake, but not for use of vitamin C-containing dietary supplements, suggesting that dietary factors associated with fruit and vegetable intake other than vitamin C are responsible for their observed reduction in breast cancer risk. Therefore, the results of this study are not supportive of a relationship between supplemental vitamin C and reduced risk of breast cancer.

In the retrospective case-control studies, the majority of the studies reported no statistically significant association between vitamin C and breast cancer risk (Katsouyanni et al., 1988; Toniolo et al., 1989; Graham et al., 1991; Gerber et al., 1991; Freudenheim et al., 1996; Rosenblatt et al., 1999; Ramaswamy et al., 1996; Bohlke et al., 1999; and Mannisto et al., 1999). Four retrospective case-control studies reported a statistically significant association between dietary vitamin C intake and decreased risk of breast cancer (Zaridze et al., 1991; Landa et al., 1994; Yuan et al., 1995; and Ronco et al., 1999). These four case-control studies that reported an association each calculated dietary vitamin C intake from information about fruit and vegetable consumption. Thus, the studies could not isolate vitamin C from other components of high vitamin C diets as being responsible for a possible association.

There is no strong, relevant, consistent body of observational evidence to infer a causal relationship between vitamin C and breast cancer risk. In fact, the evidence supports a conclusion that such a relationship is not likely. Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin C intake and a reduced risk of breast cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin C intake and reduced risk of breast cancer.

ii. Weight of the Evidence

In weighing the evidence, the agency placed the greatest weight on the most persuasive of the available evidence, i.e., the eight prospective observational studies (Enstrom, et al, 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Shibata et al., 1992; Verhoeven et al., 1997; Zhang et al., 1999; and Rohan et al., 1993). The evidence from these studies consistently shows no association between vitamin C and breast cancer risk. The lone prospective observational study reporting an association between the intake of vitamin C-containing fruits and vegetables and reduced cancer risk (Zhang et al., 1999), found that the association did not extend to vitamin C-containing dietary supplement use. These findings in Zhang et al. (1999) suggest that dietary factors associated with fruit and vegetable intake other than supplemental vitamin C intake were responsible for the reduced cancer risk observed with the intake of vitamin C-containing fruits and vegetables. Further, the results from most of the retrospective case-control studies support the conclusion from prospective studies that vitamin C is not associated with breast cancer risk (Katsouyanni et al., 1988; Toniolo et al., 1989; Graham et al., 1991; Gerber et al., 1991; Freudenheim et al., 1996; Rosenblatt et al., 1999; Ramaswamy et al., 1996; and Bohlke et al., 1999). Therefore, based on the totality of the scientific evidence, the agency concludes that the scientific evidence against a relationship between vitamin C and reduced risk of breast cancer outweighs the scientific evidence for such a relationship.

b. Vitamin E

FDA's review of the available scientific evidence did not identify any intervention trials that evaluated a possible relationship between vitamin E and reduced breast cancer risk. Without relevant data from intervention trials, the agency must evaluate the results of observational studies to determine whether there is a potential relationship between vitamin E and breast cancer risk. FDA identified nine prospective cohort studies that evaluated the relationship (Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Zhang et al., 1999; Verhoeven et al., 1997; Shibata et al., 1992; Comstock et al., 1991; Knekt et al., 1988; and Russell et al., 1988). FDA identified two prospective nested case-control studies (Rohan et al., 1993; and Dorgan et al., 1998) and fifteen relevant retrospective case-control studies (Freudenheim et al., 1996; Rosenblatt et al., 1999; Bohlke et al., 1999; Van't Veer et al., 1996; Yuan et al., 1995; Favero et al., 1998; Mezzetti et al., 1998; Ronco et al., 1999; Mannisto et al., 1999; Torun et al., 1995; Gerber et al., 1989 and 1991; Richardson et al., 1991; Basu et al., 1989; and Toniolo et al., 1989). In addition, FDA evaluated a retrospective case-control study (Chajes et al., 1996) that evaluated vitamin E levels in breast biopsy tissue.

Eight of the nine prospective cohort studies found no statistically significant association between vitamin E and breast cancer risk (Knekt et al., 1988; Comstock et al., 1991; Shibata et al., 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Verhoeven et al., 1997; and Russell et al., 1988). The single study (Zhang et al., 1999) that found a statistically significant association between total dietary vitamin E and breast cancer risk found no such statistically significant association with the use of vitamin E-containing multivitamin/mineral supplements, suggesting that dietary components of the types of foods that are high in vitamin E, but not supplemental vitamin E, were responsible for the association observed in the study. Both of the prospective nested case-control studies reported no statistically significant association between vitamin E and breast cancer risk (Rohan et al., 1993 and Dorgan et al., 1998).

The results of most of the retrospective case control studies showed no statistically significant association between vitamin E and breast cancer risk (Basu et al., 1989; Toniolo et al., 1989; Gerber et al., 1989 and 1991; Richardson et al., 1991; Yuan et al., 1995; Freudenheim et al., 1996; Rosenblatt et al., 1999; Van't Veer et al., 1996; Ronco et al., 1999; and Bohlke et al., 1999); four studies reported statistically significant decreased risk of breast cancer associated with vitamin E intake (Torun et al., 1995; Mezzetti et al., 1998; Favero et al., 1998; and Mannisto et al., 1999). Although Mannisto et al. (1999) found a statistically significant inverse relationship of breast cancer risk and dietary vitamin E intake, they found no statistically significant association for total dietary plus supplement intake. This finding suggests that the observed association may be due to factors related to a vitamin E-containing diet rather than specifically to vitamin E dietary supplements. Chajes et al. (1996) measured alpha-tocopherol in breast adipose tissue obtained from breast biopsies and compared these tissue vitamin E levels of women diagnosed with malignant breast tumors (cancer) to those diagnosed with non-malignant breast tumors. While Chajes et al. (1996) found the mean vitamin E levels in the biopsy tissue from breast cancer patients to be only 15 percent that of vitamin E levels in biopsy material from patients with non-malignant tumors, it is unknown whether the depletion of vitamin E in breast adipose tissue was a consequence of low dietary intake or of the presence of breast cancer. Chajes et al. (1996) had no data regarding nutritional status of the subjects. As such, these results are not relevant to the question as to whether there is a relationship between vitamin E intake and risk of breast cancer.

i. Consideration of Significant Scientific Agreement

Without relevant intervention trials to evaluate a possible relationship between vitamin E and reduced breast cancer risk, the agency considered whether the data from the observational studies are sufficient to establish such a relationship. The evidence from the substantial body of prospective observational studies is consistent in finding vitamin E not to be associated with breast cancer risk (Knekt et al., 1988; Comstock et al., 1991; Shibata et al., 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Verhoeven et al., 1997; Russell et al., 1988; Rohan et al., 1993; and Dorgan et al., 1998). The single prospective study (Zhang et al., 1999) that showed a possible association between breast cancer risk and dietary vitamin E intake, found no association with use of vitamin E-containing multivitamin/mineral supplements suggesting that dietary factors other than vitamin E were responsible for the association with breast cancer risk. Therefore, the results of Zhang et al. (1999) do not suggest that there is a relationship between supplemental vitamin E and reduced risk of breast cancer. Results were mixed in the retrospective case-control studies. However, most of these studies also showed no statistically significant association between vitamin E and risk of breast cancer. The three retrospective case-control studies with dietary data that reported an association between vitamin E and decreased risk of breast cancer (Mezzetti et al., 1998; Favero et al., 1998; and Mannisto et al., 1999) cannot isolate vitamin E from other substances in the diet as being responsible for a possible association. Further, the retrospective study that relied on serum vitamin E (Torun et al., 1995) cannot distinguish between influences of the disease and influences of dietary intake on serum vitamin E values. There is no strong, relevant, consistent body of observational evidence to infer a causal relationship between vitamin E and breast cancer risk. In fact, the relevant, consistent body of evidence from prospective observational studies supports a conclusion that a relationship is not likely. Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin E and reduced breast cancer risk. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin E intake and reduced risk of breast cancer.

ii. Weight of the Evidence

In evaluating the observational evidence, the agency noted that most of the prospective studies (Knekt et al., 1988; Comstock et al., 1991; Shibata et al., 1992; Hunter et al., 1993; Kushi et al., 1996; Jarvinen et al., 1997; Verhoeven et al., 1997; Russell et al., 1988; Rohan et al., 1993; and Dorgan et al., 1998) consistently show no association between vitamin E and breast cancer risk. The results of the prospective cohort study by Zhang et al. (1999), which report an association of dietary vitamin E intake and breast cancer risk, but not an association of vitamin E-containing dietary supplements and breast cancer risk, suggest that it was not vitamin E in the diet responsible for the protective association. The results of the retrospective case-control studies were consistent with the results of prospective studies. Most of the retrospective case-control studies reported no association between vitamin E and breast cancer risk (Basu et al., 1989; Toniolo et al., 1989; Gerber et al., 1989 and 1991; Richardson et al., 1991; Yuan et al., 1995; Freudenheim et al., 1996; Rosenblatt et al., 1999; Van't Veer et al., 1996; Ronco et al., 1999; and Bohlke et al., 1999). Only four retrospective case-control studies (Torun et al., 1995; Mezzetti et al., 1998; Favero et al., 1998; and Mannisto et al., 1999), found an association between vitamin E and breast cancer risk, and these studies are flawed as discussed under IV.C.2.b. and b.i. above.

FDA explained earlier in section IV.B.2. the difficulties in interpreting the results of observational studies of vitamin E and cancer. After reviewing the available data, including the limitations associated with observational data on vitamin E, the agency concludes that the quality and quantity of the available scientific evidence do not support the use of a qualified claim for a relationship between vitamin E and a reduced risk of breast cancer. Therefore, the agency is not providing for the use of a qualified claim about the use of vitamin E and reduced risk of breast cancer.

3. Cervical Cancer

a. Vitamin C

FDA's review of the available scientific evidence identified one intervention trial (Mackerras et al., 1999), one prospective nested case control study (Wideroff et al., 1998), eight retrospective case control studies (Ho et al., 1998; Ramaswamy et al., 1996; Basu et al., 1991; Ziegler et al., 1990; Verrault et al., 1989; Brock et al., 1988; VanEenwyk et al., 1991; and Herrero et al., 1991) and one cross-sectional study (Giuliano et al., 1997) that evaluated a possible relationship between vitamin C and cervical cancer risk.

In a 2-year, randomized, double-blind, placebo-controlled intervention trial, Mackerras et al. (1999) evaluated the effect of daily intakes of 500 mg vitamin C and beta-carotene, on the progression of cervical intraepithelial neoplasia (CIN) lesions. Mackerras et al. (1999) randomized 141 women diagnosed with CIN into a 2x2 factorial design with daily intakes of 500 mg vitamin C or placebo and beta-carotene or placebo (i.e., there were four groups: one group received only placebos, one received vitamin C and the placebo for beta-carotene, one received beta-carotene and the placebo for vitamin C, and one received both supplements). The investigators found no effect of vitamin C supplements on the rate of progression of CIN lesions.

CIN is a pre-cancerous stage in the process leading to invasive cervical cancer. It is well established that most of squamous cell cancers of the cervix progress through a series of well-defined pre-invasive CIN lesions (Rock et al., 2000). In the pre-invasive stages, the squamous cell dysplasia is confined within the epithelial layer of the cervix (i.e., intraepithelial neoplasia, CIN, or squamous intraepithelial lesions, CSIL). Id. When the dysplastic lesion has progressed through the entire thickness of the cervical epithelium, it is considered as carcinoma in situ. Id. Involvement of the epithelial basement membrane is the threshold distinguishing carcinoma in situ from invasive cervical cancer. Id. Progression of CIN through the pre-invasive stages is usually a protracted process. During the pre-invasive stages the disease is easily detected by Pap smear screening and can be successfully treated. The rate of CIN lesion progression is directly related to the risk of the lesion progressing to invasive cervical cancer. Id. The risk of developing invasive cervical cancer is directly related to the rate of progression of existing CIN lesions. Therefore, the finding by Mackerras et al. (1999) that vitamin C supplementation has no effect on the rate of CIN lesion progression is evidence that vitamin C supplementation does not reduce the risk of cervical cancer.

Wideroff et al. (1998), in a prospective nested case-control study, found no association between vitamin C intake and development of CIN. Three retrospective case-control studies observed a statistically significant association between dietary intakes of vitamin C and decreased risk of cervical cancer (Verrault et al., 1989; Herrero et al., 1991; and VanEenwyk et al., 1991). However, Verrault et al. (1989) found that although dietary vitamin C intake was associated with reduced cervical cancer risk, regular use of vitamin C-containing dietary supplements was not. In addition, the results from VanEenwyk et al. (1991) may be confounded by selection bias because the study had very low response rates (50-60 percent). Further, the relevance of the results from Herrero et al. (1991) are in question because of differences between the U.S. population and the population sampled by Herrero et al. (1991) (in Mexico and South America) in both nutritional status and cervical cancer etiology. Two retrospective case-control studies found no statistically significant association between dietary intakes of vitamin C and cervical cancer risk (Brock et al., 1988 and Ziegler et al., 1990). Two of the three retrospective case control studies that compared blood levels of vitamin C in cases and controls showed an inverse association between blood vitamin C levels and cervical cancer risk. (Ramaswamy et al., 1996 and Ho et al., 1998) and the remaining study found no association (Basu et al., 1991). However, the results from case-control studies which use vitamin C blood levels as a surrogate for dietary intakes are difficult to interpret since it is not possible to tell whether blood levels are low because of low dietary intakes or whether low blood levels are a result of the disease itself. Consequently, the agency gave these studies very little weight in its analysis. Guiliano et al. (1997), in a cross-sectional study in 123 non-smoking, low-income Hispanic women in the U.S., found no statistically significant association between plasma vitamin C levels and grade of cervical dysplasia.

i. Consideration of Significant Scientific Agreement

The agency considered whether the available scientific evidence could establish a relationship between vitamin C and reduced risk of cervical cancer. Recent evidence from an intervention trial shows no protective effect of vitamin C supplements against progression of cervical intraepithelial neoplastic (CIN) lesions (Mackerras et al., 1999). Similarly, a prospective nested case-control study found no association between vitamin C intake and development of CIN (Wideroff et al., 1998). The remainder of the available evidence consisted of retrospective case-control and cross-sectional studies. Five of the eight retrospective case-control studies (Verrault et al., 1989; Herrero et al., 1991; VanEenwyk et al., 1991; Ramaswamy et al., 1996 and Ho et al., 1998) found an association between vitamin C and cervical cancer risk. The results from two of these studies (VanEenwyk et al. (1991) and Herrero et al (1991)) are not reliable because of design limitations as discussed above under IV.C.3.a. Verrault et al. (1989) found no association with vitamin C-containing dietary supplements. Further, the three case-control studies that found an association (VanEenwyk et al., 1991; Herrero et al., 1991; and Verrault et al., 1989) with dietary vitamin C intake, could not isolate the effect of vitamin C from other substances in the diet as being responsible for the association. The remaining case-control studies (Basu et al., 1991; Brock et al., 1988 and Ziegler et al., 1990) found no association between vitamin C and cervical cancer risk. The cross-sectional study (Giuliano et al. (1997) found no statistically significant association between plasma vitamin C levels and grade of cervical dysplasia. The well-designed vitamin C dietary supplement intervention trial by Mackerras et al. (1999), that demonstrates no effect of vitamin C supplements on cervical intraepithelial neoplasia progression, provides clear and compelling evidence that there is no relationship between vitamin C and reduced risk of cervical cancer. Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin C and reduced risk of cervical cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin C intake and reduced risk of cervical cancer.

ii. Weight of the Evidence

In weighing the evidence, the agency considered the results of the most persuasive type of evidence available, i.e., a randomized, double-blinded, placebo-controlled clinical intervention trial, that found no effect of vitamin C supplementation on reducing cervical cancer risk. The intervention trial (Mackerras et al., 1999) results provide clear and compelling evidence against an association of vitamin C dietary supplements and reduction of cervical cancer risk, based on CIN dysplasia progression. Of the available observational evidence, the results of the prospective study (Wideroff et al., 1998) were consistent with the results of the intervention trial in finding no association of vitamin C and pre-invasive cervical neoplasia risk. Results from the retrospective case-control studies were mixed and, as noted above, some of the case-control studies that suggested an effect had serious limitations that adversely affected the reliability of their results.

The agency considers results from well-designed, large, randomized, double-blinded, placebo-controlled clinical intervention trials to be the "gold standard" of scientific evidence to establish a relationship of a nutrient and reduced disease risk. Results from such a study (Mackerras et al., 1999) show no protective effect of vitamin C supplementation and cervical cancer risk. Therefore, based on the totality of the scientific evidence, particularly the compelling evidence from a vitamin C dietary supplement intervention trial, the agency concludes that the scientific evidence against a relationship between vitamin C and reduced risk of cervical cancer outweighs the scientific evidence for such a relationship.

b. Vitamin E

FDA's review of the available scientific evidence identified no intervention trials that evaluated a possible relationship between vitamin E and cervical cancer risk. Without any relevant intervention trials, the agency evaluated evidence from observational studies to determine whether there is a relationship between vitamin E and cervical cancer risk. FDA identified two prospective nested case-control studies that evaluated the relationship (Wideroff et al., 1998 and Knekt et al., 1988), four retrospective case-control studies (Potischman et al., 1991; Verrault et al., 1989; and Cuzick et al., 1990; Ho et al., 1998), and one cross-sectional study (Giuliano et al., 1997).

Both of the prospective nested case-control studies reported no association between vitamin E and cervical cancer risk (Wideroff et al., 1998 and Knekt et al., 1988). Knekt et al. (1988) analyzed data from a cohort of approximately 15,000 Finnish women and found no association between serum vitamin E and cervical cancer risk. Wideroff et al. (1998) analyzed data from a cohort of over 17,000 Portland, Oregon area women and found no association of dietary vitamin E intake and risk of CIN.

Both of the retrospective case-control studies that evaluated serum levels of vitamin E reported an association between serum levels of vitamin E and cervical cancer risk (Cuzick et al.1990 and Ho et al., 1998). However, in retrospective studies it is not possible to determine whether lower serum levels of vitamin E are due to lower intakes or to effects of the disease. The two retrospective case-control studies that evaluated dietary vitamin E intake reported either an association of dietary vitamin E intake and cervical cancer risk (Verrault et al., 1989) or no association (Potischman et al., 1991). However, as previously noted in section IV.B.2., it is difficult to accurately estimate vitamin E intakes. Thus, in the available observational case-control studies that evaluated a possible relationship between vitamin E and cervical cancer, it was not possible to attribute any effects to vitamin E per se, in those studies that suggested such effects, or to accurately estimate vitamin E intakes.

In a cross-sectional study, Giuliano et al. (1997) found an inverse association between grade of cervical intraepithelial neoplasia (CIN) lesion and plasma vitamin E in non-smoking, low-income Hispanic women in Tucson, Arizona. However, this finding depended upon relatively few cases of higher-grade lesions (only 12 had Grade II or Grade III lesions). The authors noted that due to the low numbers of women in the higher CIN-grade categories and the lack of histological confirmation for all subjects, conclusions about the relationship between vitamin E status and CIN cannot be drawn. Additional factors urging caution in interpreting these results include the possibility that the nutritional and health status, and thus predominant disease risk factors, of the study population do not reflect that of the general population.

i. Consideration of Significant Scientific Agreement

There were no relevant intervention trials to evaluate a possible relationship between vitamin E and reduced cervical cancer risk. The evidence from the prospective nested case-control studies (Wideroff et al., 1998 and Knekt et al., 1988) showed no statistically significant association between vitamin E and cervical cancer risk. None of the available studies was able to accurately estimate vitamin E intakes, or to isolate the effects of vitamin E from other components in the diet, where the data suggested that vitamin E was responsible for a protective effect. Moreover, the results among the available observational studies were mixed and inconsistent. Therefore, there is not a body of consistent, relevant, scientific evidence upon which a relationship between vitamin E and reduced risk of cervical cancer can be causally inferred. Thus, based on its review, FDA concludes that the totality of the available scientific evidence does not support a relationship between vitamin E and reduced risk of cervical cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin E intake and reduced risk of cervical cancer.

ii. Weight of the Evidence

The available evidence consisted of only two prospective nested case-control studies, four retrospective case-control studies, and one cross-sectional study. One of the two prospective nested case-control and two of the four retrospective case-control studies measured serum vitamin E levels and the other remaining studies measured dietary vitamin E intake. The two prospective nested case-control studies found no statistically significant association between vitamin E and cervical pre-invasive neoplasia or invasive cancer risk (Wideroff et al., 1998 and Knekt et al., 1988). Further, although two case-control studies that measured serum vitamin E levels found a statistically significant association between vitamin E and cervical cancer (Cuzick et al., 1990; and Ho et al., 1998), their usefulness is limited because of the limitations imposed by serum vitamin E measurement in these type of studies and the inability to make inferences based on such measurements. The results from the one prospective cohort study (Wideroff, et al., 1998) and the two retrospective case-control studies (Verrault et al., 1989; and Potischman et al., 1991) that measured dietary intakes of vitamin E were mixed. As stated earlier, it is difficult to estimate vitamin E intakes in observational studies, and therefore, the data from these three studies that estimated vitamin E intakes are questionable at best.

In summary, the agency finds the available evidence to be limited and of low persuasiveness. The agency finds that there is an insufficient body of sound, relevant scientific evidence to support even a qualified claim about a relationship between supplemental vitamin E and reduced risk of cervical cancer in the general population. In order to make suggestions about any benefit of ingesting a substance to reduce the risk of cancer, without being false or misleading, there must be a credible scientific basis to do so. Put another way, a certain threshold level of scientific evidence supporting the purported substance-disease relationship must be met to make a claim about such a relationship, even with a disclaimer that the available evidence is inconclusive or suggestive. Below this threshold, the agency would deem any qualified claim about such a relationship to be inherently misleading because there would be an insufficient scientific basis for the claim.

Thus, the agency concludes that the available observational data do not provide a sufficient body of sound, relevant scientific evidence to support the use of a qualified claim for a relationship between vitamin E and cervical cancer risk. Therefore, the agency is not providing for the use of a qualified claim about the use of vitamin E and a reduced risk of cervical cancer.

4. Colorectal Cancer

a. Vitamin C

FDA's review of the available scientific evidence identified six intervention trials (Greenberg et al., 1994; Roncucci et al., 1993; Hofstad et al., 1998; Paganelli et al., 1992; McKeown-Eyssen et al., 1988; and DeCosse et al., 1989), four prospective cohort studies (Eichholzer et al., 1996; Bostick, et al, 1993; Enstrom et al., 1992; and Shibata et al., 1992), and eight retrospective case control studies (Ferraroni et al., 1994; Whelan et al., 1999; Enger et al., 1996; LaVecchia et al., 1997; Benito et al., 1991; West et al., 1989; LaVecchia et al., 1988; and Freudenheim et al., 1990) that investigated a possible relationship between vitamin C and reduced colorectal cancer risk.

Five of the randomized intervention trials used the incidence of recurrent colorectal adenomatous polyps, a precursor of malignant cancer, as a surrogate marker of colorectal cancer risk (McKeown-Eyssen et al., 1988; DeCosse et al., 1989; Roncucci et al., 1993; Greenberg et al., 1994; Hofstad et al., 1998). Development of colorectal cancer is a multi-step process beginning with adenomatous polyps. Most colorectal adenomatous polyps remain as small non-malignant polyps, but a small proportion grow into larger, more dysplastic polyps, which in turn evolve into malignant adenocarcinomas. Because all colorectal cancers are believed to develop from adenomatous polyps, polyp appearance (i.e., incidence) is considered a surrogate for a cancer endpoint (Einspahr et al., 1997). Furthermore, it has been established that screening for and removal of adenomatous polyps prevents the development of colorectal cancer (Winawer et al., 1993); that is, colorectal cancer does not develop in the absence of adenomatous polyps. Thus, the link between adenomatous polyps and subsequent colorectal cancer risk in humans is established.

The standard colorectal polyp prevention trial protocol begins with colonoscopy screening of prospective subjects. All detected polyps are removed, and cancer-free subjects in whom at least one initial adenomatous polyp was found are enrolled in the study. The first follow up colonoscopy examination is scheduled within one year of the initial screening examination, and any polyps detected within one year are considered as polyps missed in the initial examination rather than new polyps. A second follow up colonoscopy examination is scheduled several years later to determine the rate of polyp recurrence. Because invasive colorectal cancer begins as an adenomatous polyp, a treatment that reduces the reappearance of adenomatous polyps is considered as reducing the risk of developing invasive cancer (Einspahr et al., 1997). The expected reappearance rate of adenomatous polyps in patients having had a previous adenomatous polyp is approximately 10 percent per year (Schatzkin et al., 1994). Therefore, use of the recurrence of adenomatous polyps, in subjects who had an initial polyp detected and removed, as a clinical trial endpoint provides much greater chance of detecting treatment effects on cancer risk than would a study of the actual cancer endpoint (Schatzkin et al., 1994). The incidence of colorectal adenomatous polyp recurrence correlates with dietary factors known to influence colorectal cancer risk, e.g., total fat, fruit, vegetable, and cereal grain consumption (Platz et al., 1997 and Giovannucci et al., 1992).

Among the five colorectal adenomatous polyp prevention intervention trials, two trials reported no statistically significant protective effect of vitamin C supplements on reducing colon cancer risk (McKeown-Eyssen et al., 1988 and Greenberg et al., 1994). The Greenberg trial (Greenberg et al., 1994) was a randomized, placebo-controlled 2x2 factorial design with a beta-carotene supplement and a combined vitamins E and C supplement as treatments. They reported no treatment effects on recurrent polyp incidence after 4 years of supplementation. The Greenberg study was the largest of the vitamin C supplement polyp prevention trials; 864 subjects were enrolled and 751 subjects underwent two planned follow up colonoscopy examinations (at one year and four years). Therefore, it is the study with the most statistical power to detect an effect of the vitamin supplements. The results of this trial show that vitamin supplementation for four years with vitamins C and E did not affect the rate of adenomatous polyp recurrence, a surrogate measure of colorectal cancer risk, in subjects who had adenomatous polyps removed before entering the study. Neither was the antioxidant vitamin supplementation effective for polyp prevention in any subgroup of subjects or in any subtype of polyp defined by size or location.

The polyp prevention trial reported in McKeown-Eyssen et al. (1988) used, as the intervention, a vitamin supplement combination consisting of vitamins E and C. This intervention trial included 185 subjects with adenomatous polyps at time of initial screening, 137 of whom underwent the planned 2-year follow up colonoscopy examination. The results of this trial show no statistically significant effect of the vitamin supplement on incidence of colorectal adenomatous polyp recurrence. Thus, this study is not supportive of a relationship between vitamin C supplementation and reduced risk of colorectal cancer.

Two polyp prevention intervention trials that reported a protective effect of antioxidant vitamin supplements had design limitations that preclude reliance on their results (Roncucci et al., 1993 and Hofstad et al., 1998). The Roncucci et al. (1993) trial was an intervention trial that randomized 255 subjects into one of three treatment groups, 1) vitamin supplement of vitamins A, E, and C, 2) lactulose, or 3) no treatment. Unlike the other polyp prevention trials, the Roncucci et al. (1993) trial was not placebo controlled. The authors reported that the incidence of polyp recurrence, for subjects reexamined between 12 and 65 months after entry into the trial, was reduced in the vitamin supplement group. However, the Roncucci et al. (1993) trial was compromised by a very low follow-up rate; approximately 80 percent of the subjects dropped-out before 24 months. The high dropout rate in this study makes the results difficult to interpret and possibly introduces bias. Moreover, this study was not placebo-controlled which also introduces bias. This study protocol did not have a means of determining compliance with the vitamin supplement-dosing regimen, other than asking patients if they had adhered to the treatment schedule. In summary, because of the high attrition rate and lack of placebo controls, the results cannot be relied upon.

The polyp prevention trial reported by Hofstad et al. (1998) included as the active treatment a vitamin-mineral supplement combination consisting of calcium, selenium, beta-carotene, and vitamins E and C. This three-year, placebo-controlled intervention trial included only 93 subjects with adenomatous polyps at time of initial screening. Colorectal cancer patients and patients who had sections of their colons surgically removed were also included in the study. Both of these conditions could influence subsequent recurrent polyp development and thus bias the study results. Hofstad et al. (1998) found that the combination vitamin-mineral supplement was protective against recurrent adenomatous polyps in study subjects with a single initial polyp, but was not protective for study subjects with multiple initial polyps. One-half of the study subjects included in the recurrent polyp analysis had multiple initial polyps. Because the supplement was not protective against recurrent polyps in one-half of the subjects, i.e., in those who had multiple initial adenomatous polyps, the applicability of the results are limited. Further, the study protocol allowed subjects to continue consuming self-selected dietary supplements in addition to the study-provided supplement, which confounded the results of this trial. Due to the limitation in the design of the study and the limited applicability of the results, the agency is using caution in interpreting the conclusions of this study with respect to the proposed claim.

The fifth vitamin C colorectal polyp prevention intervention trial, by DeCosse et al. (1989), studied patients with familial polyposis who have a hereditary predisposition to developing both a profusion of polyps and colorectal cancer. All subjects had undergone complete surgical removal of the colon and a portion of the rectum as a cancer preventive procedure prior to the study. Because the familial polyposis patients in this study did not have intact colons and because the etiology of colorectal cancer in these patients is different from that of the general population, the true effect of vitamin C or of vitamin E supplementation on colorectal polyps cannot be determined. Nevertheless, no effect of vitamin C and E supplementation on polyp recurrence was detected.

The sixth vitamin C intervention trial examined effects of antioxidant vitamin intervention on subsequent in vitro epithelial cell proliferation rates in tissues obtained from rectal mucosal biopsies (Paganelli et al., 1992). However, the agency did not find this study to be relevant to its evaluation because of the uncertainties involved in the in vitro measurement of mucosal proliferation and the uncertainties about the relationship of altered mucosal cell proliferation rates and risk of colorectal cancer.

Among the five polyp prevention intervention trials, the Greenberg trial (Greenberg et al., 1994) is the most persuasive in terms of study size, duration of intervention, and completeness of follow-up. All five of the polyp prevention trials used supplements with vitamin C in combination with other nutrients in their treatments. Consequently, the two trials that reported a protective effect (Roncucci et al., 1993 and Hofstad et al., 1998) were not able to distinguish an effect of vitamin C from potential effects of the other components of the test supplements. Considered overall, the results from these adenomatous polyp prevention trials do not support a protective effect of vitamin C against the risk of colorectal cancer.

Three prospective cohort studies found colon cancer risk not to be statistically significantly associated with vitamin C (Eichholzer et al., 1996; Enstrom et al., 1992; and Bostick et al., 1993). Although Bostick et al. (1993) found a statistically significant inverse association between vitamin C supplement use and colorectal cancer risk by comparing the lowest to the highest quintile of vitamin C supplement use, after adjusting for other dietary factors in a multivariate analysis of the same data they found no statistically significant association of dietary vitamin C intake and colorectal cancer risk. This suggests that the original univariate association may have been due to dietary factors other than vitamin C. Another prospective study reported a statistically significant inverse association between both dietary vitamin C intake and vitamin C supplement use and colorectal cancer risk in women (Shibata et al., 1992). However, Shibata et al. (1992) reported no association of dietary vitamin C intake or vitamin C supplement use and colorectal cancer risk in men (Shibata et al., 1992). Four retrospective case-control studies of colorectal cancer or colorectal polyp risk reported no statistically significant association between dietary vitamin C intake and colorectal cancer risk (West et al., 1989; Benito et al., 1991; Whelan et al., 1999; and Enger et al., 1996), while four others reported a statistically significant inverse association with dietary vitamin C intake (Freudenheim et al., 1990; Ferraroni et al., 1994; and LaVecchia et al., 1988 and 1997).

i. Consideration of Significant Scientific Agreement

The agency considered whether the available scientific evidence establishes a relationship between vitamin C and reduced risk of colorectal cancer. The largest and most persuasive of the polyp prevention intervention trials, Greenberg et al. (1994) showed no statistically significant effect of vitamin C supplementation on reducing colorectal cancer risk. The smaller polyp prevention intervention trial by McKeown-Eyssen et al. (1988) also showed no statistically significant effect. Limitations in the design and conduct of the trials by Roncucci et al. (1993) and Hofstad et al. (1998) preclude their results from being considered as sound, relevant scientific evidence. Also, the Roncucci et al. (1993) and Hofstad et al. (1998) trials would not have been able to distinguish an effect of vitamin C from potential effects of the other components of the test supplement because they used supplements with vitamin C in combination with other nutrients in their treatments. The DeCosse trial (DeCosse et al., 1989) is not relevant to this evaluation because their study subjects were familial polyposis patients with complete colectomies. The Paganelli trial (Paganelli et al., 1992 ) is not relevant to this evaluation because of uncertainties about the relationship between altered mucosal cell proliferation rates and cancer risk.

Three prospective cohort studies reported no statistically significant association between vitamin C and colorectal cancer risk (Eichholzer et al., 1996; Enstrom et al., 1992; and Bostick et al., 1993). Another prospective study reported a statistically significant inverse association between both dietary vitamin C and the use of vitamin C supplements and colorectal cancer risk in women, but reported no statistically significant association in men (Shibata et al., 1992). Among the retrospective case-control observational studies, four reported no statistically significant association between dietary vitamin C intake and colorectal cancer or colorectal polyp risk (West et al., 1989; Benito et al., 1991; Whelan et al., 1999; and Enger et al., 1996), while four others reported a statistically significant inverse association with dietary vitamin C intake (Freudenheim et al., 1990; Ferraroni et al., 1994; and LaVecchia et al., 1988 and 1997).

The well-designed vitamin C dietary supplement intervention trial by Greenberg et al. (1994), that demonstrates no effect of vitamin C supplementation on colorectal adenomatous polyp recurrence, provides clear and compelling evidence that there is no relationship between vitamin C and reduced risk of colorectal cancer. Further, the results of Greenberg et al. (1994) are supported by the results of a smaller polyp prevention intervention trial (McKeown-Eyssen et al., 1988) and a body of prospective observational study evidence. Two polyp prevention intervention trials that found associations between vitamin C and colorectal cancer risk (Roncucci et al., 1993; and Hofstad et al., 1998) had major limitations that limit the relevance and reliability of their results. Therefore, based on its review, FDA concludes that the totality of available scientific evidence does not support a relationship between vitamin C and reduced risk of colorectal cancer. Accordingly, the agency concludes that there is not significant scientific agreement among qualified experts that a relationship exists between supplemental vitamin C intake and reduced risk of colorectal cancer.

ii. Weight of the Evidence

In weighing the evidence, the agency considered the results of the most persuasive type of evidence available, i.e., a well-designed, randomized, double-blinded, placebo-controlled clinical intervention trial, that found no protective effect of antioxidant vitamin supplements (i.e., vitamin C and vitamin E) against colorectal cancer risk. Results from the Greenberg et al. (1994) vitamin C intervention trial, which is the largest polyp prevention trial, in terms of subjects completing the study and in duration of intervention, and thus is the study with the most statistical power to detect any differences between the vitamin supplemented and placebo groups, provides clear and compelling evidence against a relationship of vitamin C dietary supplements and reduction of colorectal cancer risk. A smaller polyp prevention intervention trial (McKeown-Eyssen et al., 1988) also found no statistically significant effect of antioxidant vitamin supplementation on adenomatous polyp recurrence. Two intervention trials that reported a protective effect (Roncucci et al., 1993; and Hofstad et al., 1998) had design limitations which produced results that are unreliable. Thus, FDA placed less weight on these two studies. The agency did not include the remaining two intervention trials (DeCosse et al., 1989 and Paganelli et al., 1992) in its consideration of weight of the evidence because of design limitations that raised questions about the relevancy of these results to the relationship between vitamin C and reduced risk of colorectal cancer in the general population. The majority of the prospective observational studies reported no statistically significant association between vitamin C and colorectal cancer risk (Eichholzer et al., 1996, Bostick et al., 1993; and Enstrom et al., 1992) although one reported a statistically significant association in women but not in men (Shibata et al., 1992).

The agency considers results from large, well-designed, randomized, double-blinded, placebo-controlled clinical intervention trials to be the "gold standard" of scientific evidence to establish a relationship of a nutrient and reduced disease risk. Results from such a study (Greenberg et al., 1994) show no protective effect of vitamin C supplementation on colorectal cancer risk. Therefore, based on the totality of the available scientific evidence, particularly the compelling evidence from a vitamin C dietary supplement intervention trial by Greenberg et al. (1994), the agency concludes that the scientific evidence against a relationship between vitamin C and colorectal cancer risk outweighs the scientific evidence for such a relationship. Thus, the agency is not providing for the use of a qualified claim about the use of vitamin C and a reduced risk of colorectal cancer.

b. Vitamin E

FDA's review of the available evidence identified six intervention trials (Greenberg et al., 1994; Roncucci et al., 1993; Hofstad et al., 1998; Paganelli et al., 1992; McKeown-Eyssen et al., 1988; and DeCosse et al., 1989) that investigated a possible relationship between vitamin E and colorectal cancer risk. The agency also identified three post-hoc analyses of colorectal cancer risk data from the ATBC Lung Cancer Prevention Study (ATBC Study Group, 1994; Malila et al., 1999; and Albanes et al., 2000). In addition, the agency identified eight relevant prospective cohort studies (Eichholzer et al., 1996, Shibata et al., 1992; Comstock et al., 1991; Schober et al., 1987; Wald et al., 1987; Stahelin et al., 1991; Knekt et al., 1988; and Bostick et al., 1993) and six relevant retrospective case-control studies (Benito et al., 1991; Ferraroni et al., 1994; Enger et al., 1996; Freudenheim et al., 1990; LaVecchia et al., 1997; and Whelan et al., 1999).

All five of the polyp prevention trials discussed above in the vitamin C section (Greenberg et al., 1994; Roncucci et al., 1993; Hofstad et al., 1998; McKeown-Eyssen et al., 1988; and DeCosse et al., 1989), included both vitamins C and E in their dietary supplement treatments. Accordingly, the agency's discussion and conclusions about these studies provided in the previous section on vitamin C apply to this section on vitamin E. To reiterate, the most persuasive intervention trial, Greenberg et al. (1994), found no protective effect of a vitamin E and C dietary supplement on colorectal cancer risk, as assessed by adenomatous polyp recurrence, a surrogate measure of colorectal cancer risk. McKeown-Eyssen et al. (1988) also found no statistically significant effect of a vitamin E and C supplement on the incidence of colorectal adenomatous polyp recurrence, as discussed above. The two trials that reported a protect