Chapter  75:  Effect of the Supplemental Use of Antioxidants Vitamin C, Vitamin E, and Coenzyme Q10 for the Prevention and Treatment of Cancer

Selection Criteria

Trials were included in the synthesis of evidence if they focused on vitamins C or E or coenzyme Q10 as supplements for the treatment or prevention of cancer and if they presented the results of clinical trials on human subjects or were a meta-analysis or systematic review or it they provided descriptive or background information about antioxidants. Language of publication was not a barrier to inclusion.

Antioxidants and Cancer

The April 2000 National Academy of Sciences/Institute of Medicine/Food and Nutrition Board report entitled Dietary Reference Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids defined the term dietary antioxidant, provided Dietary Reference Intakes (DRIs) for the antioxidant vitamins and minerals, and reviewed the evidence supporting a role for these nutrients in preventing or treating a variety of chronic diseases. A dietary antioxidant is a substance in foods that significantly decreases the adverse effects of reactive species, such as reactive oxygen and nitrogen species, on normal physiological functions in humans.¹ The DRIs provided by the report include four values: the Recommended Dietary Allowances (RDAs), the dietary intake level that is sufficient to meet the nutritional requirements of most healthy individuals in a particular age and gender group; Adequate Intakes (AIs), which are recommended intakes based on estimated or observed intake levels when data are inadequate to set an RDA; Tolerable Upper Limits (ULs), the highest level of nutrient intake that is likely to pose no adverse health risks for the majority of healthy individuals; and Estimated Average

Intakes (EARs), the intake that is estimated to meet the requirements of half the individuals in an age/gender group.¹ The committee that authored the report found laboratory and epidemiological evidence that diets rich in fruits and vegetables (foods that are high in antioxidants) may be associated with the prevention of certain types of cancer, but found a lack of scientific basis for recommendations regarding any specific nutrient supplement.² However, they were unable to establish a direct link between the intake of dietary antioxidants and the prevention of disease, recommending that further, large-scale studies were needed. It has been in the area of prevention of cancer rather than treatment of cancer that much of the discussion and work on antioxidants has focused.

The Origin of Ideas on Diet and Cancer Prevention

The discovery of the vitamins and their requirements in human nutrition was based on findings that the omission of these substances from the diet resulted in the acute appearance of constellations of symptoms. In preventing these symptoms, vitamins functio n as coenzymes and sometimes as hormones. However, in recent years, as some disease processes have been proposed to be the result of chemical oxidation, several of the vitamins have been found to have an additional set of functions, namely that of preventing such oxidation. The interest in a link between the intake of the antioxidant vitamins and chronic disease prevention is based on this proposed role for the antioxidant vitamins in human nutrition.

It has long been argued that the adequacy of the vitamin supply to cells and tissues influences the development, progress, and outcome of cancers.³ A major challenge to the integrity and function of cells and tissues is thought to come from chemical species called free radicals (defined as any atom or molecule whose nucleus has one or more unpaired electrons). The body produces these highly reactive species, such as the oxygen radical, as byproducts of some metabolic processes, the process of destroying foreign organisms, and in response to particular types of physical stress. The physiological processes that rid the body of toxins as well as those that metabolize drugs are also believed to lead to free-radical formation.

Free radicals may, in turn, join with other free radicals or react with non-radicals to cause chain reactions and form new radicals. These radicals can then attack cell membranes or their constituent lipoproteins, which is now believed to contribute to atherosclerosis. Free radicals can also attack the DNA, the chemical constituent of genes, causing mutations.⁴ The body has evolved antioxidant defenses to protect against free radical—induced damage; for example, cells manufacture repair enzymes that can “destroy free-radical-damaged proteins, remove oxidized fatty acids from membranes, and repair free-radical damage to DNA…”⁴ and additional, extracellular, antioxidant defenses also exist. Vitamins E and C, the carotenoids, and coenzyme Q10 are potentially involved in these antioxidant defenses. However, these defenses may not be totally effective. Thus, the concentration of free radicals in the body may continue to increase, more damage can occur, and, as a result, the body suffers oxidative stress.⁵ Severe oxidative stress may result in cell injury and destruction.

Given what we now know about the potential for free radicals to cause DNA damage and other types of destruction, a theory has developed that one possible cause of diseases such as cancer may be the uncontrolled effect of free radicals. For example, if the DNA damage were not repaired, mutations would result, which theoretically might lead to the development of cancers. If diseases could be caused by free-radical damage, it is also theoretically possib le that these diseases could be prevented by increasing the dietary intake of antioxidant-rich foods, foods rich in vitamins E and C and other substances shown to reduce oxidative damage in the laboratory. Yet, as Halliwell⁴ notes, while free radicals have been implicated in over 100 human diseases, this implication does not constitute proof of their role in disease formation or proof that treatments aimed at preventing free radicals can prevent or cure disease.

Cancer Development

Research on the association between vitamins and cancer has focused largely on the potential for prevention (although some research focuses on the role of vitamins in cancer treatment). Based on animal studies, carcinogenesis (the development of cancer) has traditionally been viewed as a two-stage process. ⁶ The first stage, initiation, involves irreversible mutation of genetic sequences, resulting in permanent changes to the genotype (the sequence of DNA carried by every cell). According to this model, this stage is necessary but not sufficient for cancer to develop. The second stage, promotion, involves a change in gene expression (the process by which genes are read and the proteins they encode are synthesized, resulting in a “visible” manifestation of the genotype), which requires a promoter substance and thus is believed to be reversible. Promoters enhance the expression of the mutated gene(s) and selectively stimulate the growth of cells that express these genes.

This two-stage model is now considered overly simplistic. Instead, carcinogenesis is now seen as a multistage process⁷ however, the two-stage model conceptualizes the belief that cells must undergo a permanent genetic change followed by a stimulus that selectively allows these cells to divide and escape normal growth controls.

The genetic changes necessary for cancer initiation are thought to be caused by the actions of mutagenic chemicals or physical forces (such as high doses of radiation) acting on the DNA, processes that may require metabolic activation. As mentioned above, if the damaged DNA is not repaired quickly, it becomes a stable mutation during the next cycle of DNA replication. Alternatively, miscoding may occur during the repair process.⁵ Some mutations that may result in cancer include those that result in the activation of proto-oncogenes and those that inactivate tumor suppressor and other antimetastasis genes.

The next phase, tumor promotion, enhances the probability of additional cumulative genetic damage, including endogenous mutations, by allowing expansion of the population of mutation-containing cells. The likelihood of cancer formation is further increased at this stage by exposure to other mutagenic agents.⁶

The Antioxidant Supplements and Cancer Prevention

Oxidative processes have been implicated in both initiation and promotion of cancer. Likewise, animal studies have revealed effects of the antioxidant nutrients on both processes.

Vitamin E exhibits antioxidant properties by acting as a lipid-soluble free radical scavenger in cell membranes. Thus, vitamin E may be involved in both initiation and promotion stages. Among the other potentially anticarcinogenic effects of vitamin E are its ability to inhibit formation of the carcinogenic chemical nitrosamine from nitrites in some foods, and its ability to promote immune system function.⁶

Vitamin C (ascorbic acid) also acts as an antioxidant, and through its ability to scavenge free radicals, it may have protective effects on biopolymers such as DNA. Like vitamin E, vitamin C may be protective for both initiation and promotion of carcinogenesis. Also, like vitamin E, it is thought to prevent formation of nitrosamine (by converting nitrite to nitrous oxide)⁶ and to influence immune system function. Vitamin C has also been reported to affect liver enzymes responsible for detoxification and transformation of carcinogens.⁶

Coenzyme Q10 (also termed ubiquinone and ubidecarenone) is an endogenously synthesized chemical (called a quinone) that is also obtained through food intake. In addition to its role as an electron carrier in the mitochondrial electron transport chain, it can also function as a soluble antioxidant.⁸ No recommended daily intake has been established for Q10, no deficiency has ever been shown,⁵ and no side effects have been observed at any dose level.⁹ In addition to its role as an antioxidant (possibly in conjunction with vitamin E)10 and as a free radical scavenger, other roles have been proposed for Q10. These include acting as a nonspecific stimulant for the immune system,⁹ and playing a role in membrane stabilization, prostaglandin metabolism, inhibition of intracellular phospholipases, and stabilization of calcium-dependent slow channels.¹⁰ Decreased levels of Q10 have been noted with aging and in such disorders as congestive heart failure, cardiomyopathy, cancer, hypertension, Parkinson's disease, spontaneous abortion, male infertility, chronic hemodialysis, and periodontal disease.¹⁰

As we have briefly reviewed, vitamins C and E and coenzyme Q10 have been implicated in a variety of potential anticarcinogenic processes. This report will review the evidence that supplements of these substances have the ability to prevent some types of cancer.

Technical Expert Panel

The SCEPC is advised on CAM topics by a group of technical experts regarding the search and inclusion criteria and appropriate analyses. The technical experts represent diverse disciplines including acupuncture, Ayurvedic medicine, chiropractic, dentistry, general internal medicine, gastroenterology, rheumatology, integrative medicine (the practice of combining alternative and conventional medicine), neurophysiology, pharmacology, psychiatry, psychoneuroimmunology, psychology, sociology, botanical medicine, and traditional Chinese medicine. The technical experts assisted the project in several ways. They aided us in identifying potential topics for review, appropriate sources of relevant literature, and technical experts for peer review; assessing our search strategies; and addressing specific questions in their areas of expertise. Appendix A lists members of the expert panel along with their affiliations.

Identification of Literature Sources

Potential evidence for the report came from three areas: on-line library databases, the reference lists of all relevant articles, and other sources such as identified experts and the personal libraries of project staff and their associates. The reference librarian at RAND identified traditional biomedical databases as well as databases that focus on the condition of interest and alternative and complementary medicine (Table 1).

We conducted four searches specifically on the interventions of interest. The full search strategies are displayed in Appendix B. We utilized the National Library of Medicine's controlled vocabulary thesaurus called Medical Subject Headings or “MeSH terms.” Limiting the output to human studies, we searched using the terms coenzyme Q10, vitamin E, vitamin C, and their many pharmacological synonyms (Table 2); the condition of interest (cancer); and study design or article type (randomized controlled trials, clinical controlled trials, meta-analyses, and systematic reviews). Because this report is focused on efficacy, clinical trials are preferred since they provide control groups which account for confounding factors. These searches yielded a total of 4595 titles, many of which were duplicates, because one article would appear repeatedly as each new search was added.

Two reviewers (a physician and a PhD) independently evaluated deduplicated lists of 1079 titles that the on-line database searches generated as well as 258 additional titles from other sources, such as professional libraries and reference mining. The reviewers read the lists of titles and accepted articles that:

• focused on vitamin C, vitamin E, or coenzyme Q10 for treatment or prevention of cancer, or the modification of a known risk factor for cancer or improvement in a pre-malignant state;

• focused on controlled trials on humans;

• presented a meta-analysis or systematic review of the interventions and condition;

• presented historical or descriptive background information about antioxidants and their use.

Articles that either reviewer classified as meeting these criteria were accepted. Articles were rejected that both reviewers considered:

• focused on a disease state that was not the topic of interest;

• contained animal or in vitro data unless human clinical trial information or significant background information was also included.

Language was not considered a barrier to inclusion.

From this stage of the screening process, the reviewers requested a total of 1337 articles, of which we were able to obtain 1125. Selected articles were further evaluated to see if they met the inclusion criteria. Based on this evaluation, we selected 432 that went on to further screening.

Using Microsoft Access database software, we tracked requests for articles. We used Pro-Cite as a link to read the citations into the Access database as well as to manage our reference list. We also used the database to produce and store our data collection instruments. Table 3 summarizes the search strategy shown in Appendix B. The details of the screening process are discussed in the next section.

Risk Ratio Analysis

For each of the three outcomes (death, tumors, and colonic polyps) that a trial reported, we estimated the log risk ratio comparing the relevant treatment group to either placebo or another comparison group as appropriate. We note that occasionally death or new tumor outcomes were further subdivided, e.g., death due to different types of cancers. We note further that some studies for the same trial would present comparisons in alternative ways. For example, one study might present death data for vitamin C and placebo groups separately, while another study presented death data for a different type of cancer for the vitamin E group versus all other study groups combined. The available data thus limited our ability to evaluate different comparisons for different outcomes. We also estimated the standard error of the log risk ratio for each trial and constructed a 95% confidence interval. We conducted the analysis on the log scale to stabilize the variance. We then back-transformed the log risk ratio and its confidence interval to the risk ratio scale for interpretability.

In summary, for each trial, comparison, and outcome for which data were available, we estimated the risk ratio (RR) and its 95% confidence interval. As an example of how to interpret a risk ratio, consider the outcome of all-cancer death when comparing the treatment of beta-carotene versus placebo. A risk ratio smaller than 1 indicates that a lower risk of death is associated with beta-carotene as compared to placebo.

For the death and new tumor outcomes, the trials were considered too heterogeneous to pool meta-analytically. For these outcomes, we present trial results individually with the separate outcomes defined as they appear in specific trial reports. In particular, three large trials (ATBC and the two Linxian Trials) were significantly different from each other and from the other small trials, so that meta-analytic pooling was not advisable. The main differences were study population (primary prevention versus treatment) and length of follow-up. For these large trials, we did consider whether we could combine related outcomes within trial. We note that we distinguish pooling results meta-analytically across trials from combining outcomes within a single trial. For example, a trial may report deaths due to different types of cancers separately. If clinically appropriate, we combined these deaths across all types of cancers reported, assuming that a patient's death could not be attributed to more than one type of cancer, so that we were not double-counting deaths in the combined count. For this new combined cancer death outcome that we created, we estimated a risk ratio as described previously.

Meta-Analysis for the Colonic Polyps Outcome

The trials that examined colonic polyps as an outcome were considered clinically homogeneous enough to warrant meta-analysis. We performed meta-analysis for any subgroup of three or more trials that had similar designs and comparison groups, and that measured colonic polyps for a particular type of cancer over similar follow-up periods.

For each subgroup of trials that qualified for meta-analysis, we estimated the DerSimonian and Laird random effects¹⁹ pooled log risk ratio, and its confidence interval. We also present the chi-squared test for heterogeneity p-value.²⁰ We back-transformed the pooled result to the risk ratio scale for interpretation, and present the pooled risk ratio, its 95% confidence interval, and associated forest plot. In this plot, each individual study risk ratio is shown with its confidence interval as a box whose area is inversely proportional to the estimated study variance. The pooled risk ratio and its confidence interval are shown as a diamond at that bottom of the plot with a dotted vertical line indicating the pooled estimate. A vertical solid line at a risk ratio of 1 indicates no treatment effect.

Publication Bias

For each subgroup of studies for which we conducted a meta-analysis, we assessed the possibility of publication bias by evaluating a funnel plot of the log risk ratios graphically for asymmetry resulting from the nonpublication of small, negative studies. Because graphical evaluation can be subjective, we also conducted an adjusted rank correlation test²¹ and a regression asymmetry test²² as formal statistical tests for publication bias.

Studies Reporting on Death from the ATBC Trial

The ATBC trial randomized 29,133 male smokers from Finland to receive one of four possible regimens: placebo, alpha-tocopherol alone (AT) (50 mg/day), beta-carotene (BC) alone (20 mg/day), or both vitamins. Patients were followed for a minimum of 5 years and a maximum of 8 years. Six studies from the ATBC trial reported on death due to a variety of different cancer types and reported sufficient data for analysis. Some, in addition, reported on all-cause mortality or all cancer mortality. Results from this analysis are included in Table 4. Please note that the results of the ATBC trial are reported in two different ways in Table 4.

Two studies looked at mortality from lung cancer but reported on this outcome in slightly different ways. Albanes et al²⁴ reported on death from lung cancer in each of the four treatment arms. The risk ratio (RR) for AT only versus placebo was 0.93 (95% CI: 0.48, 2.47) and for AT combined with BC was 1.15 (95% CI: 0.91, 1.45). The second study³⁹ combined arms so that results were reported as all arms using AT versus arms without AT. For all-cause mortality, the RR was 1.02 (95% CI: 0.96, 1.08) and for lung cancer death the RR was 1.02 (95% CI: 0.87, 1.2).

Three additional studies reported on death from a variety of other tumor types by individual treatment arms. Heinonen and colleagues³⁰ reported on the mortality from prostate cancer. For AT versus placebo, the RR was 0.61 (95% CI: 0.29, 1.29), and for AT + BC versus placebo the RR was 0.67 (95% CI: 0.32, 1.38). In an article by Virtamo et al,³⁸ deaths from urinary tract cancers were reported. For urothelial cell carcinoma, the RR of the AT versus placebo comparison was 1.20 (95% CI: 0.37, 3.93) and for AT + BC versus placebo comparison was 1.6 (95% CI: 0.52, 4.89). For renal cell carcinoma, the RR's were 0.79 (95% CI: 0.36, 1.73) and 0.72 (95% CI: 0.32, 1.61), respectively. Albanes²⁵ discussed deaths due to colorectal cancer. The RR for AT versus placebo was 1.09 (95% CI: 0.48, 2.47) and for AT + BC versus placebo was 1.18 (95% CI: 0.53, 2.64).

We calculated an RR for a combined death outcome, regardless of tumor type, for the four ATBC studies that reported their results for all four arms (this is not a meta-analysis because results are not pooled across trials).^{24, 25, 30, 38} The RR for this combined outcome for AT versus placebo was 0.91 (95% CI: 0.74, 1.12) and for AT + BC versus placebo was 1.08 (95% CI: 0.89, 1.32).

Finally, a single additional study,³⁷ which reported on mortality from pancreatic cancer, combined the treatment arms and reported all interventions with AT versus all interventions without AT. The RR for this comparison was 1.44 (95% CI: 0.93, 2.23).

Studies Reporting on Death from the Linxian Trials

The first Linxian Nutrition Intervention trial enrolled approximately 30,000 members of the general population of an area of central China that had a very high incidence of carcinoma of the esophagus and stomach. These patients (the General Population Group) were randomized to receive one of eight treatments. They were given either placebo, or formula A (retinol (5000 IU) and zinc oxide (22.5 mg)), or formula B (riboflavin (3.2 mg) and niacin (40 mg)), or formula C (ascorbic acid (120 mg) and molybdenum (30 μg)), or formula D (selenium (50 μg), and beta-carotene (15 mg), and alpha-tocopherol (30 mg)). These formulas were each given in combination with one of the other formulas and all four formulas were given together. No formula was given by itself alone. Interventions using formula C (containing vitamin C) in any combination and interventions using formula D (containing vitamin E) in any combination versus placebo are the comparisons of interest for this report. From the larger population, an additional population, already at higher risk of developing upper gastrointestinal tract cancers due to prior existence of dysplasia of the stomach and/or esophagus, was segregated for a separate trial. They were randomized to receive either a complex intervention—including beta-carotene (15 mg), vitamin A (10,000 IU), vitamin E (60 IU), vitamin C (180 mg), and multiple minerals daily—or placebo.³² This trial is referred to in the published literature as the Dysplasia Group. Patients were followed for 72 months in both the general study and the dysplasia study.

Two studies from the Linxian Trials had adequate statistics for further analysis. Blot et al⁴⁰ report a RR for all-cause mortality in the general Linxian population for any group that took formula C (RR = 1.02 (95% CI: 0.93, 1.1) and formula D (RR = 0.91 (95% CI: 0.84, 0.99)).

Specifically for death due to cancer, the RR for formula C versus placebo was 1.06 (95% CI: 0.92, 1.21) and for formula D versus placebo was 0.87 (95% CI: 0.76, 1.00). This study also reported each of the combination arms individually. No comparison reported a statistically significant benefit. The lowest risk ratio reported for an individual arm was found in the Blot⁴⁰ study in the A+D arm (cancer death RR = 0.75 (0.57, 1.00)).

Li and colleagues32 reported on the effect of the supplement intervention in the Dysplasia Group. The intervention was a combination of vitamins including both vitamins C and E. All-cause mortality had a RR of 0.94 (95% CI: 0.77, 1.16). All death due to all cancer had a RR of 0.98 (95% CI: 0.74, 1.31) and for death specifically due to esophageal cancer, the RR was 0.87 (95% CI: 0.56, 1.33).

Studies Reporting on Death from Trials Using Vitamin C as Treatment

Four studies, corresponding to four trials that tested the efficacy of vitamin C for treatment of patients with cancer, had sufficient statistics to proceed for further analysis. However, these trials were not pooled due to the heterogeneous nature of their populations and interventions.

Lamm et al³¹ evaluated the effect on patients with transitional cell carcinoma of the bladder of intravesicular and/or percutaneously administered bacillus Calmetee-Guerin (BCG), a substance used to provoke an immune response, combined with either the recommended daily allowance (RDA) of a number of vitamins or a dose of the same supplements which exceeded the RDA (referred to by the authors as megadose). The daily RDA dosage included vitamin A (5,000 IU), vitamin C (60 mg), and vitamin E (30 IU), as well as other vitamins. The megadose intervention included an 800% increase in vitamin A (40,000 IU), and a 3,300% increase in vitamin C (2,000 mg), a 1,330% increase in vitamin E (400 IU) per day, as well as similar increases in other vitamins. The RR comparing the effect of RDA doses of vitamins to the megadoses of vitamins for all-cause death was 0.86 (95% CI: 0.37, 2.01). Two studies examined the effect of vitamin C alone on cancer. Investigators attempted to replicated results reported by Linus Pauling⁴¹ for the efficacy of high-dose vitamin C to prolong survival in cancer patients.³⁵ They randomized patients with advanced carcinoma of the rectum and colon to receive either placebo or ten grams per day of vitamin C by mouth. The RR for all-cause mortality in this trial (vitamin C versus placebo) was 1.04 (95% CI: 0.69, 1.57). Another trial³⁶ tested the effects of three grams per day of vitamin C or placebo in a group of 27 breast cancer patients who had already received conventional therapy. The risk ratio for all-cause mortality comparing vitamin C versus placebo is 1.52 (95% CI: 0.72, 3.23). A final trial, the Heart Protection Study Collaboration Group,³⁴ randomly assigned 20, 536 adults (ages 40–80) to receive either antioxidant vitamin supplements (600 mg synthetic vitamin E, 250 mg vitamin C, and 20 mg beta-carotene daily) or a placebo. The study was primarily designed to assess cardiovascular endpoints in a high-risk population, however, all cause mortality was reported as well. The RR reported for this study was 1.04 (95% CI: 0.97, 1.11).

Summary of Results from the Analysis of Death

The results of all of the studies included in this analysis are summarized in Table 5. For the interventions tested, in the populations described, there is no evidence for a benefit for survival.

Trials Not Included in the Death Analysis

Nine studies were considered for the preceding analysis because they appeared to report on death, but they then were excluded due to insufficient statistics or other reasons. These studies will be briefly discussed here.

Albanes et al²³ reported on a pilot of the ATBC trial. Although the main focus of the pilot study was to assess enrollment and compliance issues, dropouts due to death or cancer were mentioned. These results were not separated either by outcome or intervention and thus were not amenable to analysis. In addition, it is not clear if these subjects were then included in the main study and had results reported again. This pilot study had a Jadad score of 5.

Two articles, corresponding to four studies from the Linxian General Population and Dysplasia Trials, were excluded. The first¹¹ represented two studies because it presented the methodology for the General Population Trial as well as the Dysplasia Trial that is described earlier in this section. The results of these pilot studies demonstrated that it was feasible to conduct these trials.

The third and fourth studies reported in the article from the Linxian Trial¹² reported that the total number of deaths for the general and intervention trials combined was 2,100; of those, 37% of the total deaths were due to cancer. In addition, the authors reported that none of the four combination supplements in the General Population Trial produced a significant reduction in the prevalence of death from esophageal, gastric, or other cancer.

Three studies corresponding to three unique trials involving the use of vitamin C from the same research group^{27, 28, 33} and one additional study of the efficacy of vitamin E to treat advanced stage cancer²⁹ were not included in the analysis. These results are described briefly here.

Creagan, Moertel, O'Fallon, and colleagues²⁷ reported on a randomized, double-blinded trial of 150 patients with proven terminal cancer of a variety of types, all of whom had previously received conventional care, including cytotoxic drugs. The intervention group received high-dose vitamin C (10 g daily), whereas the control group received a placebo. The two groups did not differ significantly in survival (mean survival for both groups was approximately 7 weeks). Because only survival-curve data were provided, it was not possible to include these results in the death risk ratio analysis.

This same research team published an abstract of what appears to be a subset of the previous trial.⁴² They had enrolled the first 128 of the expected 160 patients.

Finally, Moertel, Fleming, Creagan, and colleagues³³ conducted a second prospective, randomized, double-blind, controlled trial on the effect of high-dose vitamin C (2500 mg daily) in 100 patients with advanced colorectal cancer. In contrast to the previous study by this group, none of the patients in this trial had received prior treatment with cytotoxic drugs. The patients were followed as long as they could take the oral medications or until there was evidence of marked progression of the malignant disease. They were assessed at 4 weeks and every 8 weeks thereafter. Because only total deaths were reported, the results of the study could not be included in the pooled analysis. The vitamin C therapy showed no advantage over the placebo for either tumor progression or survival, and no patient with measurable disease had objective improvement.

The final treatment trial not included in the analysis was by Gogos et al²⁹ and focused on the effect of supplementation with omega-3 polyunsaturated fatty acids plus vitamin E on immune parameters in severely ill cancer patients. In a prospective, randomized, controlled study, 60 patients with generalized solid tumors were given either 200 mg of vitamin E daily plus 18 grams of fish oil daily or placebo until their deaths. None of the patients had received chemotherapeutic or immunomodulating agents during the 4 months prior to the study, and none were being treated for their tumors at the time of the study. Because only the survival curve was reported and showed the results could not be pooled in the death analysis.

Studies Reporting on New Tumor Development from the ATBC Trial

Seven studies of the ATBC trial reported on the development of a variety of new tumors. Results of this analysis are summarized in Table 6. Details of the design of the ATBC trial are discussed earlier in the section entitled Studies Reporting on Death from the ATBC trial.

An article by Varis et al⁴⁵ reported on the risk of developing carcinoma (all cell types). They analyzed their data separately for each of the four interventions. For developing a new carcinoma, they reported a RR of 1.04 (95% CI: 0.15, 7.32) for the AT versus placebo comparison and 1.84 (95% CI: 0.34, 10.01) for the AT + BC comparison.

Two studies reported on development of new lung carcinomas in the ATBC trial but in slightly different ways. The ATBC study³⁹ analyzed the results of the ATBC trial by combining all arms that had a particular intervention. Thus, their analysis reported on the relative risk of developing a new lung cancer in groups that did or did not take alpha-tocopherol without separating the AT-only group from the AT + BC group. Additionally, in the BC versus no BC comparison, the BC + AT group and the BC groups are combined. The RR for this study was 0.98 (95% CI: 0.86, 1.11). The second study to look at the development of new lung tumors, reported on in Albanes et al,²⁴ analyzed each of the four arms separately. They reported a RR for the development of new lung cancers for the AT versus placebo comparison of 0.98 (95% CI: 0.81, 1.19) and for AT + BC of 1.16 (95% CI: 0.96, 1.39).

All of the remainder of the ATBC studies analyzed each of the four intervention arms separately and focused on the development of a variety of different tumor types. Heinonen and colleagues³⁰ reported on the development of new prostate cancers. For the AT versus placebo comparison, the RR was 0.64 (95% CI: 0.44, 0.94), and for the AT + BC versus placebo comparison, the RR was 0.84 (95% CI: 0.59, 1.19). Virtamo et al³⁸ assessed the effects of AT and BC on development of urinary tract cancers. For development of urothelial cancers, the AT versus placebo comparison had a RR of 1.27 (95% CI: 0.83, 1.95), and for AT + BC the RR was 1.14 (95% CI: 0.73, 1.77). For development of renal cell cancer, the AT versus placebo comparison had a RR of 1.00 (95% CI: 0.59, 1.70), and for AT + BC, the RR was also 1.00 (95% CI: 0.59, 1.71).

For development of new pancreatic carcinomas, Rautalahti and colleagues³⁷ analyzed data from each of four arms separately as well as by combining all AT arms together. The RR for the AT versus placebo comparison was 0.96 (95% CI: 0.56, 1.66); for AT + BC versus placebo was 1.00 (95% CI: 0.58,1.72); and for AT versus no AT was 1.34 (95% CI: 0.88, 2.04). Albanes et al²⁵ reported on the development of new colorectal cancers. The RR for AT versus placebo was 0.78 (95% CI: 0.48, 1.27) and for AT + BC versus placebo was 0.81 (95% CI: 0.50, 1.31).

We combined the tumor outcomes of the five studies that reported on results by separate arms^{24, 25, 30, 37, 38} regardless of tumor type (this is not a meta-analysis because results are not pooled across trials). The resulting RR for the AT versus placebo comparison was 0.93 (95% CI: 0.81, 1.07); and for the AT + BC versus placebo, the RR was 1.05 (95% CI: 0.92, 1.20).

Studies Reporting on New Tumor Development from the Linxian Trials

Four studies from the Linxian Trials reported on the outcome of new tumor development. One study¹² reported on the development of new gastric and esophageal cancers in the General Population Trial. Three of these studies^{12, 32, 43} reported on similar outcomes from the Dysplasia portion of the Linxian Trial. One of these studies¹² could not be included in the analysis due to insufficient statistics. The details of this analysis are summarized in Table 7. Details of the designs of the Linxian General Population and Dysplasia Group Trials are included in the earlier section entitled Studies Reporting on Death from the Linxian Trials.

Taylor and colleagues¹² did not report directly on new tumors in the entire General Population Trial sample, although they noted that a total of 1,298 incident cancers were identified in this group. Instead, they reported on new tumor development in a subset of the entire General Population Trial sample who received an endoscopy with gastric and esophageal biopsies at the completion of the 5.25-year trial. This intervention was designed to find cancers that had not been identified by usual means during the course of the trial. The RR for presence of an additional new carcinoma on esophageal biopsy for the C versus no C comparison was 1.18 (95% CI: 0.27, 5.20) and for the D versus no D comparison was 0.71 (95% CI: 0.17, 2.95). The RR for the presence of a new tumor on gastric biopsy was 2.71 (95% CI: 0.55, 13.25) for the C versus no C comparison and was 1.22 (95% CI: 0.31, 4.79) for the D versus no D comparison.

Three studies also reported data on new tumor development in the Dysplasia Trial. Taylor and colleagues¹² noted that a total of 448 incident cancers were found in the Dysplasia Trial but did not separate the results by intervention. No additional analysis was performed on this group and the results were not reported in sufficient detail to permit further analysis. Li and colleagues³² reported risk ratios for the development of either esophageal or gastric cancer after 72 months of treatment. For esophageal cancer the RR was 0.96 (95% CI: 0.76, 1.22); for gastric cancer the RR was 1.19 (95% CI: 0.89, 1.58); and for all cancers combined the RR was 1.03 (95% CI: 0.87, 1.22).

Dawsey and colleagues⁴³ reported on the results of biopsies done of the gastric and esophageal areas during the Linxian Dysplasia Trial (30 months) and at the end of the trial (6 years). The number of cancers reported is smaller than in the Li³² study, because this study does not report on all incident cancers identified during the trial, just the cases identified by biopsy.

The RR for the presence of new esophageal tumors during the midpoint of the trial was overall 0.99 (95% CI: 0.60, 1.64); for the esophagus only it was 0.78 (95% CI: 0.41, 1.49); and for the stomach only it was 1.63 (95% CI: 0.75, 3.52). At the completion of the trial, the RR for the overall rate of the presence of a new tumor was 0.92 (95% CI: 0.54, 1.57); for the esophagus alone it was 1.53 (95% CI: 0.51, 4.58); and for the stomach only it was 0.91 (95% CI: 0.49, 1.68).

Studies Reporting on New Tumor Development from Other Trials

Two trials besides the ATBC and Linxian Trials reported on the outcome of new tumor development. Lamm et al³¹ was a treatment secondary prevention trial as opposed to a primary prevention trial. It tested the effect of vitamin supplementation on the development of new bladder tumors in patients previously treated for transitional cell carcinoma of the bladder. (The details of this study design are discussed earlier in the section entitled Studies Reporting on Death from Trials Using Vitamin C as Treatment.) Compared to RDA doses of vitamins, megadoses of vitamins were associated with a reduced RR of 0.50 for new bladder tumors (95% CI: 0.32, 0.78). The Heart Protection Study Collaboration Group,³⁴ randomly assigned 20, 536 adults (ages 40–80) to receive either antioxidant vitamin supplements (600 mg synthetic vitamin E, 250 mg vitamin C, and 20 mg beta-carotene daily) or a placebo. The study was primarily designed to assess cardiovascular endpoints in a high-risk population, however, new tumor development was reported as well. The RR reported for this study was 0.98 (95% CI: 0.89, 1.08).

Summary of Results from the Analysis of New Tumor Development Outcome

The results of all of the studies included in this analysis are summarized in Table 7. For the interventions tested, in the populations described, there is no evidence for a benefit for primary prevention of new tumors except for a single arm of the ATBC trial, AT versus placebo for the development of new prostate cancers. The single treatment/secondary prevention trial discussed did report that the addition of megadose vitamins conferred a benefit. However, the ability to generalize from this finding is limited because the intervention was multicomponent, preventing attribution of efficacy to any particular component. Additionally, all groups also received BCG, a major confounder.

Studies Not Included in the New Tumor Development Analysis

Four studies from the Linxian Trial were not included in the new tumor analysis.^{11 (Studies A&B), 12 (Study A), 40, 46} Blot's study reported adjusted risk ratios for the development of gastric or esophageal cancer. Insufficient information was given to calculate unadjusted risk ratios; therefore, these data were not included in the previous analysis. However, the adjusted risk ratios for this trial (adjusted for age and sex), reported by individual arm, were for all cancer (C versus no C) 1.06 (95% CI: 0.95, 1.18), and for D versus no D 0.93 (95% CI: 0.83, 1.03). For esophageal cancer (C versus no C), the adjusted RR was 1.06 (95% CI: 0.91, 1.24) and for D versus no D was 1.02 (95% CI: 0.87, 1.19). For gastric cancer, (C versus no C) the adjusted RR was 1.10 (95% CI: 0.92, 1.30) and for D versus no D was 0.84 (95% CI: 0.71, 1.00).

Li and colleagues' article¹¹ presents the methodology for the Linxian General Population and Dysplasia Trials. These studies and their new tumor results are described in the death analysis. As noted earlier, the study by Taylor¹² on the Linxian Dysplasia Trial could also not be included due to insufficient statistics.

Meta-Analysis for Secondary Prevention of Polyp Formation

Interventions in these four trials varied. There were no trials that used either vitamins C or E as single interventions. All treatment arms were not pooled because it was felt that vitamins C and E, with and without beta-carotene or vitamin A, were not equivalent interventions. Within the groupings of vitamins C and E with and without beta-carotene or vitamin A, there were insufficient studies to perform an analysis stratified based on dosage.

Specifically, the interventions and doses for these trials are herein discussed. Greenberg and his colleagues⁴⁷ gave 864 patients either placebo or beta-carotene (25 mg daily) or vitamin C (1 gm daily) and vitamin E (400 mg daily) or all three vitamins in a two-by-two factorial design for 4 years. McKeown-Eyssen et al⁴⁹ randomly assigned 200 patients to receive either placebo or vitamin C (400 mg daily) and alpha-tocopherol (400 mg daily) for 2 years. Hofstad and colleagues⁴⁸ gave 116 patients either placebo or a mixture of beta-carotene (15 mg), vitamin C (150 mg), vitamin E (75 mg), selenium 101 μg, and calcium (1.6 gm) daily for 3 years. For the trial performed by Roncucci and his colleagues,⁵⁰ 225 individuals were given either no treatment or lactulose (20 gm/day) or a combination of vitamin A (30,000 IU), vitamin C (1 gm), and vitamin E (70 mg) per day for an average of 18 months.

Trials Featuring Combinations of Vitamins C and E Only for Secondary Prevention of Polyp Formation

Two trials^{47, 49} of the four considered for pooled analysis had treatment arms that involved the combination of vitamins C and E without beta-carotene or vitamin A. Details of these trials are summarized in the Evidence Table. Because there were only two trials, a pooled analysis was not performed. The estimated risk ratios for these two trials, along with their 95% confidence intervals, are summarized in Table 8. A lower RR in this analysis favors treatment because it represents a lower likelihood of forming new colonic polyps as compared to placebo.

The RRs for the Greenberg and the McKeowen-Eyssen trials are not significantly different from 1; therefore there is no evidence that the combinations of vitamins C and E tested are more effective than placebo in the secondary prevention of recurrent adenomatous polyps of the colon.

Trials Featuring Combinations of Vitamins C and E with Beta-Carotene or Vitamin A for Secondary Prevention of Polyp Formation

Three trials of the four considered for pooled analysis used combinations of vitamins C and E with carotenoids compared with a placebo. Two trials were placebo-controlled. Details of these trials are summarized in the Evidence Table. The three interventions were considered sufficiently equivalent to allow pooling—even though calcium, which was included in the intervention used in the Hofstad trial,⁴⁸ has activity of its own in prevention of polyp formation, and vitamin A (not beta-carotene) was used in the Roncucci trial. The estimated RRs for these three studies, along with their 95% confidence intervals and the pooled estimate, are summarized in Table 9 and in Figure 2.

The pooled estimate yields a RR of 0.6, which is clinically important but not statistically significant (p = 0.13). In addition, the chi-squared test of heterogeneity is significant (p = 0.001), indicating a high degree of heterogeneity among these trials. Sensitivity analyses to account for heterogeneity could not be performed due to the small number of trials, but a visual inspection of Figure 2 suggests that heterogeneity may be due to the differences in population size and numbers of outcomes observed between Roncucci⁵⁰ versus Greenberg⁴⁷ and Hofstad.⁴⁸ For example, Hofstad has the smallest total sample size (n = 93) but has a number of outcomes that are proportionally greater than either of the other two trials. Conversely, for a relatively large total number of patients (n = 209), Roncucci develops many fewer new colonic polyps. It is likely, therefore, that a significant amount of this heterogeneity is due to patient population selection.

Assessing publication bias with so few trials is difficult at best. The funnel plot represented in Figure 3, although limited, shows no obvious bias. In addition, formal statistical tests revealed no statistically significant bias.

Trials Reporting on Polyp Formation Not Included in the Above Analysis

Three trials were not included in the prior analysis.^52–54 The details of these trials are included in the Evidence Table, and their salient features will be briefly discussed here.

The primary prevention trial reported by Malila and colleagues,⁵² as part of the ATBC Trial, was not appropriate for pooling with other secondary prevention trials on the grounds of clinical issues and study design. The other 2 trials^53–54 study populations at extraordinarily high risk of developing polyps. They were felt to be clinically dissimilar enough to preclude pooling of their data with populations with average risk. Malila and colleagues evaluated a subgroup (15,538) of the 29,133 male Finnish smokers enrolled in the trial. Subjects who were not known to have colonic polyps at the start of the trial were randomly assigned to one of four groups: vitamin E (50 mg/day, n = 3,890); beta-carotene (20 mg/day, n = 3,883); both supplements (n = 3,878); placebo (n = 3,887). Patients were followed for an average of 6.2 years. Whereas vitamin E supplementation resulted in a statistically significant increase in the risk for development of new adenomas (RR 1.66; 95% CI: 1.19, -2.32), it did not increase the risk for developing colorectal cancer. Beta-carotene had no effect on the risk of developing adenomas (RR 0.97; 95% CI: 0.69, -1.38) or cancer.

Bussey, DeCosse, Deschner, and colleagues⁵⁴ reported on a randomized, double-blind trial of 47 patients with polyposis coli, a familial condition that causes extensive polyp formation in the colon and leads to a high risk of cancer formation, who received either vitamin C (3 gm daily) or a placebo. At 21 months, rectal mucosal biopsies were taken from 31 subjects. The results showed a reduction in the polyp area for the vitamin C arm at the 9-month follow-up (p < 0.03) and trends toward a reduction in both the number and area of rectal polyps during the middle of the trial.

DeCosse, Miller, and Lesser⁵³ reported on a randomized, double-blind trial in 58 patients with familial adenomatous polyposis (polyposis coli) who had undergone a total colectomy with ileorectal anastomosis at least 1 year prior to the commencement of the trial and had a residual section of rectum and sigmoid colon (of on average 15 cm) that would be susceptible to new tumor formation. Patients were given either vitamin C (4 gm/day) and vitamin E (400 IU/day) or high-dose fiber (22.5 gm/day), or assigned to a control group that had only low-dose fiber (2.2g/day) plus placebo. The trial lasted 4 years, during which patients received serial proctosigmoidoscopies (total of 18). All new colonic polyps (recurrence) and growth in preexisting colonic polyps (progression) were recorded. The results suggested a limited effect of the supplements on rectal polyp recurrence and progression.

Trials Reporting on Intermediate Outcomes

Seven studies corresponding to seven trials (one of these studies reports on the ATBC trial) reported on a variety of intermediate outcomes relevant to the development of cancer or improvements in the risk factors for development of a particular type of cancer.^55–61

Two trials evaluated the effects of antioxidants on the development or progression of risk factors for gastric cancer.^{56, 62} One of the widely recognized precursors to intestinal gastric carcinoma is chronic atrophic gastritis with intestinal metaplasia. Zullo, Rinaldi, Hassan, and colleagues⁵⁶ conducted a randomized, controlled trial using vitamin C in patients with chronic atrophic gastritis. Following eradication of Helicobacter pylori, 66 patients with documented metaplasia received either vitamin C (500 mg/day) or no treatment (the method of randomization was not described) for 6 months. At the end of the trial, endoscopic examination demonstrated a significant decrease in the appearance of intestinal metaplasia in the vitamin C group compared with the control group. In addition, among those who presented with chronic inactive pangastritis with widespread metaplasia at entry, less extensive antritis with intestinal metaplasia was seen in the intervention group. The researchers concluded that vitamin C helps resolve intestinal metaplasia and may, by implication, be effective for primary prevention of gastric carcinoma in this high-risk group.

In a trial of patients with two precancerous, multifocal atrophic gastritis and dysplasia,⁶² subjects (n = 631 who completed at 72 months) were randomly assigned to one of 8 arms of the study: placebo; anti-Helicobacter pylori triple therapy; beta-carotene; vitamin C; anti-Helicobacter pylori plus beta-carotene; anti-Helicobacter pylori plus vitamin C; beta-carotene plus vitamin C; anti-Helicobacter pylori, plus beta-carotene plus vitamin C. The dosage was 30 mg per day for beta-carotene and 1 g twice a day for vitamin C. The anti-Helicobacter pylori therapy consisted of amoxicillin (500 mg 3 times daily), metronidazole (375 mg 3 times daily), and bismuth subsalicylate (262 mg 3 times per day). Patients were evaluated at 36 and 72 months. All three basic interventions (H. pylori treatment, beta-carotene, and vitamin C) in this high-risk population resulted in significant increases in the rates of regression. (RRs were 4.8; 5.1, and 5.0, respectively). The authors concluded that all three treatments might interfere with a precancerous process and provide benefit in a high-risk population.

Two trials evaluated the effects of an antioxidant intervention on the development of oral leukoplakia, which is considered to be a risk factor for the development of oral cancer.^{55, 63} Smoking increases the occurrence of oral leukoplakia and increases the number of keratinized cells in the epithelium of the tongue and palate, even when the mucosa in these areas is clinically normal. However, the link between these intermediate outcomes and cancer is not as clearly established as the link with other intermediate outcomes such as the presence of adenomatous polyps with adenocarcinoma of the colon. Nevertheless, the ATBC trial by Liede, Hietanen, Saxen, and colleagues⁶³ studied the impact of vitamin E and beta-carotene on the prevalence of oral mucosal lesions in smokers in a randomized, double-blind study. A random sample of 409 participants in a cancer prevention study (the ATBC study) was chosen to receive supplementation with either vitamin E (50 mg/day), beta-carotene (20 mg/day), or a placebo for 6 years. An oral examination was performed at the end of the trial. No statistically significant differences were found among any of the groups in the prevalence of oral mucosal lesions. In fact, only 24 patients in total had leukoplakia (5.9% of the total population) and only 7 of those 24 lesions were dysplastic.

In a randomized, controlled, double-blind fashion, 532 men between 50 to 69 years of age with oral leukoplakia and/or chronic esophagitis⁵⁵ were given either riboflavin, or retinol with beta-carotene and vitamin E, both, or a placebo. A significant reduction in oral leukoplakia was noted after 6 months for those receiving retinol, beta-carotene and vitamin E. After 20 months, no effect was seen in chronic esophagitis—although the risk for progression was lower in the retinol, beta-carotene and vitamin E group. Doses of 100,000 IU of retinol, 80 mg vitamin E, and 80 mg of riboflavin were administered weekly. The complexity of the intervention and difference in response between groups make interpretation of the efficacy of any single item difficult.

A single trial evaluated the effect of an intervention vitamin C with and without beta-carotene, and all groups were compared to placebo in women with cervical abnormalities that, if untreated, may progress to cervical carcinoma.⁵⁷ Mackerras et al⁵⁷ conducted a randomized double-blind study with a placebo in 141 women with confirmed minor squamous atypia or cervical epithelial neoplasia (CIN) stage I. The subjects were assigned randomly to an oral daily dose of 30 mg beta-carotene or 500 mg vitamin C, both, or neither. Over 2 years of follow-up, there was no statistically significant difference in the regression of the lesions. Therefore, the researchers concluded that high doses of either compound increases the rate of regression or decreases the progression of minor atypia or CIN l. Again, because the outcome measured is not fully developed cancer per se, it was not included in the pooled analysis.

A different intermediate end point was used in the trial by Cahill, O'sullivan, Mathias, and colleagues.⁵⁸ Colonic crypt cell proliferation and a shift in the proliferative zone in the crypt are precursors to colon carcinoma. Following colonoscopy and colon biopsy, ten patients were randomly assigned to each of four intervention arms: arm 1 of the trial received no supplementation, arm 2 received vitamin E (160 mg/day), arm 3 received vitamin C (750 mg/day), and arm 4 received beta-carotene (9 mg/day). Twenty subjects with a normal colonoscopy and normal colonic mucosa were included as a control group. After 1 month, colonic biopsies were repeated. Both vitamin C and beta-carotene significantly reduced the total proliferation, but vitamin E had no effect. Beta-carotene reduced the colonic cell proliferation only at the base of the crypts, whereas vitamin C reduced proliferation in all crypt compartments from the apex to the base, when compared to age- and gender-matched controls.

A rising prostate specific antigen (PSA) level is a widely used diagnostic method for identifying possible presence of early prostate cancer. This is not an intermediate outcome or a risk factor for a particular cancer per se, but it does have interest as a possible way to detect early tumors. The reliability and utility of PSA to screen for prostate cancer is controversial however it is used as a tumor marker in patients who have undergone prostatectomy. In a randomized, double-blinded, crossover, prevention trial, Schroder, Kranse, Dijik, and colleagues⁶⁴ examined the impact of a dietary intervention on prostate cancer in a sample of 37 men with rising PSA levels after undergoing radical prostatectomy, radiotherapy, and pelvic node dissection, but no endocrine treatment. The dietary intervention included soy extract, tea extract, carotenoids, phytosterols, selenium, and vitamin E (doses not given) in addition to a regular diet. The effect of the diet on PSA levels was assessed relative to that of a placebo during a two-week run-in period followed by two 6-week crossover periods that alternated with two washout periods. The results showed that during the supplement treatment periods, the slope of the normal rise in PSA was decreased, which translated into an 8-week delay in the rise of the PSA with a 6-week course of supplements. The effect of such a delay on prostate cancer mortality on other clinical outcomes is not known.

Heterogeneity

A number of issues potentially limit the effectiveness of this review. Methodologically, there was marked heterogeneity in the size of the population, the intent of the trial, the types of outcomes, and follow-up times. We identified three large primary prevention trials (ATBC, Linxian General Population and Linxian Dysplasia Group) that each reported on a number of separate outcomes. The majority of remaining trials were studies of much smaller numbers of people. They included not only secondary prevention trials but also treatment trials such as vitamin C for treatment of cancer. In addition, the populations varied from representatives of the healthy population to subjects identified at high risk for the development of cancer or who in fact already had cancer. The observed heterogeneity in study populations and designs deterred us from conducting a meta-analysis in two of our three outcome domains—death and new tumors— and also excluded some studies from our colonic polyps analysis. In the face of this heterogeneity, we provide individual study risk ratios and discuss the studies descriptively. We cannot assess the relationship between the possible heterogeneity in treatment effects and study or population characteristics, due to the small numbers of studies available.

Quality of the Trials

Given the consistent positive effects from the observational literature, the number of clinical trials as well as their depth and quality was disappointing. The quality of these studies varied as well. More than half of the trials had a Jadad score of less than three. The presence of Jadad scores of less tha n three has been associated in the literature with bias.¹⁴ In addition, we excluded from consideration case series, cohort studies and epidemiologic assessments. While these would have expanded the data base, they do not provide the most direct evidence of efficacy.

Publication Bias

We conducted a comprehensive literature search, including hand-searching reference lists of identified articles, and contacting experts. We also searched databases that include unpublished studies, such as the Cochrane registry of trials and the so-called “grey literature” (unpublished data, conference proceedings, and abstracts). Although we were unable to find any unpublished literature, we were able to include a number of conference proceedings in our analysis. Research by McAuley and colleagues⁶⁵ has demonstrated that the exclusion of the “grey literature” exaggerates the estimates of the effects of interventions that are being tested. Therefore, to the degree that there is significant unpublished or non-indexed literature pertinent to this topic that we did not find, this review is limited. Language was not an exclusion criterion for our search.

We could only test formally for publication bias in the single setting for which we did a meta-analysis, and based on only three studies. In this setting, we found no evidence of publication bias but have little statistical power given the small number of studies. We acknowledge that publication bias may still exist despite our best efforts to conduct a comprehensive search, and the lack of statistical evidence of existence.

Diverse Interventions

Clinically, a number of potential limitations could be identified as well. Few studies evaluated single agents for efficacy. There was no standard amount of either vitamins C or E given, nor were the multi-vitamin formulas consistent from study to study. Some of this variation may be due to differences in the populations assessed; however, it also reflects lack of consensus on recommended doses of these vitamins to be used therapeutically. Given the small number of studies and the differences in doses and formulas, no assessment could be made regarding effectiveness of varying dosage levels or combinations of individual supplements.

However, the ability to infer from this finding is limited because the multicomponent intervention limits our ability to attribute the reported efficacy to any particular component. However, there will always be problems of confounding variables when studies are not designed to isolate effects of single nutrients. However, take single nutrients for treating cancer, making it a challenge to test the efficacy of almost any of the treatments being used in practice.

The role of these clinical trials in understanding anti-oxidants in cancer can be expressed in a series of logical steps:

• Cellular/molecular evidence has shown that oxidation is associated with DNA and other molecular damage.

• Cellular/molecular evidence relates DNA damage to the development of cancer.

• Epidemiologic data supports a relationship between consumption of diets rich in anti-oxidants and decreased rates of cancer.

• However, the randomized controlled trials (RCTs) reviewed here failed to support the hypothesis that anti-oxidant supplements help prevent cancer.

In trying to resolve the RCT results with the cellular/molecular and the epidemiologic data there are several possible explanations:

1. The type of antioxidants used in the trials was different (synthetic vs. natural).

2. The dose used was wrong.

3. The results were due to something other than the single anti-oxidants or combinations that were tested in the trials.

4. The duration of treatment was too short.

5. The observed epidemiologic association is not causal in the same way in which the recent demonstration that epidemiologic studies relating the use of hormonal replacement therapy to potential cardiac effects in post-menopausal women was not confirmed by a definitive RCT.

The conclusion of the review is that the negative results only apply to the anti-oxidants tested, in the doses tested, and for the populations tested and do not constitute “proof” that anti-oxidants do not influence cancer. However, the generally negative results from the RCTs do place the burden of proof on the proponents of anti-oxidant supplements to identify the specific supplement, the dosage and the population combination that is efficacious.

Other Limitations

In addition, it will be difficult to generalize from this body of work, because insufficient numbers of female patients were included for study. Also, few trials were conducted on cancers specific to women, such as breast or cervical cancer. Finally, the lack of any published studies on coenzyme Q10 appropriate for inclusion in this analysis limit the ability to assess this popular therapy. Unfortunately, the only literature identified involved uncontrolled clinical trials, case series and case reports.