Meta-analysis: antioxidant supplements for primary and secondary prevention of colorectal adenoma

Authors

  • G. BJELAKOVIC,

    1. The Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen University Hospital, H:S Rigshospitalet, Copenhagen, Denmark
    2. Department of Internal Medicine – Gastroenterology and Hepatology, Department of Anatomy, Medical Faculty, University of Nis, Nis, Serbia and Montenegro
    Search for more papers by this author
  • A. NAGORNI,

    1. Department of Internal Medicine – Gastroenterology and Hepatology, Department of Anatomy, Medical Faculty, University of Nis, Nis, Serbia and Montenegro
    Search for more papers by this author
  • D. NIKOLOVA,

    1. The Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen University Hospital, H:S Rigshospitalet, Copenhagen, Denmark
    Search for more papers by this author
  • R. G. SIMONETTI,

    1. The Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen University Hospital, H:S Rigshospitalet, Copenhagen, Denmark
    2. Divisione di Medicina, Ospedale V. Cervello, Palermo, Italy
    Search for more papers by this author
  • M. BJELAKOVIC,

    1. Department of Internal Medicine – Gastroenterology and Hepatology, Department of Anatomy, Medical Faculty, University of Nis, Nis, Serbia and Montenegro
    Search for more papers by this author
  • C. GLUUD

    1. The Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen University Hospital, H:S Rigshospitalet, Copenhagen, Denmark
    Search for more papers by this author

Dr G. Bjelakovic, Department of Internal Medicine – Gastroenterology and Hepatology, Medical Faculty, University of Nis, Bulevar Dr Zorana Djindjica 81, 18000 Nis, Serbia and Montenegro.
E-mail: goranb@junis.ni.ac.yu

Abstract

Summary

Background

Colorectal cancer may be prevented by reducing the development of adenomatous polyps.

Aim

To assess the benefits and harms of antioxidant supplements in preventing colorectal adenoma.

Methods

Using the Cochrane Collaboration methodology we reviewed all randomized clinical trials comparing antioxidant supplements with placebo or no intervention. We searched electronic databases and the reference lists until October 2005. Outcome measures were development of colorectal adenoma adverse events. We analysed dichotomous outcomes with fixed- and random-effects model meta-analyses and calculated the relative risk with 95% confidence interval.

Results

We identified eight randomized trials (17 620 participants). Neither fixed-effect (relative risk: 0.93, 95% CI: 0.81–1.1) nor random-effect model meta-analyses (0.82, 0.60–1.1) showed statistically significant effects of supplementation with β-carotene, vitamins A, C, E and selenium alone or in combination. Antioxidant supplements seemed to increase the development of colorectal adenoma in three low-bias risk trials (1.2, 0.99–1.4) and significantly decrease its development in five high-bias risk trials (0.59, 0.47–0.74). The estimates difference is significant (P < 0.0001). There was no significant difference between the intervention groups regarding adverse events, including mortality (0.82, 0.47–1.4).

Conclusion

We found no convincing evidence that antioxidant supplements have significant beneficial effect on primary or secondary prevention of colorectal adenoma.

Background

Adenomatous polyps are considered precursors of colorectal cancer. Colorectal cancer prevention may therefore be directed at preventing the development of adenomatous polyps.1, 2 Adenomatous polyps can be found in 33% of the general population aged 50 years3 and in 50% aged 70 years.4 The gastrointestinal tract is constantly exposed to oxidants and antioxidants.5Oxidative stress may cause gastrointestinal diseases and cancer.6–8 Dietary antioxidants, such as β-carotene, vitamin A, vitamin C, vitamin E and selenium are potential cancer preventive agents. They fight-free radicals that may cause oxidative DNA damage and cancer development.9–11

Dietary prevention of colorectal cancer still needs to be scientifically proven.12 Numerous observational studies have addressed the associations between individual nutrients and risk of colorectal cancer.13–16 Results have been inconsistent. The evidence on whether antioxidant supplements are effective in decreasing the development (occurrence or recurrence) of colorectal adenoma is also contradictory.17 Besides potential deleterious effects on human cells, reactive oxygen species may in moderate concentrations play an important role in apoptosis, phagocytosis, and detoxification.18 Dysregulation of apoptosis may cause gastrointestinal diseases, including cancer.19 In order to obtain an overview of the status on antioxidant supplements and development of colorectal adenoma, we conducted a systematic review and meta-analysis of all randomized trials assessing antioxidant supplementation in primary or secondary prevention of colorectal adenoma.

Methods

The present review was conducted using The Cochrane Collaboration methodology.20 We included all trials that had randomized participants to antioxidant supplements (β-carotene, vitamin A, vitamin C, vitamin E, and selenium) vs. placebo or no intervention irrespective of blinding, publication status, publication year, or language, and that reported the development (occurrence or recurrence) of colorectal adenoma. We included the listed antioxidant supplements no matter the used dose, duration, and route of administration, or whether they were administered separately or in combination. We analysed the antioxidants singly, combined, and mixed with other vitamins or trace elements that have no antioxidant function. We analysed the data using the intention-to-treat principle irrespective of how the original trialists had analysed the data. For trials with factorial design, we based our results on ‘at margins’ analyses,21 comparing all antioxidant-supplemented arms with arms not supplemented with antioxidants. To determine the effect of a single antioxidant we performed ‘inside the table’ analyses21 in which we compared the single antioxidant arm with the placebo/no intervention arm. In trials with parallel group design with more than two arms, we only compared the arms supplemented with antioxidants and placebo/no intervention. Collateral interventions were allowed if used equally in both intervention arms. Participants at high risk of developing colorectal adenoma (i.e. those who had already developed colorectal adenoma) as well as participants from the general population were included. The primary outcome measure was development of colorectal adenoma (recurrence or occurrence during follow-up). The secondary outcome measure was adverse events, including mortality.

We identified trials by searching the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library (Issue 3, 2005; http://www.cochrane.org), MEDLINE (1966 to October 2005), EMBASE (1985 to October 2005), LILACS (1982 to October 2005), and Science Citation Index Expanded (1945 to October 2005).22 We read reference lists and contacted manufacturers of antioxidant supplements to ask for unpublished randomized trials.

For each randomized trial, we recorded first author; country of origin; number of participants; participants age (range and mean or median); sex ratio; drop-out rate; study design; type of supplemented antioxidant; dose; regimens; duration of supplementation; duration of follow-up and additional interventions. Data from the longest follow-up period were used for analyses. We recorded the number of participants with development of colorectal adenoma in each arm of the trial.

According to empirical evidence,23–26 we assessed the risk of bias in the trials based on their reports and information from the authors on central methodological components. Trials with adequate generation of the allocation sequence, allocation concealment, blinding and follow-up were considered low-bias risk trials, i.e. of high quality.

Statistical analyses

We used The Cochrane Collaboration software (revman analyses1.0; http://www.cochrane.org) and stata 8.2 (Stata Corp., College Station, TX, USA) to analyse the extracted data. We analysed the data with both fixed-27 and random-effects model analysis.28 We calculated the relative risk with 95% confidence interval (CI). Random-effects meta-regression was used to assess which covariates influence the effect across trials.29 The included covariates were dose of the individual antioxidant supplements and bias risk. Univariate and bivariate analyses including both covariates were performed. Results were presented with regression coefficients and 95% CI.

We assessed heterogeneity with I2, which describes the percentage of total variation across studies because of heterogeneity rather than chance. Negative values of I2 are put equal to zero. I2 lies between 0% (no observed heterogeneity) and 100%. Larger values show increasing heterogeneity.30

We performed the Begg and Mazumdar-adjusted rank correlation test31 and the Egger et al. regression asymmetry test for bias.32

Results

Trial flow

Database searches yielded 124 references. After excluding duplicates and clearly irrelevant references, we retrieved 18 references. We identified additional 18 references from reference lists. Of these 36 references, we excluded 15 on non-randomized studies and four describing two randomized trials not containing relevant data. The remaining 17 references, describing eight randomized trials,33–40 fulfilled our inclusion criteria and could provide data for our analyses. One of the trials37 only reported preliminary data.

Trial characteristics

A total of 17 620 participants were randomized in the eight trials. The number of participants in each trial ranged from 116 to 15 538. Two trials did not report data on sex.37, 38 In the remainder, the proportion of males was 71%. The mean age was 59 years, ranging from 15 to 84 years.

Recurrence of colorectal adenoma in patients who had their colorectal adenoma(s) removed was the primary outcome measure in seven trials.33–39 One of these trials39 included participants with previous colorectal cancer, other cancer, or familial colorectal cancer. In six of these trials, all polyps were removed at entry before randomization;33–38 the seventh trial only removed polyps with a diameter above 9 mm.39 The smaller polyps were observed and removed at the end of this trial.39 The surveillance colonoscopy for colorectal adenoma development was performed after 6–48 months.33–39 The polyps were histologically classified according to the World Health Organisation classification41 in two trials.38, 39 In the other trials this was not defined. Grading of dysplasia was reported in one trial.36 The development of new adenoma in healthy people was the primary outcome in the eighth trial.40 This trial was performed as part of the ATBC Study42 with 29 133 participants that aimed to evaluate the effect of vitamin E and β-carotene on the incidence of lung cancer and other cancers in middle-aged male smokers. The adenoma subtrial included 15 538 participants, living within the area of three major Finnish cities. The participants did not have a colonoscopic screening performed at entry. Each participant visited his local study centre three times annually giving information about medical contacts and symptoms. Each case of colorectal adenoma was identified at the pathology laboratories located within the study region.

The route of antioxidant administration was oral in all the trials. The types, doses, dose regimens, and duration of antioxidant supplementation were as follows: β-carotene (15–25 mg), vitamin A (6000–30 000 IU), vitamin C (150–1000 mg), vitamin E (30–400 mg), and selenium (101–200 μg) daily or on alternate days for 1–6.1 years (Table 1). β-Carotene, vitamins A, C, E, and selenium were given either separately or in different combinations. Five trials used parallel group design, and three trials used factorial designs: two trials two-by-two and one trial two-by-two-by-two (Table 2). Follow-up ended at the end of intervention.

Table 1.   Recommended dietary allowance, tolerable upper intake level, experimental doses, and regimens of antioxidant supplements used in included randomized trials
Antioxidant supplementsRecommended dietary allowanceTolerable upper intake levelsExperimental dosesRegimens
MenWomen
  1. ND, not determined.

β-CaroteneNDNDND15–25 mgDaily
Vitamin A (IU)3000233099006000–30 000Daily or on alternate days
Vitamin C (mg)90702000150–1000Daily or on alternate days
Vitamin E (mg)1515100030–400Daily or on alternate days
Selenium (μg)5555400101–200Daily
Table 2.   Design of included trials, experimental antioxidant supplements used, additional therapy and incidence of colorectal adenomas (fixed-effect)
TrialTrial designExperimental antioxidant supplementsAdditional therapyColorectal adenoma
McKeown-Eyssen et al.33ParallelVitamin C, and vitamin ENone0.79, 0.53–1.2
Roncucci et al.34ParallelVitamin A, vitamin C, and vitamin ELactulose 20 g/daily0.16, 0.06–0.43
Greenberg et al.35Factorial 2 × 2β-Carotene, vitamin C, and vitamin ENone1.0, 0.82–1.3
MacLennan et al.36Factorial 2 × 2 × 2β-CaroteneLow fat; bran1.3, 0.89–1.8
Ponz de Leon and Roncucci37ParallelVitamin A, vitamin C, and vitamin EN-acetyl cysteine 600 mg0.76, 0.34–1.7
Bonelli et al.38ParallelVitamin A, vitamin C, vitamin E, and seleniumZinc 30 mg0.69, 0.41–1.2
Hofstad et al.39Parallelβ-Carotene, vitamin C, vitamin E, and seleniumCalcium carbonate 1.6 g/daily0.58, 0.39–0.87
Malila et al.40Factorial 2 × 2β-Carotene and vitamin ENone1.5, 0.97–2.2

Seven trials used placebo and one trial used no intervention in the control groups. For the factorial-designed trials,35, 36, 40 we conducted our analyses based on ‘at margins’ analyses, i.e. all participants receiving antioxidant supplements vs. participants not receiving them.

Five trials tested additional interventions. Two three-armed trials with parallel group design tested lactulose or N-acetylcysteine in one arm.34, 37 The two-by-two-by-two factorial design trial tested two nutritional interventions – wheat bran fibre and fat reduction.36 Two parallel group trials used zinc (30 mg/daily)38 and calcium carbonate (1.6 g/daily)39 in the antioxidant arm (Table 2).

Methodological quality of included trials

A total of three of eight trials35, 36, 40 (37.5%) had low risk of bias, having adequate generation of allocation sequence, allocation concealment, blinding and follow-up.23–26 The remaining trials had high risk of bias (Table 3).

Table 3.   Methodological quality and bias risk of randomized clinical trials comparing antioxidant supplements vs. placebo/no intervention
 Generation of the allocation sequence* Allocation concealment* Blinding* Follow-up*Bias risk of the trialSample sizeIntention -to-treat
  1. * Quality components used for assessing the risk of bias.

  2. † Not obtainable from the authors.

McKeown-Eyssen et al.33Not reported†Not reported†Not reported†AdequateHighNoNo
Roncucci et al.34Not reported†Not reported†InadequateAdequateHighNoYes
Greenberg et al.35AdequateAdequateAdequateAdequateLowYesNo
MacLennan et al.36AdequateAdequateAdequateAdequateLowYesYes
Ponz de Leon and Roncucci37Not reported†Not reported†InadequateInadequateHighNoNo
Bonelli et al.38Not reported†Not reported†AdequateAdequateHighYesYes
Hofstad et al.39Not reported†Not reported†AdequateAdequateHighYesNo
Malila et al.40AdequateAdequateAdequateAdequateLowYesYes

Effect of antioxidant supplements

Neither the fixed-effect (relative risk: 0.93, 95% CI: 0.81–1.1) nor the random-effect model meta-analyses (0.82, 0.60–1.1) showed statistically significant effects of supplementation with β-carotene, vitamins A, C, E and selenium alone or in combination vs. placebo or no intervention on the development of colorectal adenoma in all eight trials. Heterogeneity was significant (I2 = 74.9%).

Analysed with a fixed-effect model, the three low-bias risk trials showed a tendency of antioxidant supplements towards increasing the development of colorectal adenoma (1.2, 0.99–1.4) with moderate and no statistically significant heterogeneity (I2 = 28.4%; Figure 1). In the five high-bias risk trials, the antioxidant supplements significantly decreased the development of colorectal adenoma (0.59, 0.47–0.74), but heterogeneity was substantial and statistically significant (I2 = 57.3%; Figure 1). The difference between the estimates is significant (P < 0.0001). Analysed with a random-effects model, the three low-bias risk trials showed a tendency of the antioxidant supplements towards increasing the development of colorectal adenoma (1.2, 0.95–1.5). In the five high-bias risk trials, antioxidant supplements significantly decreased the development of colorectal adenoma (0.59, 0.40–0.87).

Figure 1.

 Effect of antioxidant supplements vs. placebo/no intervention on incidence of colorectal adenomas stratifying trials according to methodological quality. CI, confidence interval; n, number of patients with outcome; N, number of participants at risk; d.f., degrees of freedom; χ2, chi-squared statistic; I2, the percentage of total variation across studies that is due to heterogeneity rather than chance; RR, relative risks. Relative risks are plotted (black squares with area proportional to the amount of statistical information in each trial) comparing outcome among participants allocated to antioxidant supplements with those allocated to placebo, alongside with their 95% CI (horizontal lines). For particular subtotals, the result and its 95% CI are represented by a diamond, with the relative risk (95% CI) and its statistical significance given alongside. Squares or diamonds to the left of the solid vertical line indicate benefit with antioxidant supplements, but this is conventionally significant (P < 0.05) only if the horizontal line or diamond does not overlap the solid vertical line.

We found significant bias in the included trials by Begg and Mazumdar-adjusted rank correlation test31 (P = 0.035) and the Egger et al.32 regression asymmetry test (P = 0.024; Figure 2).

Figure 2.

 The relative risk ratio (RR) of antioxidant supplements vs. placebo or no intervention on development of colorectal adenoma – funnel plot.

Effect of antioxidant supplements in trials using colonoscopic surveillance

Only seven of the trials performed colonoscopic examination at the end of follow-up to examine for the primary outcome measure, while the eighth trial40 used the development of colorectal adenoma as primary outcome measure. One of the peer-reviewers of this publication requested us to perform a post hoc sensitivity analysis in which this trial was excluded.

Meta-analysis of the seven trials using colonoscopic examination showed that antioxidant supplements significantly reduced the development of colorectal adenoma in fixed-effect model analysis (0.85, 0.73–0.98), but not in random-effect model analysis (0.75, 0.54–1.0). Heterogeneity was substantial and significant (I2 = 73.6%).

In two low-bias risk trials, the antioxidant supplements did not significantly affect the development of colorectal adenoma neither with the fixed-effect (1.1, 0.90–1.3) nor the random-effect model meta-analyses (1.1, 0.90–1.3) without statistically significant heterogeneity (I2 = 4.0%).

Effect of the different antioxidant supplements

The effects of the different antioxidant supplements on the development of colorectal adenoma are shown in Table 4, disregarding the bias risk of the trial.

Table 4.   Relative risks with 95% confidence intervals (fixed- and random-effects models)
Experimental antioxidant supplementsColorectal adenoma
Fixed-effectRandom-effect
  1. Intervention effect of different antioxidant supplements vs. placebo or no intervention on the development of colorectal adenoma, according to the antioxidant supplements, administered in the individual trials, disregarding the bias risk of the trials.

  2. ND, not determined as the result is equal to the fixed-effect model due to only one included trial.

β-Carotene35, 36, 401.1, 0.87–1.31.1, 0.87–1.3
Vitamin E401.7, 1.1–2.8ND
β-Carotene and vitamin E401.6, 1.0–2.6ND
Vitamin C and vitamin E33, 351.0, 0.80–1.30.97, 0.70–1.3
Vitamin A, vitamin C and vitamin E34, 370.33, 0.18–0.600.36, 0.07–1.8
Vitamin A, vitamin C, vitamin E and selenium380.69, 0.41–1.2ND
β-Carotene, vitamin C and vitamin E351.0, 0.75–1.3ND
β-Carotene, vitamin C, vitamin E and selenium390.58, 0.39–0.87ND

The intervention effect of the antioxidant supplements when compared with placebo/no intervention on the development of colorectal adenoma, stratifying trials according to used antioxidant, irrespective of other antioxidants used, is shown in Table 5, disregarding the bias risk of the trials.

Table 5.   Relative risk, 95% confidence interval (fixed- and random-effects models)
 Fixed-effectRandom-effectHeterogeneity I2 (%)
  1. Intervention effect of antioxidant supplements vs. placebo/no intervention on incidence of colorectal adenoma stratifying trials according to administered antioxidant (either singly or combined with other antioxidants), disregarding bias risk of the trials.

β-Carotene35, 36, 39, 400.97, 0.81–1.20.93, 0.67–1.365.1
Vitamin A34, 37, 380.49, 0.33–0.720.47, 0.20–1.174.0
Vitamin C33–35, 37–390.76, 0.64–0.910.67, 0.46–0.9873.6
Vitamin E33–35, 37–400.88, 0.74–1.00.77, 0.52–1.178.3
Selenium38, 390.63, 0.46–0.870.62, 0.45–0.850

Univariate meta-regression analyses showed that vitamin A (coefficient: −0.00004; 95% CI: −0.000079 to −0.000018) and bias risk in the trial (coefficient: 0.95; 0.46–1.4) have significant influence on the development of colorectal adenoma. None of the other antioxidant supplements (β-carotene; vitamin C; vitamin E; selenium) was significantly associated with the development of colorectal adenoma.

In the bivariate meta-regression analyses with two covariates (dose of the individual antioxidant supplement and bias risk in the trial), none of the examined antioxidants was significantly associated with the development of colorectal adenoma. In the bivariate meta-regression analyses, high-bias risk trials were significantly associated with the estimated intervention effect of β-carotene (P = 0.001), vitamin A (P =0.008), vitamin C (P < 0.001), vitamin E (P = 0.002) and selenium (P = 0.003).

Adverse events

Mortality was only reported in three trials33, 35, 38 and antioxidant supplements showed no significant effect on mortality in fixed- (0.82, 0.47–1.4) and random-effects model meta-analyses (0.80, 0.46–1.4). There was no significant heterogeneity (I2 = 0%).

Several adverse events were recorded in the antioxidant-supplemented group. Patients supplemented with β-carotene, vitamin C, vitamin E, and selenium experienced constipation, diarrhoea, and bloating, while patients supplemented with vitamin A, vitamin C, and vitamin E experienced pruritus. These increases were, however, not statistically significant.

Discussion

We did not find convincing and valid evidence that β-carotene, vitamin A, vitamin C, vitamin E, and selenium given in different combinations and doses prevent colorectal adenoma. These findings support our earlier observations that the discussed antioxidant supplements administered separately or in certain combinations vs. placebo for a period of 2–12 years had no significant influence on the development of colorectal cancer.43 We, therefore, question the hypothesis that antioxidants may inhibit the development of colorectal adenoma and prevent cancer.

We observed that the estimated effect of the antioxidant supplements seems to depend significantly on the risk of bias of the trial. In the trials with low risk of bias (high methodological quality), antioxidant supplements showed a tendency towards increasing the development of colorectal adenoma. Heterogeneity was not significant. In the trials with high risk of bias (low methodological quality), antioxidant supplements had significant beneficial effect as regards development of colorectal adenoma, but with significant heterogeneity. Such heterogeneity ought to question the results.20 Bias risk of the trials significantly influence the estimated intervention effect of antioxidant supplements on the development of colorectal adenoma in both univariate and bivariate meta-regression analyses. Our present results mirror our previous results on colorectal cancer.43 Our results are also in accordance with several studies linking unclear/inadequate methodological quality (i.e. high risk of bias) with significant overestimation of beneficial intervention effects.23–26 We found statistically significant bias risk in the trials included in our meta-analysis. These estimated intervention effects could be predicted by dividing the trials according to the methodological components.

When comparing the intervention effects of different types of antioxidants, we found no significant heterogeneity in the effects of supplements on colorectal adenoma in the trials with low risk of bias, neither when a single antioxidant was assessed nor when a combination of antioxidants was assessed. When we stratified the trials, irrespective of bias risk, according to the antioxidant used, vitamin A, vitamin C, and selenium in fixed-effect model meta-analyses and vitamin C and selenium in random-effects model meta-analyses seemed to have significant beneficial effects on the development of colorectal adenoma. However, heterogeneity was statistically significant for all examined antioxidants except for selenium. Because of the methodological shortcomings of these trials, the results are likely biased.

Some of the included trials have not taken into account the recommended daily allowances or upper tolerable intake levels of antioxidant vitamins (Table 1).44, 45 Dietary references intakes come from observational and experimental randomized and non-randomized studies in humans.44, 45 The recommended daily allowance is intended as recommendation for daily intake. It may not be suitable for choosing the dose of antioxidant supplementation in people without special needs or in people who receive antioxidants through their food.43, 46 Previous trials have shown that antioxidant supplements in doses above the recommended ones might not be beneficial in preventing specific diseases.43, 47

Our results show that β-carotene does not posses significant preventive properties regarding colorectal adenoma. This observation is further supported by our previous systematic review assessing the influence of β-carotene supplements on gastrointestinal cancers.43 Recent studies have discussed the possible harms of β-carotene.48 Furthermore, β-carotene can lead to an increased risk of lung cancer according to randomized trials.42, 49

Vitamin A demonstrated a significant beneficial effect on colorectal adenoma development in the univariate meta-regression analysis. However, this effect disappeared in the bivariate meta-regression analysis also taking bias risk into consideration.

Recent studies have demonstrated that vitamin C could act as both a pro-oxidant and as an antioxidant in vivo50 and that dietary vitamin C may cause DNA damage and promote carcinogenesis.51 Connelly et al.52 examined associations between high doses of vitamin C and apoptosis in colonic mucosa. They found that high vitamin C uptake was inversely associated with apoptosis among participants with adenoma. Another study suggested an association between inhibition of apoptosis and colorectal cancer development.53 This may explain the tendency towards harmful effects observed in three of eight polyp prevention trials.35, 36, 40

Vitamin E alone showed a significantly harmful influence on the development of colorectal adenoma.40 It should be emphasized that vitamin E supplementation did not affect incidence of colorectal cancer in the same study.54 This lack of detrimental effect may be due to a type II error.

Selenium in combination with β-carotene, vitamin A, vitamin C, and vitamin E was tested in two trials38, 39 with high risk of bias. The development of colorectal adenoma was significantly reduced in one of the trials.39 Another trial with the same combination of antioxidants, but without selenium35 found no significant effect. Multiple studies have examined whether selenium protects against colorectal adenoma and cancer. Early et al.55 found that patients with colorectal adenoma or cancer are not selenium-deficient. In contrast, two studies demonstrated an inverse association between serum levels of selenium and adenoma risk.56, 57 Reid et al.58 evaluated the association of selenium supplementation and colorectal adenoma in a sample of participants of Nutritional Prevention of Cancer trial and found reduced risk among participants with low baseline selenium level. Furthermore, our systematic review suggested that selenium supplementation may be beneficial in preventing gastrointestinal cancers, particularly liver cancer.43 Although several models have been explored, molecular mechanisms involved in the possible cancer preventive role of selenium are largely unknown.

Several non-serious adverse events from antioxidant supplements have been reported in the included trials, but the differences between the experimental and control groups were not significant. We observed no significant effect of antioxidant supplements on mortality, but the confidence limits were wide. We have previously observed that the antioxidant supplements may significantly increase mortality.43 We cannot exclude a type II error in our present analysis.

Certain potential limitations of this review warrant consideration. The included trials may have inherent weaknesses. In a number of trials the drop-out rate was high and compliance was low. Diagnostic criteria and timing of screening differed among the trials or were not always well defined. Bias risks in the included trials were significant. This may have influenced the results of the individual trials and thereby our meta-analyses. The majority of the trials examined the recurrence of colorectal adenoma. Therefore, our results may not be translated to the development of colorectal adenoma in the general population. Finally, supplementation only lasted from 1 to 6 years, and for drawing definitive conclusions it can be considered a too short duration in comparison with the long time it may take to develop colorectal adenoma.

We would like to draw attention to the fact that colorectal adenoma is considered a prototype surrogate outcome in the development of chemopreventive drugs for the adenoma-carcinoma sequence in the colon.59, 60 However, potential pitfalls of adenoma-carcinoma sequence should be considered. Hypothetical benefit or harm may occur later than the surrogate outcome.61 Even more, some evidence suggest potential limitations of adenoma-carcinoma sequence. The proportion of colorectal cancers that develop within pre-existing adenoma is not known. The behaviour of colorectal adenoma is highly unpredictable and the period needed for malignant transformation depends on different clinical conditions. A certain proportion of patients with colorectal adenoma will never develop invasive cancer.62

Our present results are in accordance with the results of our systematic review on the role of antioxidant supplements for preventing gastrointestinal cancers43 as well as the British Nutrition Foundation Recommendations,63, 64 and United States Preventive Services Task Force Recommendations65, 66 for the use of vitamin supplementation to prevent cancer. These studies and guidelines collectively suggest that antioxidant supplements may not convey health benefits. Basic issues on both the efficacy and safety of the use of relatively high doses of antioxidant supplements need to be resolved. Further evidence is needed regarding the mechanisms of action, biotransformation, bioavailability, safety, and appropriate dose of antioxidants in order to determine whether they can offer disease prevention.63, 67, 68The result of our meta-analysis suggests that at present there is no convincing evidence that antioxidant supplements have beneficial effect in the primary or secondary prevention of colorectal adenoma.

Authors contribution

Goran Bjelakovic conceived the idea for and drafted the manuscript. Aleksandar Nagorni revised the manuscript. Dimitrinka Nikolova developed the search strategy, checked data extraction, and revised the manuscript. Rosa Simonetti checked data extraction and revised the manuscript. Marija Bjelakovic revised the manuscript. Christian Gluud provided strategy for data analyses and interpretation, and revised and drafted the manuscript. All co-authors contributed with comments.

Conflict of interest statement

This study was supported by The Copenhagen Trial Unit, Centre for Clinical Intervention Research, H:S Rigshospitalet, Copenhagen, Denmark.

Acknowledgements

The authors are grateful to Maurizio Ponz de Leon, John A. Baron and Leila A. Mott who kindly responded to our requests for further information on the trial they were involved in. The funding source had no involvement in the collection, analyses, interpretation of data, writing of the report and in the decision to submit the paper for publication.

Ancillary