Cruciferous Vegetable Consumption and Lung Cancer Risk: A Systematic Review

Introduction

Lung cancer is the leading worldwide cause of cancer death (1). Cigarette smoking accounts for ∼85% of the population burden of lung cancer in developed countries such as the United States, but selected dietary factors may modulate lung cancer risk (2). A major report of the World Cancer Research Fund and the American Institute for Cancer Research concluded that the evidence was “limited suggestive” that vegetable intake is inversely associated with lung cancer (3). However, the associations with lung cancer may vary according to the specific class of vegetable considered.

In particular, cruciferous vegetables (broccoli, cabbage, cauliflower, Brussels sprouts, kale) have been hypothesized to have anticancer properties that may contribute to reduced risk for lung cancer. Cruciferous vegetables are a rich source of isothiocyanates. Isothiocyanates may inhibit the bioactivation of procarcinogens found in tobacco smoke such as polycyclic aromatic hydrocarbons and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (ref. 4). Isothiocyanates may also enhance excretion of carcinogenic metabolites before they can damage DNA (5, 6). Furthermore, sulforaphane, a major isothiocyanate found in broccoli, can induce cell cycle arrest and apoptosis (6, 7).

The mammalian glutathione S-transferase (GST) is a superfamily of phase II genes that are broadly categorized into four classes: Alpha, Mu, Theta, and Pi (7). Each class consists of genes that encode a variety of isoenzymes (e.g., GSTA1, GSTM1, GSTT1, and GSTP1), which play an important role in xenobiotic metabolism (7), as well as isothiocyanate metabolism (8). Variants of several GSTs have been studied in relation to dietary isothiocyanate intake and cancer risk, but GSTM1 and GSTT1 are the two variants most frequently studied that may modify the association between cruciferous vegetable intake and lung cancer risk (6, 9). A common polymorphism in the GSTM1 and GSTT1 genes results in gene deletion, and individuals with homozygous deletions are devoid of the respective enzyme activity (7). Individuals with homozygous deletion of GSTM1, GSTT1, or both may metabolize isothiocyanates less efficiently and may be more intensely exposed to them after consumption of cruciferous vegetables. For this reason, individuals with the null GSTM1 or GSTT1 genotypes may have a lower risk for lung cancer when exposed to isothiocyanates (10-12).

There has been a rapid accumulation of epidemiologic evidence on the association between cruciferous vegetable intake and lung cancer (6, 9, 13). No systematic reviews have previously been conducted to thoroughly unify this information and formally assess this evidence. To address this information gap, we did a meta-analysis of the evidence on this topic, including the potential gene-diet interaction between cruciferous vegetable intake and GSTM1 and GSTT1.

Materials and Methods

This systematic review stemmed from a project funded by the World Cancer Research Fund–American Institute for Cancer Research to develop a report entitled “Food, Nutrition, Physical Activity and the Prevention of Cancer: a Global Perspective” (3). All work adhered to a standardized protocol developed by World Cancer Research Fund (http://www.wcrf.org/research/second_wcrf_aicr_report.lasso; ref. 3).

Search Strategy

For the World Cancer Research Fund report, we sought all evidence on the associations between dietary intake, physical activity, or anthropometric measures and lung cancer that were reported in randomized clinical trials, cohort studies, and case-control studies. We used the search strategy for dietary factors as described previously in (14), adapted for the outcome of lung carcinoma as described in ref. 15. The following electronic databases were searched: PubMed, Embase, Pascal, ISI Web of Science, the Cochrane Library, Biological Abstracts, Cumulative Index to Nursing and Allied Health Literature, National Institute on Alcohol Abuse and Alcoholism–Alcohol and Alcohol Problems Science Database, Agricola, CINAHL-EBSCOhost, Index Medicus for WHO Eastern Mediterranean Region, Index Medicus for South East Asian Region, and Latin American and Caribbean Center on Health Sciences Information. The search included all studies published up to April 2006. We also hand-searched references in the relevant review articles from the bibliographic database search and those cited in the 1997 World Cancer Research Fund report (2) or chosen for data abstraction. After this original search, we extended the PubMed search through December 2007. There were no language restrictions. If a published article was in a language that was beyond the expertise of our research team, World Cancer Research Fund had the article (16) translated into English.

Study Selection

The following exclusion criteria were applied to the screening of articles for the World Cancer Research Fund report: (a) no original data (reviews, editorials, meta-analyses); (b) studies not addressing the association between dietary intake, physical activity, or anthropometric measures and lung cancer; (c) studies not in humans; and (d) case reports and case series. The eligibility of each abstract or full-text article was assessed independently by two reviewers.

For the present report, we further limited the studies to those that reported on the association between cruciferous vegetable intake and lung cancer. Cruciferous vegetables were measured in different ways, including (a) total cruciferous vegetable intake, (b) total isothiocyanate intake, and (c) intake of specific individual cruciferous vegetables (e.g., broccoli, cabbage, or cauliflower). When measures of association or variability were not reported or could not be calculated using the data provided, we excluded the articles from the formal meta-analysis but discussed the findings of the article qualitatively. We made no systematic effort to contact authors. If separate reports from the same study were published, the report with the most updated data was selected for inclusion.

Data Abstraction. For each eligible article, two reviewers abstracted the data into an electronic database created by World Cancer Research Fund. The data abstraction was done serially, with any disagreements between reviewers resolved by consensus. Each reviewer classified the vegetables studied in each article into classes (e.g., cruciferous vegetables, Allium vegetables) according to the World Cancer Research Fund protocol (3). If a specific vegetable was not listed in the protocol, a nutritionist (L.E. Caulfield) assigned the appropriate vegetable subgroup. In the World Cancer Research Fund protocol, broccoli, cabbage, turnip or mustard greens, kale, sauerkraut, and cauliflower were classified as cruciferous vegetables. One article (17) reported the results from two different cohort studies; we treated the studies as unique, and our reviewers abstracted the data for each study separately.

To assess study quality, we adapted the criteria used by Longnecker et al. (18) for observational studies.

Statistical Analysis

The primary quantitative analyses focused on total dietary intake of cruciferous vegetables. Total cruciferous vegetable consumption was typically defined as a combination of at least three cruciferous vegetables, which usually included broccoli and cabbage plus other cruciferous vegetables. We also analyzed the associations between specific cruciferous vegetables and lung cancer risk. Separate meta-analyses were conducted for case-control and prospective cohort studies, by smoking status (never smokers or ever smokers) and by ethnicity (Western or Asian).

Studies varied in how frequency of cruciferous vegetable intake was measured, either as “times eaten” or “servings.” Thus, the pooled estimates entailed combining results of studies based on either the number of times eaten or servings. This is a potential source of heterogeneity across studies. Furthermore, the time units ranged from amount consumed per week to per month.

For all studies, odds ratios or relative risks and their respective 95% confidence intervals (95% CI) were abstracted. When a study reported several relative risk estimates, we abstracted the one adjusted for the most covariates. Two studies investigated the interaction between isothiocyanates and genotypes of GSTM1 and GSTT1 but did not report estimates for the main effects association between dietary isothiocyanate intake and lung cancer risk (19, 20). For these studies, we used the raw data presented in the articles to calculate the unadjusted odds ratios and 95% CIs. Meta-analyses were done with and without these studies included. Pooled odds ratio and relative risk estimates were obtained using inverse-variance weights in random effects models. Statistical heterogeneity was assessed using the DerSimonian and Laird's Q statistic and the I² statistic.

Sensitivity analyses to examine the influence of each individual study were conducted by excluding each study from the meta-analysis and comparing the point estimates including and excluding the study. Metaregression was used to explore for sources of heterogeneity. Publication bias was examined using funnel plots.

When sufficient data were presented in the original publication (≥3 exposure categories of intake frequency along with the numbers of cases and controls within each category), we assessed for the presence of a dose-response trend (21). For the dose-response analyses, we converted all results to amount consumed per day.

To assess for interaction between GST genotypes and cruciferous vegetable intake on the risk for lung cancer, we estimated the association between total cruciferous vegetable intake and lung cancer stratified by GSTM1 and GSTT1 status. With the exception of one cohort study (22) that used a genotyping assay that could differentiate between three GSTM1 and GSTT1 genotypes, two possible genotypes (present and null) were reported for GSTM1 and GSTT1 in all case-control studies. Data analyses were thus stratified as follows: GSTM1 present or null, GSTT1 present or null, and GSTM1/GSTT1 double present or double null. Only one study reported results for GSTM1/GSTT1 null/present or present/null genotypes, so this category was not included in the meta-analyses. We tested for within-study effect modification by GST genotype by calculating the difference in (log) odds ratios between GST subgroups within each study (23, 24). We then obtained a summary interaction P value by pooling these differences across studies.

Results

Search Results

We identified 37 studies that quantified the association between cruciferous vegetable consumption and lung cancer risk (Fig. 1 ). Of these, we excluded three early reports from studies that subsequently published updated data (25, 26) or with larger sample size (27), two that reported on individual cruciferous vegetables but did not report the total number of cases (28, 29), one that reported on interaction between cruciferous vegetables and GST genes reported genotyping data that were not comparable and did not provide sufficient data to calculate SEs to be included in the meta-analysis (22), and one that reported only on a biomarker (rather than dietary intake) of isothiocyanates (10). Of the 30 studies included in the meta-analyses, 18 studies [6 cohort (17, 30-33) and 12 case-control (16, 19, 20, 34-42)] reported on total cruciferous vegetable intake. One cohort study (43) and 11 case-control studies (44-54) reported on intake of individual cruciferous vegetables. Quality assessment for the studies reported on total cruciferous consumption and lung cancer is summarized in Supplemental Fig. 1.

Total Cruciferous Vegetables

Six prospective cohort studies (Table 1 ) and 12 case-control studies (Table 2 ), representing a total of 8,227 lung cancer cases, reported associations between total cruciferous vegetable intake and lung cancer risk. The studies were carried out in Europe (8 studies; refs. 16, 30, 31, 35, 37, 39-41), the United States (6 studies; refs. 17, 19, 20, 32, 33), Asia (2 studies; refs. 34, 38), Canada (1 study; ref. 42), and Australia (1 study; ref. 36). The duration of follow-up ranged from 4 to 12 years in the six prospective cohort studies (17, 30-33). All prospective studies adjusted for smoking status. Of the 12 case-control studies, 5 were confined to never smokers (34, 37, 39, 41, 42). Of the seven studies that included ever smokers, five reported odds ratios adjusted for smoking (16, 35, 36, 38, 40). The remaining two (19, 20) were studies that examined GST-diet interaction and thus estimates for the association between dietary isothiocyanates, and lung cancer risk were calculated unadjusted estimates based on published numbers of cases and controls.

Compared with those in the lowest categories of total cruciferous vegetable intake, the risk for lung cancer among those in the highest consumption categories was 22% lower (random-effects pooled odds ratio, 0.78; 95% CI, 0.70-0.88; P heterogeneity = 0.53; I² = 0%) in case-control studies and 17% lower in cohort studies (pooled relative risk, 0.83; 95% CI, 0.62-1.08; P heterogeneity = 0.02; I² = 62.8%; Fig. 2 ). With the exclusion of the two case-control studies (19, 20) whose unadjusted odds ratios and confidence intervals were calculated from published data, the pooled odds ratio was attenuated only slightly at 0.81 (95% CI, 0.70-0.94; P heterogeneity = 0.48; I² = 0%). Likewise, the results were the same when restricted to studies on never smokers (refs. 20, 34, 35, 37-42; pooled odds ratio, 0.78; 95% CI, 0.64-0.95; P heterogeneity = 0.54; I² = 0%; Fig. 3 ).

All prospective cohort and 10 case-control studies reported at least three categories of total cruciferous vegetable intake (Table 3 ). Of the six prospective cohort studies, three studies were compatible with an inverse dose-response association between cruciferous vegetable intake and lung cancer risk (17, 31, 32), whereas three others showed no evidence of a dose-response trend (17, 30, 33). Eight case-control studies provided dose-response data in sufficient detail to be included in dose-response meta-analysis. The pooled odds ratio for lung cancer associated with an increase of one cruciferous vegetable serving per day was 0.74 (95% CI, 0.73-0.75). Two additional case-control studies (20, 35) reported frequency of intake (that is, low, medium, high) that could not be combined with the other studies. Of the two studies, one (35) was compatible with a decrease in lung cancer risk with increasing consumption of cruciferous vegetables whereas the other (20) showed no discernable pattern.

Broccoli and cabbage as individual cruciferous vegetables were also inversely associated with lung cancer risk. In data generated from case-control studies, the pooled odds ratios for lung cancer risk comparing the highest versus lowest categories of intake were 0.53 (95% CI, 0.34-0.83; P heterogeneity = 0.14; I² = 43.0%) for broccoli (42, 45, 47-49) and 0.70 (95% CI, 0.54-0.91; P heterogeneity = 0.02; I² = 54.9%) for cabbage (40, 42-44, 46, 51-54).

Results Stratified by GST Genotypes

Five case-control studies (refs. 19, 20, 38-40; n = 3,715 lung cancer cases) reported on the association between cruciferous vegetable consumption and lung cancer risk stratified by GSTM1 and/or GSTT1 genotypes (Table 4 ). Of these, one reported only on GSTM1 (39), whereas four assessed GSTM1 and GSTT1 (19, 20, 38, 40). Cruciferous vegetable consumption was measured by food frequency questionnaire in all five studies, and exposure was quantified either as intake of isothiocyanates (n = 2 studies; refs. 19, 39) or of total cruciferous vegetables (n = 3 studies; refs. 20, 38, 40).

In these five studies, the pooled odds ratio of lung cancer for the highest versus lowest category of cruciferous vegetable intake was 0.74 (95% CI, 0.66-0.84; Fig. 4 ). When stratified by genotype, the inverse association between cruciferous vegetable intake and lung cancer was stronger in those who were null for GSTM1 and GSTT1 (pooled odds ratio, 0.41; 95% CI, 0.26-0.65; P heterogeneity = 0.64; I² = 0) than in those with the GSTM1 and GSTT1 present genotype (odds ratio, 0.75; 95% CI, 0.62-0.91; P heterogeneity = 0.43; I² = 0). This gene-diet interaction was statistically significant (P for interaction = 0.01). In the lone study to characterize the association between cruciferous vegetables and lung cancer among those with GSTM1 present/GSTT1 null or GSTM1 null/ GSTT1 present genotypes, the highest versus lowest intake of cruciferous vegetables was associated with a nonsignificant 20% reduction (odds ratio, 0.80; 95% CI, 0.60-1.08) in lung cancer risk (40), an association more similar to that among those with the double present genotypes (odds ratio, 0.88; 95% CI, 0.65-1.21) than among those with the double null genotypes (odds ratio, 0.28; 95% CI, 0.11-0.67).

Heterogeneity and Publication Bias

Metaregression results showed that study size, geographic location, or gender could not explain the observed heterogeneity for cohort studies. Sensitivity analysis results showed that the exclusion of individual studies did not substantially alter the pooled relative risks, which ranged from 0.75 to 0.91. All funnel plots to assess for possible indication of publication bias for meta-analyses of cohort and case-control studies, in addition to subgroup analyses by GST status, seemed symmetrical (data not shown).

Discussion

There is growing interest in the potential association between cruciferous vegetables and lung cancer. For example, 72% of the studies ascertained in this review were published in 2000 onward. By objectively synthesizing the complete body of epidemiologic evidence on this topic, the enhanced signal-to-noise ratio allows the overall associations to be seen more clearly across a range of different study approaches. An additional contribution of the present report is that pooling the evidence across studies allows the results within important subgroups to have much greater statistical precision than any individual study. For this specific topic, there is particular value in investigating the association across subgroups defined by cigarette smoking status and GSTM1/GSTT1 genotypes.

The results of this systematic review, which included 18 studies, indicate that cruciferous vegetable intake is inversely associated with lung cancer risk. Compared with those who consumed the least amount of cruciferous vegetables, the lung cancer risk among those who consumed the most cruciferous vegetables was 22% lower in case-control studies (statistically significant) and 17% lower in prospective cohort studies (not statistically significant). Furthermore, case-control studies showed a significant inverse dose-response trend. Cohort studies provided only equivocal support for the presence of a dose-response trend. Intake of individual cruciferous vegetables (e.g., broccoli, cabbage) was strongly inversely associated with lung cancer risk.

A unique characteristic of cruciferous vegetables is that they are a rich source of glucosinolates (13). The anticarcinogenic properties of cruciferous vegetables may be attributable to isothiocyanates derived specifically from glucosinolates (13, 57). Several experimental and mechanistic studies support a potential anticancer role of isothiocyanates (58, 59). Sulforaphane, an isothiocyanate found in broccoli, is involved in several pathways, including induction of detoxifying genes, cell cycle control, and apoptosis, acting as an antioxidant (59) and inhibiting histone deacetylase (60). These experimental findings buttress the biological plausibility of the association between cruciferous vegetable intake and lung cancer risk. However, the epidemiologic evidence considered in this systematic review does not allow inferences to pinpoint isothiocyanates as the key protective constituent of cruciferous vegetables because other nutrients and phytochemicals (e.g., folate, flavonols, and carotenoids) found in cruciferous vegetables may also contribute to the inverse association with lung cancer.

Genetic factors related to isothiocyanate metabolism have been hypothesized to contribute to interindividual differences in the degree of protection conferred by cruciferous vegetable consumption (9). Specifically, individuals with GSTM1 and GSTT1 null genotypes metabolize isothiocyanates less efficiently, permitting isothiocyanates to remain biologically active for a longer period (9, 61). In our meta-analysis, a gene-diet interaction was present; when stratified by GSTM1 and GSTT1 variants, the inverse associations between cruciferous vegetable intake and lung cancer risk were more marked in those with the double null genotype. Corroborative findings were also reported in a nested case-control study carried out in China, in which the significant inverse association between urinary isothiocyanate levels and lung cancer risk was stronger among men with the GSTM1 and GSTT1 double null genotype (10). The lone cohort study (22) to report on the potential interaction between cruciferous vegetables and GSTM1 on lung cancer risk only reported results in the text that were not detailed enough to be included in the formal meta-analyses. The reported P value for interaction was not statistically significant (P = 0.38). The presence of a potential gene-diet interaction adds internal consistency to the overall body of evidence on the association between cruciferous vegetables and lung cancer, and takes a step toward addressing the causal criteria of biological plausibility and coherence. The prevalence of the GSTM1 and GSTT1 homozygous deletion genotypes ranges from 42% to 60% and 24% to 51%, respectively. The high prevalence of these putative low-risk (in this setting) genotypes, coupled with the possibility that those with the null genotypes may derive greater benefit from an isothiocyanate-rich diet, underscores the potential public health importance of this gene-diet interaction.

Any consideration of a dietary factor in relation to lung cancer needs to carefully evaluate the potential confounding role of cigarette smoking. Cigarette smoking is the principal cause of lung cancer and cigarette smokers tend to eat less healthful diets than nonsmokers (62). Thus, even studies that statistically adjusted for cigarette smoking may show associations because of residual confounding (63). In the present report, several lines of evidence suggest that the inverse association between cruciferous vegetable intake and lung cancer may be independent of cigarette smoking. First, the inverse association between cruciferous vegetable intake and lung cancer remained robust even in analyses limited to never smokers. Second, the summary association was barely affected by the inclusion or exclusion of estimates from studies that were not adjusted for cigarette smoking. Third, residual confounding by smoking is unlikely to explain the interaction between cruciferous vegetable intake and GST genotypes.

A weakness of the evidence that comprises this systematic review is the measurement error inherent in the use of dietary questionnaires. The measurement error is of less concern if it is nondifferential, but the evidence from case-control studies may be prone to differential reporting because of recall bias, whereby the dietary recall of lung cancer cases may differ from controls (64). That the observed summary estimate for case-control studies was so similar to that for cohort studies suggests that in this instance recall bias does not play a prominent role. Another weakness in the evidence is that the biological potency of cruciferous vegetables may differ, depending on whether the vegetables are consumed raw or cooked, because this influences the bioavailability of isothiocyanates (65). The lack of information about cooking methods thus introduces a potential source of heterogeneity in the results.

In this systematic review, higher intake of cruciferous vegetables was modestly inversely associated with lung cancer risk. The data accumulated to date are consistent across study designs and the different populations studied. The evidence is robust that this association is independent of cigarette smoking. The association is biologically plausible, and the case for plausibility is strengthened by the stronger inverse association observed among individuals with the null genotype for GSTM1 and GSTT1. This line of inquiry seems to hold promise for lung cancer prevention.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.