Colorectal cancer (CRC) is the third most common cancer worldwide, with an estimated number of new cases and deaths in 2008 of 1·2 million and 608 000 (accounting for 8 % of all cancer deaths), respectively( Reference Ferlay, Shin and Bray 1 ). CRC incidence rates vary markedly worldwide, with rates per 100 000 from 4·1 to 59·1 for males and from 3·6 to 39·5 for females( Reference Center, Jemal and Smith 2 ). These large variations have been attributed to different environmental, lifestyle and dietary factors. In particular, diet and nutrition are estimated to explain as much as 50 % of the worldwide incidence of CRC( Reference Vargas and Thompson 3 ).
Among different foods, garlic (Allium sativum) has received a particular attention in recent years because of its high content of organosulfur compounds and flavonoids( Reference Iciek, Kwiecień and Włodek 4 ). The allyl sulfur constituents in garlic, which comprise ~1 % of its dry weight, seem to be responsible for its chemopreventive effects( Reference El-Bayoumy, Sinha and Pinto 5 ). In particular, these compounds have been shown to reduce the formation of aberrant crypt foci and to prevent carcinogen-induced colon cancer in different animal models( Reference Ross, Finley and Milner 6 ). The mechanisms by which sulfur compounds inhibit cancer cell growth have been the topic of intense research over the last two decades and include: activation of metabolizing enzymes that detoxify carcinogens; reduction of DNA adducts; antioxidant effects; regulation of cell-cycle arrest; induction of apoptosis and differentiation; histone modification; and inhibition of angiogenesis and invasion( Reference Powolny and Singh 7 , Reference Yi and Su 8 ). On the other hand, conclusions from epidemiological research aimed to find an association between garlic consumption and CRC risk are inconsistent or even contradictory( Reference Ngo, Williams and Cobiac 9 , Reference Alpers 10 ). In a previous meta-analysis, a risk reduction of 31 % could be observed between CRC and garlic intake (raw and cooked, excluding garlic supplements; RR=0·69; 95 % CI 0·55, 0·89)( Reference Fleischauer, Poole and Arab 11 ). Since then several other studies have been published on this topic with contrasting results. Therefore we conducted a meta-analysis for deriving a more precise estimation of this association.
Materials and methods
Our literature search was aimed at identifying available research studies that examined the effects of garlic on CRC. Studies included in our meta-analysis were identified by searching, without restrictions, multiple literature databases including ISI Web of Knowledge, MEDLINE and EMBASE, and selecting all articles published up to September 2014. The following strings were used for the search: (‘garlic’ OR ‘allium sativum’) AND (‘cancer’ OR ‘neoplastic disease’ OR ‘neoplasm’) AND (‘colon’ OR ‘colorectal’ OR ‘rectal’). Additionally, we also checked the reference lists of retrieved papers and recent reviews. After removing duplicates in the primary research we identified 439 studies. Although useful to have background information, reviews and meta-analyses were excluded.
We systematically reviewed and selected the studies that met all of the following criteria: (i) the study (cohort or case–control) must have had garlic consumption assessed; (ii) it must have provided a risk estimate (hazard ratio, relative risk or odds ratio) for colorectal, colon or rectal cancer incidence as well as its 95 % confidence interval; and (iii) it must have provided information on adjustment for confounding factors. Two investigators reviewed the eligibility of all studies according to the predetermined selection criteria independently. From the results of the selected studies, we extracted the risk estimate of the highest relative to the lowest intake for the analysis. For the overall estimation the hazard ratio and relative risk were taken to be approximations to the odds ratio, and the meta-analysis was done as if all types of ratio were odds ratios. The combined risk estimate was calculated using a random-effects model in which the effect measures were odds ratios or relative risks. In this analysis data from both females and males, and from colon, rectal and colorectal cancer, as independent populations, were included.
Assessment of study quality
The study quality was assessed by a system based on the Newcastle–Ottawa Scale method( Reference Wells, Shea and O’Connell 12 ). Two investigators (R.F. and M.C.) assessed the quality of each selected study and discrepancies were addressed by a joint re-evaluation of the original article with a third reviewer. The full score was 9 and a total score ≥7 was used to indicate a high-quality study. To avoid selection bias, no study was excluded because of these quality criteria.
Heterogeneity between studies was assessed using the Cochrane Q test and I 2 score. The χ 2-based Cochran’s Q statistic and the I 2 statistic were used to quantify evaluated heterogeneity( Reference Higgins and Thompson 13 ). The I 2 statistic yields results ranging from 0 to 100 % and I 2 > 50 % represents substantial heterogeneity( Reference Higgins, Thompson and Deeks 14 ). Results of the meta-analysis may be biased if the probability of a study being published is dependent on its results. We used the methods of Begg and Mazumdar( Reference Begg and Mazumdar 15 ) and Egger et al.( Reference Egger, Davey Smith and Schneider 16 ) to detect publication bias. Both methods test for funnel plot asymmetry, the former being based on the rank correlation between the effect estimates and their sampling variances, and the latter on a linear regression of a standard normal deviate on its precision. If a potential bias was detected, we further conducted a sensitivity analysis to assess the robustness of combined effect estimates and the possible influence of the bias and to have the bias corrected. We also conducted a sensitivity analysis to investigate the influence of each single study on the overall risk estimate by omitting one study in turn. We considered the funnel plot to be asymmetrical if the intercept of Egger’s regression line deviated from zero with a P value of less than 0·10. We should note that this test for asymmetry possesses relatively low power to detect a real publication bias when the total number of studies included in the meta-analysis is small (twenty-five or fewer), which is the case in the current review. The ProMeta Version 2·0 statistical program (Internovi) was used for the analysis. All reported P values are from two-sided statistical tests and differences with P≤0·05 were considered significant.
The flowchart of the study selection process is shown in Fig. 1. After the analysis of titles and abstracts, we identified fourteen studies on garlic consumption and CRC risk in man. In addition, five studies, identified through the reference lists of recent relevant reviews and already selected articles, were included for the analysis. After the full-text assessment five studies were excluded from the analysis as follows: three were intervention trials that tested the effect of aged garlic extracts on colorectal adenoma occurrence( Reference Tanaka, Haruma and Kunihiro 17 – Reference Ishikawa, Saeki and Otani 19 ), one was a case–control study not reporting the risk estimate( Reference Kotzev, Mirchev and Manevska 20 ) and one was a case–control study on adenomatous polyps( Reference Witte, Longnecker and Bird 21 ) (Fig. 1). Therefore only fourteen studies met the inclusion criteria: seven were case–control( Reference Hu, Liu and Yu 22 – Reference Karagianni, Merikas and Georgopoulos 28 ) and seven were cohort studies( Reference Giovannucci, Rimm and Stampfer 29 – Reference Meng, Zhang and Giovannucci 35 ) (Table 1). Among the cohort studies two considered garlic supplements only( Reference Dorant, van den Brandt and Goldbohm 31 , Reference Satia, Littman and Slatore 33 ), while among the other five studies two considered garlic supplements in addition to dietary intake( Reference McCullough, Jacobs and Shah 34 , Reference Meng, Zhang and Giovannucci 35 ). Table 1 summarizes the detailed characteristics of included studies. Three studies reported findings only for females( Reference Hu, Liu and Yu 22 , Reference Steinmetz, Kushi and Bostick 30 , Reference Sellers, Bazyk and Bostick 32 ), one only for males( Reference Giovannucci, Rimm and Stampfer 29 ), eight for females and males together( Reference Iscovich, L’Abbé and Castelleto 23 , Reference Levi, Pasche and La Vecchia 25 – Reference Karagianni, Merikas and Georgopoulos 28 , Reference Dorant, van den Brandt and Goldbohm 31 , Reference Satia, Littman and Slatore 33 , Reference McCullough, Jacobs and Shah 34 ) and three presented findings for males and females separately( Reference Le Marchand, Hankin and Wilkens 24 , Reference McCullough, Jacobs and Shah 34 , Reference Meng, Zhang and Giovannucci 35 ). Three studies reported results of risk for rectal cancer( Reference Hu, Liu and Yu 22 , Reference Dorant, van den Brandt and Goldbohm 31 , Reference Meng, Zhang and Giovannucci 35 ), six for colon cancer( Reference Iscovich, L’Abbé and Castelleto 23 , Reference Giovannucci, Rimm and Stampfer 29 – Reference Sellers, Bazyk and Bostick 32 , Reference Meng, Zhang and Giovannucci 35 ) and eight for CRC( Reference Le Marchand, Hankin and Wilkens 24 – Reference Karagianni, Merikas and Georgopoulos 28 , Reference Satia, Littman and Slatore 33 – Reference Meng, Zhang and Giovannucci 35 ).
HR, hazard ratio; RR, relative risk; F, females; M, males; Ref., reference category; NA, not assessed; PA, physical activity; TEI, total energy intake; WC, waist circumference; NSAID, non-steroidal anti-inflammatory drugs; PMH, postmenopausal hormones.
* Garlic, onion and red pepper.
† Garlic and onion.
Study-specific quality scores of each study are summarized in Supplemental Table S1 and Supplemental Table S2 for case–control and cohort studies, respectively (see online supplementary material). The ranges of quality score were from 5 to 8 (median: 6) and from 6 to 9 (median: 8) for case–control and cohort studies, respectively. High-quality studies (i.e. those studies that had a score ≥7) included three case–control( Reference Galeone, Pelucchi and Levi 26 – Reference Karagianni, Merikas and Georgopoulos 28 ) and six cohort( Reference Giovannucci, Rimm and Stampfer 29 – Reference McCullough, Jacobs and Shah 34 ) studies.
The analysis of the fourteen studies pooled together yielded a combined risk estimate of 0·93 (95 % CI 0·82, 1·06; P=0·281) and test of heterogeneity Q=176·85 (I 2= 83·6 %, P≤0·001). Publication bias was investigated by a funnel plot (Fig. 2). Bias detection revealed a significant effect (P≤0·001) using the method of Begg and Mazumder( Reference Begg and Mazumdar 15 ), while no bias was detected by the Egger test( Reference Egger, Davey Smith and Schneider 16 ) (P=0·121). Sensitivity analyses investigating the influence of each single study on the overall risk estimate by omitting one study in turn suggested that the overall risk estimates were not substantially modified by any single study, with a range from 0·87 (95 % CI 0·73, 1·04) to 0·96 (95 % CI 0·84, 1·08). Of note, the heterogeneity was still observed after omitting each study in turn. Further analyses were performed by stratifying the data on the basis of study type. The forest plots are reported in Fig. 3(a) (case–control studies), Fig. 3(b) (cohort studies) and Fig. 3(c) (supplement studies). The results showed that only in the case–control studies was there a statistically significant reduction (37 %) of cancer risk in association with garlic intake, with a risk estimate of 0·63 (95 % CI 0·48, 0·82; P=0·001; Table 2). Table 2 also reports the results of both heterogeneity and publication bias tests. For the case–control studies only, a high significant heterogeneity was observed while publication bias was significant by the Egger test. A sensitivity analysis excluding Karagianni et al.’s study( Reference Le Marchand, Hankin and Wilkens 24 ), which caused asymmetry of the funnel plot, yielded a combined risk estimate of 0·65 (95 % CI 0·51, 0·84; P≤0·01) with Q=24·45 (I 2=75·5 %, P≤0·001), and P=0·054 and P=0·176 for publication bias by the Begg and Egger methods, respectively. We further separately analysed studies according to cancer sites (colon, rectum, colorectal) and sex (female, male, both). No significant effects were observed in all cases (Table 2).
F, females; M, males.
The results of the current meta-analysis indicate that, when all selected studies (n 14) were considered, garlic consumption was not associated with CRC risk. A small reduction of risk was observed (7 %) but this effect was not statistically significant. Our results are in contrast with a previous meta-analysis based on seven studies, four case–control( Reference Hu, Liu and Yu 22 – Reference Levi, Pasche and La Vecchia 25 ) and three cohort( Reference Giovannucci, Rimm and Stampfer 29 – Reference Dorant, van den Brandt and Goldbohm 31 ), which suggested a preventive effect and estimated that high consumption of garlic decreases the risk of CRC by 30 %( Reference Fleischauer, Poole and Arab 11 ). In comparison with the previous meta-analysis, our updated search identified and included seven more studies( Reference Galeone, Pelucchi and Levi 26 – Reference Karagianni, Merikas and Georgopoulos 28 , Reference Sellers, Bazyk and Bostick 32 – Reference Meng, Zhang and Giovannucci 35 ) which may be responsible for these discrepancies. In addition, our analysis included the estimated risks associated with the use of garlic supplements which were excluded in the previous meta-analysis( Reference Fleischauer, Poole and Arab 11 ). Stratification of the sample on the basis of cancer sites (colon, rectal and colorectal) and sex (female, male and both) also revealed no statistically significant effects on CRC risk. However, when separately analysed on the basis of study type, we found that garlic was associated with a significant reduction (37 %) of CRC risk in the case–control studies while no effects were observed for both cohort and supplement studies. This discrepancy between the case–control and cohort studies may be due to several reasons. Case–control studies have several weaknesses and critical points, which can lead to incorrect conclusions. They are particularly susceptible to recall and selection bias which may produce misclassification of exposure between case and control groups. Moreover, the control group may not be representative of the general population as a consequence of various degrees of selection bias among healthy subjects( Reference Grimes and Schulz 36 ). In addition, some case–control studies included in the current meta-analysis did not adjust for important confounding factors such as red meat, energy intake, alcohol and others that have been consistently associated with CRC risk( Reference Win, Macinnis and Hopper 37 ). Therefore, as discussed above, findings derived from retrospective studies should be interpreted with caution while cohort studies certainly have a greater degree of reliability. However, it should be noted that six cohort studies out of seven were performed in the USA, so suggesting that geographical differences may exist. Similarly to cohort studies, also the results on supplements suggest that use of garlic supplements was not able to prevent CRC. Our results are in agreement with a recent meta-analysis, published during the preparation of this manuscript, investigating the association between high intake of allium vegetables and CRC risk( Reference Zhu, Zou and Qi 38 ). In that study, the stratified analysis showed a not statistically significant increase of CRC risk in association with garlic intake in cohort studies (OR=1·11; 95 % CI 0·95, 1·29)( Reference Zhu, Zou and Qi 38 ). In addition, the use of garlic supplements was associated with a significant increase of CRC risk (OR=1·18; 95 % CI 1·02, 1·36)( Reference Zhu, Zou and Qi 38 ). The small difference with our data may be due to the inclusion in our analysis of two studies( Reference Giovannucci, Rimm and Stampfer 29 , Reference Steinmetz, Kushi and Bostick 30 ) which were excluded in the above reported meta-analysis( Reference Zhu, Zou and Qi 38 ). Moreover, in the analysis on supplements we used a random-effects model and excluded data from past use of garlic supplements( Reference McCullough, Jacobs and Shah 34 ).
Recently, two meta-analyses have been published showing the effects of garlic intake on gastric( Reference Zhou, Zhuang and Hu 39 ) and prostate cancer risk( Reference Zhou, Ding and Liu 40 ). Similarly to our results, a preventive effect of garlic on gastric cancer was observed in three hospital-based case–control studies (OR=0·57; 95 % CI 0·34, 0·80) and in eight population-based case–control studies (OR=0·52; 95 % CI 0·37, 0·67), while one cohort study showed an increased risk (OR=1·28; 95 % CI 0·45, 3·66)( Reference Zhou, Zhuang and Hu 39 ). Furthermore, the results obtained on prostate cancer were similar to our results in showing that meta-analysis from the case–control studies suggested a significant reduction in risk (OR=0·77; 95 % CI 0·64, 0·91) while the results from the cohort studies were null (OR=0·96; 95 % CI 0·89, 1·05)( Reference Zhou, Ding and Liu 40 ). Several mechanisms have been suggested to participate in the potential anticancer effects of garlic and its components. Garlic is rich in organosulfur compounds and flavonoids, which have been reported to exert chemopreventive effects in animal and in vitro studies by different mechanisms including modulation of carcinogen-metabolizing enzymes, cell-cycle arrest, induction of apoptotic cell death and/or differentiation, suppression of oncogenic signal transduction pathways, and inhibition of neoangiogenesis( Reference Antony and Singh 8 , 41 ). It should be noted that all of these effects have been evidenced at high doses of compounds which may be not easily reached with the normal human diet. This is particularly evident for garlic, which is generally used in low amounts.
The present meta-analysis provides evidence that consumption of garlic is not associated with a reduced CRC risk. The preventive effect suggested by the case–control studies may be due to potential confounding factors and exposure misclassification. Further studies will be needed to clarify these discrepancies; in particular, cohort studies should be carried out in the continents of Asia and Europe to confirm the US findings.
Financial support: All work was completed at the University of Perugia, Italy. The authors thank their home institution for financial support. The University of Perugia had no role in the design, analysis or writing of this article. Conflict of interest: None. Authorship: Study concept and design: M.C., L.M. and R.F. Acquisition of data: M.C. and R.F. Analysis and interpretation of data: M.C., L.M. and R.F. All authors contributed substantively to this manuscript, were involved with critical revisions to the manuscript and provided approval for its publication. Ethics of human subject participation: Ethical approval was not required.
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1368980015001263