Meat, fish, and esophageal cancer risk: a systematic review and dose-response meta-analysis

bs_bs_banner

Lead Article

Meat, fish, and esophageal cancer risk: a systematic review and dose-response meta-analysis Maryam Salehi, Maziar Moradi-Lakeh, Mohhamad Hossein Salehi, Marziyeh Nojomi, and Fariba Kolahdooz Risk factors for esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC) are well defined, while the role of diet in these conditions remains controversial. To help elucidate the role of particular dietary components, major bibliographic databases were searched for published studies (1990–2011) on associations between esophageal cancer risk (EC) and consumption of various types of meat and fish. Random-effects models and dose-response meta-analyses were used to pool study results. Subgroup analyses were conducted by histological subtype, study design, and nationality. Four cohorts and 31 case-control studies were identified. The overall pooled relative risk (RR) of EC and the confidence intervals (CIs) for the groups with the highest versus the lowest levels of intake were as follows: 0.99 (95% CI: 0.85–1.15) for total meat; 1.40 (95%CI: 1.09–1.81) for red meat; 1.41 (95%CI: 1.13–1.76) for processed meat; 0.87 (95%CI: 0.60–1.24) for poultry; and 0.80 (95%CI: 0.64–1.00) for fish. People with the highest levels of red meat intake had a significantly increased risk of ESCC. Processed meat intake was associated with increased risk of EAC. These results suggest that low levels of red and processed meat consumption and higher levels of fish intake might reduce EC risk. © 2013 International Life Sciences Institute

INTRODUCTION Esophageal cancer is the eighth most common cancer type and the sixth leading cause of cancer mortality worldwide.1 Despite treatment improvements, early diagnosis is difficult and the first-year survival rate remains low at 10–13%.2 Thus, identification of modifiable factors contributing to the etiology of esophageal cancer would lower the burden of this fatal disease. Some risk factors, such as body mass index (BMI), age, smoking, and alcohol consumption, are shared by the two main histological subtypes of esophageal cancer (EC), i.e., esophageal

adenocarcinoma (EAC) and esophageal squamous cell carcinoma (ESCC), although some other risk factors are specific to either ESCC or EAC.3 The reported effects of other factors (including diet) in EC are presently inconsistent in the available literature. Meat consumption has been reported as a potential risk factor for breast,4 ovary,5 prostate,6 gastric,4 and colorectal cancers.7,8 Meat is a source of N-nitroso compounds that have been associated with cancer risk.9 In addition, red meat contains heme iron that can physically abrade and irritate the epithelial lining of the digestive tract causing cellular damage.10 Compared to white meat,

Affiliations: M Salehi is with the Department of Community Medicine, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran, and Research Center for Patient Safety, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran. M Moradi-Lakeh is with the Department of Community Medicine, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran, and the Gastrointestinal and Liver Disease Research Center, Firoozgar Hospital, Tehran University of Medical Sciences, Tehran, Iran. MH Salehi is with the Cancer Research Center, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran. M Nojomi is with the Department of Community Medicine, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran. F Kolahdooz is with the Department of Medicine, University of Alberta, Edmonton, Alberta, Canada Correspondence: F Kolahdooz, Aboriginal and Global Health Research Group, Department of Medicine, University of Alberta, Unit 5-10, University Terrace, 8303-112 Street, Edmonton, AB, Canada T6G 2T4. E-mail: [email protected]. Phone: +1-780-492-3214. Fax: +1-780-492-3018. Key words: diet, esophageal cancer, meat intake doi:10.1111/nure.12028 Nutrition Reviews® Vol. 71(5):257–267

257

red meat is a richer source of heme iron. Thus, white meat’s association with cancer may be attributed to its comparative lack of heme.11 Cooking meat at high temperatures contributes to the formation of carcinogenic and mutagenic heterocyclic amines and polycyclic aromatic hydrocarbons.12 One study13 reviewed the epidemiological data from case-control studies and reported a link between gastric cancer and nitrite and nitrosamine intake, but evidence of an association with EC was insufficient. While the 1997 World Cancer Research Fund/American Institute for Cancer Research report concluded there was inadequate evidence to suggest an association between meat intake and EC risk,14 in 2007, based on 18 case-control studies, the World Cancer Research Fund reported that red meat and processed meat were associated with a “limited suggestive increased risk” for EC.4 These reports did not separate results for ESCC and EAC, yet there is strong evidence that these subtypes have distinct etiologies.15 Since 2007, more studies have been published; thus, the present review was conducted using meta-analysis as a systematic approach to update and expand upon the scientific evidence for an association between the consumption of total meat, processed meat, red meat, white meat, poultry, and fish and the risk of two histologic subtypes of EC. The underlying hypotheses were as follows: 1) intake of red and processed meats is positively associated with a risk of EC, 2) intake of poultry is inversely associated with EC risk, 3) intake of fish is inversely associated with EC risk, and 4) ESCC and EAC have different risk estimates. METHODS The meta-analysis was conducted according to the Metaanalysis of Observational Studies in Epidemiology (MOOSE) guidelines for reviews of observational studies.16 Eligible studies were identified by two authors (MS and FK) who searched the Medline and Embase databases as well as the Web of Knowledge. Reports eligible for inclusion were observational studies published in English from January 1990 through January 2011 that presented risk estimates for an association between EC (ESCC and/or EAC) and consumption of total meat, red meat (beef, pork, lamb), processed meat (bacon, sausage, hot dogs, salami, ham), white meat, poultry, and fish. Cured, preserved, and salted meats were considered to be equivalent to “processed meat.” The following medical subject heading (MeSH) terms and/or text words were used: “meat” or “foods” or “diet” combined with “esophageal cancer,” “esophageal neoplasm,” “esophagus cancer,” or “esophagus neoplasm.” The search was repeated using British spelling. References cited in the identified articles were also manually searched. Studies that were clearly not 258

relevant based on their titles and abstracts were excluded automatically. The remaining articles were fully read to determine their compatibility with the inclusion criteria and if they reported relative risks (RRs) or odds ratios (ORs) and related confidence intervals (CIs) or required data to compute them. To reduce the possibility of missing relevant papers, two authors (MS and FK) checked the results and assessed the quality of studies independently using the Critical Appraisal Skills Programme (CASP). The following information was extracted from relevant studies: first author’s name, publication year, country in which the study was performed (nationality), study design, sample size, year of follow up (for cohort studies), cancer histologic subtypes included, dietary questionnaire used, food items investigated along with intakes evaluated, confounders taken into account, and risk approximations for comparisons between the highest and lowest categories of intakes. Extracted data were inspected for concordance by two authors (FK and MS). Odds ratio estimates from case-control studies and risk or rate ratios from cohort studies were all assumed to be valid approximations of RR.17 A weighted average of the logarithms of RR was calculated using the DerSimonian and Laird method, while random effects were considered in order to pool RR estimates.18 For estimating overall RR in studies that gave separate estimates for ESCC and EAC, the estimates, weighted by the inverse of the variance, were pooled within each study. Cochran’s Q-test and Begg’s rank correlation test were used to assess heterogeneity for each pooled estimate, and Egger’s regression model was used to assess publication bias.19 Heterogeneity was considered to exist for P ⱕ 0.1 for the Q statistic. As recommended in the MOOSE guidelines,16 analyses stratified by key study design aspects were also performed in order to assess the effect that varying study quality may have on the results. Stratified analyses were conducted of histological subtypes of EC (ESCC, EAC) and nationality subtypes (Asian, American, and European); potential confounders were controlled for where data were available. A sensitivity analysis was conducted to assess the contribution of each study to the overall effect. For the dose-response meta-analyses, intake stated in “times” or “servings of intake” was converted into grams (g) using 50 g as a standard portion size for processed meat, 120 g for total meat or red meat, and 100 g for fish.4 When reported in articles, mean or median intake categories were used. If unavailable, midpoints were used as the relative risk of the corresponding category. When the lowest category was unrestricted, zero consumption was considered as the boundary and the amplitude of the lower closest category was used when the highest category was unrestricted. All analyses were conducted using Stata 10 software (StataCorp LP, College Station, TX, USA). Nutrition Reviews® Vol. 71(5):257–267

RESULTS The literature search identified 247 potentially relevant publications. Five additional articles were identified by searching the references of papers. After screening titles and abstracts, 189 publications were considered irrelevant and excluded. The remaining 58 articles were retrieved for additional review, and 44 of them reported data on an association between meat/fish intake and EC. Five of these studies20–24 were secondary reports from a study that was already included and were consequently excluded (Table 1). An additional four studies were excluded from the main analyses because they presented insufficient information for calculating CIs25–27 or reported only doseresponse ORs.28 The final analyses included 35 articles; one article29 reported results from two individual populations (i.e., low- and high-risk areas in China). The characteristics of the four cohort studies30–33 and 31 case-control studies3,9,29,34–60 included in the meta-analysis are presented in Tables 2 and 3, respectively. The final analyses included a total of 35 articles related to meat/fish consumption and risk of EC: 21 for total meat,29–32,37–40,43,44,46,48,49,51,52,54,55,58,59,61 14 for red meat,31,33,37,39,42,43,45,47,50,51,53,56,60,61 17 for processed meat,9,31,33,35–37,39–41,43,44,47,48,50,51,54,61 four for white meat,33,42,51,61 nine for poultry,31,37,39,43–45,47,56,60 17 for fish,29,30,34,37,39,42–45,47,49,53,54,58,59,61 and three for barbecued meat.3,56,57 To estimate the overall RR for EC in two studies that provided separate estimates for ESCC and EAC,33,52 estimates within each study were pooled using a fixed effect model. Three studies3,33,52 reported the results of dietary analyses for ESCC and EAC separately, 14 studies32,36–38,42,46–50,53,56,57,59 focused on ESCC, five studies31,39,43,51,61 focused on EAC, and 13 studies9,29,30,34,35,40,41,44,45,54,55,58,60 did not distinguish between ESCC and EAC. The 35 articles used for the analyses here represent a global perspective: 14 were published in the United States,9,33,35,39,42,43,46,48,50,51,53,56,57,60 11 in Europe,31,34,37,38,41,45,47,49,52,54,61 nine in Asia,29,30,32,36,40,44,55,58,59 and one in Australia.3 The study population included men and women in 32; three studies were conducted among men only 30,36,50; one study was adjusted for age and sex only 32; in all other studies the potential confounders were age, sex, alcohol consumption, and smoking for ESCC and age, sex, BMI, and smoking for EAC.3 Table 4 shows the overall results of the meta-analysis for each category of meat and the subgroup analyses. The overall pooled RR for EC was 0.99 (95% CI: 0.85–1.15) with significant heterogeneity for individuals in the highest category evaluated against those in the lowest category of total meat intake (P for heterogeneity: 0.005; I2: 49.7%). After stratifying by study type and cancer type Nutrition Reviews® Vol. 71(5):257–267

(Figure 1a), the pooled RR did not change materially, whereas in the analysis stratified by nationality, the pooled estimate fell to 0.77 for Asians and became statistically significant (Table 4). In a dose-response metaanalysis of nine studies,32,37,38,40,46,54,58,59,61 there was no evidence of an association between intake of total meat (100 g/day) and risk of EC (RR: 1.01; 95% CI: 0.99–1.01; I2: 33.84; P for heterogeneity: 0.02). Similarly, the trend estimates for both ESSC and EAC or mixed histological subtypes were null (data not shown). When the highest red meat intake was compared with the lowest red meat intake, the pooled RR for EC was 1.40 (95% CI: 1.09–1.81), with significant heterogeneity (P for heterogeneity: 0.001). Following stratification by study type, the pooled RRs were null for populationbased case-control studies, whereas the estimate for cohort studies (RR: 1.32; 95% CI: 1.03–1.71; P for heterogeneity: 0.5; I2: 0.0%) and hospital-based case-control studies (RR: 1.66; 95% CI: 1.02–2.69; P for heterogeneity: 0.001; I2: 82%) showed a significantly increased risk. For red meat intake, seven studies reported results for ESCC and six studies reported results for EAC. The corresponding summary RR was associated with an increased risk for ESCC (RR: 1.63; 95% CI: 1.00–2.63; P for heterogeneity < 0.0001; I2: 79%) but not for EAC (Figure 1b). When the estimate for total red meat was replaced with the estimate for fresh red meat for one population-based case-control study on EAC that reported these separately, the overall pooled RR for EAC increased to 1.29 and became statistically significant (95% CI: 1.03–1.63) with no significant heterogeneity (P for heterogeneity: 0.3; I2: 19%). Among the 14 studies that investigated red meat, only two35,61 were included in the dose-response analysis of red meat intake and risk of EC and the result was null (data not shown). Meta-analysis of 17 studies that evaluated processed meat consumption in relation to EC risk, gave a significant positive association with an overall summary RR of 1.41 (95% CI: 1.13–1.76) for the highest intake category compared with the lowest intake category (P for heterogeneity < 0.0001; I2: 62%). However, when analyses were stratified by study design, the results remained statistically significant only for population-based case-control studies (RR: 1.32; 95% CI: 1.01–1.72; P for heterogeneity: 0.4; I2: 0%) and became borderline significant for hospital-based case-control studies (RR: 1.43; 95% CI: 0.99–2.06; P for heterogeneity < 0.0001; I2: 62%).Analyses by cancer type showed a significant increase in risk for EAC only (RR: 1.37; 95% CI: 1.05–1.78) (Figure 1c). Further analyses revealed there was no association between EC and processed meat intake among three Asian studies, but there were significant increases in risk among the highest intake groups in studies conducted in both Europe and the United States (Table 4). In a dose259

260

Nutrition Reviews® Vol. 71(5):257–267

a

Only male subjects included.

Hung (2004)21 Taiwan

ESCC

284/480a

66/393

282/678

EAC

Hospital-based case-control studies EAC/ESCC De Stefani (1999)23 Uruguay

206/678

ESCC

Silvera (2008)28USA

211/633

EAC/ESCC

Xibib (2003)25 China

Total meat; tertile 3 compared with tertile 1 Red meat; tertile 3 compared with tertile 1 Processed meat; tertile 3 compared with tertile 1 White meat; tertile 3 compared with tertile 1 Salted meat; tertile 3 compared with tertile 1 Cured meat; ⱖ1 compared with