Consumption of hot beverages and foods and the risk of esophageal cancer: a meta-analysis of observational studies

Previous studies have mostly focused on the effects of specific constituents of beverages and foods on the risk of esophageal cancer (EC). An increasing number of studies are now emerging examining the health consequences of the high temperature of beverages and foods. We conducted a meta-analysis to summarize the evidence and clarify the association between hot beverages and foods consumption and EC risk. We searched the PubMed, Embase, and Web of Science databases for relevant studies, published before May 1, 2014, with the aim to estimate the association between hot beverage and food consumption and EC risk. A random-effect model was used to pool the results from the included studies. Publication bias was assessed by using the Begg test, the Egger test, and funnel plot. Thirty-nine studies satisfied the inclusion criteria, giving a total of 42,475 non-overlapping participants and 13,811 EC cases. Hot beverage and food consumption was significantly associated with EC risk, with an odds ratio (OR) of 1.82 (95% confidence interval [CI], 1.53–2.17). The risk was higher for esophageal squamous cell carcinoma, with a pooled OR of 1.60 (95% CI, 1.29–2.00), and was insignificant for esophageal adenocarcinoma (OR: 0.79; 95% CI: 0.53–1.16). Subgroup analyses suggests that the association between hot beverage and food consumption and EC risk were significant in Asian population (OR: 2.06; 95% CI: 1.62-2.61) and South American population (OR: 1.52; 95% CI: 1.25-1.85), but not significant in European population (OR: 0.95; 95% CI: 0.68-1.34). Hot beverage and food consumption is associated with a significantly increased risk of EC, especially in Asian and South American populations, indicating the importance in changing people’s dietary habits to prevent EC.


Background
Esophageal cancer (EC) is the eighth most common cancer in the world and ranks six among all cancers in mortality [1]. Many studies have shown that dietary habits are significantly correlated with the occurrence of EC [2,3], most of which linking specific constituents of beverages and foods to EC. For example, Polyphenols in green tea was found to inhibit esophageal tumorigenesis [4], whereas maté infusion and caffeine appeared to induce mutagenic effects [5]. An increasing number of studies have investigated the possible relationship between the temperature of beverages and foods and EC risk [6][7][8], since recurrent thermal injuries to the esophageal mucosa owing to the consumption of hot drinks or foods has long been considered a risk factor for EC [9].
Hot beverage consumption could substantially increase the intraesophageal temperature, depending on the initial drinking temperature. An animal study showed that the structure and the function of the esophageal epithelium were damaged by heat stress even [10]. However, epidemiological evidence on the causal relationship between the temperature of beverages and foods and EC is not well established. Research on the relationship was often done as a component of larger studies that focused on specific beverage or food gradients, and the results varied greatly across studies. Some studies found no association between hot beverages and foods and EC risk [11][12][13], arguing that the oral cavity could modulate the heat, and the temperature could fall too rapidly to cause injury to the esophageal mucosa [14]. But many other studies reported that the intake of hot beverages and foods increased EC risk [11,15,16].
In 2009, Islami and colleagues [9] reviewed fifty nine studies and found that over half of the studies showed statistically significant increased risk of EC associated with higher temperature of beverage and food intake. However, the authors did not use quantitative techniques to compute summary estimates of the risk, and the review is outdated. Therefore, we conducted this meta-analysis to ascertain the association between hot beverage and food consumption and EC risk more precisely, relying on all available evidence up-to-date, and to identify the potential factors affecting this association.

Search strategy
This meta-analysis was conducted according to the checklist of the Meta-Analysis of Observational Studies in Epidemiology Guideline [17]. We searched PubMed, Embase, and Web of Science databases from inception to May 1, 2014 for all epidemiological studies on hot beverage and food consumption in relation to EC risk, using the string '(esophageal OR oesophageal) AND (cancer OR carcinoma OR neoplasm) AND (tea OR maté OR coffee OR beverage OR liquid OR alcohol OR food OR diet)'. In addition, we scrutinized the reference lists from retrieved articles to identify other relevant studies.

Inclusion criteria
Studies were considered eligible for inclusion if they met the following criteria: (1) the study was a case-control or cohort study design, (2) it was published in English, (3) the exposure was hot beverage or food consumption, (4) the outcome of interest was EC, and (5) the study reported the odds ratio (OR) or relative risk (RR) with 95% confidence intervals (CIs) for the association between hot beverages or foods and EC risk or provided sufficient data to calculate them.

Date extraction
We extracted the following data from each retrieved article: name of first author, publication year and country of study, study design, specific outcomes, characteristics of study population, number of cases and participants, exposure type, exposure measurement, outcome assessment, comparison categories, OR or RR and corresponding 95% CI, and confounding factors adjusted in the analyses. Data from included studies were independently extracted by two authors (Y.W.C and Y.C), and disagreements were resolved through discussion with the third reviewer (Z.X.L).

Quality assessment
Two independent reviewers (Y.W.C and C.Y) evaluated the quality of the included studies by the Newcastle-Ottawa Scale [18], which was a nine-point scale that allocated points based on the selection process (0-4points), the comparability (0-2 points), and the assessment of outcomes of study participants (0-3points). We assigned scores of 0-3, 4-6, and 7-9 for low, moderate, and high quality of studies, respectively.

Statistical analysis
Random-effects model was used to estimate the summary ORs or RRs for the association between hot beverage and food consumption and EC risk. Taking the subjectivity of differentiating between hot and very hot into account, we used the specific OR for standardized category (hot and very hot) versus reference category (cold and warm) of beverage and food consumption. We defined exposure as hot beverages and foods (standardized category, preference for high-temperature foods and drinks, often consuming of them) versus non-hot beverages and foods (all other combinations). If studies had partly overlapped subjects, only the one with a larger sample size was selected for the analysis. If a study reported results for different beverages and foods separately, those beverage/food specific results were regarded as separate reports on the relationship between temperature and EC risk. One study [11] contained 4 kinds of drinks, and was, therefore, accounted as four independent reports. Another study [19] reporting tea, water and food was regarded as three reports. Two studies [13,20] conducted in two different areas of China were considered as two reports respectively, and another study [16] including two large multicenter case-control studies was treated as two reports.
Statistical heterogeneity among studies was evaluated using the I 2 statistic, where values of 25%, 50% and 75% represent cut-off points for low, moderate and high degrees of heterogeneity, respectively [21]. To assess the heterogeneity across all included studies, the study location (Asia, South America, Europe, Africa), study setting (populationbased, hospital based), study quality (≥7,<7),type of EC (esophageal squamous cell carcinoma (ESCC), esophageal adenocarcinoma (EAC)), and sample size (≥1000,<1000) were further examined using meta-regression. In sensitivity analyses, we conducted leave-one-out analyses [22] for each study to examine the magnitude of influence of each study on pooled risk estimates. Subgroup analyses by age, sex, study location, hot beverage and food categories, study quality, smoking and alcohol intake, study setting, outcome assessment and exposure assessment were conducted to examine the robustness of the primary results. Publication bias was assessed using the Begg test [23], the Egger test [24] and funnel plot. All statistical analyses were performed using STATA version 11.0 (Stata Corp, College Station, Texas, USA). All tests were two sided with a significance level of 0.05.

Literature search
The search identified a total of 3780 unique articles from PubMed, Google scholar, and Web of Science databases, of which 189 articles were identified as potentially relevant. After retrieving and reviewing the full text, we determined that 39 studies met our inclusion criteria. The process of study selection is shown in Figure 1. Table 1 shows the main characteristic of the 39 included studies. These studies were published between 1979 and 2014, all of which with case-control design. The sample sizes of studies ranged from 143 to 4,118 with a total of 42,475 subjects. The number of EC cases diagnosed in the studies ranged from 47 to 1,310, with a total of 13,811 reported EC cases. Seventeen studies were conducted in     China [13,19,20,[25][26][27][28][29][30][31][32][33][34][35][36][37][38], six in Uruguay [11,16,[39][40][41][42], three in Argentina [11,16,43], three in Brazil [11,16,44], three in Paraguay [11,16,45], three in India [46][47][48], three in Iran [15,49,50], two in British [8,51], one in Australia [12], one in Sweden [7], one in Greece [6], one in Kenya [52], and one in Japan [53]. Thirty studies reported results for men and women together, four reported the results for men and women separately, and three reported results for men only and two for women only. Three studies reported results separately by type of EC. Two studies were deemed high quality, 36 moderate quality studies, and one low quality study. The average quality score for all included studies was 5.00.

Hot beverage and food consumption and the risk of esophageal cancer
The results from the random-effects meta-analysis of hot beverage and food consumption and the risk of EC were shown in Figure 2. Thirty-two of 47 independent reports from 39 studies suggested a positive relation between hot beverage and food consumption and EC risk.  Table 2 showed the results based on subgroup analyses, which were to examine the stability of the primary results and explore the resources of potential heterogeneity. The associations between hot beverage and food consumption and the risk of EC were similarly significant in subgroup analyses, with the exception of EAC (OR = 0.79, 95% CI = 0.53-1.16, I 2 = 50.30%, P = 0.110) and European population (OR = 0.95, 95% CI = 0.68-1.34, I 2 = 62.40%, P = 0.031).

Sensitivity analysis and meta-regression
We excluded each study in turn and pooled the results of the remaining included studies. The positive association was not materially changed upon the exclusions, with a pooled OR range from 1.75 (95% CI, 1.47 to 2.07; P = 0.001) to 1.87(95% CI, 1.58 to 2.20; P = 0.001), which indicates that the overall result was not significantly influenced by any individual studies. Our meta-regression analysis reveals that the study location (P = 0.001), the type of EC (P = 0.047) and sample size (P = 0.033) were significant sources of heterogeneity. Study location alone explained 34.39% of the τ 2 in the meta-regression; type of EC explained 12.97%; and sample size explained 8.99%. The results were shown in Table 3.

Publication bias
Visual inspection of funnel plot did not identify substantial asymmetry (see Figure 3). The Begg rank correlation test and the Egger linear regression test indicated no evidence of publication bias across included studies (Begg test Z = 0.59, P = 0.557; Egger test t = 1.58, P = 0.121).

Discussion
In this large pooled analysis of 42475 participants (13811 EC cases) from 39 case-control studies, we confirmed a positive association between hot beverage and food consumption and EC risk. Individuals who usually have beverages and food served very hot or hot were almost twice likely to develop EC than individuals who usually have beverages and foods served warm or cold. Our subgroup analyses show that the results held true across various populations despite significant heterogeneity.
Our meta-analysis shows that the consumption of hot beverages and foods are significantly associated with ESCC (OR, 1.60; 95% CI, 1.29-2.00) but not with EAC (0.79, 95% CI, 0.53-1.16). A large body of observational evidence suggests that the risk factors for ESCC and EAC may be different. For example, alcohol intake is a strong and well established risk factor for ESCC but it is not associated with EAC [54]; a high body mass index (BMI) is associated with an increased risk of EAC but a decreased risk of ESCC [55]; ESCC is strongly associated with high-level exposure to tobacco smoking in Western populations [54,56], whereas EAC is associated with gastro-esophageal reflux disease and Barrett's esophagus [57]. More studies are needed to explore why hot beverage and food consumption is associated with an increased risk for ESCC but not EAC.
Another notable finding is that hot beverage and food consumption appears not to be a risk factor for EC in European population (OR, 0.95; 95% CI, 0.68-1.34). The result might be ascribed to the small sample size (3,728 participants and 1,039 EC cases) or the unique dietary habits of Europeans. A previous study noted that Europeans tend to add cold milk to the exposure beverages, tea or coffee before consumption [12], which may cause people say they drink hot actually only warm and result in substantial difference between the temperature perceived by drinkers and the actual temperature of their drinks.
It is conceivable that hot beverages and foods may cause thermal injury to the esophageal mucosa, and there are several biological mechanisms through which thermal injury in general could increase the risk of EC. Inflammatory processes associated with chronic irritation of the esophageal mucosa caused by local hyperthermia could stimulate the endogenous formation of reactive nitrogen species and nitrosamines [58]. This hypothesis is supported by a high rate of somatic G to A transitions in CpG dinucleotides of the TP53 gene in Figure 2 Forest plot of odds ratios from 39 studies linking hot beverage and food consumption and the risk of esophageal cancer. esophageal tumor samples from geographical areas in which drinking hot beverages is considered an important risk factor for EC [59][60][61][62]; these mutations may indicate increased nitric oxide synthase activity in tumors [63]. The barrier function of the esophageal epithelium can be impaired by thermal injury, which may increase the risk of damage from exposure to intraluminal carcinogens [10], such as polycyclic aromatic hydrocarbons. Elevated temperatures could also accelerate metabolic reaction, including those with carcinogenic substances in tobacco and alcohol [64]. In fact, the association between consuming hot drinks and the occurrence of precancerous lesions of the esophagus has been repeatedly reported [65][66][67]. In addition, dietary deficiencies may weaken the esophageal tissue because of the constant irritation, which may act as a predisposing factor for EC [47]. It has also been postulated that contact of hot liquid and food with the esophageal mucosa could increase gastric reflux, causing further damage from gastric acid [68]. One review proposed that the overproduction of prostaglandin E 2 and leukotriene B 4 as well as overexpression of their receptors are major factors in exacerbating inflammation and oxidative stress, which is the main pathogenesis associated with EAC [57]. The result from our meta-analysis of epidemiological studies is consistent with these biomedical research findings and postulations.   All the original studies used in our meta-analysis are of case-control study design, which is particularly vulnerable to potential biases (both selection bias and information bias). The included studies were conducted among different populations, mostly along with various categorizations of beverages and foods, which could confound our analysis on the specific link between the temperature of beverages and foods and the risk of EC. Lastly, the study relied on self-reported consumption of hot beverages and foods; as a result, the categorization of "hot or very hot" versus "cold or warm" is subject to reporting bias. In addition, the limited information provided in the included studies ruled out the possibility of conducting a doseresponse analysis. Nonetheless, this is the first metaanalysis to systematically quantify the association between hot beverage and food consumption and EC risk, and the results of our study are of broad interest to medical science and the public since consumption of beverages such as tea, coffee, and maté are prevalent worldwide [64,69,70] and many people prefer to drink them at a high or very high temperature [15,71].
In the light of our findings, certain factors should be considered in future studies. Large prospective studies are needed to investigate the association of hot beverage and food consumption with both EC risk and the type of EC, not only because of the different ESCC and EAC risk factors but also the rapid changes in incidence of EAC [12]. In addition, measuring the actual temperature of hot beverage and food would provide dose-response data that would allow for evaluation of the relationship with EC risk more precisely. Finally, confounding factors, such as BMI, smoking, alcohol intake, and socioeconomic status, should be adjusted to allow dissection of the actual influence of hot beverage and food on EC, thereby providing provide stronger research-based evidence.

Conclusions
In summary, our meta-analysis shows that hot beverage and food consumption is associated with a significantly increase in the risk of EC, especially in Asian and South American populations. Given that hot beverages and foods are prevalent in modern society, the results of our meta-analysis have important implications for cancer etiology research as well as applications in health education and clinical practice.

Ethics approval
Ethical approval is not required for this review.