Results of the present study suggest that among macronutrients, consuming more carbohydrate, dietary fiber and (n-3) fatty acids and among micronutrients, higher intakes of folate, vitamin E and selenium have the most protective effects against ESCC in a high-risk population in Iran. An increased ESCC risk estimate was observed among those with highest intakes of SFA, cholesterol, discretionary calorie, sodium and total fat. Most importantly, being in the highest tertile of joint folate and vitamin E intake was associated with 98% reduction in the ESCC risk.
This is the first study in a high-risk population to evaluate the impact of a wide range of macronutrients, minerals and vitamins on risk of ESCC. Similar to previous case-control studies, we found a decreased ESCC risk associated with higher intakes of nutrients with plant origin and an increased risk for intake of several nutrients found primarily in animal-based foods [28, 29, 58–61]. In addition to the differences in nutritional composition of animal- and plant-based foods that contribute to this effect, heterocyclic amines that are potent mutagens, and animal carcinogens formed during cooking of meats are also responsible . The highest level of mutagenic activity is produced during frying, broiling and barbecuing animal products , which could potentially injure the esophageal mucosa .
Epidemiological studies have shown a positive association between total fat, cholesterol and SFA intakes with ESCC and esophageal adenocarcinoma risk [33, 39, 57, 64–67]. In the present study, more than one-thirds of total energy intake among ESCC cases was derived from dietary fat and those with higher intakes of total fat, SFA, cholesterol and discretionary calories had an increased risk of ESCC. According to the Dietary Guidelines for Americans 2005 , 12-20% of total energy intake could be taken from discretionary calories, while in the present study more than 50% of total calories consumed were from discretionary calories, which is of concern since those in higher tertiles of discretionary calorie intake had about 1.5 times higher risk of ESCC. Since all our analyses were adjusted for usual adult BMI, the risk-enhancing effect of high fat diet on ESCC observed in the present study was independent of adiposity, which is a strong risk factor for carcinogenesis. Further effect modification by BMI revealed that although ESCC risk was higher among those with higher BMI values, p for interaction was not significant (data not shown). However, this effect is likely to have been underestimated since ESCC patients tend to decrease their dietary fat intake in an effort to prevent exacerbation of reflux symptoms, and hence our result is likely to have been distorted through underestimation of magnitude of true association for dietary fat. Although some studies have shown an inverse association between EC risk and intakes of added oil and PUFAs , we failed to show a significant relationship.
Findings of previous studies have been inconclusive regarding the role of protein intake in esophageal cancer risk with some classic studies suggesting an inverse relationship [28, 69]. We observed a positive association between protein intake and ESCC risk, which is in line with more recent studies [33, 38, 39, 57, 67, 70]; however, this effect was only statistically significant in the age- and sex-adjusted model.
High intake of dietary fiber in the present study decreased ESCC risk by about 70%. Although few studies have questioned the role of dietary fiber in cancer protection [71–74], most have proved a strong inverse association between fiber intake and risk of ESCC, esophageal adenocarcinoma and stomach cancer [33, 34, 38, 57, 59, 60, 67, 70, 75, 76]. The role of carbohydrate intake in esophageal cancer risk is not yet clear with some studies showing a negative association , but not all [77–79]. In the present study, participants with higher carbohydrate intakes had markedly reduced ESCC risk. However, carbohydrate intake was also negatively correlated with fat intakes (correlation coefficient = -0.615) and hence a higher percentage of carbohydrate may just reflect lower intakes of fat and explain its inverse association with ESCC . In addition, higher consumption of carbohydrate could be reflection of more plant-based food intakes, and especially fruit and vegetable; although in the present research, the correlation between carbohydrate and fruits and vegetable intakes was not significant (r = 0.064).
It has been suggested that the cancer-protective effects of fruits and vegetables intake is mediated through their several antioxidants and dietary components, such as folate, vitamin A, β-carotene, vitamin C and dietary fiber [64, 69, 80–83]. Previously we showed an inverse association between fruit and vegetable consumption and risk of ESCC in the same population , and in the present study, after adjustment for fruit and vegetable intake, the association of dietary folate, vitamin E and selenium with ESCC risk remained significant, suggesting an independent protective role for these nutrients.
Epidemiological evidence regarding the role of folate intake in ESCC risk are scanty [21, 67, 84, 85]. Findings of the present research are in agreement with those of the previous studies showing a strong inverse association between dietary folate intake and risk of ESCC [21, 22, 57, 67, 85]. However, it is likely that we have more clearly observed an inverse relationship between folate intake and ESCC compared to other studies , since in Iran there is no mandatory folate fortification and the use of dietary supplements is very uncommon; hence, folate is mainly taken from diet in this population. In populations with mandatory folate fortification and frequent supplement use, most of the population may have sufficient folate intakes to prevent cancer from these sources and little further benefit may be seen for dietary folate intake. On the other hand, the marked cancer-protective effect we observed for high folate intakes could be partly attributed to the comparatively high rates of folate intake deficiency, as more than 90% of cases and 50% of controls in this study had intakes below the Recommended Dietary Allowances (RDA) (data not shown) . Folate is an important cofactor in DNA metabolism and its deficiency has been linked to higher risk of epithelial tumors [22, 23, 25, 70, 87]. Several mechanisms have been proposed to explain the protective effect of folate, which are mainly focused on prevention of hypomethylation and maintenance of the DNA repair system by influencing the nucleotide pool for DNA replication and repair [88–90].
Similar to folate, vitamin B2, B6, B12 and methionine have major roles in one-carbon metabolism. Previous studies in Iran have reported inadequate riboflavin intakes among patients with esophageal cancer . This is in line with findings from this research and those of previous studies, which have shown protective effects for this nutrient against the risk of EC [28, 69, 82, 91]. Inadequate vitamin B6 intakes among ESCC cases also, might leads to chromosome breakage , defective DNA synthesis and methylation and could increase the ESCC risk [35, 67, 85].
In this study, we failed to observe a significant association between vitamin B12 intake and ESCC risk. However, it has been suggested that the positive relationship between vitamin B12 and EC risk observed in some studies , could be explained by the fact that vitamin B12 is derived exclusively from animal sources and hence may be simply a marker for consumption of animal protein and other factors or nutrients in these foods . It has also been documented that individuals living in the high ESCC risk regions have significantly lower vitamin B9 and B12 intakes, compared to those living in the low risk areas  and those with vitamin B12 deficiency disorders (e.g. pernicious anemia) are at greater risk of EC [94, 95]. Overall, causal relationship between vitamin B12 intake and ESCC risk is not yet clear and evidence in this regard is lacking. In the present study, high methionine intake was inversely associated with ESCC risk, which might be explained through its involvement in SAM production, which is necessary for retaining folate in body 
Higher intakes of vitamin E and vitamin D in this study were associated with about 90% and 70% ESCC risk reduction, respectively. The finding of a strong inverse association between ESCC with vitamin D intake in this study is consistent with some studies [97, 98], although in contrast with others . This contradiction could be explained by the fact that dietary and supplemental vitamin D intakes only comprise a relatively small proportion of the variation in 25-hydroxy vitamin D levels in the body, and sunlight exposure, skin pigmentation, geographic region of residence, season, BMI, and differences in vitamin D receptor expressions (genetic differences) are the major predictors of 25(OH) D levels in the body [99, 100].
Dietary antioxidants (vitamin C, β-carotene and vitamin E) have been shown to decrease the EC risk [27, 28, 33, 38, 39, 65, 82, 98, 101] through several mechanisms such as deactivating excited oxygen molecules and preventing lipid peroxidation. [27, 28, 38, 61, 65]. Dietary antioxidants also play major roles in prevention of damage to the mucosa of the upper aerodigestive tract caused by oxidative stress of smoking and alcohol consumption. In the Linxian China trial, supplementation with a combination of vitamin E, β-carotene, and selenium reduced the incidence of esophageal/gastric cardia cancer by 6% ; this is consistent with our findings of a strong inverse association between vitamin E and selenium with the risk of ESCC. Dietary selenium is inversely associated with cell cycle predictors of neoplastic progression and the ESCC risk [103–105]. The potential of combined supplementation in the Linxian trial in reducing the EC risk has been mainly attributed to the effect of selenium, which has a highly deficient intake in Chinese population. This is in agreement with results of our study in which none of the cases had adequate selenium intakes, and higher intake of selenium was associated with 85% reduction in ESCC risk.
Vitamin A and β-carotene were not significantly associated with ESCC risk in our population, which is in line with several studies [57, 69, 81, 98, 106] and in contrast with others [107, 108]. Our inability to detect a significant relationship may be due to considering vitamin A intakes from both plant and animal origins together, while it has been suggested that vitamin A of plant origin is associated with decreased ESCC risk, whereas vitamin A of animal origin increases the level of risk [67, 109]. In addition, the protective effect of high carotene intake observed in some of the previous studies could have been mediated through high intakes of plant-based foods, which contain different micronutrients and hence contribute to the general effect [59, 61, 98]. High intake of vitamin C in this study was associated with more than 60% reduction in ESCC risk. Vitamin C, as an important antioxidant and inhibitor of endogenous synthesis of N-nitroso compounds [109–111] prevents carcinogenesis of esophageal cells . However, an intervention trial in China failed to show a reduction in esophageal cancer incidence and mortality in individuals taking 120 mg/day vitamin C for 5.25 years .
In the present study, higher sodium intake was associated with almost 50% increase in the ESCC risk. Some epidemiological studies have suggested a role for higher salt intake in carcinogenesis. Although salt is not a carcinogen per se, it acts as an irritant to the esophageal protective mucosal layer, which results in inflammatory regenerative response, increased DNA synthesis and cell proliferation  and may also enhance carcinogenesis induced by other carcinogens .
Deficiency of several vitamins/minerals has been associated with higher EC incidence, with the most pronounced effect observed in the developing countries . Previous studies in Iran have reported high rates of vitamin/mineral deficiencies among EC patients [42–44]; which is in line with previous research showing deficiency of zinc , calcium  and potassium  to be widespread among EC patients. Calcium intake from foods in this study was associated with about 50% reduction in ESCC risk. However, supplemental calcium intake has previously failed to show beneficial effects on EC risk, which could been explained by the confounding effect of higher calcium supplement intakes in the form of GERD medication by cases compared to the controls .
This study has several limitations. Firstly, as with other case-control studies, recall bias and selection bias were inevitable. In case-control studies, there is the possibility that cases may recall their diets differently after a cancer diagnosis. However, our participants were generally of low literacy and socioeconomic status with little knowledge about the role of diet and nutrients in the cancer risk, which should have reduced the possibility of recall bias. Moreover, using hospital controls and administering validated FFQs by trained interviewers in a hospital setting might have further reduced the recall bias and improved comparability of information of cases and controls [115, 116]. With regards to the selection bias, high participation rates (94% among cases and 91% among controls) in this study minimized the potential for selective participation according to the lifestyle practices (such as diet).
Not having data on participants' alcohol consumption was yet another barrier. Our subjects refrained from reporting their alcohol intake due to the fact that consuming alcohol is legally prohibited in Iran [117, 118]. In addition, since the import of alcoholic beverages is banned in Iran, the contents of alcoholic beverages that Iranians consume may differ from those consumed in other countries [117, 118]. On the other hand, the Iranian FCT does not provide data on any type of alcoholic beverages . Opium use was also not answered by our participants due to the cultural barriers and sensitivity of this issue among Iranian population , which could have resulted in confounded estimates in the present study due to the possible role of opium in ESCC risk. However, it has been suggested that opium contributes to ESCC development only in a subgroup of patients and not in the majority [6, 19]. In contrast to low-incidence regions for EC, a much smaller proportion of esophageal cancer cases are attributed to alcohol, tobacco and opium use in high-risk regions [9, 13, 19, 119]. This suggests a more prominent role for nutritional deficiencies in EC development in high-risk areas, such as Iran and China, where a larger number of cases could be attributed to insufficient nutritional intakes [19, 28, 42, 44, 75, 102, 120].
The third limitation is the possibility of some micronutrient misclassification due to not having data on supplement use. However, it has been suggested that dietary supplements and fortified foods, mask the beneficial effect of food intakes in reducing the cancer risk  and some of them have independent positive association with esophageal cancer risk . In addition, supplement intake is very uncommon among our population since the majority of participants in the present study were rural dwellers with little or no education and low socioeconomic status. The average monthly family income in Iran is about 975 US$  while in the present study only one of the controls had an income higher than this average and there were no significant differences between cases and controls (160.91 US $ in cases and 189.69 US $ in controls). This might relate to the sampling method in the present research which was performed in general hospitals of one of the high-risk regions of Iran. Generally, lack of information on supplement use would most likely have negligible effects on our estimates and if anything, would likely results in underestimation of association between ESCC and nutrient intakes.
Another limitation of the present study was using a semi-quantitative FFQ, which despite its common use for characterizing the habitual dietary intakes, is well-recognized for its weakness in quantification of nutrient intakes . Using a semi-quantitative dietary assessment tool limits our conclusions mostly to comparisons between cases and controls and hence conclusions about adequacy of diet are relative and should be interpreted by caution, since these types of comparisons generally overestimate the true effect of exposure on the outcome. However, the theoretical basics that formed the food frequency method have been based on the good correlation of "frequency" of food intake and the "total weights" of the same foods consumed over a several-day period [124, 125]. However, the potential source of error in the use of FFQ in the present study results from lack of a standardized Iranian FCT , although we employed the same FCTs used for validation of the Iranian FFQ [51, 52]. The fundamental concept behind the calculation of nutrients from FCTs is that the nutrient contents of specific foods are relatively constant, and their variability might not be large enough to distort calculations . In addition, much of the errors relating to sample-to-sample variation in nutrient composition of foods are reduced by using the estimates of long-term nutrient intakes obtained from the FFQs .
Another drawback of this study was using nutrient values in the statistical analyses without directly referring to the foods which contributed most to the nutrient intake and its variation. According to Willet, an optimal approach to epidemiologic analysis is to employ both foods and nutrients to represent diets ; in this way the case for causality is strengthened when an association is observed both with the overall intake of a nutrient and also with more than one source of that nutrient, especially when the food sources are different. Previously we evaluated the role of food group intakes in the etiology of esophageal squamous cell carcinoma (ESCC) among the same population ; by conducting the present study, we aimed at providing preliminary evidence on the extent to which certain nutrients could influence the ESCC risk in such a high-risk region.
Another potential source of error is that several naturally continuous variables (e.g. BMI, physical activity) were categorized for the purpose of analysis, which might have increased the possibility of residual confounding and decreased the precision and power of the study. However, we compensated for this limitation by choosing categories with multiple sufficiently narrow intervals to decrease the residual confounding and heterogeneity of subjects within intervals.
Additionally, availability of B-vitamins is influenced by diet, supplement use, alcohol consumption and generic polymorphism and B-vitamins are all involved in one-carbon metabolism which requires vitamin B2, B6, B9 and B12. This complexity added to the problems of using estimated intake of nutrients obtained at one point in time must be considered when interpreting the results of this study. Small sample size is also a limitation which might have resulted in unstable results and extreme relative risk estimates observed in some of the subgroups, although this is one of the largest sample series in an Iranian population and a number of strong consistent findings have emerged from this sample [45, 50].
This study has several strengths. Firstly, eliminating information bias associated with use of proxy data enabled us to consider numerous potential cofounders. Detailed assessment and adjustment for several important confounders and total energy intake are other important aspects of this study. We attempted to reduce the measurement error and false-positive effect by calculating nutrient intake residuals standardized for total energy intake rather than reporting absolute nutrient values. This further accounted for the confounding effect of total energy intake on nutrient intake estimates .
In the present study, a validated FFQ was used which provided subjects with the option of answering in terms of day, week or month which enhanced reporting precision considering the fact that frequency of consumption is a truly continuous variable . In addition, we asked incident ESCC patients diagnosed within 6 month of the interview to recall their diets from 1 year before diagnosis in order to capture a full cycle of seasons so that responses should not be dependent on the time of the year and be representative of habitual long-term intakes . This is of note since short FFQs that have been previously used for collecting dietary data are considered the main reasons for the contradictory findings on the role of nutrients in cancer risk . In addition, 24-hour dietary recalls which have been used in several studies to assess recent or current diets in relation to cancer risk, could severely compromise the accuracy of mean intake estimates due to the day-to-day variation in dietary intakes . Moreover, study design further limits the reliability of short-term recalls in case control studies, since dietary recall provides information on post-diagnosis diet, while the relevant exposures have occurred earlier . Given the long latency period of cancer, remote dietary intakes are far more important than the recent diet in cancer incidence studies since current dietary intake underestimates the true role of diet in cancer etiology .
Finally, evaluating the nutrient-cancer relationship in a population without mandatory nutrient fortification, where supplement use is uncommon and nutrient intakes are low has enabled us to capture the association of nutrients and ESCC more clearly compared to previous studies ; as such, results of the present study should be considered representative of the influence of vitamins and minerals that are found naturally in foods.