Dietary Determinants of Esophageal Cancer in Arsi Zone, Central Ethiopia: A Case-Control Study

Background: Esophageal cancer is ranked 5 th of all types of malignancies in Ethiopia following uterine, cervical, breast, and colorectal cancers. The ndings regarding the dietary risk factors associated with an increased incidence of esophageal cancer were inconsistent. Methods: A matched case-control study was conducted from from June 1, 2019, to June 30, 2020. A total of 104 cases and 208 controls were involved. Cases were consecutively recruited from higher referral hospitals and matched to two population-based controls (1:2 ratio) on age, sex, and residence/altitude. Data were collected using structured questionnaires, coded and entered into the EPI info version7, and transported to SPSS software version 23. Binary and multiple logistic regressions were conducted to check the association between independent and dependent variables. Adjusted odds ratios and the corresponding 95% condence intervals were estimated to assess the strength of association. P-values <0.05 were used to declare statistical signicance. Results: Cooking foods in a living room (AOR=2.8,[95% CI:( 1.11,7.38), teeth loss ( AOR= 4.4, [95%CI: (1.87,10.56),consumption of very hot porridge (AOR=4.2 [95% 8.72),drinking very hot coffee (AOR=3.9 [95% CI:(1.70,8.93),drinking large volume of coffee at a time (AOR=6.3 [95% CI:(2.93,13.78), eating porridge fast (AOR=9.2[95% CI:(2.98, 28.88) were positively associated with esophageal cancer. Conclusion: Multiple dietary practices were associated with an increased risk of esophageal cancer. The ndings imply the need for behavior change communication targeting cooking and consumption behaviors to curb the of esophageal cancer in the study

total of 104 cases and 208 controls were involved. Cases were consecutively recruited from higher referral hospitals and matched to two population-based controls (1:2 ratio) on age, sex, and residence/altitude. Data were collected using structured questionnaires, coded and entered into the EPI info version7, and transported to SPSS software version 23. Binary and multiple logistic regressions were conducted to check the association between independent and dependent variables. Adjusted odds ratios and the corresponding 95% con dence intervals were estimated to assess the strength of association. P-values <0.05 were used to declare statistical signi cance.

Background
Human diets have been linked to about twenty percent of all cancers in developing countries (1). Diets have decisive roles as a protective or in the initiation and progression of chronic diseases (2,3). Nutrients in the diets control the transcription factors that transform the gene expression while human genetic make can delineate susceptibility to diet-dependent health disorders (4).
Epidemiologic studies found strong signi cant associations between diets and esophageal cancer (EC) (5). Consumptions of red and processed meat, saturated fat, discretionary calorie in the highest category of intakes were signi cantly associated with increased risk of esophageal cancer (5,6). Similarly, an elevated risk of EC was found among high-risk populations who consumed hot coffee, mate, tea, barbecued, and fried meat (7)(8)(9)(10)(11)(12). The other dietary related practices associated with an increased risk of EC were eating salted sh, fried takeaway foods, food eating speed, and teeth loss (13,14).
The associations between hot foods consumption patterns and risk of EC were not always consistent.
The study in Sweden identi ed no signi cant association between drinking hot beverages and EC either independently or after adjusting for other causes (15). According to Wu et. al(2006) opinion, the oral cavity could adjust the heat so that the hot liquid foods temperature could fall too rapidly to cause injury to the esophageal mucosa (16).
Micronutrients and antioxidant substances are protective against esophageal cancer (17,18). Previous studies discovered the strongest inverse relationships between the consumption of vitamins, beta carotene from raw fruits, dark green leafy, and cruciferous vegetables and esophageal carcinoma (19,20). Likewise, an inverse signi cant relationship was found between higher dietary calcium intake and the risk of esophageal cancer (21).
Earlier studies in Africa revealed that the rise of EC incidences in endemic areas over the past decades were attributed to the consumption of crops that had degenerative effects (22). An increased risk was observed among populations that consumed maize (corn) and wheat-based staple foods compared to those who consumed diversi ed and nutritious foods (23). However, recent studies reported Polycyclic Aromatic Hydrocarbons(PAHs) emitted from partially combustible source fuels used for cooking foods, cooking places (24), consumption of hot foods as independent determinants of esophageal cancer among populations living in high-risk areas in Africa (10,(25)(26)(27).
Esophageal cancer is ranked 5th of all types of malignancies in Ethiopia following uterine, cervical, breast, and colorectal cancers (28). Arsi Zone is one of the EC endemic areas in Ethiopia (29). The ndings regarding the dietary risk factors associated with an increased incidence of esophageal cancer were inconsistent. For example, in a small size pilot study, eating salty diets and inadequate vegetable intakes were identi ed as the independent predictors of risk of esophageal cancer (30). A case-control study by  (31,32). Furthermore, no study exists whether a fermented homemade traditional beverage that contains a high concentration of the carcinogenic chemical(acrylamide) would contribute to an increased risk of esophageal cancer (33). The present study was carried out to identify the dietary determinants of EC in the Arsi Zone.

Methods
Arsi Zone is located in Oromia National Regional State in central Ethiopia. The zone lies between 60 45' N to 80 58'N and 380 32 E to 400 50' E. Assela is the capital town of the Zone located at 175 km from Addis Ababa. Barley, wheat, and maize are the pre-dominant cereals and among pulses, horse beans, and eld peas are grown widely. Vegetables, root crops, and stimulants are also grown (34).

Study design
A case-control study was employed from June 1, 2019, to June 30, 2020.

Sampling procedures
Cases were endoscopically examined and histologically con rmed EC patients who attended referral hospitals. They were consecutively recruited from Asella Teaching and Referral Hospital in Arsi Zone and other higher referral hospitals mentioned in a previous epidemiological study (35). The control were healthy individuals (absence of any symptom of cancer during data collection) and who lived in that community for at least 5 years. Controls were recruited from the same kebeles(smallest administrative unit) where the cases have emerged. Lists of eligible controls were prepared and those who gave consent were selected by the lottery method. A ratio of cases to controls of 1:2 ratios was used to select the sample. Further matching of cases and controls was done by age and sex. Three cases (two females and one male) were excluded from the study because of serious illness and their unwillingness to give information. The nal sample size was 104 cases and 208 controls.

Data Collection and Measurement
Data were collected by ve trained BSc nurses using interviewer-administered questionnaires. The data collection tool for dietary practices was adapted from a validated Food Frequency Questionnaire (FFQs) (36)(37)(38) to local food items by researchers. The questionnaire comprised of socio-demographic and economic characteristics, habitual dietary practices, and food cooking places, source of fuels, and the presence of ventilation in a cooking area, food consumption temperatures, and volume of hot drinks. The volume of coffee drinking volume at a time was checked in a survey preceding this study (unpublished data). The smallest coffee drinking (coffee cup) contains 80 ml to 140 ml of coffee and is labeled as a low volume coffee intake. Glasses, beakers, and gourd bottles (Quluu in the local language) contain about 300 ml of coffee and categorized as high volume coffee intake. Furthermore, all participants were requested about the speed of eating porridge with a group of people taking into account an anecdotal report of consumption of hot porridge in a shared manner from a serving utensil. Fast eater was de ned as a person who is rst to nish when eating porridge with a group of people, normal eater if neither rst nor last to nish, and slow eater if the person last to nish when eating food with a group of people respectively (8). Finally, drinking of alcohol and non-alcohol beverages, and use of any forms of tobacco, and history of cancer in a family were collected and their responses were categorized as yes or no. Participants were asked their usual dietary habits and frequency of consumption within a day, within weeks, and within months over the past ve years before developing dysphagia. Cases were interviewed at the oncology department, separate café, and home while the interviews for controls took place in subjects' homes.

Quality assurance
The adapted questionnaires were prepared in English, translated to local language (Afaan Oromo), and later back to English by two different experts quali ed in MSc and uent in local languages. Two days of training were provided for data collectors and supervisors regarding study objectives and interview techniques. Pretesting was conducted on 5% of the proposed sample size and amendments were made accordingly. The supervisors strictly followed the data collection procedures and feedback was given daily.

Data processing and Analysis
Data were coded and checked for completeness, consistency and entered into the EPI info version7, and transported to SPSS software version 21 for data processing and analysis. The frequencies of dietary intakes were converted into a continuous variable and computed as mean dietary intake per week. Furthermore, dietary patterns were computed using principal component analysis (PCA). The criteria used for selection variables for PCA analysis were sample size greater than 50, a ratio of cases to variables of 5 to 1 or larger, the correlation matrix for the variables that contain 2 or more correlations of 0.30 or greater, sampling adequacy greater than 0.50 in anti-image correlations and Bartlett test of sphericity is statistically signi cant(P < 0.05), the commonality of the components greater than 0.50, variables that had no loadings, or correlations, of 0.40 or higher for more than one component, components that had no only one variable in it (39). Descriptive statistics were computed and presented in frequencies and percentages for categorical variables, and means with standard deviations for continuous variables. Binary and multiple logistic regressions were conducted to check the association between each independent and dependent variable. Independent variables that have associations with the outcome variable in the bivariate logistic regression and those with a p-value of ≤ 0.25 were considered a candidate for the nal Logistic regression model. The Hosmer Lemeshow test was used to check the goodness of model tting. Multicollinearity was checked using standard error < 2.0. The multivariable logistic regression model was adjusted for the confounding effects of independent variables. Odds ratios (OR) and the corresponding 95% con dence intervals (CI) were estimated to assess the strength of association. P-values < 0.05 were used to declare statistical signi cance. All analyses were performed using SPSS for windows version 23.0 (SPSS, Illinois Chicago, USA).

Dietary intake patterns
In PCA analysis, the dietary patterns of the study subjects were loaded onto two components. The components were grouped as diversi ed diets (legumes, egg, vegetables, and milk) and cereal foods consumed with fats (butter) and oils. Accordingly, 73(70.2%) cases and 131(63%) controls consumed diversi ed foods. The proportion of cereal foods with fats and oils consumption was lower among the cases (29.8%) compared to controls (37.0%). The mean (± SD) frequency of cereal food consumption per week for cases and controls was 4.9(1.1) and 4.5(1.2) respectively. The frequency of vegetable consumption was comparable between cases and controls. Cases consumed meat more frequently than controls. The mean weekly intakes of legumes and pulses, milk and dairy products, fats and oils, sweets, Garlic, and onions consumptions were lower for cases than controls (Table 3).  12(5.8%) controls have ever used some forms of tobacco (Fig. 1).

Determinants of esophageal cancer
Independent variables were sequentially computed in a block to nd the model that best predicts the determinants of esophageal cancer. As a result, ethnicity and family size from the background variables; a place for food preparation, source fuels for cooking foods, types of porridge, inputs added into porridge, porridge temperature, speed during porridge consumption, coffee drinking patterns, coffee temperature, the volume of coffee drunk, dinner time, and nonalcoholic homemade drinks had a p-value < 0.25. The interaction between dietary practices, alcohol drinking, tobacco use, and khat chewing was checked but no signi cant association was observed. The nal model was selected based on the theoretical and statistical signi cance of the predictors. Hence, cooking foods in a living room (AOR = 2.8, [95% CI: (

Discussion
The study ndings showed variables including food cooking place, teeth loss, porridge consumption temperature, coffee drinking temperature, the volume of coffee drank and speed during porridge consumption were signi cantly associated with increased risk of EC after adjusting for other variables.
The place of cooking foods was signi cantly associated with an increased risk of esophageal cancer. The nding is consistent with the study in Malawi that identi ed cooking foods in a living room as a signi cant predictor of esophageal cancer (40). Food cooking places were reported as one of the causes of indoor air pollutants and carcinogens found in foods (40,41). For instance, a study in Iran reported contamination of foods with PAHs as a risk factor of EC among the populations living in the EC endemic area (42). The nding has practical applicability as the majority of populations in Ethiopia use their living houses for a cooking place (43).
Another remarkable nding in this study was that teeth loss appears to be a risk factor for esophageal cancer. The nding is similar to previous studies that reported teeth loss as an independent determinant of esophageal cancer (8,25,(44)(45)(46)(47). Measuring the effects of tooth loss after the development of EC may be prone to reverse causation. However, the persistent signi cant association of teeth loss with an increased risk of EC after adjusting for potential confounders in this study and pooled meta-analysis study conducted among other populations (46) make it an important risk factor of esophageal cancer.
Several mechanisms were postulated on the role of teeth loss in the pathogenesis of cancer. Chronic in ammations and periodontitis caused by bacterial infections may create a pocket of toxic metabolites while poorly masticated hard foods may injure the esophageal linings during swallowing (45). Besides, individuals who lost teeth may avoid the consumption of hard texture foods such as fruits and vegetables which are protective against cancer (44). On the other hand, individuals who lost teeth may modify their diets to soft foods such as hot porridge and soup that can induce thermal injury to the esophagus.
Other independent determinants of EC in the present study were food and beverage consumption temperatures. The odds of developing EC were higher for very hot porridge consumers compared to those who consumed hot porridge. The association between porridge consumption temperature and the risk of EC is consistent with the study ndings in Ethiopia (32,48).
The strongest association was observed between the speed of consuming porridge and the risk of esophageal cancer. Consequently, fast eaters were 9.2 times more likely to develop EC compared to slow eaters. The nding supports the study that found fast eating as a signi cant predictor of esophageal cancer (8). Case report studies revealed severely damaged esophageal linings after individuals swallowed large bolus of hot foods (49,50). The reason could be fast eaters may swallow a very hot bolus of porridge without moderating the temperature through the air or by mixing with saliva in the oral cavity.
Coffee drinking temperature demonstrated a signi cant positive association with an increased risk of esophageal cancer. The likelihood of developing EC was 3.9 times higher for very hot coffee drinkers compared to hot coffee drinkers. The nding is alike to studies that reported positive associations between coffee drinking temperatures and risk of esophageal cancer (10, 51)but contrary to a study in Europe that did not nd a signi cant association between drinking hot coffee and esophageal cancer (52).
The disparity between the study ndings can be explained by the fact that populations in Europe usually add cold milk to hot coffee before drinking it. Besides, there are remarkable differences in the histological types and etiological factors of EC across the geographical locations and racial patterns (52)(53)(54). The pathophysiological processes that linked hot foods to the risk of EC were described in numerous experimental and observational studies (55). Consumption of foods at an elevated temperature has been linked to the formation of endogenous reactive nitrogen species, nitrosamines, TP53 gene mutations, the diminished barrier function of the esophageal epithelium to carcinogenic materials (7,(56)(57)(58).
In this study, the risk of developing EC was further increased with the volume of coffee consumed. As a result, drinkers of a large volume of coffee at a time were 6.3 times more likely to develop EC compared to small volume coffee drinkers. The ndings regarding coffee drinking volume and risk of EC were inconsistent. In a systematic review, only three of twenty studies showed positive associations (7) while a meta-analysis study among East Asian populations did not nd a relationship between coffee drinking volume and risk of esophageal cancer (51). Whereas, an experimental study con rmed a raised intraesophageal temperature with a volume of coffee consumed than by coffee temperature (57).

Strength and limitation of the study
The strength of this study is that it is the rst case-control study conducted among a study population entirely represented from EC endemic area in Ethiopia. The study revealed multiple dietary practices associated with EC that may contest the overriding hypothesis that linked porridge consumption as the only dietary risk factors associated with an increased risk of esophageal in the Arsi Zone. The unavoidable limitations of this study are recall and information biases because of collecting data based on past experiences and participants' self-reported practices.

Conclusion
Multiple dietary practices are associated with an increased risk of esophageal cancer. Food cooking place, teeth loss, porridge consumption temperature, coffee drinking temperature, the volume of coffee drank, speed during porridge consumption were independent determinants of risk of esophageal cancer.