Risk factors for esophageal cancer in a high-incidence area of Malawi

To explore associations of nutritional, infectious, and lifestyle factors with esophageal cancer (EC) occurrence in a high-risk area of Malawi. This case–control study was performed with 227 patients undergoing endoscopy for dysphagia or other upper gastrointestinal complaints. Data on clinicopathological characteristics and risk factors were collected using a questionnaire developed for this study specifically. Ninety-eight blood samples were collected and the prevalence of antibodies against human immunodeficiency virus, herpes simplex virus, cytomegalovirus, Epstein–Barr virus, varicella-zoster virus, and Helicobacter pylori were determined serologically. The tumor and control groups comprised 157 (69.2%) and 70 (31.8%) patients, respectively. Patients with tumors were significantly older than controls (55.5 vs. 43.5 years, p < 0.001). The male/female ratio did not differ between groups (59% and 54% male, respectively; p = 0.469). EC was associated with smoking (p < 0.001), and alcohol consumption (p = 0.020), but 43% of patients with tumors did not smoke or drink. EC was associated with the consumption of hot food and tea (p = 0.003) and smoked fish (p = 0.011). EC was not associated with any serologically investigated infectious agents. In an age adjusted binary logistic regression analysis of all nutritive factors, only locally made alcohol was significant [odds ratio (OR), 9.252; 95% confidence interval (CI), 1.455—58.822; p = 0.018]. Apart from alcohol consumption and smoking, the consumption of hot food or tea and smoked fish are associated with EC. Locally distilled alcohol consumption increases the EC risk in Malawi.


Introduction
Esophageal cancer (EC) is the eighth most common cancer diagnosis, with an incidence of about 604,000 cases worldwide. It shows great geographical variation, with a high frequency of esophageal squamous cell carcinoma (ESCC) occurrence in areas deemed to be at high risk, including the "EC belt" in south-central Asia and in parts of southern and eastern Africa [1]. Malawi has one of the highest incidence rates of EC worldwide, with 24.2 cases per 100,000 inhabitants, in contrast to 6.6 and 3.2 cases per 100,000 inhabitants in the United States and United Kingdom, respectively [2].
Causative factors for this high incidence have not been identified conclusively. Tobacco use and alcohol consumption, the main risk factors in Western countries [3], have been associated with EC in African countries [4]. Thus, we investigated the association of alcohol and tobacco use with EC in Malawi, although the consumption in this country is similar to or lesser than that in other African countries with much lower EC incidence rates [5,6]. We also investigated nutritional risk factors, which are common in the population of Malawi, where hot food and tea are commonly taken straight from the fire. In addition, in contrast to other African countries with lower incidence rates of EC, maize consumption is high in Malawi [7].
Other diseases with high incidence rates in Africa may be risk factors for EC in Malawi. Chronic infections are known to cause more than 15% of malignant neoplasms worldwide [8][9][10][11]. Seventy percent of patients with human immunodeficiency virus (HIV) infection live in sub-Saharan Africa. HIV is known to be associated with the occurrence of lymphomas and cervical and anal carcinomas [12]. High prevalence rates in Africa have also been reported for cytomegalovirus (CMV), herpes simplex virus (HSV), and Helicobacter pylori infection [13][14][15]. Human papillomavirus (HPV) has been found in only a subset of patients (15% of those with ESCC) in Malawi [16].
Thus, data on risk factors for EC in sub-Saharan Africa and Malawi are limited. The aim of this case-control study was to investigate the associations of EC occurrence in Malawi with nutritional, infectious, and lifestyle factors. As no curative treatment for EC is currently available in Malawi and most low-and middle-income countries worldwide, information about risk factors is needed to develop prevention strategies.

Materials and methods
This study was performed with data from patients who underwent upper-gastrointestinal endoscopy at Zomba Central Hospital and Queen Elizabeth Hospital, Blantyre, Malawi, in 2010 and 2014-2016. 136 patients (59.9%) were recruited in Zomba (117 tumor patients and 19 controls), whereas 91 patients (40.1%) underwent endoscopy at Queens in Blantyre (40 tumor patients and 51 controls). All patients provided written consent (by signature or fingerprint) prior to endoscopy. The study was approved by the Research and Ethics Committee of the College of Medicine Blantyre (no. P.04/10/930). All patients were Malawians, mainly originating from the Southern and Central region of the country and belonging to the Chewa, Yao or Ngoni tribes. Data on patients' clinicopathological characteristics and risk factors were collected by a questionnaire developed by our group for this study specifically. Questionnaire categories were general characteristics (e.g., age, sex, weight, height), symptoms (e.g., dysphagia), history of infectious diseases (e.g., tuberculosis), nutritive factors (e.g., consumption of fish, smoked fish, hot food and tea, maize porridge), and smoking and alcohol use. Daily maize consumption was measured as the consumption of nsima maize porridge (in 250-g increments), and the consumption of vegetables and fish was measured as the number of times consumed per week. Patients with tumors were asked to indicate the amounts of maize porridge, vegetables, and fish consumed before their symptoms had started.
Endoscopy was performed on all patients with dysphagia, and on patients with other symptoms. EC was diagnosed macroscopically during endoscopy and confirmed histologically in a subset of 40 patients. Patients with EC were assigned to the tumor group, and those without tumor were assigned to the control group.

Serological analysis
For serological analysis, blood samples were collected from 98 patients before endoscopy. They were analyzed at the Department of Medical Microbiology, University Hospital of Jena, Germany.
Serological status for HIV, HSV, CMV, Epstein-Barr virus (EBV), and varicella-zoster virus (VZV) was determined using commercially available enzyme-linked immunosorbent assays [ELISAs; Enzygnost® HIV Integral II, anti-HSV/immunoglobulin (Ig)M and anti-HSV/IgG, anti-CMV/IgM and anti-CMV/IgG, anti-EBV/IgM II and anti-EBV/IgG, anti-VZV/IgM and anti-VZV/IgG; Siemens, Erlangen, Germany] in a BEPIII system (Siemens) according to the manufacturer´s protocol. Anti-H. pylori IgG ELI-SAs were also performed. In brief, sera were diluted 1:201 with dilution buffer. Then, 100 µl diluted serum and 100 µl undiluted calibrator were placed in a plate and incubated for 30 min at room temperature on a microplate shaker. The microtiter plate was washed three times with wash buffer, 100 µl IgG conjugate was added, and the plate was covered and incubated at room temperature for 30 min on a microplate shaker, followed by three washes. Then, 100 µl Tetramethylbenzidine was added to each well of the microtiter plate and the plate was incubated 10 min at room temperature on a microplate shaker. Finally, 100 µl stop solution was added and extinction was measured by photometer.

Statistical analysis
To assess relationships between the epidemiological and clinical data, Fisher's exact test was used for nominal variables and the Mann-Whitney U test was used for continuous variables. Medians and interquartile ranges (IQRs) were calculated to describe the variance of the metric variables. Odds ratios and 95% confidence intervals (CIs) were determined for each parameter. A multivariate binary logistic regression analysis and calculation of odds ratios were performed with all risk factors, including smoking, alcohol consumption, and nutritive factors. To avoid bias, the alcohol (former or current) and self-made 1 3 cigarette consumption variables were excluded, as alcohol consumption was covered by beer/spirits/locally made alcohol variable and cigarette smoking was covered by the smoking (former or current) variable. P values < 0.05 were considered to be significant. The statistical analysis was performed with IBM SPSS Statistics (ver. 22; IBM Corporation, Armonk, NY, USA).

General characteristics
In total, 227 patients (157 in the tumor group, 70 in the control group) were included in this study. Some patients of both groups could not provide responses for all items. The study results are summarized in Tables 1 and 2. Ninety-three (59.2%) patients in the tumor group were male; sex was not associated with tumor presence

Infectious diseases
Compared with the control group, the presence of EC was not associated with a history of oral thrush (10.

Smoking and alcohol consumption
Compared with no exposure, smoking exposure (former and current smoking) was associated significantly with the presence of EC [50.7% (tumor group) vs. 24.6% (control group), p < 0.001]. Significantly more patients with tumors than controls smoked self-made cigarettes (28.8% vs. 12.3%, p = 0.022). Among smokers, patients in the tumor group had smoked significantly longer than those in the control group

Binary logistic regression results
The multivariate binary logistic regression model included age, smoking, and the consumption of beer/spirits/locally made alcohol, hot tea and food, spicy food, smoked fish, fish, vegetables, and maize porridge. Table 2 locally made alcohol was the only factor associated positively with EC (odds ratio, 9.252; (95% CI, 1.455 -58.822; p = 0.018). The weekly consumption of large amounts of fish had a protective effect (odds ratio, 0.702; (95% CI, 0.529 -0.932; p = 0.014).

Discussion
Malawi has been designated as a country whose population is at high risk of EC development [1,2]. The reason for this elevated risk has not been determined, and data on risk factors in sub-Saharan Africa are limited. As EC treatments will not be available soon in low-and middle-income countries, the exploration of risk factors is needed to identify subpopulations at high risk of EC development, and to develop prevention strategies to reduce morbidity and mortality from this disease in the long term. In this study, we investigated associations of EC occurrence in Malawi with nutritional, lifestyle, and infectious factors. We collected data on risk factors in 227 patients, of whom 157 presented with endoscopically proven EC. All biopsies that were taken showed histopathological characteristics of ESCC; no adenocarcinoma was found. We observed 100% accordance between the macroscopic appearance and histological diagnosis. These findings are consistent with those of another study conducted in Malawi, in which no esophageal adenocarcinoma was found in 82 EC biopsy samples taken between 2004 and 2008 [17].
In this study, the presence of EC was associated with age, smoking, and the consumption of alcohol, hot food or tea, and smoked fish. In the age adjusted binary logistic regression analysis, the consumption of locally made alcohol remained the only significant factor.
The association of age with the presence of EC is not surprising and is consistent with other data from Africa. ESCC incidence rates increase with age, and approximately 80% of people with EC in eastern Africa are at least 50 years of age [4]. On the other hand, 18.7% of patients in the tumor group in this study were aged ≤ 40 years. Thus, it is unlikely that the high occurrence rates of EC in Malawi can be explained primarily by age.
In Europe and the United States, alcohol and tobacco consumption are the predominant risk factors, with a strong synergistic effect on ESCC risk [18]. Smoking and alcohol consumption were also identified as major risk factors for EC in a high-incidence area in South Africa [19]. In this study, alcohol consumption and smoking were associated with EC, but 43.5% of patients with tumors consumed neither tobacco nor alcohol. Similarly, the consumption of alcohol and tobacco is less common in high-incidence areas in Asia than in Western countries, and seems to play a lesser role in the etiology of ESCC [18]. Tran et al. [20] reported relative risks of 1.34 for cigarette exposure (ever-smoking) and 0.92 for alcohol consumption in a high-risk area in China. In a high-risk area in Iran, the history of alcohol consumption was negligible (1%) and the use of cigarettes was limited (27%) among patients with ESCC [21]. Thus, other environmental factors also should be considered in high-risk areas.
Associations of ESCC with other risk factors have been discussed in the literature. Environmental factors such as the consumption of hot food and beverages and poor diet have been associated with elevated risks of EC and ESCC in highrisk areas [18]. The association of EC occurrence with the consumption of hot tea and food in this study is similar to the increased risk of EC associated with this factor in high-risk areas in Iran [22], South America [23], and China [24]. Hot beverages can harm the esophageal epithelium via thermal effects and chemical constituents [25]. Local hyperthermia impairs the barrier function of the esophageal epithelium [26], and mild and moderate esophagitis has been associated with the consumption of boiling-hot beverages [27]. In this study, we found no association of EC with the consumption of fish or vegetables. Studies conducted in high-risk areas in Iran and China revealed an inverse association of ESCC with the consumption of fresh vegetables [28][29][30] and fish [31]. However, our results are consistent with those of a case-control study conducted in Zambia, which revealed no significant difference in the consumption of fruits, vegetables, and fish between EC cases and controls [32].
We investigated additional factors that are common in southern Malawi, such as the consumption of smoked fish and maize porridge; the presence of EC was associated with smoked fish consumption. Carcinogenic substances such as polycyclic aromatic hydrocarbons (PAHs) have been found in smoked fish, especially fish smoked heavily in traditional kilns [33]. High exposure to PAHs in the general populations of high-risk areas in Iran [34,35] and China [36] have been reported. PAH-DNA adducts have been detected in esophageal biopsy samples [37], and greater PAH exposure of the non-tumoral esophageal epithelium was detected in patients with ESCC than in controls in a high-risk area in Iran [38]. Foods containing protease inhibitors, such as beans, which are frequently consumed in southern Africa, have been associated with EC [39]. Furthermore, EC was associated with the consumption of purchased maize in South Africa [40], possibly due to nutritional deficiencies [2], fungal contamination [41], or increased intragastric production of prostaglandin E 2 , which leads to increased cell proliferation [42,43]. However, we found no association between the consumption of maize porridge and EC in this study.
HIV, CMV, HSV, and H. pylori infection are prevalent in Africa [12][13][14][15]. Our serological investigation revealed no association of EC presence with positive HIV serology, consistent with the finding of another study from Malawi [44], or with positive serology for CMV, EBV, HSV, or VZV. CMV is known to have onco-modulatory properties [45], and EBV has been detected, for example, in patients with nasopharyngeal carcinoma [46]. Consistent with our findings, no association between CMV and EC was found in studies of tissue samples from China [47,48]. Various prevalence rates of EBV in tissue samples have been reported; EBV DNA was found only in subsets of ESCC tissue samples [47,[49][50][51]. Similarly, various prevalence rates of HSV in EC tissue samples in high-risk areas of China have been reported [47,48,52]. The role of H. pylori in gastric malignancies has become evident [46], but its role in ESCC remains unclear [53,54]. We found no association between antibodies against H. pylori and the presence of EC in this study, in contrast to the decreased risk of ESCC in individuals with H. pylori infection observed in other case-control studies [55,56].
The lack of association between tumor development and the infectious agents investigated in this study may be attributable to high seroprevalence among control patients and low prevalence among patients with tumors, or to the small number of blood samples investigated. The latter can be considered a study limitation. Another technical limitation is that we investigated associations of infectious agents with EC only serologically, and not using tumor tissues. A further limitation of our study is that we used a self-generated questionnaire and not a validated questionnaire, as no such instrument was available for the Malawian setting in 2010, when the study began.
In our sample, many control patients reported the same nutritional habits as did patients with tumors. High PAH levels in urine samples from healthy subjects in high-risk areas in China, Iran, and Brazil have been reported [18]. Thus, genetic susceptibility also needs to be considered in future studies. The risk of ESCC was elevated among patients with family histories of upper-gastrointestinal cancer and EC in a high-risk area of China [56]. Furthermore, EC is more common in certain ethnic groups in high-risk areas, such as Kazakhs in China [57,58], and the genetic background of migrants in China appears to influence the incidence of ESCC [59]. Variants in several chromosome regions were associated with an increased risk of EC in China [60]. Genetic diversity appears to be greater in African populations than in populations living on other continents [61]. Data on the genetic susceptibility to EC in Africa are limited. In South Africa, certain variants of alcohol dehydrogenase were found to be associated with an increased risk of EC development [62]. Bye et al. [63] suggested that the genetic contribution to ESCC differs in South African and Chinese populations due to differences in genetic architecture. A recently published transcriptomic analysis conducted with patients with ESCC in Malawi revealed genetic aberrations similar to those found in Asian and North American cohorts, such as TP53 mutations [64]. Mutational signature analysis showed common signatures associated with aging and cytidine deaminase activity; a third signature was of unknown origin, and signatures of inhaled tobacco use, aflatoxin, and mismatch repair were notably absent. Thus, additional factors that are characteristic in Malawi should be investigated and the genetic susceptibility to EC should be considered in further studies.

Conclusion
The main finding of this study is that the consumption of locally distilled alcohol is associated with EC in Malawi. This finding suggests that the same pathophysiology underlies ESCC development in Malawi and Western countries. Considering that alcohol consumption is negligible among patients with ESCC in many other highly endemic areas, such as Iran, other environmental factors (e.g., consumption of hot food/tea) might have amplifying effects. The consumption of hot food/tea, beer and local alcohol had high odds ratios in our binary regression analysis. Maize is the main component of locally made alcohol, beer and nsima, the staple food, in Malawi. Further studies with large numbers of patients and sound designs are needed to clarify these potential associations.
The results of this study should be viewed as only one component of a systematic understanding of risk factors contributing to the high incidence of EC in southern Malawi. Investigations of fungal co-infection and toxins, maize contamination, and the association with HPV are in progress or have been published [16]. The performance of epidemiological studies in low-resource settings is always challenging. In particular, we included a small number of control patients in this study, which is a major limitation. Thus, our results should be interpreted with caution, and more standardized investigations with larger patient samples are needed. However, our results suggest that the production and consumption of locally distilled alcohol should be abolished or reduced as much as possible within a short period of time. More attention should be devoted to the development of primary prevention strategies for non-communicable diseases in sub-Saharan Africa. This effort may entail the addressing of other challenges, such as the shortage of healthcare workers; lack of infrastructure; insufficient access to early diagnosis, treatment, and palliative care; and poor public awareness of cancer [65].