Background
Leukemia is the most frequent type of childhood cancer [
1]. Acute lymphoblastic leukemia (ALL), representing 80% of diagnoses, is the main subtype of childhood leukemia followed by acute myeloid leukemia (AML) and the chronic subtypes of leukemia [
1]. Lower incidence of childhood leukemia has been documented among Chinese, Japanese, and Koreans than among Caucasians [
2]. Several risk factors are thought to play a role in the hematopoietic carcinogenesis. These factors include environmental factors (e.g., benzene, ionizing radiation, nonionizing electromagnetic fields, pesticides, occupations, and parental occupations), medical events (e.g., radiation therapy, and chemotherapy agents), and familial and genetic factors. However, not all leukemia cases are explainable by these factors [
3‐
5].
Nutrition has been previously implicated in playing a complex role in cancer etiology. Carcinogen exposure may occur through preservation methods and high temperature cooking [
6]. Chinese-style salted fish is characterized as a Group 1 carcinogen by the International Agency for Research on Cancer [
7]. Consumption of cured/smoked meat and fish, which contains
N-nitroso precursors, leads to the formation of carcinogenic
N-nitroso compounds in the acidic stomach [
8]. Antioxidants, such as vitamin C, E, flavones and flavanones in fresh fruits, vegetables, green tea, and soybeans were found to block the nitrosation reaction [
9‐
11]. Epidemiologic studies have suggested that increased consumption of cured meat is associated with a higher risk of colorectal cancer and stomach cancer [
12]; while consumption of fresh fruits and vegetables is associated with a decreased risk of breast, colon, lung, pancreas, bladder, larynx, stomach, esophageal, and oral cancers [
13‐
16].
Compared to the solid tumors, few studies have reported the effects of diet in relation to childhood leukemia and have inconsistent results. In addition, all of them were conducted in Western countries [
17‐
19]. To assess children's diet and the risk of childhood leukemia, we performed an analysis in a population based case-control study in a Han Chinese population.
Methods
Study population
This population-based case-control study was conducted in metropolitan Kaohsiung, southern Taiwan. Details of the study area as well as the cases and controls recruitment criterion were described previously [
20,
21]. Briefly, cases and controls were selected from the non-agricultural areas in Kaohsiung to avoid potential confounding effects of pesticides. Incident leukemia cases were identified on the basis of International Classification of Disease 9
th Revision (ICD-9) criteria, codes 204–208. Cases were histologically confirmed to have an incident, primary diagnosis of leukemia (International Classification of Disease 9
th Revision criteria, codes 204–208), less than 20 years old, diagnosed between November 1997 and December 2005 and residents of the study area at the time of enrollment. Patients with secondary or recurrent tumors were excluded. Cases recruitment was performed through two systems including the rapid case ascertainment system from the large referral hospitals in the study area and computer files abstracted from records of the mandatory national health insurance system operated by the Department of Health of Kaohsiung, Taiwan. The former was set up by the Kaohsiung Medical University Hospital to obtain information regarding newly diagnosed cases from the large referral hospitals in the study area (Kaohsiung Medical University Hospital, Kaohsiung Chang-Gung Memorial Hospital, and Kaohsiung Veterans General Hospital). All citizens who have established a registered domicile for at least 4 months in Taiwan are covered by a National Medical Insurance System and have access to these referral hospitals. With these two case ascertainment combined, all of the cases occurring in metropolitan Kaohsiung have been identified. All case diagnoses were confirmed by an experienced pathologist in the Kaohsiung Medical University Hospital.
After each incident case was identified, matched controls were randomly selected using a personal identification number assigned by the Household Registration Office from the population registry data of the chosen study area. This number assignment is independent of current residence and would not bias the control selection by residence. Individuals with any known malignancy were excluded from the controls of the study. Cases and controls were matched on age (± 1 year) and gender using a 1:3 matching ratio.
Totally 190 eligible cases and 842 eligible controls were identified. Among them, 179 cases (94%) and 475 controls (56%) agreed to participants. The reasons for nonparticipation for the eligible cases included refusal (2 percent); could not be contacted (1 percent); missing address information (2 percent); and parents were divorced or widowed (1 percent). Reasons for nonparticipation for the eligible controls included refusal (22 percent); could not be contacted (14 percent); missing address information (6 percent); and parents were divorced or widowed (2 percent). Therefore, not all of the cases had 3 matched controls. The differences between participants and non-participants by residence are not statistically significant [
21].
The study protocol was approved by the Institutional Review Boards of the Harvard School of Public Health and Kaohsiung Medical University. All study subjects ≥ 18 years of age consented; or their parents (if < 18 years of age) assented to participate in the study.
Questionnaires
A trained interviewer conducted an in-person interview immediately after participants were recruited. Follow-up phone interviews were administered occasionally if required. Either the subject's biologic mother or the subject himself/herself completed the questionnaire. A modified questionnaire originally developed by Children's Cancer Group and the Pediatric Oncology Group Study [
22] was used to obtain information on the subjects' socio-demographic characteristics, medical history, residential history up to 2 years prior to birth, occupational history (if the subject was ≥ 16 years of age), cigarette smoking, alcohol consumption, diet, and exposure to various hazardous agents. The questionnaire was translated from English into traditional Chinese, and back-translated independently to assure accuracy. Some modifications were done for better cultural adaptation. For example, Chinese style sausage, salted fish, and dried salted duck were added in the food group of cured/smoked meat or fish. A portion of the habitual diet questionnaire concentrated on consumption of the following foods/food groups: fruits, vegetables, bean-curd foods (to-fu, dried to-fu, to-fu pudding, etc), tea (black tea, green tea, and oolong tea), alcohol-containing beverages (beer, spirits, and wine), cured/smoked meat or fish (Chinese style sausage, salted fish, preserved meat, bacon, ham, hot dog, dried salted duck), and pickled vegetables (pickled Chinese cabbage, pickled potherb mustard, salted cabbage, etc). Interviewers asked the cases to estimate their usual intake of these food groups prior to diagnosis. Controls were asked to estimate their usual intake before the recruitment. For each food/food group, participants reported their frequency of intake. The usual dietary intakes listed in the questionnaire were in frequency categories (as the original food groups in Table
1). The frequency categories for the responses were never, > = 1/week for more than 6 months (ever) for tea and alcohol-containing beverages; < 1/week, 1/week, 2–3/week, 4–6/week, few times/day for fruits and vegetables; never, few times/year, < 4/month, 1–2/week, 3–4/week, > = 5/week for bean-curd foods, cured/smoked meat and fish, and pickled vegetables. Portion size was not measured.
Table 1
Definition of the regrouped food frequency categories for data analysis based on original categories obtained from the child's diet questionnaire
Cured meat/fish | Never; few times/year;< 4/month | 1–2/week; 3–4/week; > = 5/week |
Pickled vegetables | Never; few times/year | < 4/month; 1–2/week; 3–4/week; > = 5/week |
Bean-curd foods | Never; few times/year | < 4/month; 1–2/week; 3–4/week; > = 5/week |
Vegetables | < 1/week; 1/week; 2–3/week; 5–6/week | few times/day |
Fruits | < 1/week; 1/week; 2–3/week; 5–6/week | few times/day |
A similar questionnaire for the mother was used with additional questions added to obtain information on maternal reproductive history, alcohol-containing beverages and supplement use (vitamins and iron supplements) and/or medication usage 21 months before delivery to the study subject's date of birth or the date breast-feeding stopped.
Statistical analyses
In the analyses, the frequency categories of food consumption on the questionnaire were collapsed into simpler categories for sufficient numbers of cases and controls for each category in each food/food group. The cut-points for the regrouped categories were based on the median consumption frequencies in the controls. New categories were used to test if the childhood leukemia risk is associated with frequent cured/smoked meat or fish and pickled vegetables consumption (above median) and the rare and occasional fruits, vegetables, and bean-curd foods consumption (below median). The regrouped food frequency categories are summarized in table
1.
The data was analyzed using the SAS
® software version 9.1 (SAS Institute, Cary, North Carolina). Among the identified 179 acute leukemia cases (136 ALL and 43 AML) and 475 controls, we restricted our analysis to Han Chinese participants to reduce confounding by ethnicity (158 cases and 404 controls). In addition, since it is less likely for infants to have habitual cured or pickled foods intake and the potential different risk factors for infants who develop leukemia [
23], cases and their respective controls under 2 years of age were excluded. 145 acute leukemia cases (112 ALL and 33 AML) and 370 matched controls are in the analyses. Selected principal characteristics were evaluated by χ
2, Fisher exact, and
t tests, as appropriate.
Analyses of the frequency of consumption of each food/food group and risk of leukemia were performed using logistic regression models with adjustments made for age, sex, maternal age, birth weight, breastfeeding, parental education levels, parental and subjects' smoking history, maternal vitamins and iron supplements intake status, and all the other food items. As no confounding was seen by these variables alone or together on the associations of interest, the results presented are only adjusted for the matching factors including age and sex. A Wald's test based on the original ordinal dietary variables obtained from the child's diet questionnaire was used to test for trend. Stratified analyses were performed to estimate the combined dietary effects.
All reported p values are from two-sided tests. An adjusted p-value less than 0.05 was considered to be statistically significant.
Discussion
In this study, we found that the consumption of cured/smoked meat and fish is associated with an increased risk of acute leukemia; while bean-curd foods and vegetables is associated with a reduced risk of childhood acute leukemia. We found no apparent associations for consumption of pickled vegetables, tea, fruits and risk of acute leukemia. To our knowledge, this is the first study that addressed the association of child's diet and the leukemia risk in a Chinese population.
Our study confirmed previous findings conducted in California, USA by Peters et al [
18]. In their study, cured/smoked meat consumption was associated with a higher risk of developing childhood leukemia (OR = 9.5; 95% CI: 1.6–57.6) for child's consumption of 12 or more hot dogs a month versus none during the reference period. Consumption of cured/smoked meat and fish, which contain
N-nitroso precursors, could lead to the formation of carcinogenic
N-nitroso compound in the acidic stomach [
8]. However, no significant association between cured/smoked meat consumption and leukemia risk was reported in two other studies [
17,
19]. In these two studies, a smaller proportion of children reported frequent cured/smoked meat consumption [
17,
19]. In our study, 24% of the controls and 35% of the acute leukemia cases consumed cured/smoked meat and fish more than once a week. In addition, differences in the food preparation methods among studies and over time may result in varied amounts of contaminants.
The traditional Chinese diet is characterized by a high intake of bean curd foods and vegetables, but a low intake of fried or grilled foods. Our study found that, compared to subjects with rare consumption of bean-curd foods and vegetables, individuals with frequent consumption have a lower risk of acute leukemia. Vegetables provide nitrosation-inhibiting antioxidants such as contain vitamin C and E, carotenoids, tocopherols, and selenium [
25]. The findings of the consumption of vegetables associated with a reduced risk of childhood leukemia are consistent with previous studies [
19,
26,
27]. Epidemiological studies have demonstrated an association between the consumption of soybean and a reduced risk for breast cancer [
28,
29], prostate cancer [
30,
31], stomach cancer [
32], nasopharyngeal carcinoma [
33], cardiovascular disease [
34], and atherosclerosis [
35]. Soy isoflavones and soy protein are mainly found in soybeans and soy foods. Many studies have focused on their antiestrogenic properties in possibly preventing hormonally mediated cancers. Consumption of soy foods may be associated with a reduced risk of non-hormonally dependent cancers as well as hormonally dependent cancers. In addition to the antiestrogenic properties, studies have reported that soy isoflavones are associated with inhibition of cancer initiation by inducing detoxifying enzymes [
36,
37] and its antioxidant [
38] and antitumor activities, including inhibition of angiogenesis [
39], topoisomerase [
40], and tyrosine kinase [
41]. Experimental studies have suggested that consumption of dietary antioxidants can block endogenous nitrosation [
9,
10]. Our study found that, among the frequent cured/smoked meat and fish consumers, the risk of the participants rarely consumed vegetables or bean-curd foods is about 3 times or higher in comparison to the participants frequently consumed vegetables or bean-curd foods (Table
4).
There are several strengths of this present study including: a) the selection of controls from a general population registry; b) an adequate sample size, considering that leukemia is a relatively rare disease and we studied a relatively small geographic area of a single ethnic group; c) comprehensive information was collected to adjust for several potential confounders/effect modifiers reported by previous studies, including maternal age, birth weight, breastfeeding, parental and subjects' smoking history, and parental education levels.
Several potential limitations should be considered when interpreting our study results. Molecular studies indicate that early childhood leukemia can originate prenatally [
42‐
44]. Studies have shown that maternal consumption of vegetables, fruits and vitamin supplement during pregnancy may reduce the risk [
23,
45,
46]. Maternal consumption of vegetables and fruits during pregnancy were not measured in our study. Moreover, lacking portion size and information on food preparation method could result in different levels of contaminants.
In addition, since the primary focus of the hypothesis is the effect of
N-nitrosamines dietary exposure in relation to acute leukemia, only these specific food groups were collected in the questionnaire. Several other uncollected dietary factors may confound or modify the association of leukemia risk. For example, the cured/smoked meat and fish are often cooked or consumed with garlic (Allium sativum), which contains several anti-cancer compounds [
47]. Although misclassification by dietary exposure to garlic is likely to be non-differential and may attenuate the odds ratio of cured/smoked meat and fish, we cannot rule out that our finding may be due to other unknown dietary or lifestyle factors.
Another potential problem is recall bias, where differential misclassification of dietary intake between cases and controls may overestimate the association between diet and cancer [
48]. Our cases and controls were identified between 1997 and 2005. The food supply and dietary patterns may have changed during the 8-year period (1997–2005). However, the controls were matched at age and time of diagnosis with incident cases. Furthermore, the cases and controls were interviewed in the similar time points and conducted immediately after the subjects recruited to reduce potential measurement errors from recent dietary changes after diagnosis of cancer as well as misclassification from recall bias. In addition, there was no public perception at the time of data collection that child's diet is associated with leukemia risk. The effects of changes in the food supply and the transitions in dietary pattern by time should be similar among cases and controls. It is thus unlikely that the association observed was biased by time. Therefore, it is unlikely that the observed association was a result of exposure misclassification. We believe our findings justify further testing of this hypothesis with other studies.
Acknowledgements
This research was supported by National Institutes of Health (ES09723, ES00002).
The authors gratefully acknowledge Chien-Chin Chou for questionnaire data verification, Maureen Convery, and Chu-Ling Yu for technical assistance.
The members of the Kaohsiung Leukemia Research Group were as follows: Kaohsiung Medical University Chung-Ho Memorial Hospital – Tai-Tsung Chang, Sheng-Fung Lin, Shyh-Shin Chiou, Ren-Chin Jang, Hui-Hua Hsiao, Ta-Chih Liu, and Pei-Chin Lin; Kaohsiung Chang Gung Memorial Hospital – Chih-Cheng Hsiao, Jiunn-Ming Sheen, Ching-Yuan Kuo, Ming-Chung Wang, Cheng-Hua Huang, and Chung-Bin Huang; and Kaohsiung Veterans General Hospital – Yuk-Cheung Wong, Hung-Bo Wu, Shyh-Jer Lin, Yu-Ming Sun, Kai-Sheng Hsieh, and Yu-Hsiang Chang.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
CYL, MTW, PCP, CKH, and DC developed the design of the study. CYL, YHH, and DC did the data analysis and interpretation of data. CYL drafted the manuscript and all other authors were involved revising the manuscript critically. All authors have given final approval of the version to be published.