INTRODUCTIONOver the past decade, as the leading cause of death inKorea, cancer burden has still been the country’s major public health concern.1 The age-standardized cancer incidence rateduring 2005 and 2014 has been increased from 240.8 to 270.7 per 100,000.2, 3 Although theattributable cause of cancer in Korea has been conducted by estimatingpopulation attributable fraction for several factors, there is the lack ofresult in term of dietary factor.4 Dietary intake and cancer risk in Korean populationhave been investigated in many studies.
Woo et al. conducted the meta-analysisof 10 individual studies to evaluate the association between dietary intake andcancer risk in Korean population.5 Overall, soy foods, fruits, and vegetableswere associated with a decreased risk of cancer. In contrast, high salt foodsincreased the risk of gastric cancer and meat intake increased the risk ofcolorectal cancer.
The odds ratios (ORs) from individual studies werecalculated based on frequency of food intake only and there is an inconsistenceof categories among individual studies also.5Thus, identifying the nutritional behavior in cancersurvivors based on details of daily calories intake of different types of foodgroup plays an important role in evaluating dietary lifestyle and cancer risk.Therefore, we conducted this study to evaluate the nutrition intake behavior of cancer survivors based onmain sources of energy intake from different food groups in comparison withnon-cancer individuals. Also, we investigated the nutrition intake behaviorby five main types of cancer to identify whether there was the effect of cancertype on nutrition intake behavior. MATERIALSAND METHODS1.
Study populationKorean Genome Epidemiology Study (KoGES)is an ongoing consortium project from 2001, which consists of six population-basedand gene-environment prospective cohort studies.6 This study used the data extractedfrom the Health Examinee (HEXA) cohort, which is a part of KoGES project.6 In the HEXA cohort study, theparticipants which aged 40-69 were recruited from 38 centers.7 The staffs were trained andcollected information on socio-demographic characteristics, history of disease,medication, and family, lifestyle behavior of alcohol, tobacco, and diet aswell as biological specimens by using a standardized questionnaire for Korean.7 The KoGES and HEXA cohort are alsodescribed in details at the Korea National Institute of Health website.8Participants reporting the history ofcancer were diagnosed with any type of cancer by using the guideline ofDeSantis et al, which identified as any person who had been diagnosed with anytype of cancer, either who are under treatment or who had completely recovered.9 As a result, a total of … cancer survivors and … non-cancer individualswere identified.
After individual matching with 1:1 ratio by age (±2 years),gender (male and female), and enrollment center location (38 centers), 5,269 cancer survivorsand 5,219 non-cancer subjects were included in the final analysis (Figure 1).2. Nutrition behaviorThe nutrition behavior was assessed by using avalidated and reproductive food frequency questionnaire.10 Based on age, gender,height, weight, type of food intake, frequencies of servings, and portionsizes, the daily energy intake of 671 food combinations according to 17 foodgroups was measured by using nutrient software CAN-Pro 4.
0 (The KoreanNutrition Society). Nutrition behavior was determined by calculating theproportion of calories intake of each food group: Cereals and grain; potatoesand starches; sugars and sweets; legume and their products; seeds and nuts;vegetables; mushrooms; fruits; meat, poultry, and their products; eggs;fishes and shellfishes; seaweeds; milks and dairy products; oils and fats; beverage;seasonings; and others.3.
Statistical analysisAll statistical analyses were performed by usingpackage software SAS version 9.4 (SAS Institute, Cary, NC, USA), and R version3.2.2. From the total matchedindividuals, …subjects did not complete the dietary questionnaires, leaving missing valuesfor nutrition intake (…cancer survivors and …non-cancer subjects). After excluding these participants, nutrition intakebehavior was compared between 5,269 cancer survivors and 5,219 non-cancer subjectsby using the proportion of total energy intake for 17 different food groups.
To investigate the association between severalbaseline characteristics with risk of cancer, we performed Chi-square test and multipleconditional logistic regression, which adjusted for marital status, educationlevel, income level, employment status, smoking, drinking, physical activity,and body mass index. The mean of proportion of energy intake between cancer andnon-cancer group was compared by using two independent samples test (t-test).Also, we conducted subgroup analysis of dietary behavior for single cancer typewhere 300 or more cases were available by using analysis of variance (ANOVA)with a post hoc comparison (Tukey’s test). The Bonferroni adjustment was appliedfor multiple comparisons with p-value is set to 0.003. As the result, gastriccancer (N=731), colon cancer (N=378), breast cancer (N=875), cervical cancer(N=609), and thyroid cancer (N=888) were included in the analysis. Additionally,to evaluate how much difference in the proportion of nutrition intake between eachcancer type and non-cancer group according to 17 types of food group,regression coefficient and 95% confidence intervals (CIs) were calculated byusing multiple linear regression adjusted for baseline characteristicvariables. RESULTS1.
Baseline characteristics of study populationTable 1 shows the sociodemographic characteristics of5,269 cancer survivors and 5,219 non-cancer subjects from the HEXA cohort.Because of the 1:1 matched design, age and gender were equally distributed incancer and non-cancer group. In the total study population, employment status,smoking, drinking, and physical activity were significantly associated withrisk of cancer (p<0.001).
The significant difference was further analyzed bylogistic regression. While there was an increased risk of cancer withunemployment status (OR=1.37, 95% CI=1.23-1.54), former smoking (OR=1.43, 95%CI=1.16-1.77), former drinking (OR=2.
94, 95% CI=2.27-3.83), and physicalactivity either <150 minutes/week (OR=1.30, 95% CI=1.
04-1.63) or ?150minutes/week (OR=1.26, 95% CI=1.15-1.42), current drinking was observed todecrease risk of cancer (OR=0.60, 95% CI=0.53-0.
67).2. Proportion of calories intake in cancer survivors and non-cancerindividualsThe mean of proportion of calories intake in cancersurvivors and non-cancer individuals was compared in Table 2. Overall, cancersurvivors spent their diet on potatoes and starches; legume and their products;seeds and nuts; vegetables; mushrooms; fruits; and seasonings more than non-cancer individuals (p<0.
05). In contrast, cereals and grain;sugars and sweets; meat,poultry, and their products; and oils and fats accounted the lowerproportion in cancer survivors’ diet than in those of non-cancer individuals(p<0.05).In the subgroup analysis by type ofcancer, most of the results appeared to be similar with overall cancer.
Forexamples, seeds and nuts tended to be used more in gastric cancer, breastcancer, and thyroid cancer subjects than non-cancer individual (p<0.003).Seasonings were observed to much more intake in gastric cancer, colon cancer,and breast cancer patients (p<0.003). Also, there was a significant lower consumeof meat, poultry, and their products in all fivetypes of cancer (p<0.003). 3.
Multiple linear regression for calories intake proportion among cancersurvivors and non-cancer individualsThe differences of calories intakeproportion between cancer survivors and non-cancer individuals were provided inTable 3. Compared with non-cancer individuals, the proportion of potatoes and starches(0.26%); legume and their products (0.29%); seeds and nuts (0.24%); vegetables(0.17%); mushrooms (0.01%); fruits (1.
26%); and seasonings (0.20%) intakes weresignificant higher in cancer survivors. However, the significant lower intakesof sugars and sweets(0.10%); meat, poultry, and their products (1.55%);and oils and fats (0.17%) were obtained in cancer survivors after adjusting forbaseline characteristic variables (p<0.05).
When we conducted the subgroup analysisfor five main types of cancer individuals, these intake patterns were more prominently observed in breastcancer and gastric cancer individuals compared to healthy people (p<0.05).DISCUSSIONThis study aimed to investigate and compare thenutrition behavior between cancer survivors and non-cancer individuals inKorean population. The results from both ANOVA test and multiple linearregression model suggested the more prefer intake of potatoes and starches;legume and their products; seeds and nuts; vegetables; mushrooms; fruits; and seasonings and the less preferintake of sugars and sweets;meat, poultry, and their products; and oils and fats in cancersurvivors. These significantly different terms were most observed in breastcancer and gastric cancer subjects.Dietary factor in term of food item and cancer riskhas been mentioned in the report of World Cancer Research Fund (WCRF) andAmerican Institute for Cancer Research (AICR). In general, fruits andvegetables have the protective effect on different types of cancer, and saltyfood and red meat are associated with an increased risk of cancer.
11 The result from the current study suggeststhat the participants follow the dietary guideline recommended by WCRF andAICR. To the best of the authors’ knowledge, this is one ofthe few studies which investigated the difference in nutrition behavior betweencancer survivors and non-cancer individuals, especially in Korean population. Basedon a large, prospective cohort study in Korea, survivors of five major cancersincluding gastric, colon, breast, cervical, and thyroid cancer1 were compared with non-cancer subjects.Additionally, the daily calories intake was calculated by a specific nutritionsoftware and update version CAN-Pro 4.
0, which is developed for Korean subjectsbased on food culture of Korea.Despite many strengths, the current study has somelimitations. First, the identification of cancer and non-cancer survivors wasbased on self-reported questionnaire, which might lead to misclassification.However, previous study reported the high validation for self-report of cancerhistory,12 thus the risk ofmisclassification bias could be minimal. Secondly, although age and gender wereused for matching, we also include these variables as well as other baselinesocio-demographic variables as the covariates in the regression model, whichmight cause over-adjustment bias or unnecessary adjustment.13 Also, because socio-demographiccharacteristics were considered to be confounders for the association betweendiet and cancer risk based on literature review […], thus we adjusted for allthese variables in the regression model. Thirdly, this is a healthexamination-based study, which has the lower power of generalization thanpopulation-based study. Patients enrolled in this study might be healthier thangeneral cancer patients and non-cancer individuals tended to concern more abouttheir dietary lifestyle.
Fourthly, the large number of participants providedthe high power of determining the statistical difference between groups.Therefore, although the significant difference of calories intake proportionwas observed in many food groups, the absolute difference was still limited,only over 1% was obtained in term of fruits and meat, poultry and their productsintake. CONCLUSIONThe dietary behavior measured by main sources of energy intake in overall cancer survivors as well as main types of cancer is different from healthyindividuals in terms of several food groups. The guideline of balanced nutritious behaviorshould be established among cancer survivors.