PDF
ePub
Citation
Print
INTRODUCTION
Cancer is a non-communicable disease and is the leading cause of death globally [1]. In the United States, it is the second leading cause of death, following cardiovascular diseases. In Korea and Japan, it is the first leading cause of death [234]. Causes of cancer vary widely and many remain unknown; however, the most significant known causes are smoking and diet [5]. Doll and Peto previously estimated that each of these 2 causes accounted for about one-third of cancer causes [6]. In a recent large prospective study, cancer-specific mortality was found to be reduced to 61% in a population that adhered to the cancer prevention recommendations by the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) [7], aligning with the estimates of the research conducted by Doll and Peto.
The method of controlling the risk factor for smoking is simply cessation. However, identifying the foods beneficial for cancer prevention and those that are carcinogenic is more complex. A variety of studies on the relationship between cancer and food, from laboratory to large-scale population studies, have been published [891011]. However, the results of these studies were diverse, and the studies differed from each other in terms of design; moreover, large differences were found according to region and culture. Furthermore, commercial information with inadequate scientific evidence has been widespread and has led to confusion in the general population.
Food is consumed daily by everyone in every socioeconomic class. Obtaining evidence-based information about foods that can prevent cancer among people without nutritional and medical expertise is difficult. Therefore, establishing an intuitive anti-cancer or carcinogenic food index is very useful in promoting cancer-related health.
Based on these needs, we developed the anti-cancer food scoring model (ACFS) 1.0 [12]. The ACFS 1.0 model assigned points to 22 food elements using a systematic method based on literature evaluation. The points were added up based on composition ratios, and unhealthy cooking methods (high salt or oil) were taken into account, to arrive at any of the 6 anti-cancer grades. This model has the advantage of easy-to-understand points and grades, integrating a vast amount of information from the literature.
This study aimed to apply the previously constructed ACFS algorithm to various international recipes, including Mediterranean, vegetarian, Chinese, Western, and Korean diets, and to verify the nutritional validity of the model using a computer-aided nutrition analysis based on the obtained information.
MATERIALS AND METHODS
Calculation model of ACFS
Here, we introduce the calculation model of ACFS briefly, as described in more detail in our previous article [12]. First, we set the 22 ACFS food element codes that are diverse enough to evaluate the majority of Western and Asian recipes. They include whole grain (WG), red meat (RM), green leafy salad (GLS), fish (FISH), garlic (Ga), soy food (SF), cruciferous vegetable (CV), allium vegetable (AV), cheese (Ch), seaweed (SW), fruit (FR), non-starchy vegetable (NSV), white meat (WM), carotene-rich vegetable (CRV), processed meat (PM), selenium-rich food (SRF), milk (Mi), egg (Egg), refined grain (RG), legume (Le), chili (Chili), and potato (Pot).
For each code, studies on 5 major cancers (i.e., breast, colorectal, stomach, lung, and liver) with high global mortality and affected by diets were analyzed. Literature evaluation was based on the 2nd expert report of WCRF/AICR [13]. In the 2nd expert report of WCRF/AICR, the level of evidence was classified into 4 stages (convincing, probable, limited-suggestive, and limited-no conclusion). Level of evidence was determined by the number of cohort or case-control studies, quality of the studies, heterogeneity among the studies, and biological plausibility. In ACFS, code grade A is allotted for convincing or probable, code grade B for limited-suggestive, and code grade C for limited-no conclusion. Then, 10, 5, and 2 points were assigned to code grades A, B and C, respectively. Assigned code points according to each of the 5 cancers, defined as “cancer specific grades”, were summed to yield “ACFS code grades” from A to E.
The meals analyzed were divided into constituent components and matched with the ACFS codes. The “ingredient score” was calculated by multiplying the ratio (using food exchange unit, FEU) of each component in the meal with the code grade point (A, B, C, D, and E correspond to 5,4,3,2,1 points respectively). FEU was used instead of the weight of the ingredient because of its similarity to serving size, which was the more commonly used measurement than weight in reference studies [1250]. FEUs for Korean and Chinese recipes were based on the Korean Diabetes Food Exchange Table. FEUs for Western, vegetarian, and Mediterranean recipes were calculated from the American Dietetic Association Food Exchange Table. The grade level, out of 5 grade levels, was determined from the ingredient scores and the meal was downgraded 1 level if it was cooked using an unhealthy method (> 2 g of salt or > 20 g of oil used in cooking). The final grade was named as the ACFS grade and was interpreted as follows:
Selection of recipes and nutritional analyses
For Chinese recipes, we searched for ‘Representative Chinese cuisine’ in the People's Network (kr.people.com.cn) and obtained the ingredient information from the websites (https://www.allrecipes.com/, http://cookingsimplechinesefoodathome.com/, http://www.people.com.cn/, all accessed between August 10 and 17, 2018) for the top 10 meals. For 20 Mediterranean and 10 vegetarian meals, commonly considered as healthy diets, the recipes were obtained from “The New Mayo Clinic Cookbook” [14]. For 18 Western meals, we selected the commonly consumed routine daily life meals and obtained the recipes from the following cooking portals, referring to author discussions (https://www.mrbreakfast.com/, https://www.tasteofhome.com/, https://www.seriouseats.com/, all accessed between August 10 and 17, 2018). Fifteen Korean home-dining and 15 Korean outdining meals were chosen from the frequently consumed meals in the 7th Korea National Health and Nutrition Examination Survey (KNHANES VII-1), 2016, 24-hour dietary recall data. [15]. For the recipes of the Korean meals, we referred to “The guidelines of using the Korean Diabetes Food Exchange Table” [16], CAN-Pro 5.0 (Computer Aided Nutritional Analysis Program; The Korean Nutrition Society, Seoul, Korea), and the recipe list book published by the Institute of Traditional Korean Food [17].
The nutritional value of the meals was analyzed using CAN-Pro 5.0. Food elements that were too exotic or regional to be analyzed using CAN-Pro were substituted with the most similar elements in the CAN-Pro database (e.g., raspberries were substituted by cranberries, cannellini beans by horse beans, and fennel or finocchio by onion). Carbohydrate, lipid, and protein rates denote the contributing proportion in total calories but not the proportion of component weight.
Ethical approval and informed consent
Ethical approval and consent to participate are not applicable as this study did not involve humans or other living organisms. All investigations conformed to the principles outlined in the Declaration of Helsinki.
Statistical analyses
Univariate analyses were performed using the Jonckheere-Terpstra test to identify trends among ACFS grade subgroups. The variables with P values of < 0.05 in the Jonckheere-Terpstra test were included in the multivariate analyses. Among the variables, possibly duplicated variables (e.g., total fat and animal fat, total fiber and water-soluble fiber, calcium and plant calcium) were filtered to be included in the multivariate analysis using the following criteria prioritized in numerical order: 1) the variable with higher statistical significance; 2) the specific variable rather than the general one. Simple nutritional or proportional values including total calories and carbohydrate, lipid, and protein rates contributing to the total calories of meals were not included in the multivariate analysis. Multivariate analyses were performed via multiple regression analyses, using the backward elimination method. The probabilities of removal and entry were 0.10 and 0.05, respectively. Standardized β coefficients and non-standardized β coefficients were presented to identify the comparative significance of each variable. All statistical tests were performed using SPSS version 23.0 (IBM Corp., Armonk, NY, USA).
RESULTS
Nutrition profiles and ACFS scores
A total of 88 international recipes, including 15 Korean home-dining, 15 Korean out-dining, 10 Chinese, 20 Mediterranean, 10 vegetarian, and 18 Western recipes, were included in this study and analyzed. The number of recipes in the highest anticarcinogenic category (S grade) was 13 (14.8%), followed by 20 in A grade (22.7%), 21 in B grade (23.9%), 15 in C grade (17.0%), 16 in D grade (18.2%), and 3 in E grade (3.4%). All 13 S grade recipes belonged to vegetarian or Mediterranean recipes. The median ACFS grades, derived from ingredient scores after considering the harmful cooking methods, were B, C, B, A, S, and D for Korean home-dining, Korean out-dining, Chinese, Mediterranean, vegetarian, and Western recipes, respectively.
For all the 88 recipes, the ingredient scores ranged from 132 to 487 with a median value of 320 and the energy values ranged from 137 to 1,643 kcal with a median value of 484 kcal. The median ingredient scores of Korean home-dining, Korean out-dining, Chinese, Mediterranean, vegetarian, and Western recipes were 310 (range: 192–407), 298 (200–368), 359 (132–487), 367 (229–443), 432 (394–459), and 201 (141–321), respectively; median energy values in kcal were 708 (350–1643), 498 (274–1364), 335 (146–582), 504 (180–1233), 298 (248–406), and 555 (137–1347), respectively.
The names of the recipes, ingredient scores, ACFS grades, calories, and brief nutritional profiles are presented in Table 2. The full nutritional profile is described in Supplementary file S1.
Nutritional analyses
For univariate analysis, we categorized the 6 ACFS grades into 3 groups: S and A; B and C; D and E. In the analysis, the following factors were correlated with the ACFS grade groups (P < 0.05): total calories, total fat, animal fat, animal protein, total protein, vitamin D, riboflavin, niacin, vitamin B12, pantothenic acid, sodium, animal iron, zinc, selenium, and cholesterol (negative trends); carbohydrate rate, fiber, water-soluble fiber, vitamin K, vitamin C, and plant calcium (positive trends). The results of the univariate analysis are summarized in Table 3. Multivariate analysis was performed on factors found to be significant in the univariate analysis. The following variables were included: animal fat, animal protein, vitamin D, riboflavin, niacin, vitamin B12, pantothenic acid, sodium, animal iron, zinc, selenium, cholesterol, fiber, vitamin K, vitamin C, and plant calcium. For multivariate analysis, the 6 ACFS grades and the quantitative amounts of nutrition profiles were used. Six profiles were found to be statistically significant: animal fat (standardized β = −0.330, P = 0.020), animal iron (standardized β = −0.345, P = 0.021), and niacin (standardized β = −0.359, P = 0.004) (negative trends); fiber (standardized β = 0.183, P = 0.034), animal protein (standardized β = 0.525, P = 0.006), and vitamin C (standardized β = 0.420, P < 0.001) (positive trends). Two profiles showed non-significant negative trends (P < 0.1): pantothenic acid (standardized β = −0.192, P = 0.062) and sodium (standardized β = −0.175, P = 0.077). Vitamin B12 showed a non-significant positive trend (standardized β = 0.193, P = 0.074). Each standardized β value represented the relative significance of each variable. These results are summarized in Table 4.
DISCUSSION
The present study demonstrated that the ACFS calculation model was significantly associated with known healthy and unhealthy nutritional factors. Univariate analysis showed that the ACFS model was strongly negatively correlated with factors such as animal fat (P < 0.001), sodium (P < 0.001), and cholesterol (P < 0.001), which are generally considered to be harmful to health, and strongly positively correlated with factors such as fiber (P < 0.001), vitamin C (P < 0.001), and plant calcium (P < 0.002), which are considered beneficial.
Multivariate analysis showed that the ACFS model is negatively correlated with animal fat, animal iron, and niacin and positively correlated with fiber, vitamin C, and animal protein (Ps < 0.05). Pantothenic acid and sodium had non-significant negative trends, while vitamin B12 showed a positive trend (Ps < 0.1).
In the recently updated WCRF and AICR third expert report, red meat consumption showed strong evidence of increased risk of colorectal cancer and limited evidence of increased risks of nasopharyngeal, lung, and pancreatic cancers [18]. The most well-known causes of cancer related to red meat consumption are heterocyclic amines and polycyclic aromatic hydrocarbons, which are carcinogens that are produced on high-temperature cooking [19]. Besides, saturated fat itself has been known to be related to the risk of cancers, including breast, lung, and colorectal cancers [202122]. This evidence aligns with our study results concerning animal fat. The iron components in animal products were found to be associated with the ACFS grade in our study. Heme iron, which is abundant in red meat, was known to be related to tumorigenesis by stimulating the endogenous formation of N-nitroso compounds [19]. Body iron stores, which were assessed by serum iron and transferrin saturation of total iron-binding capacity, were reported to increase the incidence of cancer in large population studies [2324], although these results need to be verified by studies with more nutritional perspectives.
Animal protein was found to have a positive relationship with the ACFS grade in multivariate analysis, which reflects its anticarcinogenic potential. Interestingly, animal protein showed a negative trend in the univariate analysis but reversed to have a positive trend in the multivariate analysis. We hypothesized that animal protein from poultry and fish might affect these results and the carcinogenic effect of meat might be more related to other components such as fat or iron, although further studies are warranted. Poultry intake was reported to be inversely associated with colorectal cancer risk in a recent meta-analysis, with a risk ratio (RR) of 0.89 (95% confidence interval [CI], 0.81–0.97) for a 50 g/day increase uptake of poultry [25]. A meta-analysis study reported an inverse relationship between poultry intake and lung cancer (RR 0.91, 95% CI: 0.85–0.97) [26], and a population-based case control study showed a similar relationship between pancreatic cancer and poultry intake (odds ratio: 0.7, 95% CI: 0.5–1.0) [27]. Fish intake was reported to have limited-suggestive evidence of decreasing the risk of liver and colorectal cancers in the third expert report of the WCRF and the AICR [18]. In a meta-analysis, the pooled RR of the highest fish intake category compared to the lowest fish intake category was 0.88 (95% CI: 0.78–1.00) for colorectal cancer incidence [28]. For liver cancer, the pooled RR for the highest fish intake quartile was 0.82 (95% CI: 0.71–0.94), as per a recent meta-analysis study [29]. Unfortunately, we cannot fully explain the results of the meta-analyses with regard to the mechanisms, as the mechanisms are not fully understood yet and are masked by confounding factors [2530]. We hope that future studies investigating the cancer protective mechanisms of poultry and fish can explain these results and show a correlation with ACFS.
Foods containing dietary fiber had strong protective evidence for colorectal cancer in the third expert report of the WCRF and the AICR [31]. In humans, fibers can be fermented and metabolized by the colonic microflora, forming short-chain fatty acids, such as butyrate, which have an anti-proliferative effect on colon cancer. Additionally, fibers can reduce the transit time of fecal bulk and lessen the interaction between fecal mutagens and colonic mucosa [3233]. Dietary fiber was also associated with decreased risk of breast cancer in a meta-analysis, with an RR of 0.95 per 10 g/day intake [34]. The European Prospective Investigation into Cancer and Nutrition study, which included a prospective cohort of > 500,000, reported that colorectal, breast, and liver cancers were inversely associated with dietary fiber intake [35]. Vitamin C has been known to prevent cancer and proven to have a protective effect on esophageal, laryngeal, oral cavity, pancreatic, stomach, rectal, breast, and cervical cancers, which are strongly consistent with the results of the previous meta-analyses [3637]. It was also reported to have limited -suggestive evidence of decreasing the risks of colorectal and lung cancers in the third expert report of the WCRF and the AICR [31].
The relationship between niacin intake and cancer risk is largely unexplored. Although there is some in vivo evidence that niacin status influences carcinogenesis in a tissue-specific manner, evidence from human studies is lacking [38]. In a recent clinical study, niacin intake was inversely associated with squamous cell carcinoma of the skin but showed a positive relationship with basal cell carcinoma of the skin [39]. Since ACFS is based on meal recipes (not nutritional components), correlation with niacin, which is abundant both in plant foods (e.g., peas and barley have high ACFS score) and red meats (low ACFS scores), might not be robust. Future clinical studies evaluating the anti-cancer or carcinogenic effects of niacin are necessary.
Among the factors with non-significant trends, vitamin B12 was shown to have an inverse relationship with cancer risk in recent studies. A dose-response relationship between dietary intake of vitamin B12 and decreased risk of colorectal cancer was shown in a recent meta-analysis (pooled RR: 091, 95% CI: 0.86–0.98) [40]. A recent case-control study revealed that low vitamin B12 concentration was associated with a 5.8-fold higher risk of non-cardia gastric cancer [41]. Another study also found a significant association between low levels of plasma vitamin B12 and hepatocellular carcinoma risk (odds ratio: 2.01, 95% CI: 1.02–3.98) [42]. Sodium was also negatively associated with the ACFS grade with non-significant trends. The positive association between sodium and cancer is most well-known for gastric cancer [43], and the third expert report by the WCRF and the AICR also reported strong evidence between Cantonese-style salted fish and nasopharyngeal cancer [44]. Although relevant effects of pantothenic acid have not been extensively studied in the literature, Hutschenreuther et al. [45] reported that the aggressiveness of tumor cells might be related to pantothenic acid in a cell-line study. This component should be investigated in future to explain our results.
It has long been necessary to categorize and quantify factors that can prevent or increase the risk of cancer. The International Agency for Research on Cancer systematically classifies substances that may or may not be carcinogenic [46]. The expert reports of the WCRF and the AICR, probably the most comprehensive review regarding cancer prevention, diet, and lifestyle, have proved their clinical efficacy in recent prospective trials [7]. Our ACFS model systematically analyzed the literature regarding diet and cancer prevention, based on the classification system from the WCRF and AICR [47], and enabled quantitative anti-cancer scoring of a wide-range of everyday meals. Consequently, our model allows everyone without medical or nutritional expertise to obtain an evidence-based anti-cancer or carcinogenic score of their meals.
Although several models have already been published to assess the health index of food, our model is unique because of the following reasons. Scoring methods such as aggregate nutrient density index (ANDI) [48] and NuVal® [49], which are developed in the US and commonly used, are composed of assessments focusing on overall health, while the ACFS focuses on cancer prevention or carcinogenicity. Additionally, the ACFS evaluates complete meals actually consumed in daily life, while the ANDI and NuVal evaluate food components. Healthy Eating Index (HEI) [50] and Dietary Quality Index (DQI) [51] are tools for evaluating meals rather than individual food components. However, these tools do not focus on cancer but on overall health or known nutritional expertise. Hence, the ACFS model is unique in evaluating daily meals, yielding indexes focused on cancer prevention and carcinogenicity.
The purpose of the present study was to provide a nutritional basis for the efficacy of the ACFS calculation model. Although ACFS was based on a comprehensive systematic review, there is a lack of substantive evidence to demonstrate its efficacy. This study could facilitate the design of research demonstrating the clinical utility of ACFS. Although we included 88 recipes from 6 international categories, probably the number of samples should be further increased. Representing each diet category with 10–20 recipes also has limitations. We performed fairly objective selections of representative foods for Asian recipes, but we were forced to rely mainly on the author's discussion in the corresponding websites for the selection of Western meals. Along with the clinical study design, to enhancement of nutritional reliability using larger validation with more number of international recipes will be the subject of our future studies.
We have shown that the ACFS grade generally corresponds with commonly known anti-cancer or carcinogenic factors. Although the effect of some factors needs to be elucidated through additional studies, the ACFS grade has been well correlated with factors such as fiber, vitamin C, vitamin B12, sodium, animal fat, and iron. Hence, our calculation model could be named ACFS 2.0, reflecting the newly drawn reliability of this study. Future clinical studies along with wider validation are warranted to firmly prove the utility of the model.
XML Download