This study included cases of newly diagnosed colorectal cancer patients at the Center for Colorectal Cancer of the National Cancer Center, Korea, between August 2010 and August 2013. Of the 925 patients who agreed to participate in the study and completed the questionnaires, two patients were excluded due to implausible energy intakes. In total, 923 patients were included in the analysis. The control participants were recruited between October 2007 and December 2014 among individuals visiting the Center for Cancer Prevention and Detection at the same hospital for a health check-up program provided by the National Health Insurance Cooperation, which covers the entire Korean population. Of the 9,157 healthy subjects who agreed to participate in the study and completed the questionnaires, 120 subjects were excluded due to the implausible energy intake. Of the remaining 9,037 individuals, two controls per case were randomly selected and frequency-matched by gender and 5-year age group. Additionally, individuals with missing blood samples and genotyping data were excluded. Ultimately, 632 colorectal cancer patients and 1,295 healthy controls were included in the final analysis (S1 Fig.).

Information regarding the participants’ demographic and lifestyle risk factors (e.g., smoking, alcohol consumption, and regular exercise) was collected by in-person interviews or structured questionnaires at initial recruitment, prior to cancer diagnosis. The dietary intake was assessed using a validated 106-item semi-quantitative food frequency questionnaire specifically developed for this population encompassing commonly consumed food items. The validity and reproducibility of the questionnaire have been previously reported [15]. Each participant provided their average frequency of eating and typical portion sizes in the year preceding the interview. These values were converted to obtain daily nutrient intake values using a scale with nine frequency categories and three portion size categories. We calculated the dietary inflammatory index (DII), and the calculation of the DII has been previously described [16,17]. In the present study, 35 food components were included to calculate DII: protein, fat, carbohydrate, fiber, monounsaturated fatty acid, saturated fatty acid, polyunsaturated fatty acid, n-3 fatty acid, n-6 fatty acid, cholesterol, thiamin, riboflavin, niacin, vitamin B₆, vitamin B₁₂, vitamin C, folic acid, vitamin A, vitamin D, vitamin E, β-carotene, iron, magnesium, selenium, zinc, garlic, ginger, onion, green tea, flavan-3-ol, fla-vone, flavonol, flavanone, anthocyanidin, and isoflavone.

The single-nucleotide polymorphisms (SNPs) were selected as previously described [18]. Briefly, 30 SNPs were selected from the literature based on their association with colorectal cancer in previous genome-wide association studies [18]. Details regarding the selected SNPs are provided in S2 Table.

The genotyping was conducted as described below. Genomic DNA was extracted using the MagAttract DNA Blood M48 Kit (Qiagen, Hilden, Germany) and BioRobot M48 automatic extraction equipment (Qiagen) according to the manufacturer’s instructions. The genotyping was performed using a MassARRAY iPLEX Gold Assay (Agena Bioscience, San Diego, CA). To control the genotyping quality, duplicate samples of 3% of the subjects were included in our initial genotyping analysis; the rate of discordance was < 1%.

To construct the genetic risk score, the independent association between 30 SNPs and colorectal cancer was initially examined using a logistic regression under allele-based model. Among them, 13 SNPs were significantly associated with the risk of colorectal cancer in the present study. Therefore, these SNPs were used to calculate the genetic risk scores. We constructed both unweighted and weighted genetic risk scores [19]. The unweighted genetic risk scores of individual participants were created by adding the number of risk alleles (i.e., 0, 1, or 2) at each SNP. Additionally, we calculated the weighted genetic risk score by multiplying by the effect size (the SNP’s β coefficient from the current logistic regression model) in the current case-control study. The 13 products were then summed to create a weighted genetic risk score for each study participant.

We generated a modifiable lifestyle index based on a priori knowledge of risk factors for colorectal cancer and public health recommendations. The lifestyle factors were characterized as harmful based on their independent association with colorectal cancer. First, both current and previous (prior 2 years) body mass indexes (BMIs) were calculated because the current BMI could be affected by the disease status in the case-control studies [20]. The BMI was categorized according to the World Health Organization criteria for the Asian population (< 25 kg/m² vs. ≥ 25 kg/m²). We compared the current and prior BMI, and found that the prior BMI is more significantly associated with colorectal cancer in the present study. Therefore, the prior BMI was used as an index for obesity. Second, physical activity was defined as conducting exercise regularly. Third, an unhealthy diet was defined based on the level of DII. Diet was classified as a low or high inflammatory diet based on the median intake levels of DII scores in the control group. Fourth, cigarette smoking was classified as ever or never. Smoking was chosen based on previous evidence and public health recommendations [7,21]. Fifth, people who drank more than 30 g of alcohol/day were considered heavy drinkers based on the association between alcohol consumption and colorectal cancer risk in the current study.

Finally, we assigned a lifestyle risk score (0 for healthy and 1 for unhealthy) to each participant and summed the scores of the five lifestyle factors; consequently, the lifestyle index ranged from zero (healthiest) to five (least healthy) points [11].

The differences in the demographic and lifestyle factors between the cases and controls were analyzed using the chi-squaqre test for categorical variables and Student’s t test for continuous variables.

The association between the risk factors (genetic and lifestyle risk factors) and colorectal cancer risk was analyzed using unconditional logistic regression models. The genetic risk score was calculated using 13 SNPs and divided into tertiles based on scores among controls: low (20%-30%), intermediate (50%-60%), and high (20%-30%). Because some of participants had the same risk scores, we could not divide subjects into the same proportion. The combined lifestyle factors were also divided into tertiles: low (defined as no or only one lifestyle risk factor), intermediate (defined as 2 or 3 lifestyle risk factors), and high (defined as four or five lifestyle risk factors). The lowest genetic or lifestyle risk factors were considered references. The multivariable model was adjusted for age, sex, family history of colorectal cancer, and education and additionally adjusted for the five modifiable risk factors (i.e., prior BMI, physical activity, DII, smoking, and alcohol consumption) as needed. Furthermore, we conducted a stratified analysis of the association between the lifestyle risk factors and colorectal cancer risk according to the genetic risk category. We used the results based on the unweighted and weighted risk score in further analyses. A multinomial logistic regression model was used in the analyses stratified by anatomical location (i.e., colon and rectal cancer). Interactions between genetic risk scores and lifestyle risk scores were assessed using the likelihood ratio test by comparing the model with the interaction term, with the model containing only main effects.

All statistical analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC) and STATA ver. 14 software (Stata Corp., College Station, TX). A two-sided p-value of less than 0.05 indicated statistical significance.

All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board of the National Cancer Center (IRB No. NCCNCS-10-350 and NCC2015-0202).