Ultra-Processed Food Intake and Smoking Interact in Relation with Colorectal Adenomas

Smoking and ultra-processed foods (UPFs), a substantial part of the western diet, have been suggested to have a potential carcinogenic effect, though epidemiologic data are lacking. We aimed to examine the association between high UPF intake and colorectal adenomas, and to test the interaction with smoking. In a case-control study among consecutive subjects undergoing colonoscopy in a tertiary center during 2010–2015, UPF intake and smoking were compared between cases with colorectal adenomas and controls. Within 652 participants (cases, n = 294 and controls, n = 358), high UPF intake (defined as percent of kcal from UPF above the study sample upper tertile) was positively associated with adenomas (Odds ratio (OR) = 1.75, 95% Confidence interval (CI) 1.14–2.68), advanced and proximal adenomas (OR = 2.17, 1.29–3.65 and OR = 2.38, 1.37–4.11) among the whole study sample; and with adenomas (OR = 3.54, 1.90–6.61), non-advanced adenomas (OR = 2.60, 1.20–5.63), advanced adenomas (OR = 4.76, 2.20–10.30), proximal adenomas (OR = 6.23, 2.67–14.52), and distal adenomas (OR = 2.49, 1.21–5.13) among smokers. Additionally, a dose-dependent association was observed between tertiles of UPF intake and adenomas only among smokers (p for trend < 0.001). A significant interaction between smoking and high UPF intake was detected (p for interaction = 0.004). High intake of UPFs is strongly and independently associated with colorectal adenomas, especially advanced and proximal adenoma, and interacts with smoking. Results highlight smokers as more susceptible to the negative health effects of UPF consumption on colorectal neoplasia.


Introduction
Colorectal cancer (CRC) is the world's third most common cancer and the fourth most deadly cancer, with almost 900,000 deaths annually [1,2]. In addition to a genetic background, it has been shown that metabolic profile [3], lifestyle [4], dietary habits [5,6], and smoking [4,7] are strongly and independently associated with colorectal polyps, the direct precursors of CRC. Moreover, smoking and dietary factors have been reported to interact in their association with colorectal neoplasia [8][9][10].
A major component of the western diet is processed and ultra-processed food (UPF), contributing up to 79% of the mean daily calories [11]. According to the Food and Agriculture Organization of the United Nations, food processing includes physical, biological, and chemical techniques, used to prepare ready to eat, drink, or heat foods and beverages. UPF is energy dense; high in unhealthy types of fat, refined starches, sugar, salt, and artificial additives; as well as a poor source of protein, fiber, and micronutrients [12]. The most common UPFs are snacks, drinks, and ready-to-eat meals.

Definition of Cases and Controls
Indication for colonoscopy was defined as screening (due to age), colonoscopy for alarming symptoms (such as rectal bleeding, unintentional weight loss etc.), or surveillance (due to a personal or family history of colorectal polyps or a family history of CRC past the age of 70 years).
Polyp histology was reviewed by a gastro-intestinal (GI) pathologist. Adenomas were classified as advanced or non-advanced according to the guidelines of the United States (US) Multi-Society Task Force on CRC, the American Cancer Society CRC Advisory Group, the US Multi-Society Task Force, and the American College of Radiology CRC Committee [18,19]. Advanced adenomas were defined as large adenomas (>10 mm), with features of high-grade dysplasia (HGD) or villous histology, multiple (≥3) non-advanced adenomas. Non-advanced adenomas were defined as adenomas <10 mm, without features of HGD or villous histology. Cases with more than one polyp were defined according to the polyp of highest neoplastic potential. Controls were patients with no colorectal polyps detected in their current or past colonoscopies.
Polyp location was defined as proximal adenomas (cecum, appendix, ascending colon, hepatic flexure, transverse colon, and splenic flexure) and distal adenomas (descending colon, sigmoid colon, and rectum) based on International Statistical Classification of Diseases 10th edition for CRC [20].

Data Collection
Within 2 months after undergoing a colonoscopy, all participants were requested to undergo a medical interview, anthropometric measurements, and blood tests and answer questionnaires on lifestyle and diet. Participants were face-to-face interviewed for their medical history, demographic characteristics, lifestyle, and dietary intake. Blood pressure, weight, height, and hip and waist circumference were measured using a uniform protocol. Body mass index (BMI) was calculated as weight (kilograms)/height 2 (meters). Participants' blood tests were obtained following a 12-h fast, and were analyzed at a single lab.
Smoking was defined as ever (past/present) smoking vs. never smoking.Mets was defined according to the American Heart Association (AHA) criteria [21] if patients were diagnosed with ≥3 of the following criteria: impaired fasting glucose (fasting glucose>100 mg/dl), hypertension (systolic blood pressure/diastolic blood pressure >130/85 mmHg and/or medication), low high-density-lipoprotein (HDL) (HDL <40/50 mg/dL among men and women, respectively), high triglycerides (triglycerides >150 mg/dL or medications), and abdominal obesity (waist circumference >88/102 cm among women and men, respectively). A healthy dietary index was defined according to the AHA healthy diet components [22,23].

Evaluation of UPF Intake
Evaluation of dietary intake was performed using a structured detailed semi-quantitative FFQ, assembled by the Food and Nutrition Administration, the Israeli Ministry of Health, that has been validated for the Israeli population [24], and comprised of 117 food items with specified serving sizes. All participants were asked to refer to the past year when filling in the FFQ, and were kept blinded to the study hypothesis. The classification of the processing level of foods was determined based on the NOVA classification [25]. UPFs were classified as foods that are industrial formulations that are typically comprised of many ingredients, particularly substances not commonly found in natural food products, such as hydrolyzed protein, modified starch, hydrogenated or inter-esterified oils, and additives, such as colorants, flavoring, non-sugar sweeteners, emulsifiers, humectants, sequestrants, firming, bulking, de-foaming, anti-caking, and glazing agents [25] (Table S1).
Total calories from UPF items were summed, and the proportional caloric intake of UPFs from total caloric intake, and separately for 6 food groups was calculated: (1) bread, pastries, and starch; (2) snacks; (3) beverages; (4) oils and spreads; (5) dairy; and (6) meat, poultry, and fish [13]. The proportional caloric intake of UPFs from total calories, and from each food group were categorized into tertiles according to the consumption of the study sample. High UPF intake was defined as percent of kcal from UPF above the study sample upper tertile (≥ 44.8% of total kcal).

Statistical Analysis
All statistical analyses were performed using SPSS version 25.0 for Windows (SPSS Inc., Chicago, IL, USA). Continuous variables are presented as means ± SD and nominal variables as proportions. Pearson Chi-Square test was used to test the association between nominal variables. The independent samples t-test was used to compare between cases and controls. One-way Anova was used to test the difference between three groups of proportional UPF intake. Multivariate logistic regression analysis was used to test the association between high UPF intake and colorectal adenomas, controlling for potential confounders as demographics and variables that distributed differently between cases and controls and may be related with dietary characteristics: age, gender, aspirin use, indication for colonoscopy, BMI, total kcal, and the Mets. The proportional caloric intake of UPFs from total calories, and from each food group was categorized as tertiles. For all UPF in the diet, the 2nd tertile (30.4-44.7% of total kcal) and the 3rd tertile (≥44.8% of total kcal) were compared to the 1st tertile (≤30.4% of total kcal). An interaction between high UPF intake and smoking status was assessed by the interaction term in binary logistic regression, adjusting for potential confounders, smoking status, and high UPF intake. p < 0.05 was considered statistically significant for all analyses.

Characteristics of the Study Population and Comparison between Cases and Controls
We included 652 participants in this analysis, 294 cases with colorectal adenomas and 358 controls with no past/present polyps (mean age ± standard deviation (SD) 58.5 ± 6.6 years, 50.8% men, mean BMI 28.2 ± 5.4 kg/m 2 ). Mean daily caloric intake was 2043.7 ± 692.1, of which the mean proportional caloric intake from UPFs was 38.2 ± 16.2%. Intake of UPF was mostly derived from 'snacks', 'beverages', 'oils and spreads' and 'dairy products' (Figure S1). High UPF intake (third tertile vs. first tertile) was significantly associated with male gender. Unhealthy lifestyles, including smoking, physical inactivity, and obesity, tended to be higher, and the consumption of a healthy diet tended to be lower among participants consuming high UPF, but these differences were not statistically significant. As expected, participants with high UPF intake had a higher caloric intake, higher proportional carbohydrates and SFA intake, and lower protein intake (Table S2).

The Association between UPF Intake and Colorectal Adenomas
Cases with adenomas were older, had a higher mean BMI, and included a higher proportion of males, smokers, aspirin users, participants with the Mets, and participants undergoing surveillance colonoscopy (Table 1).
Cases and controls did not differ in dietary intake of total calories, calories from different food groups, proportion of calories from macronutrients, and major nutrients, as saturated fatty acids (SFA), fiber, and sodium, but cases had lower intake of mono-unsaturated/SFA ratio. Cases had a higher proportional caloric intake of UPFs, mainly contributed by the following food groups: oils and spreads, dairy products, snacks, and beverages. These differences were also seen for subgroups of cases with advanced adenomas and proximal adenomas ( Table 2).
The proportional caloric intake of UPFs was significantly associated with adenoma stage, with a significant positive trend, among the total population and among smokers but not among never-smokers ( Figure 1). a Low socio-economic status-A combination of low education (<12 years of school education) and low income (< the group 1st Q of monthly household income) b Physical inactivity-Reported no intentional exercise or less than 20 min/week of exercise, which leads to increased heart rate and/or sweating. Abbreviations: BMI-Body Mass Index  In addition, only among smokers, there was a positive trend in the association between tertiles of proportional caloric intake of UPFs and adenomas, particularly advanced adenomas (Figure 2). In addition, only among smokers, there was a positive trend in the association between tertiles of proportional caloric intake of UPFs and adenomas, particularly advanced adenomas (Figure 2).

Association between UPF Intake and Colorectal Adenomas and Its Interaction with Smoking
In multivariate analysis, a high UPF intake (third tertile vs. first tertile) was positively associated with adenomas (odds ratio (OR) = 1.75, 95% confidence interval (CI) 1.14-2.68), advanced adenomas (OR = 2.17, 95%CI 1.29-3.65), and proximal adenomas (OR = 2.38, 95%CI 1.37-4.11) among the total study population. A positive dose-response association was detected for adenomas, advanced adenomas, and proximal adenomas. Stratified by smoking status, significant positive dose-response associations between high UPF intake (third tertile vs. first tertile) and colorectal adenomas, nonadvanced, advanced, proximal, and distal adenomas were observed only among smokers, and not among never-smokers. There was a significant interaction between smoking and high UPF intake (third tertile vs. first tertile) in relation with adenomas (p for interaction = 0.004), non-advanced

Association between UPF Intake and Colorectal Adenomas and Its Interaction with Smoking
In multivariate analysis, a high UPF intake (third tertile vs. first tertile) was positively associated with adenomas (odds ratio (OR) = 1.75, 95% confidence interval (CI) 1.14-2.68), advanced adenomas (OR = 2.17, 95%CI 1.29-3.65), and proximal adenomas (OR = 2.38, 95%CI 1.37-4.11) among the total study population. A positive dose-response association was detected for adenomas, advanced adenomas, and proximal adenomas. Stratified by smoking status, significant positive dose-response associations between high UPF intake (third tertile vs. first tertile) and colorectal adenomas, non-advanced, advanced, proximal, and distal adenomas were observed only among smokers, and not among never-smokers. There was a significant interaction between smoking and high UPF intake (third tertile vs. first tertile) in relation with adenomas (p for interaction = 0.004), non-advanced adenomas (p for interaction = 0.019), advanced adenomas (p for interaction = 0.007), proximal adenomas (p for interaction = 0.026), and distal adenomas (p = 0.004) ( Table 3). ORs are adjusted for age, gender, BMI, total kcal, aspirin use and indication for colonoscopy. a Smoking is defined as ever (past/present) smoking. b Ultra-Processed Foods (UPFs) intake was defined as the proportional caloric intake of UPFs from total caloric intake. The 2nd tertile of UPF (30.4-44.7% of total kcal) and the 3rd tertile (≥44.8% of total kcal) were compared to the 1st tertile (≤30.4% of total kcal). c The interaction between UPF intake and smoking status, adjusted for all parameters of the model, smoking and high UPF intake. Abbreviations: OR-Odds ratio, CI-Confidence interval, UPF-Ultra-Processed Food Among all food groups, high UPF intake (third tertile vs. first tertile) from 'dairy products' and 'snacks' among smokers and high UPF intake from 'oils and spreads' in both smokers and never-smokers were independently associated with colorectal adenomas in multivariate analysis (Table S3).

The Association between UPF Intake and Colonic Adenomas as Compared with Other Major Risk Factors
The association between high intake of UPF and colorectal adenomas, advanced, and proximal adenomas was as strong as that of established risk factors for colorectal cancer, such as smoking and the Mets, in a multivariate model with all risk factors adjusted for confounding factors and for one another (Figure 3). never). b Metabolic syndrome (yes vs. no). c UPF intake was defined as tertiles of the proportional caloric intake of UPFs from total caloric intake. The 2nd tertile of UPF (30.4-44.7% of total kcal) and the 3rd tertile (≥44.8% of total kcal) were compared to the 1st tertile (≤30.4% of total kcal). All ORs are adjusted for age, gender, BMI, total kcal, aspirin, indication for colonoscopy, smoking, metabolic syndrome, and high UPF intake. Abbreviations: OR-Odds ratio, CI-Confidence interval, UPF-Ultra-Processed Foods

Discussion
Processed food, a key component of the western diet, is prevalent among various food groups [13], and has been shown to contribute up to 60% of the daily kcal intake of adults [26], 40% of children [27], and 20% of infants [28]. UPF intake has been studied in association with various health outcomes, including cancer, and specifically, CRC [16], but evidence regarding its association with pre-malignant colorectal polyps is lacking. Our study findings show a positive independent association between UPF intake and colorectal adenomas.
We observed strong positive dose-response associations between high UPF intake and colorectal adenomas of both types and of both colorectal locations. These associations were observed only among smokers, and a significant interaction was detected between smoking and high UPF intake in relation to all colorectal adenoma categories. The association between high UPF intake and colorectal adenomas was as strong as the association between smoking or the Mets and adenomas, well-established risk factors for colorectal carcinogenesis, which have been previously suggested as CRC screening referral factors [29].
As expected, participants who consumed higher proportions of UPF also consumed significantly more calories, SFA, and carbohydrates. Western lifestyle characteristics, such as smoking, physical inactivity, obesity, and the consumption of a healthy diet, were different between participants characterized by different intake levels of UPF, but these differences were not significant. Cases and controls differed in UPF intake but not in intake of other dietary components of the western diet, such as fiber, sodium, and fat. Additionally, the associations between UPF intake and adenomas were consistent across types and locations of adenomas, independent of total caloric intake and BMI, which were adjusted for in multivariate analysis. Therefore, it is reasonable to assume that the association of UPF with adenomas is not just a reflection of an association with an unhealthy western diet. These results strengthen the independent association between UPF intake and colorectal neoplasia. This association is supported by biological mechanisms linking CRC development and growth, with various synthetic components of processed food. These include nitrates and nitrites [30] in processed meats, and other less studied food additives, such as monosodium glutamate, titanium dioxide [31], high-fructose corn syrup [32], and synthetic dyes [33]. Indeed, the major sources of UPFs that were found to be related with colorectal adenomas were 'snacks', 'oils and spreads', and 'dairy products'; all contain significant amounts of the above-mentioned components and food additives. Surprisingly, the 'dairy products' food group, which is considered an important part of the Dietary Approach to Stop Hypertension (DASH) diet and the Mediterranean diet [34], included 40% of the kcal from ultra-processed dairy products. Cases with colorectal adenomas consumed higher proportions of ultra-processed dairy products, and dairy products were positively associated with colorectal adenomas among smokers in multivariate analysis. Previous reports have shown dairy products to be protective of colorectal neoplasia, but results have been inconclusive [34,35]. Given that none of these previous reports had differentiated between processed and non-processed dairy products, these inconsistencies may be attributed to this potentially important distinction. Among our study population, the intake of ultra-processed 'meat, poultry, and fish' was about 10% of the total kcal, nearly a third of the intake reported in other cohorts of the western world [36]. This may explain the finding that high intake of processed 'meat, poultry, and fish' was not significantly associated with colorectal adenomas in this cohort, unlike the findings of others [37]. Another explanation may be that the intake of UPF as a whole is the exposure of importance, regardless of the food group the UPF is oriented from.
The positive association between UPF intake and adenomas was stronger with advanced adenomas, and may reflect a potential role of UPFs in colorectal neoplasia progression. The association with proximal adenomas differs from previous reports that link diet and smoking with distal adenomas [7,8]. Interestingly, the associations persisted following adjustments for various potential confounding factors among smokers but not among non-smokers, yielding a significant interaction. Importantly, in our study, smoking and other unhealthy lifestyle characteristics were not associated with UPF intake, therefore potential confounding by these factors is unlikely. Additionally, a dose-dependent association was detected between UPF intake and colorectal adenomas, with a significant positive trend seen only among smokers. Previous studies have demonstrated interactions between diet and smoking in their association with colorectal neoplasia, but these have mainly been attributed to meat intake [38,39] and plant-based diets [40]. To our knowledge, none of these studies have addressed UPFs. Therefore, we believe that this is the first report of an association between UPF intake and colorectal adenomas, and of its interaction with smoking. One potential mechanism to explain this interaction is the joint effects of smoking and diet on the gut microbiome [41], which can in turn influence CRC risk. The main proposed mechanisms linking the microbiome with colorectal neoplasia include secretion of oncogenic microbial metabolites, such as secondary bile acids, and activation and promotion of an inflammatory response in the gut mucosa, which further promote the growth of tumor cells, and inhibit apoptosis [42][43][44] The limitations of this study include the lack of temporal sequence, which does not permit a causal inference. Cases and controls were recruited from the same population and had comparable socioeconomic characteristics, thus minimizing a potential selection bias. In terms of external validity, this study population was intentionally selected to represent a population with low to medium risk for colorectal polyp; thus, our findings cannot be generalized to high-risk populations. We attempted to minimize a potential confounding effect and elaborate on effect modification by stratification across smoking status, which is strongly associated with CRC, and adjustment in multivariate analysis. Still, residual confounding may still exist. Nutritional data were collected within a single country, which may impact on the generalizability across populations with different diets. Information bias, and particularly recall bias, on lifestyle characteristics may exist, due to the retrospective nature of the study. This was minimized by a uniform structured lifestyle and dietary questionnaire, which was assessed in the same manner among cases and controls, all blinded to the study hypothesis to prevent differential bias.

Conclusions
Among smokers, high UPF intake is strongly and independently associated with colorectal adenomas, especially for advanced and proximal adenoma. The results highlight the need to further investigate the role of UPF as an independent risk factor for colorectal neoplasia, and its interaction with smoking, in larger prospective studies, incorporating microbiome analysis. As UPF intake increases worldwide, especially among children and young adults, its negative implications should be targeted for CRC prevention.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/12/11/3507/s1, Table S1. Un-processed and Processed vs. Ultra-processed Food groups, Figure S1. The proportional caloric intake of UPFs in the total diet and within food groups; Table S2. The association between high UPF intake (≥44.8% of total kcal) and lifestyle characteristics; Table S3. The adjusted association between high UPF intake from food groups, and colorectal adenoma as compared to controls, stratified by smoking status.