Even-Chain UFA Pattern, n-3 Long-Chain PUFA Pattern and Risk of Esophageal Squamous Cell Carcinoma in Southeast of China

Objective: To characterize and examine associations between dietary fatty acid intake patterns and the risk of esophageal squamous cell carcinoma (ESCC). Methods: A total of 422 patients and 423 controls were recruited. Dietary fatty acids, as a percentage of total fat, were entered into a factor analysis. Multinomial logistic regression and restricted cubic spline were used to evaluate the risk of ESCC specic for different dietary fatty acid patterns (FAPs). A forest plot was applied to show the potential effect modication. Results: The factor analysis generated 4 major fatty acid patterns: a long-chain SFA (LC-SFA) pattern; an even-chain unsaturated fatty acid (EC-UFA) pattern; a short and medium-chain SFA (SMC-SFA) pattern, and an n-3 long-chain polyunsaturated fatty acid (n-3 LC-PUFA) pattern. In the multivariate-adjusted model, the odds ratios (ORs) with 95% condence interval (CI) of ESCC were 2.069(1.314,3.257), and 0.525 (0.340–0.811) for the highest versus the lowest tertile of EC-UFA pattern and n-3 LC-PUFA pattern, respectively. The LC-SFA, SMC-SFA patterns were not associated with ESCC. There existed a nonlinear positive association between the EC-UFA pattern and the risk of ESCC (p for nonlinearity <0.05), nevertheless, there was a nonlinear negative association between the n-3 LC-PUFA pattern and the risk of ESCC (p for nonlinearity <0.001). Multiplicative interaction between fried food, pickled food, hard food, tobacco smoking, alcohol drinking, and four FAPs was also observed. Conclusions: Our study indicates that EC-UFA pattern, n-3 LC-PUFA pattern intake are associated with ESCC, which might provide a potential dietary intervention for ESCC prevention.


Introduction
Esophageal carcinoma has the 7th highest incident and constitutes the 6th leading cause of cancer deaths worldwide by GLOBOCAN 2018. Asia accounted for about 78% of all esophageal carcinoma cases, while 49% of cases occurred in China [1,2] . As the fth-most common and the fourth-most deadly cancer [3] , the age-adjusted 5-year survival of esophageal carcinoma is relatively poor, in the range of 20-30% [4,5] . More than 90% of esophageal cancer cases are ESCC in China [6] .
The incidence and prognosis of ESCC are affected by many different factors. Tobacco smoking, alcohol drinking, environmental carcinogen exposure are the major recognized risk factors for ESCC [7,8] .
Remarkably, a growing body of research has highlighted the key role of nutrition in this cancer [9,10] . Both theoretically and empirically, due to the limitations of a single nutrient or food approach, dietary pattern assessment has become an alternative method for measuring dietary exposure in nutritional epidemiology [11] . In the past several years, dietary guidelines for fat intake have been centered on reducing total fat or saturated fatty acid (SFA) intake. Whereas, recently fat quality has come into focus [12] . Combinations of multiple fatty acids may in uence ESCC risk more than single fatty acids. A direct association between total fat intake and ESCC outcomes is insu cient, so, it is important to consider the intake of fatty acid patterns.
At present, some studies have found that fatty acid patterns are closely related to the risk of certain diseases [13,14] . However, as far as we know, the association of a combination of fatty acids with incident ESCC has not been evaluated. Therefore, the present study aimed to de ne speci c dietary fatty acid patterns and then to investigate the relations between the generated fatty acid patterns and ESCC.

Study Design and Subjects
A hospital-based case-control study was conducted in Fujian province, China. Brie y, a total of 422 cases of ESCC were recruited from Fujian Provincial Cancer Hospital (FPCH) (176 cases), and the First A liated Hospital of Fujian Medical University (246 cases) in Fuzhou City during the period from June 2014 through December 2020 The inclusion criteria for cases were: (1) newly diagnosed primary patients who were histologically or cytologically diagnosed with ESCC (International Classi cation of Diseases 10th revision); (2) all cases were diagnosed with macroscopic type con rmed; (3) Chinese Han population who resided in Fujian Province at least past 10 years. Meanwhile, 423 controls were randomly chosen from community residents ordered health examination, with non-neoplastic conditions. This study was approved by the Institutional Review Board of Fujian Medical University (Fuzhou, China). Written informed consent was obtained from all participants before they participated in the study. All investigations performed in this study were conducted in accordance with the guidelines of the 1975 Declaration of Helsinki. A standard questionnaire was administered to cases and controls by specially trained interviewers. Questions covered demographic characteristics (e.g. gender, age, education level, job, marital status), dietary habits, lifestyle habits such as tobacco smoking and alcohol drinking, personal medical history, family history of cancer, and dietary factors(e.g. pickled food, fried food, hot food, hard food).

Assessment of Dietary Intake
The usual diet was assessed by a food frequency questionnaire (FFQ). The contents of the questionnaire mainly included: 1) grain, 14 items;2) Beans and their products, 8 items;3) Vegetables, 52 items;4) Fruit, 30 items;5) Animal food, 49 items;6) Bacteria, algae, and nuts, 8 items;7) Beverages, drinks, and soups, 10 items. There are seven categories of food in total, and the total number of food items is 171. The dietary data during 1 year before the diagnosis for cases or the year before the interview for controls were selected. Food items from the FFQ are converted into dietary fatty acids. The content of fatty acids and energy reference to 2018 "China Food Composition Tables-Standard edition". To make the data follow the normal distribution, the natural logarithm (LN) conversion was carried out for the total energy and the daily intake of various dietary fatty acids, and the residual method was used to correct the total energy for the average daily intake of various dietary fatty acids [15] . Missing values in the fatty acid data were replaced by the variable mean.

De nition of Variables
Subjects who had smoked at least one cigarette per day, lasting 6 months, were considered tobacco smokers [16] . Subjects who had at least one drink per week for at least 6 months were alcohol drinkers [17] . Tea drinkers were de ned as drank at least one cup of tea per week continuously for more than 6 months [18] . All participants were additionally informed that the temperature of the food was classi ed on a scale ranging from 0 to 10, and ≥ 7 was de ned as hot food. And then participants were asked to estimate the frequency that they burn their mouths when consuming hot food. Hard food was de ned as food that was mechanically irritating and unpleasant to swallow, such as nuts, dried fruit, or ultra-dried meat. As for dietary data, participants were asked how often intake each food item according to the following options: ≥5times per week, 3-4 times per week, 1-2 times per week, less than one time per week, or not at all.

Statistical Analyses
After the investigation of all case and control subjects, the data were recorded using Epidata software (v3.1), with double-entry veri cation. The distributions of demographic characteristics, substance uses (tobacco, alcohol, and tea) of ESCC, and controls were compared by Chi-square test. The Mann-Whitney U test was used to test whether there was any difference in dietary fatty acid content distribution between the case group and control group. For descriptive purposes, we generated a hierarchical cluster tree to visually evaluate the correlations between individual fatty acids [19] . Pearson correlation coe cients between fatty acids were also calculated.
A factor analysis [11] was carried out to derive dietary fatty acid patterns (FAPs). Oblique rotation was used to derive dietary FAPs to obtain a simpler structure with greater interpretability. For deriving FAPs, the percentages of energy from the 36 FAs were used as input variables to adjust for an individual's daily energy intake. Finally, four FAPs were extracted considering eigenvalue (> 1), scree plot, and variance contribution. Participants were grouped into tertiles (T) according to the factor score of each pattern. The lowest score groups of each FAP were used as the reference. Using the Spearman correlation analysis, the associations between each pattern score with the intakes of 4 food groups (cereals, meat, freshwater sh, deep-sea sh), 3 oils (peanut oil, animal oil, blend oil) were assessed.
Three multivariable logistic regression models were applied to further estimate the OR value and 95%CI between dietary fatty acid pattern score and ESCC risk. Model 1 adjusted for gender, age, education level, marital status, family tumor history, occupation, tobacco smoking, alcohol drinking, tea consumption. To further control the in uence of dietary habits on outcomes, hot food, hard food, pickled food, and fried food were included in model-2 additionally. Model 3 adjusted for model 2 and the other three FA scores.
The tting performance of the three models was evaluated by Akaike Information Criterion (AIC). The restricted cubic splines were used to visualize the trend of dietary fatty acid scores with the risk in ESCC.
Finally, we used the forest plot to demonstrate the interaction of latent exposure factors on the potential correlation. All analyses were performed using R 4.0.3 software, with α two−sided = 0.05. Table 1 presents the distribution of demographic characteristics, lifestyle risk factors in patient groups, and controls. Cases and controls had similar distributions of sex, alcohol drinking, and family history of cancer (P > 0.05). However, age, education level, marital status, occupation, income, tobacco smoking, and tea consumption were signi cantly different between these two groups (P < 0.05). S1 Table shows the difference in fatty acid intake between the case group and the control group after adjustment for energy.

The fatty acid-factor loadings of the 4 major factors
The correlation matrix of 36 fatty acids is in Fig. 1. Factor analyses, including 36 major fatty acids, identi ed 4 factors that explained 60.8% of the variation in these variables in the study population. A similar pattern was identi ed in cluster analysis, as fatty acids adjacent in the tree had similar loading values ( Fig. 1). Four FA factor scores were extracted to construct the FAP score of dietary fatty acids. Based on the major contributors to each pattern (Table 2), the most powerful factor comprised mainly positive loadings from 20:0, 16:0, 18:0, 17:0. This factor was called the "LC-SFA" factor. The second most powerful factor comprised positive loadings from EC-UFA such as 22:6, 24:1, 20:5, 20:1, and 20:3. We characterized the third FA pattern as an "SMC-SFA pattern", with high factor loading for 6:0. 10:0, and 4:0.

Correlation between dietary fatty acid scores and food groups
The correlations of the four FACP scores with intakes of food groups and oil groups are shown in the S2 Table. The LC-SFA pattern score was positively correlated to the intakes of "animal oil" (r = 0.124 p = 0.002). The EC-UFA pattern represented a diet relatively high in "peanut oil"(r = 0.189, p < 0.001). The "n-3 LC-PUFA pattern" score was positively correlated to the intakes of "deep-sea sh" (r = 0.099 p = 0.015).
3.4 Association of the fatty acid pattern score with the incidence of ESCC.
The model t performance was evaluated according to the AIC, model 3 had the lowest AIC and the best tting effect. In multivariable analyses, the n-3 LC-PUFA pattern was associated with a lower likelihood of ESCC after adjustment for all covariates (OR: 0.525, 95% CI: 0.340,0.811, P = 0.003). After adjustment for potential confounding variables, the EC-UFA pattern was positively associated with ESCC (OR: 2.069, 95% CI: 1.314,3.257, P = 0.002). The LC-SFA pattern and SMC-SFA pattern were not signi cantly associated with ESCC. (Table 3).

Linear trend of dietary fatty acid score and incidence of ESCC
The dose-response relationship between four FAPs intakes and the risk of ESCC is shown in Fig. 2. There existed a nonlinear positive association between the EC-UFA pattern and the risk of ESCC (p for nonlinearity < 0.05), nevertheless, there was a nonlinear negative association between the n-3 LC-PUFA pattern and the risk of ESCC (p for nonlinearity < 0.001) (Fig. 2).
3.6 Cross-strati ed heterogeneity test between dietary fatty acid score and ESCC incidence After adjusting for potential confounding factors, there was a multiplicative interaction between the LC-SFA pattern and the fried food (P inter action = 0.024). we discovered an increased ESCC associated with the EC-UFA pattern was detected in EC-UFA pattern interacted with hard food (P interaction =0.022), pickled food (P interaction = 0.122), and alcohol drinking (P interaction =0.002) on the development of ESCC at the multiplicative scale of the standard model. Multiplicative interaction between the n-3 LC-PUFA pattern and ESCC risk across alcohol drinking (P interaction = 0.001), pickled food(P interaction 0.001), fried food(P interaction 0.001) were obtained (Fig. 3).

Discussion
In this hospital-based case-control study, four main dietary patterns were identi ed, i.e. long-chain SFA (principal component) [20,21] , while factor analysis as a posteriori method allows the study of synergy among nutrients and consequence of the interactions between them [22] .To our knowledge, there has been no attempt to assess the effect of dietary FAPs on ESCC risk, but some components in the dietary patterns of this study were similar to those of the patterns de ned in some other studies, although not all components were identical. The "LC-SFA pattern" "SMC-SFA pattern", and "n-3 LC-PUFA pattern", were similar to the FAPs from Korea [13] , Uppsala [23] .
The "LC SFA pattern" and "SMC-SFA pattern" in this study-which was characterized by a high intake of SFA-were no signi cantly associated with the risk of ESCC. SFAs are major sources of energy [24] , attenuate weight gain [25] , and have strong antibacterial effects [26] , but there is a lack of evidence for the effect of SFA on ESCC risk. In a case-control study with adults in Iran, higher levels of SFA were associated with a lower risk of ESCC [27] . However, In a study in Korea, the "short & medium-chain SFA pattern" was associated with an increased risk of hyper-LDL cholesterolemia in men [13] . Another study highlighted that Saturated fat intake was associated with higher cancer mortality (highest vs. lowest quintile [Q5 vs. Q1]: HR: 1.26, 95% CI 1.20-1.32) in a prospective cohort study of 521,120 participants, with 16 years of follow-up [28] . High intake of saturated fat (but not total, monounsaturated or polyunsaturated fat intake) was associated with increased risk of breast cancer (Q5 vs. Q1: HR 1.13, 95% CI 1.00-1.27) in a large European multicentre prospective study (519,978 participants) [29] . Then in this study, SMC-SFA and LC-SFA patterns were not found to be associated with ESCC risk.
Nevertheless, those adhering more to the "even-chain UFA pattern" were found to be at a higher risk of ESCC. The "even-chain UFA pattern" had a high factor loading of even-chain UFA such as DHA(C22:6), nervonic acid(C24:1), EPA(C20:5), eicosenoic acid(C20:1), eicosatrienoic acid(C20:3), AA(C20:4). Some studies have shown that dietary unsaturated fatty acids are associated with an increased risk of cancer [30] . Especially, AA in the even chain fatty acid pattern is a precursor to pro-in ammatory molecules [31] . In humans, the even-chain UFA was found to be correlated with composite in ammation measures and may thus in uence the risk of cancer [32] . In ammation is a crux of the development of many chronic diseases, including cancer [33] . An in ammatory microenvironment is an important part of the tumor microenvironment [34] . Chronic in ammation is the cause of tumor transformation [35] . More studies are essential for exploring their association.
The last pattern of n-3 LC-PUFA was characterized by higher intake of DPA(22:5), docosatrienoic acid(22:3). Many previous studies have proved that the n-3 series of unsaturated fatty acids are mainly derived from sh [36] , and our study also found a positive correlation between the fourth FAP and the intake of deep-sea sh. N-3 LC-PUFA pattern, an important fatty acid that may play a role in preventing some cancers [22] . Furthermore, the n-3 LC-PUFA pattern, which has pleiotropic effects and enhances cancer cell apoptosis, modulates various eicosanoid pathways leading to reduced in ammation, such as suppressing cyclooxygenase-2 synthesis and the inhibition of arachidonic acid-derived eicosanoids [37] . Animal studies and human observational studies have demonstrated that the n-3 LC-PUFA pattern may reduce the risk of cancers such as breast, colon, and prostate [37][38][39] . In this study, the n-3 LC-PUFA pattern was also found to reduce the risk of ESCC.
When exploring the linear relationship between dietary fatty acid patterns and the risk of ESCC. As we can be seen from Fig. 2, with the increase of EC-UFAs intake, the risk of ESCC increases. However, as the intake of n-3 LC-PUFAs improved, the risk of ESCC decreased. They were associated with a dose-response risk of ESCC. N-3 polyunsaturated fatty acids (PUFAs) express anti-in ammatory properties and prevent tumor progression [40] , which is similar to our result.
Strati ed by demographic characteristics and life exposure factors, we found the association between dietary fatty acid pattern and ESCC risk could be modi ed by smoking, drinking, pickled food, hard food, and fried food. Pickled food is often preserved with the addition of nitrates or nitrites, which increases the formation of N-nitroso compounds (NOCs), which were considered to be animal carcinogens and possible human carcinogens [41] . In addition, high concentrations of salt may increase ESCC risk. Salt might directly damage the esophageal mucosa, leading to susceptibility to esophagitis and an increased risk of ESCC [42] . In this study, there was heterogeneity in the relationship between dietary FAP and the risk of ESCC with the use of preserved foods, and the protective effect of the n-3 LC-PUFA pattern was reduced with the use of preserved foods.
There have been reports that a signi cant dose-response relationship between the intake frequency of fried food and the risk of ESCC [43] . Cooking meat at high temperatures produces large amounts of polycyclic aromatic hydrocarbons, and also high levels of heterocyclic amines [44] . Both groups of chemicals have been suggested to increase the risk of ESCC [45,46] . After strati cation by fried foods, we found that the protective effect of the n-3 LC-PUFA pattern was weakened in those who regularly consumed fried foods and differed from those who did not regularly consume fried foods.
The role of alcohol use in the etiology of carcinoma of the esophagus is well established [47] . Lots of studies conducted in Kenyan [48] , Japan [49] , have provided further evidence of the close and independent role of alcohol in the etiology of ESCC. Many epidemiological studies have shown that alcohol is associated with tumor suppressor gene promoter hypermethylation and global DNA hypomethylation in several cancers, including esophageal [50] . In this study, there was heterogeneity in the association between the LC-UFA pattern and ESCC, as shown in the previous study, alcohol consumption would increase the risk of pattern 2, and the association was distinct.
To the best of our knowledge, this is the rst study that factor analysis has been used to reveal a causal relationship between dietary FAPs and ESCC risk in the Chinese population. In our daily life, people eat a diet made up of a variety of fatty acids, not just one kind of fatty acid. Therefore, it is important to consider the FAP analysis because it can re ect the actual dietary quality and summarize the effects of various dietary FAs. As compared with the traditional approach of analyzing single FA, factor analysis allows investigating the relationship between dietary habits and cancer accounting for complex interactions between dietary components.
Whereas, several limitations should be acknowledged in our study. Selection bias may exist in any hospital-based case-control study. However, all subjects were recruited from two hospitals according to strict criteria, which may minimize the selection bias. The study data were obtained from interviews and might lead to recall bias which may limit the accuracy of our results. To alleviate this effect, we performed face-to-face interviews and given the de nitions of variables. Notwithstanding these limitations, to our knowledge, this is the rst study to examine the effects of FACPs on the risk of ESCC.

Conclusions
We found that higher dietary intake of even-chain UFA pattern was associated with a higher risk of developing ESCC. On the contrary, a combination of individual fatty acids, characterized by n-3 LC-PUFA pattern, was associated with a lower incidence of ESCC. Further prospective studies with larger sample sizes are needed to con rm this association. Availability of data and materials

Abbreviations
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Consent for publication
All authors read and approved the nal manuscript.

Competing interests
The authors state that there are no con icts of interest to declare.

Authors' contributions
The authors' responsibilities were as follows-HZJ and HCC: designed the study; LZ, LJB and CYM: supervised the data collection; TXW and SJY: analyzed the data; HCC, LZ and LZQ: contributed to the data interpretation and manuscript writing; HZJ: had primary responsibility for the nal content.    Figure 1 The correlation matrix of 36 fatty acids is in Fig 1. Factor analyses, including 36 major fatty acids, identi ed 4 factors that explained 60.8% of the variation in these variables in the study population. A similar pattern was identi ed in cluster analysis, as fatty acids adjacent in the tree had similar loading values (Fig 1).

Figure 2
There existed a nonlinear positive association between the EC-UFA pattern and the risk of ESCC (p for nonlinearity <0.05), nevertheless, there was a nonlinear negative association between the n-3 LC-PUFA pattern and the risk of ESCC (p for nonlinearity <0.001) (Figure 2).