Study population
The current study is based on a hospital-based case-control study which was performed during 2000–2012. The design of the study, data collection methods, and primary purposes have been described elsewhere [17]. In brief, only incident cases of CRC were identified and recruited in the cooperation with the [blinded for the review]. Those should be histologically confirmed adenocarcinomas either of colon (ICD-X: C18) or rectum (ICD-X: C20). Only sporadic cancers met eligibility criteria, meaning all CRC cases suspected to be hereditary cancers were excluded (ICD-X: C18.9, D12.6, Q85.8, Z80.0, Z80.3). Additionally, to limit the possibility of genetically determined CRC for the current study, cases younger than 40 years of age were also excluded. Controls were patients admitted to emergency rooms and next hospitalized in the University Hospital in Krakow and the Narutowicz Municipal Hospital in Krakow, Poland. Primarily, the controls were individually matched to cases on age (range +/− 5 years) and sex with the frequency ratio 1:2 meaning we tried to identify two controls for a case being the same sex and at similar age. The study flow is presented on Fig. 1. Among exclusion criteria for both, cases and controls, were: age over 75, cognitive limitations and verbal communication problems which caused reviewers could not proceed a recall, a diagnosis of secondary cancer, or CRC being a metastasis in colon or rectum, and any other than CRC cancer (current or in the past), any type surgery of gastrointestinal tract in the past, current or past diagnosis of chronic gastrointestinal disease, and any other disease requiring dietary limitations (like diabetes, renal failure, hepatic insufficiency), and additionally a presence of prolonged (lasting longer than a month) gastrointestinal symptoms were verified.
The study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving research study participants were approved by the Jagiellonian University Ethical Committee (IRB no 1072.6120.347.2020). Written informed consent was obtained from each study participant.
Data collection and assessment
The data used for the presented study was gathered by standardized questionnaire. Trained interviewers collected information about socio-demographic characteristic, lifetime smoking habits, lifetime physical activity and diet during face-to-face recall. In details, a semi-quantitative validated food frequency questionnaire prepared in the cooperation with the German Cancer Research Centre, Heidelberg, Germany, during the preparation phase for the European Prospective Investigation into Cancer and Nutrition study [18, 19] was implemented to assess dietary habits. In total, 148-dietary items (food and drinks), were used. There were questions about consumption of cereals, dairy products, bread, type and cuts of meat and fish, (including preparation technique), fresh fruit (during summer and winter time separately), salads, fresh and cooked vegetables, rice and pasta, soups, sweets, baked goods, and others. Study participants estimated commonly consumed portion size, for each food and beverage, using standardized photographs, and next, a frequency of consumption was reported. Dietary data covered the period of 1 calendar year. Case-participants were questioned about the year which took place 5 years prior to the onset of gastrointestinal symptoms (if symptoms were present) or prior to the beginning of a diagnostic process. Control patients reported their usual diet which took place 5 years before the interview. Next, to get data on macro and micronutrient intake, the Polish food composition tables were used. As the content of selenium was not available in the primary tables, for the current study information about intake of each dietary item was additionally used to calculate the content of dietary selenium. The concentrations of selenium, including losses on cooking, for each dietary item assessed, have been presented in the additional materials (see Supplementary materials).
Covariates
Several covariates which might contribute to colorectal cancer odds have been collected. These included adult lifetime leisure time physical activity, which was assessed by a recall as the average weekly time spent on different type activities (including walking or hiking, bicycling, gardening, practicing sports and household activities) during summer and winter season separately. Additionally, respondents were asked about amount of time spent on recreational activities requiring at least moderate effort. Next, to get metabolic equivalents, the reported time for each activity was multiplied by its typical energy expenditure requirement as published in the 2011 Compendium of Physical Activities [20]. Considering participants were not asked about brakes in their activities, it was assumed that 70% of reported time was spent effectively and these was presented and used as a covariate in analyses. Among other covariates respondent’s age, sex, body mass index calculated from the weight and height measured at the admission to a clinic or hospital, the exposure to cigarette smoke (categorized as non-smoker, past smoker, current smoker) were considered. Moreover, the set of dietary covariates were used, and these included total dietary fibres, intake of dietary vitamin c and vitamin e, fish consumption (categorized yes/no; “yes” for participants reported consumption of any fish, having a portion size of at least 20 g of canned fish or 45 g of cooked fish, monthly), and taking mineral supplements (categorized yes/no).
Statistical analysis
There were two groups included in study analyses, colorectal cancer cases and controls, presented and compared in order to assess the role of dietary selenium. The groups were characterized provided means and standard deviation, and additionally, as majority of variables had skewed distributions (tested by the Shapiro-Wilk test of normality), median and interquartile range was provided in the descriptive part. The study groups were compared using the chi-squared test for categorized variables (all the expected cell values in the analyses fulfilled the assumption of being over 5), and the U-Mann-Whitney test (as the variable distribution in groups compared were skewed). To assess the role of dietary selenium intake on the CRC odds the logistic regression was used. We have decided to run 3 consecutive models: univariable one, next multivariable adjusted for the main confounders as age, sex, body mass index (BMI), average adult lifetime leisure-time physical activity, and alcohol consumption, and smoking. The third model additionally included as covariates main dietary components as total dietary fibres, dietary content of vitamin c and vitamin e, taking mineral supplements, dietary calcium intake, and fish consumption. There were models to assess odds associated with selenium considered as a continuous variable (results presented for an increase by 10 μg/day of dietary selenium) and also across categories of selenium intake (< 60 μg/day – the reference group; 60 to < 80 μg/day and ≥ 80 μg/day). We have decided to use all three models to verify a stability of selenium effect estimates. Finally, to verify the presence of an additional role of calcium, all regression analyses have been repeated across lower (< 1000 mg/day) and higher (≥1000 mg/day) dietary calcium intake subgroups. And, as we observed effects of selenium which varied across different dietary calcium intake levels, the last part of the analyses done assessed the odds of CRC across different categories created by both, the consumption of selenium and consumption of calcium. There were 9 categories created by the cut-offs of 60 μg/day and 80 μg/day of selenium and 1000 mg/day and 1500 mg/day of calcium. Subsequently the logistic regression model adjusted for all considered covariates with the reference category of < 60 μg/day of selenium and < 1000 mg/day of calcium was run. This had been done to show a possible modification effect of calcium in the selenium effect. Missing data has been reported in the descriptive part, and the pair-wise procedure was used in the multivariable analyses.