Dietary Pattern in Relation With CRC Risk Among Moroccan Population; A Multicenter Case-Control Study
Background: Colorectal cancer (CRC) is a global public health problem, an estimated of 1.4 million cases were diagnosed worldwide in 2012. Studies in health and nutrition confirmed that dietary factors were strongly associated with CRC risk. Aim: The objective of this empirically study was to reveal unobserved dietary profiles that were associated favorably or unfavorably with CRC risk in Moroccan study population. Methods: This case-control study included a total of 2906 participants in five centers, 1453 cases and 1453 controls, and was gender, age and center matched. Statistical exploratory data reduction methods were performed in this study population based on a specific scientific hypothesis linking dietary behavior and colorectal cancer risk. Principal component analysis (PCA) was applied separately in cases and in controls as individuals and with FFQ nutritional group's heads items as variables. The correlation matrix of food variables was examined to explain most of the variation in the data, reducing a large number of food variables to a smaller set that captures the major dietary factors differences in Moroccan population. Results: Three alimentary profiles were identified for controls based on three principal component analysis, which the first one was highly positive with high cereals, fruits and nuts, legumes, fish, olive oil, dairy products and legumes consumption, and was highly negative with an increasing consumption of poultry and red meat. This component explained 26.5% of the variance in initial data and described a healthy pattern characterized with high fiber intake. In opposite, five principal components were identified for cases that indicated five nutritional profiles with a predominance of dairy products, nuts, fish consumption and low legumes, olive oil and fruits intake: its explained 15.37% of total variance. Conclusion: PCA analysis is a multidimensional factor analysis method that was used in this epidemiologic study to describe the variance in our big database in relation with CRC risk among Moroccan people. This method needs a supervised analysis such as linear discriminant analysis (LDA) to give interpretation and prediction models of CRC risk related to nutritional behavior among this study population.