The Association between Diet and Diabetes: Comparing Dietary Factors Derived by Principal Component Analysis and Reduced Rank Regression
Open AccessBackground: As a major nutrition-related disease, the association between Type 2 diabetes and diet has been examined extensively using the dietary pattern approach. Previous studies identified dietary patterns (also referred to as dietary factors) on the food group- or the nutrient-level using either principal component analysis (PCA) or reduced rank regression (RRR). However, none of these studies explored the relation between dietary patterns on different levels or used both PCA and RRR to compare dietary patterns identified in a nationally representative sample of adults in the US. Objective: This study aimed to identify and compare dietary factors on the food group- and nutrient-level using both PCA and RRR in a representative sample of US adults in 2007-2010, and to examine the association between these dietary factors and diabetes status in 2011-2014.Methods: Two 24-hour dietary recalls were used to collect dietary and supplement data on adults aged 20 years and over in the National Health and Nutrition Examination Survey. Diabetes was defined as self-reported diagnosis, or positive fasting blood glucose or HbA1c test. Nutrient-and-phytochemical-based dietary factors were derived using PCA (NP-PCA), and food-group-based dietary factors were derived using PCA (FG-PCA) and RRR (FG-RRR). Correlation between dietary factors was evaluated using Spearman’s rank correlation coefficient (rs). The association between dietary factors and diabetes status was examined using Poisson regression adjusting for potential confounders. The complex survey design was incorporated in all analyses.Results: Six dietary factors were identified and retained. NP-PCA1, characterized by high intakes of B group vitamins and minerals, showed an inverse trend with diabetes status, but the association was not statistically significant (PR 0.83, 95% CI 0.58-1.18) when comparing extreme factor score quintiles. NP-PCA2, characterized by high intakes of protein, fatty acids, cholesterol and low intake of carbohydrates, was associated with increased prevalence of diabetes (PR 1.99, 95% CI 1.22-3.24). FG-PCA1, characterized by high intakes of fruits, vegetables, whole grains, nuts and seeds, dietary supplement, and low intake of solid fats, was associated with lower prevalence of diabetes (PR 0.67, 95% CI 0.47-0.95). FG-PCA2, characterized by high intakes of whole grains, cured meat, solid fats, non low-fat milk, dietary supplement, and low intakes of vegetables and legumes, was not associated with diabetes status. FG-RRR1, characterized by high intakes of eggs, nuts and seeds, oils and solid fats, was associated with higher prevalence of diabetes (PR 1.62, 95% CI 1.07-2.45). FG-RRR2, characterized by high intakes of fruits, whole grains, nuts and seeds, low-fat milk, and dietary supplement, was associated with reduced diabetes prevalence (PR 0.63, 95% CI 0.45-0.89). When all PCA and RRR factors were included in the same model, only the detrimental effect of FG-RRR1 remained statistically significant (PR 1.44, 95% CI 1.02-2.04). Nutrients loaded on the NP-PCA factors were indicative of food groups loaded highly on the FG factors. However, the correlations between NP-PCA factors with FG-RRR factors were higher than that with FG-PCA factors (rs 0.51-0.74 and -0.11-0.36, respectively). Among food groups involved, fruits, whole grains, nuts and seeds, and solid fats had high loadings on both FG-PCA and FG-RRR factors.Conclusion: Among the six dietary factors identified in this study, FG-RRR factors performed the best in predicting diabetes prevalence, followed by NP-PCA factors and FG-PCA factors. The decision on which method to use should be based on the purpose and design of the study. NP-PCA may be preferred to the other two methods in studies exploring the nutrient mechanism related to diet and diabetes. FG-RRR may be preferred in studies identifying beneficial or adverse diet composition associated with diabetes. When possible, FG-RRR and FG-PCA should be used concurrently, as the comparison of these factors help identify food groups that are involved in diabetes etiology as well as consumption behaviors. Identified food groups may also be the primary targets in public health recommendations regarding healthy diet.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.