Comparison between compositional data analysis and principal component analysis for identifying dietary patterns...

Dietary patterns play an important role in regulating serum uric acid (SUA) levels in the body. Recently, compositional data analysis (CoDA) has been proposed as an alternative technique in identifying dietary patterns. However, the relative advantages of CoDA, particularly in identifying dietary patterns associated with hyperuricemia have not been investigated. We evaluated and compared CoDA, including compositional principal component analysis (CPCA) and principal balances analysis (PBA), with the most commonly used principal component analysis (PCA) in determining dietary patterns associated with hyperuricemia.

The 3 day 24-h dietary recall method was used to estimate dietary data from 3,954 study participants of the China Health and Nutrition Survey (CHNS). Dietary patterns were constructed using PCA, CPCA and PBA. These methods were compared based on the performance to identify plausible patterns associated with hyperuricemia.

PCA, CPCA and PBA all identified three dietary patterns, with a common "traditional southern Chinese" pattern high in rice and animal-based foods and low in wheat products and dairy. Only this pattern was positively associated with risk of hyperuricemia [PCA: OR (95%CI) = 1.29 (1.15-1.46); CPCA: OR (95%CI) = 1.25 (1.10-1.40); PBA: OR (95%CI) = 1.23 (1.09-1.38)].

All three dietary patterns methods in our study identified that a "traditional southern Chinese" dietary pattern was associated with increased risk of hyperuricemia, suggesting a robust and consistent finding.