An application of partial least squares for identifying dietary patterns in bone health

Abstract

In a large cohort of older women, a mechanism-driven statistical technique for assessing dietary patterns that considers a potential nutrient pathway found two dietary patterns associated with lumbar spine and femoral neck bone mineral density. A "healthy" dietary pattern was observed to be beneficial for bone mineral density.

Introduction: Dietary patterns represent a broader, more realistic representation of how foods are consumed, compared to individual food or nutrient analyses. Partial least-squares (PLS) is a data-reduction technique for identifying dietary patterns that maximizes correlation between foods and nutrients hypothesized to be on the path to disease, is more hypothesis-driven than previous methods, and has not been applied to the study of dietary patterns in relation to bone health.

Methods: Women from the Aberdeen Prospective Osteoporosis Screening Study (2007-2011, n = 2129, age = 66 years (2.2)) provided dietary intake using a food frequency questionnaire; 37 food groups were created. We applied PLS to the 37 food groups and 9 chosen response variables (calcium, potassium, vitamin C, vitamin D, protein, alcohol, magnesium, phosphorus, zinc) to identify dietary patterns associated with bone mineral density (BMD) cross-sectionally. Multivariable regression was used to assess the relationship between the retained dietary patterns and BMD at the lumbar spine and femoral neck, adjusting for age, body mass index, physical activity level, smoking, and national deprivation category.

Results: Five dietary patterns were identified, explaining 25% of the variation in food groups and 77% in the response variables. Two dietary patterns were positively associated with lumbar spine (per unit increase in factor 2: 0.012 g/cm² [95% CI: 0.006, 0.01]; factor 4: 0.007 g/cm² [95% CI: 0.00001, 0.01]) and femoral neck (factor 2: 0.006 g/cm² [95% CI: 0.002, 0.01]; factor 4: 0.008 g/cm² [95% CI: 0.003, 0.01)]) BMD. Dietary pattern 2 was characterized by high intakes of milk, vegetables, fruit and vegetable juices, and wine, and low intakes of processed meats, cheese, biscuits, cakes, puddings, confectionary, sweetened fizzy drinks and spirits while dietary pattern 4 was characterized by high intakes of fruits, red and white meats, and wine, and low intakes of vegetables and sweet spreads.

Conclusion: Our findings using a robust statistical technique provided important support to initiatives focusing on what constitutes a healthy diet and its implications.

Keywords: Bone mineral density; Dietary patterns; Partial least-squares; Postmenopausal women.

Fig. 1

Factor loadings of food groups from dietary patterns significantly associated with lumbar spine or femoral neck bone mineral density derived using partial least-squares

Fig. 2

Mean and standard deviation of lumbar spine and femoral neck bone mineral density for quartiles of dietary pattern factors 2 and 4