Researchers have previously looked at the relationship between individual foods and nutrients and health. Today, there is interest in looking at how combinations of foods and beverages intake, or dietary patterns, influence health by applying different scientific methods. A statistical method called reduced rank regression analysis can be used to describe the patterns of foods and beverages people eat based on a set of “response variables” that are known to be related to the health outcome of interest. This summary of a NEL review presents what research evidence currently exist when reduced rank regression analysis is the method used to study the dietary patterns of groups of people and their likelihood of becoming overweight and risking obesity.

Conclusion

There were a number of methodological differences among the studies examining the relationship between dietary patterns derived using reduced rank regression and body weight status. The disparate nature of these studies made it difficult to compare results, and therefore, no conclusions were drawn.

What the Research Says

Six studies looked at dietary patterns found using reduced rank regression analysis and the risk of becoming over weight and obese. However, these studies had some key issues that make it hard to make any recommendations:

• There were few studies available.
• There were many differences in how the studies were done.
• The populations studied were different between studies.

Background

The goal of this systematic review project is to identify patterns of food and beverage intake that promote health and prevent disease. Historically, most dietary guidance has been based on research conducted on individual food components or nutrients. Dietary patterns are defined as the quantities, proportions, variety, or combination of different foods, drinks, and nutrients (when available) in diets, and the frequency with which they are habitually consumed. Reduced rank regression (RRR) is a statistical method that determines dietary patterns (combinations of food intake) that explain as much variation as possible among a set of response variables related to a health outcome of interest. It is an a posteriori method since it uses both existing evidence and exploratory statistics. The objective of this systematic review was to assess the relationship between patterns of food and beverage intake derived by using reduced rank regression, and body weight or risk for developing obesity.

Conclusion Statement

There are a number of methodological differences among the studies examining the relationship between dietary patterns derived using reduced rank regression and body weight status. The disparate nature of these studies made it difficult to compare results, and therefore, no conclusions were drawn. 

(Grade: Not Assignable)

Methods

Literature searches were conducted using PubMed, Embase, Navigator (BIOSIS, CAB Abstracts, and Food Science and Technology Abstracts), and Cochrane databases to identify studies that evaluated the association between dietary patterns derived using reduced rank regression analysis and risk of obesity. Studies that met the following criteria were included in the review: Human subjects; Ages: 2 years and older; Populations: Healthy and those with elevated chronic disease risk; individuals with chronic disease; published in English in a peer-reviewed journal; Sample size: Minimum of 30 subjects per study arm; Dropout rate less than 20 percent; Study assesses dietary intake using reduced rank regression analysis; study considered body weight and risks of obesity; subjects from countries with high or very high human development (based on the 2011 Human Development Index).

The date range for the conduct of the studies was unlimited. The results of each included study were summarized in evidence worksheets (including a study quality rating) and evidence table. A group of subject matter experts were involved in a qualitative synthesis of the body of evidence, development of a conclusion statement, and assessment of the strength of the evidence (grade) using pre-established criteria including evaluation of the quality, quantity and consistency, magnitude of effect, and generalizability of available evidence.

Findings

• There were six prospective cohort studies that examined dietary patterns derived using reduced rank regression analysis and their association with body weight and risks of obesity. The studies ranged in size from 141 to 24,958 subjects. Three studies were conducted in children, two in adults, and four of the studies included both females and males, while one study included only girls. One study each was conducted in the United States, Korea, United Kingdom, Iran, and Germany. The follow-up for these studies ranged in duration from 22 months to 8 years.
• The response variables and dietary assessment methods used varied widely by study and did not allow conclusions to be drawn across studies. Dietary patterns are based only on foods that are actually consumed, therefore results or outcomes from extracted patterns may be only confirmatory.
• Food groups were not examined in relation to weight status in these studies, and many of the studies used “roughly similar response variables” but generated patterns that were varied. Body weight was determined on the basis of observed food component intake, and as such, evaluating food patterns with respect to body weight change is highly dependent on the preceding analysis of carbohydrate, fat, and fiber density. Thus, results for these food patterns need replication using independent data sets.
• There was lack of consistency in the extracted patterns resulting in a mix of food groups within extracted patterns that are known to be harmful or those known to be protective. This may present limitations in attempting to address issues related to whether food groups within these patterns may act independent of each other (i.e., what role would whole grains and vegetables play in a generated pattern that also contains high fat?)
• The populations studied also varied widely by country of study, as well as region of the world. Diets consumed by these populations also varied by country making it difficult to translate and generalize the findings.

Discussion

The ability to draw a gradable conclusion was limited due to the following issues:

• Each study used different response variables in the reduced rank regression analyses. Two studies used biomarkers including change in BMI, percent body fat, bone mineral content, and bone mineral density as response variables. The second study included fat and bone mass as response variables. Four other studies used nutrients: dietary energy density, fiber density, and percent of energy as fat; total fat, carbohydrate, and fiber; and one used fat, PUFA: SFA ratio, calcium, cholesterol, and fiber as response variables. In reduced rank regression, the dietary patterns identified are those that explain the most variation in the response variables chosen. Therefore, because the studies included in this review used different response variables, the dietary patterns derived may not be comparable.
• Different weight-related outcomes were examined across the studies. The most common outcomes considered in at least four studies were body mass index and fat mass or percentage. Two studies examined incidence of overweight or obesity and excess adiposity and one study examined waist circumference. This variability made it difficult to identify themes within this body of evidence.
• Dietary assessment methods were different across the studies. Of the six studies, three used diet, one used 24-hour recalls, another used both a 24-hour recall and a diet record, while the sixth used a food frequency questionnaire. It is unclear what impacts different dietary assessment methods have on the derivation of dietary patterns using reduced rank regression.
• The studies were not consistent in their use of confounders in analyses. In particular, physical activity was not included as a confounder in the analyses in two studies.