Increase Font Size Decrease Font Size View as PDF Print

What is the relationship between adherence to dietary guidelines/recommendations or specific dietary patterns, assessed using reduced rank regression analysis, and measures of body weight or 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; therefore, no conclusions were drawn.
 

Grade

Not Assignable (IV)

 

Key Findings

  • Six positive quality prospective cohort studies that used reduced rank regression to examine the relationship between dietary patterns and body weight status were included in this review. However, differences in methodologies used and populations studied prevented comparison across studies, and conclusions could not be drawn.
  • Further research is needed to examine dietary patterns and body weight status using reduced rank regression, preferably with standardized methods and response variables.


Evidence Summary Overview

Description of the Evidence

Six prospective cohort studies that used reduced rank regression analysis (see Appendix A) to examine the relationship between dietary patterns and body weight were included in this systematic review (Ambrosini, 2012; Johnson, 2008; Noh, 2011; Schulz, 2005; Sherafat-Kazemzadeh, 2010; Wosje, 2010). Ambrosini (2012) and Johnson (2008) represent the UK Avon Longitudinal Study of Parents and Children (ALSPAC) cohort, but due to variations in methodology used and subjects examined, they are described separately in this review. All of the studies were rated as positive quality. Two studies were conducted in the United Kingdom, and one study each was conducted in the United States, Korea, Iran and Germany. The sample sizes for these studies ranged from 141 participants to 24,958 participants (two studies had less than 200 participants, one study had less than 500, one study had less than 1,000, one study had less than 6,500 and one study had more than 24,000). Four studies were conducted in children and two in adults. Five of the studies included both females and males, while one study included only girls (Noh, 2011).

The studies in this review used a range of different dietary assessment methods. Three studies used three-day diet records (Johnson, 2008; Ambrosini, 2012; Wosje, 2010), another used a 24-hour recall and a two-day diet record (Noh, 2011), one used an FFQ (Schulz, 2005), and one used two 24-hour dietary recalls (Sherafat-Kazemzadeh, 2010). The studies also varied in terms of weight-related outcomes examined. Ambrosini (2012) examined fat mass index, excess adiposity, BMI and weight status; Johnson (2008) examined fat mass, fat mass index, BMI, weight status, body fat percentage and excess adiposity; Noh (2011) examine BMI and body fat percentage; Schulz (2005) looked at annual change in body weight; Sherafat-Kazemzadeh (2010) examined BMI, waist circumference and waist-to-hip ratio; and Wosje (2010) examined fat mass.

The independent variables in all six studies were dietary patterns determined with reduced rank regression analysis; however, the response variables used to identify the dietary patterns differed by study. Two studies used biomarkers as response variables (Noh, 2011; Wosje, 2010), and four studies used nutrients as response variables (Ambrosini, 2012; Johnson, 2008; Schulz, 2005; Sherafat-Kazemzadeh, 2010). The response variables used and dietary patterns extracted for each study are described in more detail below.

Evidence Summary Paragraphs

Johnson, 2008 (positive quality) selected dietary energy density (DED), fiber density (FD) and percentage of energy intake from fat as response variables. Three dietary patterns were extracted from a group of children ages five years to nine years (a random sub-sample of the larger ALSPAC cohort) in the United Kingdom. Pattern 1 explained the most variation in the response variables (47%) at ages five years and seven years. Patterns 2 and 3 explained less than 20% of the variation and were not used in subsequent analyses. Pattern 1 at five years of age was characterized by higher intakes of lower fiber bread, crisp and savory snacks, chocolate and confectionery, high-fat milk and cream and cheese and cheese dishes, and lower intakes of fresh fruit, vegetables, boiled or baked potatoes, high-fiber bread and high-fiber breakfast cereal. Pattern 1 at seven years of age was characterized by higher intakes of crisps and savory snacks, chocolate and confectionery, low-fiber bread, biscuits and cakes and processed meat, and lower intakes of fresh fruit, vegetables, high-fiber breakfast cereals, boiled or baked potatoes and high-fiber bread.

Ambrosini, 2012 conducted another study in the ALSPAC cohort, this time using data from the full cohort, collected over a longer follow-up period. Ambrosini (2012) used the same response variables [dietary energy density (DED), fiber density (FD), and percentage of energy intake from fat]. Three dietary patterns were extracted from a group of children ages seven years to 15 years in the United Kingdom. Pattern 1 explained the most variation in the response variables (45%) at all ages. Patterns 2 and 3 explained less than 15% of the variation and were not used in subsequent analyses. Similar to the previous report, Pattern 1 was characterized by higher intakes of chocolate and confectionery, lower fiber bread, cakes and biscuits, crisps and full-fat milk, and lower intakes of fresh fruit, raw/boiled vegetables, high-fiber breakfast cereal, boiled potatoes and high-fiber bread.

Noh, 2011 (positive quality) selected change in BMI, change in percent body fat, change in bone mineral content, and change in bone mineral density as response variables. Four dietary patterns were extracted. Patterns 1 and 2 explained more variation in the response variables (14%) than Patterns 3 and 4, so these two patterns were used in subsequent analyses. Pattern 1 was characterized by higher intakes of eggs and rice, and lower intakes of nuts and seeds, processed meats, potatoes and eastern grains. Pattern 2 was characterized by higher intakes of fruits, nuts and seeds, milk and dairy products, other beverages, eggs, fruit juices and eastern grains, and lower intakes of vegetables mushrooms and kimchi.

Schulz, 2005 (positive quality) selected the nutrient densities (gram per 1MJ) of the dietary variables total fat, total carbohydrates and fiber as response variables. Three dietary patterns were extracted. Pattern 1 explained the most response variation (53% of total variation). Patterns 2 and 3 explained only 21% and 10%, respectively, of the variation and were not further used in subsequent analyses. Pattern 1 included foods such as whole-grain bread, fresh fruit, fruit juices, grains, cereals and raw vegetables.

Sherafat-Kazemzadeh, 2010 (positive quality) selected fat, polyunsaturated to saturated fat ratio, calcium, cholesterol and fiber intake as response variables. Five dietary patterns were extracted, and all five patterns were used in subsequent analyses. Pattern 1 explained 39%, Pattern 2 explained 19%, Pattern 3 explained 13%, Pattern 4 explained 9% and Pattern 5 explained 5% of total variation. Pattern 1 (“traditional pattern”) included sources of hydrogenated and saturated fat, egg, red and processed meat, refined carbohydrates, vegetables and whole grain and starchy vegetables. Pattern 2 (“fiber and PUFA pattern”) included plant oils, starchy vegetables, legumes, other vegetables, salty snacks and fruit and nuts, with negative loadings for dairy products. Pattern 3 (“fiber and dairy pattern”) included fruits and vegetables, dairy and whole grain, as well as negative loadings for plant oil and egg. Pattern 4 (“dairy pattern”) included dairy, egg and plant oil, with negative loadings for saturated and trans fat, refined carbohydrates, vegetables and fruit. Pattern 5 (“egg pattern”) included egg, fruit and salty snacks, with negative loadings for dairy, plant and saturated oil and red meat.

Wosje, 2010 (positive quality) selected fat mass and bone mass as response variables. Two patterns were extracted. Pattern 1 explained 13% to 19% of variation in the response variables, and Pattern 2 explained 11% to 18% of the variation. Pattern 1 included foods such as whole grains, cheese, processed meats, eggs, fried potatoes, discretionary fats and artificially sweetened beverages. Pattern 2 included food such as dark-green vegetables, deep-yellow vegetables and processed meats.

Table 4-C-III-1. Studies Examining What Combinations of Food Intake (assessed using reduced rank regression) explain the most variation in risk of obesity)

Assessment of the Body of Evidence

This review included six positive-quality prospective cohort studies. However, while a sufficient quantity and quality of studies were potentially available, the studies varied substantially in methodology used and populations considered, which resulted in insufficient information from which to draw conclusions about the relationship between dietary patterns derived using reduced rank regression and body weight status.

Limitations of the Evidence

Methodological Differences
  • Each study used different response variables in the reduced rank regression analyses. Two studies used biomarkers as response variables. Noh (2011) used change in BMI, percent body fat, bone mineral content, and bone mineral density as response variables, and Wosje (2010) included fat and bone mass as response variables. Four studies used nutrients as response variables; Ambrosini (2012) and Johnson (2008) used dietary energy density, fiber density and percent of energy as fat; Schulze (2005) used total fat, carbohydrate, and fiber; and Sherafat-Kazemzadeh (2010) used fat, PUFA: SFA, calcium, cholesterol and fiber. 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 were body mass index (Ambrosini, 2012; Johnson, 2008; Noh, 2011; Sherafat-Kazemzadeh, 2010) and fat mass or percentage (Ambrosini, 2012; Johnson, 2008; Noh, 2011; Wosje, 2010). Two studies examined incidence of overweight or obesity and excess adiposity (Ambrosini, 2012; Johnson, 2008). Only one study examined waist circumference (Serafat-Kazemzadeh, 2010). 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 records (Ambrosini, 2012; Johnson, 2008; Wosje, 2010), one used 24-hour recalls (Sherafat-Kazemzadeh, 2010), another used a 24-hour recall and a diet record (Noh, 2011), and one used an FFQ (Schulz, 2005). 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 by Johnson (2008) or Noh (2011).
Population Differences
  • Each study was conducted in a different country (United States, Korea, United Kingdom, Iran and Germany) and represented populations in different regions of the world, which prevented the ability to compare and interpret the results
  • The studies were conducted with different age groups, four with children (Ambrosini, 2012; Johnson, 2008; Noh, 2011; Wosje, 2010) and two with adults (Schulz, 2005; Sherafat-Kazemzadeh, 2010). Even among the studies with children, the age groups were significantly different.

Research Recommendations

More research using reduced rank regression should be conducted. Additionally, standardization in methodology, particularly in response variables used, is needed.

References

  1. Ambrosini DL, Emmett PM, Northstone K, Howe LD, Tilling K and Jebb SA. Identification of a dietary pattern prospectively associated with increased adiposity during childhood and adolescence. International Journal of Obesity. 2012; 36, 1299–1305; doi:10.1038/ijo.2012.127; published online 7 August 2012.
  2. Johnson L, Mander AP, Jones LR, Emmett PM, Jebb SA. Energy-dense, low-fiber, high-fat dietary pattern is associated with increased fatness in childhood. Am J Clin Nutr. 2008 Apr;87(4):846-54. PMID: 18400706.
  3. Noh HY, Song YJ, Lee JE, Joung H, Park MK, Li SJ, Paik HY. Dietary patterns are associated with physical growth among school girls aged 9-11 years. Nutr Res Pract. 2011 Dec;5(6):569-77. Epub 2011 Dec 31. PMID: 22259683.
  4. Schulz M, Nöthlings U, Hoffmann K, Bergmann MM, Boeing H. Identification of a food pattern characterized by high-fiber and low-fat food choices associated with low prospective weight change in the EPIC-Potsdam cohort. J Nutr. 2005 May;135(5):1183-9. PMID: 15867301.
  5. Sherafat-Kazemzadeh R, Egtesadi S, Mirmiran P, Gohari M, Farahani SJ, Esfahani FH, Vafa MR, Hedayati M, Azizi F. Dietary patterns by reduced rank regression predicting changes in obesity indices in a cohort study: Tehran Lipid and Glucose Study. Asia Pac J Clin Nutr. 2010;19(1):22-32. PMID: 20199984.
  6. Wosje KS, Khoury PR, Claytor RP, Copeland KA, Hornung RW, Daniels SR, Kalkwarf HJ. Dietary patterns associated with fat and bone mass in young children. Am J Clin Nutr. 2010 Aug;92(2):294-303. Epub 2010 Jun 2. PMID: 20519562.



Research Design and Implementation
For a summary of the Research Design and Implementation results, click here.