![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
* Jean Mayer United States Department of Agriculture Human Nutrition Research Center on Aging at Tufts University, Boston, MA and Institute of Environmental Medicine, Division of Nutrition, Karolinska Institute, Stockholm, Sweden
2 To whom correspondence should be addressed. E-mail: pknewby{at}post.harvard.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
0.90, P < 0.0001, for all associations). The patterns derived in this study were similar to those derived in other studies, indicating reproducibility of food patterns across populations. Our study suggests that food patterns measured by the use of confirmatory factor analysis are reproducible over time, and weaker correlations observed may reflect natural changes in eating behavior, the food supply, and/or perceptions of what is considered healthy. Testing whether patterns measured over a long time period can be used in analytic studies is the next step in assessing the validity of this method.
KEY WORDS: • dietary patterns • reproducibility • stability • exploratory factor analysis • confirmatory factor analysis
Over the past several years, the use of dietary pattern methods in nutritional epidemiologic research has increased substantially. Patterning methods consider multiple foods, beverages, and/or nutrients and therefore create dietary variables that more realistically resemble actual eating behavior. As a result, these methods have grown in popularity as a valuable complement to the traditional approach. In addition, research using the traditional approach, which often studies single nutrients or foods, is limited because of collinearity among nutrients (1) and the inability to detect small effects from single nutrients (2).
Studies using empirically derived patterns are based only on dietary intake data and therefore are better poised to provide an understanding of actual diets. Factor analysis is a data aggregation procedure used to reduce dietary data into meaningful food patterns based on intercorrelations between dietary items. The factors are then named, usually according to those foods that most heavily contribute to the pattern, and the patterns can then be used as the primary exposure variables in dietary studies. Exploratory factor analysis does not require a theoretical basis and uses only the data to derive food patterns empirically; most applications in nutritional epidemiology use principal components analysis. Confirmatory factor analysis, by contrast, may be guided both by results from an exploratory factor analysis and by a priori knowledge of nutritional behavior.
In the past decade, many more researchers have begun to use patterning methods, and factor analysis is the most commonly used method (3–5). However, several fundamental methodologic questions still require investigation. First, although >50 nutritional studies have used factor analysis (5), there are limited data on the reproducibility of this method (6–8). Studying the reproducibility of patterns derived by the use of factor analysis is an important step in establishing the validity of this method. In addition, none of these studies examined the long-term reproducibility, or stability, of patterns over time. Understanding whether empirically derived patterns are stable over time will improve our comprehension of factor analysis using dietary data, hence its application in nutritional epidemiologic research.
A second important question in dietary pattern research is the utility of confirmatory factor analysis. This method is intuitively appealing because it may be based in theory and also reduces some of the subjectivity involved in exploratory procedures. However, few studies have used confirmatory factor analysis in nutritional epidemiology (8–11). More research is required to understand how this method is used in nutritional epidemiology, as well as how solutions derived from the use of confirmatory factor analysis differ from those using exploratory factor analysis.
The main goal of this study was to further our knowledge of factor analysis methods in nutritional epidemiologic research. Our primary objective was to explore the stability of food patterns derived from the use of confirmatory factor analysis. Our secondary objective was to compare factor solutions from confirmatory factor analysis with those derived from exploratory factor analysis.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
Data analyzed were from the Swedish Mammography Cohort (SMC),3 a population-based mammography screening program introduced in 1987–1990 to all women born between 1914 and 1948 in Västmanland and Uppsala counties in central Sweden (hereafter referred to as 1987).
Of the women invited to the screening, 66,651 (74%) agreed to participate in the study, as described in detail elsewhere (12). A second questionnaire was mailed to 58,055 participants in the SMC in October 1997, of whom 38,894 (67%) responded. The study was approved by the ethics committees of Karolinska Institute and Uppsala University Hospital.
To obtain a healthy cohort at baseline, we first excluded women with a cardiovascular condition of angina, diabetes, coronary disease, stroke, or a previous cancer diagnosis (n = 2431). Cardiovascular conditions were based on hospital diagnoses obtained from the Swedish Inpatient Registry, and cancer diagnoses were obtained from the Swedish Cancer Registry. Of the remaining study participants, 1749 study participants with missing height or weight data and 160 participants with implausible values for weight, height, or BMI were excluded. The final exclusions were based on completion of the food frequency questionnaire (FFQ) (further described in the next section). Specifically, participants who did not complete at least 75% of the food item questions on either the 1987 or the 1997 questionnaire (n = 373) or participants who reported implausible total daily food intakes (<2,508 kJ/d or >16,720 kJ/d) (n = 431) were excluded. After all exclusions, 33,840 women were available for analysis.
Dietary assessment
Diet was assessed with a FFQ at both time points. The first FFQ contained 67 food items, and participants were asked how often, on average, they had consumed the foods in the past 6 months. For several items, standard household serving sizes were used (e.g., a slice of bread, a glass of milk). The majority of food items were measured using 8 frequency consumption categories, ranging from "never/seldom" to "4 or more times per day"; for these foods, no serving size was specified. The remaining 6 items asked qualitative questions concerning the types of bread consumed and fats used on breads and in cooking; these items were not included in our analysis because we were measuring food intakes quantitatively in servings per day. Further, questions about frequency of bread and fat consumption also appeared in the food frequency section, thus we were still able to include these foods in the 1987 food groups.
The FFQ used in 1997 included 97 food items. Intakes were assessed using frequency categories, as in 1987. Slight modifications to the consumption categories were made (e.g., the first category was "0 times per month" rather than "seldom/never," and the highest category was "3 or more times per day" rather than "4 or more times per day."). As in 1987, qualitative questions about types of fats (e.g., butter, margarine) used in dressing, cooking, and sandwiches (e.g., whether spreads of butter were "thick" or "thin") were also asked. However, unlike in 1987, food items such as butter and margarine were not individually quantified in the food frequency section, thus we did not include these foods in the 1997 food groups.
Food intakes from both questionnaires were converted from the 8 frequency consumption categories to daily servings. Foods were then grouped on the basis of fat and fiber content, culinary use, and previous food grouping schemes used in the SMC (13) and in a study of food patterns and BMI used in the Baltimore Longitudinal Study of Aging (14) (Appendix). When possible, we also tried to create similar food grouping schemes to ensure comparability of measurements. For example, we included nuts in the salty snacks group for both FFQs because nuts were included together with other snack foods in 1987 and thus could not be separated into their own group. Also, questions about pizza, soy foods, and sauces (e.g., crème fraiche, mayonnaise) were added to the FFQ in 1997; thus, these food groups could not be measured in 1987. In total, 29 food groups were created for 1987, and 32 food groups were created for 1997. Each food item and food group was checked for plausibility, and unreasonable intakes (>3 SD above the mean) were recoded to the highest plausible value for a given food or food group. Energy and nutrient intakes were derived from reported food intakes, in which we used age-specific serving sizes based on a mean value of 5922 d of weighed food records kept by 213 women randomly selected from the SMC study population. Nutrients were calculated using the database from the Swedish Food Administration.
Statistical analysis
Food patterns were derived by use of both exploratory and confirmatory factor analysis at each time point, with each set of food groups being used as input variables. Both procedures are described further below.
Exploratory factor analysis. Exploratory factor analysis was performed according to the PROC FACTOR procedure in SAS (version 8.2, SAS Institute), which uses principal components; we derived uncorrelated factors using orthogonal rotation (varimax option in SAS). Solutions for 2 through 8 factors were derived and rotated, and the scree plots and the factors themselves were observed to see which solution was most meaningful, in light of both the patterns themselves and previous literature (4,5). We repeated the same procedure at both time points, and in both cases a 6-factor solution was selected for further analysis. Patterns were named according to those foods that had the highest factor loadings, as well as the general nutritional content of foods that loaded highly on the pattern.
Confirmatory factor analysis. Confirmatory factor analysis is used when researchers have a working hypothesis about the underlying factor structure (in this case, food patterns). The hypothesis may be derived from prior nutritional knowledge or from exploratory analyses. In this study, we tried to confirm factors that were suggested both by the exploratory factor analysis and by our understanding of the current literature (5).
In confirmatory factor analysis, the observed covariance matrix, which takes into account the correlations, means, and variances of all food groups, is used to calculate confirmed factor loadings. Explored factors from each time point were confirmed using LISREL software (version 8.54, Scientific Software International, Inc). In contrast to exploratory analysis, only food groups with factor loadings above a defined cut-point are retained to measure a food pattern in order to isolate the "core" of the pattern. We tested each of the explored factors in separate confirmatory models, in which food groups with factor loadings |0.20| from the exploratory analysis were entered as independent terms. For example, we entered the processed meat, meat, liver, refined grains, legumes, potatoes, eggs, reduced-fat dairy, and cereal variables into the model to test the Western/Swedish factor in 1987. Schumacher and Lomax (15) recommend that only factors with >3 dominant factor loadings should be confirmed, so we did not attempt to confirm the High-fat Dairy or coffee patterns derived in 1987 or the Reduced-fat sauces or Grains/High-fat dairy/Coffee patterns derived in 1997. Thus, we attempted to confirm the Healthy, Western/Swedish, Alcohol, and Sweets pattern at each time point.
Because factor loadings for a given variable on one particular factor are affected by how that variable is correlated with the other factors, we tested models simultaneously. Models for the 4 major food patterns were built by first including all variables from the exploratory factor analysis with a factor loading |0.20|. Next, decisions must be made as to which variables should be retained in each confirmed factor. We decided to retain all food group items with a factor loading |0.15| for any of the confirmed factors to retain the significant core of the pattern, because several food groups with factor loadings <|0.20| were consistent with our a priori knowledge of extant food patterns and appeared reasonable. For example, we elected to retain cereal and whole grains in the Healthy pattern in 1997.
For both explored and confirmed factors, factor scores were calculated for each participant for each of the factors at each time point. The standardized intakes of each of the food groups (mean = 0, SD = 1) were weighted by their factor loadings and summed (16). Thus, each individual received a factor score for 4 unique food patterns (i.e., explored in 1987, explored in 1997, confirmed in 1987, and confirmed in 1997).
Additional analyses.
To better understand the stability of eating habits, we calculated mean ± SD intakes for each of the food groups at both time points and calculated Spearman correlation coefficients to compare intakes, because this test does not assume linearity of the variables, and food groups are not usually normally distributed. We also used Spearman correlations to examine the associations of factors with nutrient and total energy intakes. Pearson correlation coefficients were used to examine the stability of food patterns by comparing the factors in 1987 with those derived in 1997. In secondary analyses, we compared how well the explored factors correlated with the confirmed factors at both time points (e.g., whether the Healthy pattern measured using exploratory factor analysis was correlated with the Healthy pattern measured using confirmatory factor analysis). We calculated 2-sided P values, and 
was set at 0.05 for all analyses. All statistical analyses were conducted using SAS (version 8.2, SAS Institute).
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
|
|
|
|
0.90 for all, P < 0.0001; data not shown).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
Our study thus raises interesting questions about the long-term stability of eating habits over time. The loading of soy products on the Healthy pattern in 1997, but not in 1987, provides an interesting example. First, soy products could not load on the pattern in 1987, given that these foods were not queried on the 1987 FFQ; thus, the ability to capture eating patterns is most obviously limited by what data are collected in the primary dietary assessment method, in this case the FFQ. Second, soy products have become more prevalent on supermarket shelves over the past decade; thus, even if included in the 1987 questions, this food group still may not have loaded heavily, given that soy foods were not as readily available hence consumed. Third, women have become increasingly aware of the health benefits of soy products in recent years; thus, these foods would have been more likely to be incorporated into a "healthy" diet in 1997 than 1987, and this is especially true among older women who may have heard about the estrogenic properties of soy (18,19). Therefore, changes in the dietary assessment method, changes in the food supply, changes in diet over the lifespan, and changes in what is perceived as "healthy" all contribute to natural and expected changes in dietary patterns. We observed stronger stability correlations for the Alcohol and Sweets patterns over the time period, perhaps because what is defined and perceived as "alcohol" (e.g., wine, beer, and liquor) or "sweet" (e.g., baked goods) is more stable over time, and such foods are consistently available for consumption and are often measured in dietary studies.
It is important to further note the potential effect of changes in the dietary assessment method when interpreting changes in food patterns. Specifically, changes in mean food group intakes over time may reflect not only changes in eating behavior but also changes to the FFQ itself. Increasing the number of food items on an FFQ can lead to higher mean estimates of food and nutrient intakes, although questionnaires are still able to discriminate among individuals, and similar associations with disease risk are observed (20). Our study shows higher mean intakes for fruit and vegetables in 1997 compared with 1987, for example, which may reflect that more individual food items were added to the FFQ in 1997 for these groups (see Appendix). However, our results also indicate lower mean intakes in 1997 than in 1987 for food groups in which the number of food items was increased (e.g., whole grains). Likewise, higher mean intakes in 1997 were noted for other food groups that included the same number of items at both time points (e.g., beer). Therefore, although the increase in mean intakes we observed may be somewhat inflated because of changes in measurement, it is likely that most of these changes are real.
Dietary patterns are characterized by nutrient composition as well as by food (group) contributors. Each of the patterns we derived was related differently to nutrient intakes, suggesting that our 4-pattern (confirmed) factor solution yielded different, meaningful patterns. Whereas all of our correlations were highly significant, which was likely a reflection of our large sample size, the observed associations with macro- and micronutrients were in the expected direction and were similar to those seen in other studies. For example, the correlation between our Healthy pattern and fiber (r = 0.41 and r = 0.69 in 1987 and 1997, respectively, P < 0.05 for both) was of similar magnitude to a plant-based pattern observed among men and women in the Baltimore Longitudinal Study of Aging (r = 0.39, P < 0.05) (21), among women in the EPIC-Potsdam Study (r = 0.45, P < 0.05) (22), and among men in the Health Professionals Follow-Up Study (r = 0.41, significance not reported) (6). Correlations between our Healthy pattern and ß-carotene, folate, vitamin C, and vitamin B-6, however, were stronger than those observed in other studies (6,22,23), likely because our pattern was more heavily dominated by plant-based foods. The changes we observed in the correlations between the patterns and nutrients over time are also reasonable, given the changes in the major food contributors to the pattern. Following our example, the Healthy pattern was characterized by fish and poultry in 1987, as well as plant foods, whereas in 1997 this pattern was characterized only by plant foods; these changes explain the stronger correlations with fiber and micronutrients in 1997 compared with 1987.
A secondary aim of our study was to compare solutions from exploratory factor analysis with those from confirmatory factor analysis. We confirmed only patterns that had >3 items with factor loadings |0.20|, thus 4 of 6 patterns were confirmed at both time points. The correlations between the explored and confirmed factors were high (r 0.90 for all patterns) in both 1987 and 1997 and were similar in magnitude to those shown in a population in Denmark (8). Although high correlations are expected because the significant food items in both analyses remain the same, it is not yet clear what benefit confirmatory factor analysis provides over exploratory factor analysis in dietary studies. As mentioned earlier, confirmatory factor analysis is intuitively appealing because patterns can be hypothesized and tested a priori without the use of exploratory factor analysis, thus reducing some of the subjective decisions that must be made in exploratory analysis. Our confirmatory factor analysis was driven both by our exploratory results and by prior knowledge. For example, we included cereals and whole grains in our Healthy pattern model, even though those variables did not load highly in the confirmed analysis, because previous research shows that cereals and whole grains often contribute to a Healthy diet (5), and these items were also major contributors in the explored analysis. It may be that as more is learned about food patterns derived by the use of exploratory factor analysis, researchers may be less reliant on this approach and more able to use extant nutrition knowledge in a confirmatory factor analysis.
An ongoing question of pattern analysis is whether patterns are reproducible across populations. We derived Alcohol and Sweets patterns, and a comprehensive review reported that these patterns are highly reproducible, appearing in 26 and 19 separate populations, respectively (5). Our Healthy pattern was dominated by vegetables and fruit, along with fish in 1987 and legumes and soy products in 1997. Healthy patterns are also commonly reported (9,10,24–31), which sometimes include fish (32–35). The food composition of so-called Healthy patterns may vary according to population differences, food grouping schemes, or both.
Less healthy patterns (e.g., Western, Traditional) are also commonly observed and are usually dominated by meat, processed meat, potatoes, and refined grains(33,34,36–38). Traditional food patterns observed in Danish (39) and Dutch populations (40) also share common elements with the Western pattern. Conversely, whereas the Western-type pattern among Greeks (35) was high in meat, eggs, and potatoes, it also had moderate factor loadings for vegetables and fruit, possibly reflecting the fact that fruit and vegetables are more ubiquitously consumed in Mediterranean diets.
Differences observed in Western patterns likely come from the melding of occidental dietary habits with traditional food habits, which is why we chose to name our Western pattern Western/Swedish. For example, our Western pattern included legumes in 1987, a food that commonly contributes to "healthy" dietary patterns. This may be because one of the food components of the legumes food group was pea soup, which is a traditional Swedish food. On the other hand, legumes were in fact a major contributor to the Healthy pattern in 1997, often possibly reflecting that legumes received more media attention as a healthful food during the study period. Traditional diets often become Westernized over time, just as Healthy diets evolve with current nutrition knowledge. Iizumi and Amemiya (41) observed a tendency for the staple food pattern to become more Westernized over time among Japanese adults, suggesting that evolving Western diets may retain traditional foods, which are often culture-specific.
In conclusion, food patterns measured by the use of confirmatory factor analysis were generally stable, despite expected changes in eating behavior, the food supply, and/or perceptions of what is considered healthy over time. The patterns derived were similar to those described in other studies, suggesting good reproducibility and generalizability across populations. In addition, strong correlations between explored and confirmed factors were observed. Testing whether pattern variables measured over a long time period can be used in analytic studies of diet and health is the next step in assessing the validity of this method.
|
APPENDIX |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
|
1 Information on these food groups was collected at only one time point.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
3 Abbreviations used: FFQ, food frequency questionnaire; SMC, Swedish Mammography Cohort.
Manuscript received 30 September 2005. Initial review completed 26 October 2005. Revision accepted 28 December 2005.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIXLITERATURE CITED |
|---|
1. Wirfalt AK, Jeffery RW. Using cluster analysis to examine dietary patterns: nutrient intakes, gender, and weight status differ across food pattern clusters. J Am Diet Assoc. 1997;97:272–9.[Medline]
2. Sacks FM, Obarzanek E, Windhauser MM, Svetkey LP, Vollmer WM, McCullough M, Karanja N, Lin PH, Steele P, et al. Rationale and design of the Dietary Approaches to Stop Hyptertension Trial (DASH): A multicenter controlled-feeding study of dietary patterns to lower blood pressure. Ann Epidemiol. 1995;5:108–18.[Medline]
3. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol. 2002;13:3–9.[Medline]
4. Kant AK. Dietary patterns and health outcomes. J Am Diet Assoc. 2004;104:615–35.[Medline]
5. Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev. 2004;62:177–203.[Medline]
6. Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, Sampson L, Willett WC. Reproducibility and validity of dietary patterns assessed with a food- frequency questionnaire. Am J Clin Nutr. 1999;69:243–9.
7. Khani BR, Ye W, Terry P, Wolk A. Reproducibility and validity of major dietary patterns among Swedish women assessed with a food-frequency questionnaire. J Nutr. 2004;134:1541–5.
8. Togo P, Heitmann BL, Sorensen TI, Osler M. Consistency of food intake factors by different dietary assessment methods and population groups. Br J Nutr. 2003;90:667–78.[Medline]
9. Tseng M, Breslow RA, DeVellis RF, Ziegler RG. Dietary patterns and prostate cancer risk in the National Health and Nutrition Examination Survey Epidemiological Follow-up Study cohort. Cancer Epidemiol Biomarkers Prev. 2004;13:71–7.
10. Togo P, Osler M, Sorensen TI, Heitmann BL. A longitudinal study of food intake patterns and obesity in adult Danish men and women. Int J Obes Relat Metab Disord. 2004;28:583–93.[Medline]
11. Maskarinec G, Novotny R, Tasaki K. Dietary patterns are associated with body mass index in multiethnic women. J Nutr. 2000;130:3068–72.
12. Michels KB, Holmberg L, Bergkvist L, Wolk A. Coffee, tea, and caffeine consumption and breast cancer incidence in a cohort of Swedish women. Ann Epidemiol. 2002;12:21–6.[Medline]
13. Terry P, Suzuki R, Hu FB, Wolk A. A prospective study of major dietary patterns and the risk of breast cancer. Cancer Epidemiol Biomarkers Prev. 2001;10:1281–5.
14. Newby PK, Muller D, Hallfrisch J, Qiao N, Andres R, Tucker KL. Dietary patterns and changes in body mass index and waist circumference in adults. Am J Clin Nutr. 2003;77:1417–25.
15. Schumacher RE, Lomax RG. A beginner's guide to structural equation modeling. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., 1996.
16. Kim J-O, Mueller C. Factor analysis: statistical methods and practical issues. Newbury Park, CA: Sage Publications, Inc., 1978.
17. Terry P, Hu FB, Hansen H, Wolk A. Prospective study of major dietary patterns and colorectal cancer risk in women. Am J Epidemiol. 2001;154:1143–9.
18. Adlercreutz H, Mazur W. Phyto-oestrogens and Western diseases. Ann Med. 1997;29:95–120.[Medline]
19. Knight DC, Eden JA. A review of the clinical effects of phytoestrogens. Obstet Gynecol. 1996;87:897–904.[Abstract]
20. Willett WC. Nutritional epidemiology. 2nd ed. New York: Oxford University Press, 1998.
21. Newby PK, Muller D, Hallfrisch J, Andres R, Tucker KL. Food patterns measured by factor analysis and anthropometric changes in adults. Am J Clin Nutr. 2004;80:504–13.
22. Schulze MB, Hoffmann K, Kroke A, Boeing H. Dietary patterns and their association with food and nutrient intake in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study. Br J Nutr. 2001;85:363–73.[Medline]
23. McCann SE, Marshall JR, Brasure JR, Graham S, Freudenheim JL. Analysis of patterns of food intake in nutritional epidemiology: food classification in principal components analysis and the subsequent impact on estimates for endometrial cancer. Public Health Nutr. 2001;4:989–97.[Medline]
24. Tseng M, DeVellis RF. Fundamental dietary patterns and their correlates among US whites. J Am Diet Assoc. 2001;101:929–32.[Medline]
25. Masaki M, Sugimori H, Nakamura K, Tadera M. Dietary patterns and stomach cancer among middle-aged male workers in Tokyo. Asian Pac J Cancer Prev. 2003;4:61–6.[Medline]
26. Kumagai S, Shibata H, Watanabe S, Suzuki T, Haga H. Effect of food intake pattern on all-cause mortality in the community elderly: a 7-year longitudinal study. J Nutr Health Aging. 1999;3:29–33.[Medline]
27. Gerdes LU, Bronnum-Hansen H, Osler M, Madsen M, Jorgensen T, Schroll M. Trends in lifestyle coronary risk factors in the Danish MONICA population 1982–1992. Public Health. 2002;116:81–8.
28. Schulze MB, Hoffmann K, Kroke A, Boeing H. Risk of hypertension among women in the EPIC-Potsdam Study: comparison of relative risk estimates for exploratory and hypothesis-oriented dietary patterns. Am J Epidemiol. 2003;158:365–73.
29. Markaki I, Linos D, Linos A. The influence of dietary patterns on the development of thyroid cancer. Eur J Cancer. 2003;39:1912–9.[Medline]
30. Handa K, Kreiger N. Diet patterns and the risk of renal cell carcinoma. Public Health Nutr. 2002;5:757–67.[Medline]
31. Whichelow MJ, Prevost AT. Dietary patterns and their associations with demographic, lifestyle and health variables in a random sample of British adults. Br J Nutr. 1996;76:17–30.[Medline]
32. Williams DE, Prevost AT, Whichelow MJ, Cox BD, Day NE, Wareham NJ. A cross-sectional study of dietary patterns with glucose intolerance and other features of the metabolic syndrome. Br J Nutr. 2000;83:257–66.[Medline]
33. Fung TT, Willett WC, Stampfer MJ, Manson JE, Hu FB. Dietary patterns and the risk of coronary heart disease in women. Arch Intern Med. 2001;161:1857–62.
34. Slattery ML, Boucher KM, Caan BJ, Potter JD, Ma KN. Eating patterns and risk of colon cancer. Am J Epidemiol. 1998;148:4–16.
35. Costacou T, Bamia C, Ferrari P, Riboli E, Trichopoulos D, Trichopoulou A. Tracing the Mediterranean diet through principal components and cluster analyses in the Greek population. Eur J Clin Nutr. 2003;57:1378–85.[Medline]
36. Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D, Willett WC. Prospective study of major dietary patterns and risk of coronary heart disease in men. Am J Clin Nutr. 2000;72:912–21.
37. Kerver JM, Yang EJ, Bianchi L, Song WO. Dietary patterns associated with risk factors for cardiovascular disease in healthy US adults. Am J Clin Nutr. 2003;78:1103–10.
38. Sanchez-Villegas A, Delgado-Rodriguez M, Martinez-Gonzalez MA, De Irala-Estevez J. Gender, age, socio-demographic and lifestyle factors associated with major dietary patterns in the Spanish Project SUN (Seguimiento Universidad de Navarra). Eur J Clin Nutr. 2003;57:285–92.[Medline]
39. Togo P, Osler M, Sorensen TI, Heitmann BL. Food intake patterns and body mass index in observational studies. Int J Obes Relat Metab Disord. 2001;25:1741–51.[Medline]
40. van Dam RM, Grievink L, Ocke MC, Feskens EJM. Patterns of food consumption and risk factors for cardiovascular disease in the general Dutch population. Am J Clin Nutr. 2003;77:1156–63.
41. Iizumi H, Amemiya T. Eleven-year follow-up of changes in individuals' food consumption patterns. Int J Vitam Nutr Res. 1986;56:399–409.[Medline]
This article has been cited by other articles:
|
|
 |
|
  S. A. McNaughton, K. Ball, G. D. Mishra, and D. A. Crawford Dietary Patterns of Adolescents and Risk of Obesity and Hypertension J. Nutr., February 1, 2008; 138(2): 364 - 370. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Li and Y. Wang Tracking of Dietary Intake Patterns Is Associated with Baseline Characteristics of Urban Low-Income African-American Adolescents J. Nutr., January 1, 2008; 138(1): 94 - 100. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Akesson, C. Weismayer, P. K. Newby, and A. Wolk Combined Effect of Low-Risk Dietary and Lifestyle Behaviors in Primary Prevention of Myocardial Infarction in Women Arch Intern Med, October 22, 2007; 167(19): 2122 - 2127. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E Kesse, F Clavel-Chapelon, and M. Boutron-Ruault Dietary Patterns and Risk of Colorectal Tumors: A Cohort of French Women of the National Education System (E3N) Am. J. Epidemiol., December 1, 2006; 164(11): 1085 - 1093. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. K. Newby, C. Weismayer, A. Akesson, K. L. Tucker, and A. Wolk Longitudinal Changes in Food Patterns Predict Changes in Weight and Body Mass Index and the Effects Are Greatest in Obese Women J. Nutr., October 1, 2006; 136(10): 2580 - 2587. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1.5% fat), light sour milk/yogurt, light hard cheese (10–17%)