Secondary Logo

Journal Logo

A Prospective Study of Dietary Patterns and Mortality in Chinese Women

Cai, Hui*; Shu, Xiao Ou*; Gao, Yu-Tang; Li, Honglan; Yang, Gong*; Zheng, Wei*

doi: 10.1097/01.ede.0000259967.21114.45
DIET: Original Article
Free

Background: Many foods and nutrients have been suggested to influence life expectancy. However, previous studies have not examined the relationship between dietary patterns and cause-specific mortality. Our study prospectively examines the relationship of dietary patterns with total mortality and cause-specific mortality in a population-based cohort study of Chinese women.

Methods: The Shanghai Women's Health Study is a population-based cohort study of 74,942 women age 40 to 70 years at the time of recruitment (September 1996 to May 2000). Detailed dietary information was collected using a validated, quantitative food frequency questionnaire. The cohort has been followed using a combination of in-person interviews and record linkage with various registries. Dietary patterns, derived from principal component analysis, were examined for their relation to total mortality and cause-specific mortality using Cox regression models.

Results: After an average of 5.7 years of follow-up (423,717 person-years of observation), there were 1565 deaths. We derived 3 major dietary patterns (vegetable-rich, fruit-rich, and meat-rich). The adjusted hazard ratios for the fruit-rich diet were 0.94 (95% CI = 0.89–0.98) for all causes of death and 0.89 (0.81–0.99), 0.79 (0.69–0.91), and 0.51 (0.39–0.65) for death caused by cardiovascular disease, stroke, and diabetes, respectively. The meat-rich diet was associated with increased risk of diabetes (HR = 1.18; 95% CI = 0.98–1.42) and a slightly elevated risk of total mortality.

Conclusion: In general, a fruit-rich diet was related to lower mortality, whereas a meat-rich diet appeared to increase the probability of death.

From the *Department of Medicine, Vanderbilt Epidemiology Center, Vanderbilt University Medical Center and Vanderbilt-Ingram Cancer Center, Nashville, TN; and †Department of Epidemiology, Shanghai Cancer Institute, Shanghai, People's Republic of China.

Submitted 10 July 2006; accepted 1 January 2007.

Supported by the US National Institutes of Health (RO1 CA070867).

Correspondence: Hui Cai, Vanderbilt Epidemiology Center, 6000 MCE, Vanderbilt University, 1215 21st Avenue South, Nashville, TN 37232-8300. E-mail: hui.cai@vanderbilt.edu.

Certain dietary patterns, such as the so-called Mediterranean diet characterized by high intake of fruits and vegetables, moderate to high intake of fish, low intake of saturated fat, and low intake of meat and poultry,1 can help to maintain health and prolong life. Traditionally, investigations of associations between diet and health have focused on single nutrients or foods, such as fruits, vegetables, or red meat.2,3 Even though these studies provide valuable information for specific food items related to the risk of disease and mortality, such one-to-one relationships have limitations. For example, foods are combined in complex ways. In a single meal, people generally consume a combination of meats, vegetables, and drinks. Also, diets may vary day-to-day even though most individuals consume certain dishes or foods more regularly than others. Because dietary variables are often highly intercorrelated, it can be difficult to determine the effects of single dietary components.4 Therefore, research methods that examine the effects of overall diet on human health are important epidemiologic tools.

Dietary pattern analysis has been used to overcome the methodologic limitations of previous dietary studies.5–7 This approach has proved informative and is used increasingly in studies of Western populations,8–13 However, relatively little work has been done on the investigation of dietary patterns and their relationships with health outcomes in non-Western populations. Such investigations may prove important, because dietary traditions and cultural and social norms in different populations lead to distinct dietary patterns.14 The Shanghai Women's Health Study is a large population-based prospective cohort study of Chinese women. In this study, we derived dietary patterns from a baseline food frequency questionnaire (FFQ) using principal component analysis. We then investigated the prospective relationship of these patterns with total mortality and cause-specific mortality.

Back to Top | Article Outline

METHODS

Subjects

The Shanghai Women's Health Study is an ongoing, population-based, prospective study conducted in 7 communities in urban Shanghai. All permanent female residents between 40 and 70 years of age in the study communities (n = 81,170) were recruited between September 1996 and May 2000. In-person interviews were conducted for 74,942 (92%) women during this period. This baseline survey included information on socioeconomic status, living habits, history of chronic disease, physical activity, and dietary habits. A FFQ was administered to assess usual dietary intake over the 12 months before the interview. The major reasons for nonparticipation were refusal (3%), absence during the enrollment period (3%), and other miscellaneous reason such as health, hearing, or speaking problems (2%). Details of the baseline survey have been reported elsewhere.15

Back to Top | Article Outline

Data Collection

The FFQ was based on a similar dietary questionnaire used in previous epidemiologic studies of cancer in Shanghai. A total of 71 foods and food groups were included in the questionnaire, which covered about 86% of commonly consumed foods in urban Shanghai. For each food or food group, subjects were asked how frequently (daily, weekly, monthly, yearly, or never) they consumed the food or food groups over the preceding year, followed by a question on the amount consumed in lians (a unit of weight equal to 50 g) per unit of time. During the baseline survey, approximately 1000 participants in each season were asked for information on the number of months per year each seasonal food was consumed. The seasonal consumption information was then regressed on the age, education, and income of the subgroup of study participants, and the regression coefficients were used to weight the derived consumption of seasonal foods for the whole cohort of women. The validation study indicated that the FFQ can reliably and accurately measure usual intake of major nutrients and food groups among women in Shanghai.16

Since baseline recruitment, the cohort has been followed using a combination of biannual in-person interviews and record linkage with the tumor and death registries maintained by the Shanghai Center for Disease Control and Prevention. The first in-person follow-up for all living cohort members was conducted from 2000 to 2002. Follow-up of disease or death outcomes was completed for 74,764 cohort members, a response rate of 99.8%. The second follow-up survey was launched in May 2002 and completed in December 2004, with a response rate of 99.1%. Only 10 women were lost to follow-up.

Back to Top | Article Outline

Dietary Pattern Derivation

To perform a dietary pattern analysis using principal component analysis,17 we included 71 individual foods or food groups from the FFQ in the analysis as absolute weight in grams per day. Before analysis, all dietary variables were adjusted for energy intake using the residual approach.

We used the PROC FACTOR procedure in SAS (version 9.1; SAS Institute, Cary, NC) to perform the analysis. The procedure uses principal component analysis and orthogonal rotation (the varimax option in SAS) to derive noncorrelated factors and to render results more easily interpretable. To determine the number of factors to retain, we examined both the scree plots and the factors themselves to see which sets of factors most meaningfully described distinct food patterns. From these analyses, 3 main factors were identified. Factor loadings were calculated for each food or food group across the 3 factors. Factors were thereby interpreted as dietary patterns and each pattern was named after the food group with the highest loading (absolute value of loading >0.30). These loadings can be considered correlation coefficients between food groups and dietary patterns; they take values between −1 and +1. We then calculated for each study participant a factor score for each of the 3 factors; the standardized intake of each of the 71 foods or food groups was weighted by its factor loading and summed. The sums were then standardized (mean ± SD = 0 ± 1).

Back to Top | Article Outline

Risk Analysis

Factor scores were used for comparison with other lifestyle factors and for estimating associations with total mortality and cause-specific mortality. These scores were categorized into quartiles based on their distribution in the study population. To determine the association between dietary patterns and main causes of death, we estimated the adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for each quartile compared with the lowest quartile of each dietary pattern, using Cox proportional hazard models. We adjusted for the following potential confounders in the multivariable models: age (4 categories); education (primary school or lower, middle and high school, college and above); marriage status (yes, no); income per person (low, middle, high); smoking status (never, ever); alcohol consumption (never, ever); tea consumption (never, ever); ginseng use (never, ever); physical activity energy expenditure18 including leisure-time physical activity, house work, walking, cycling, etc. (MET-hours/wk, quartile); and body mass index (BMI, kg/m2; 4 categories). We tested linear trends across categories of dietary patterns by modeling the category value of each participant as a continuous variable. These hazard ratios were then used to determine risk in the corresponding dietary patterns. The interaction effect between food patterns and the presence of chronic disease at baseline was also calculated for cause-specific mortality using cross-product terms in the model. In our study, women who died of causes other than the one under study were censored from the analysis at the time of death.

Back to Top | Article Outline

RESULTS

The study included 74,942 women aged 40 to 70 years. During average of 5.7 years of follow-up, there were 1565 deaths. The baseline characteristics of both surviving and deceased women are shown in Table 1. As expected, deceased women were older, less likely to have been married, more likely to have lower levels of education, income and energy expenditure, and a higher rate of ginseng use. Also, they were more likely to be smokers than surviving women, although their levels of BMI and waist-hip ratio were almost the same as surviving women.

TABLE 1

TABLE 1

The scree plot of eigenvalues depicted 3 major dietary patterns; the factor-loading matrices for those dietary patterns are listed in Table 2. The larger the loading of a given food item to the factor, the greater the contribution of that food item to the specific factor. Negative loading indicates a negative association with the factor. The first dietary pattern was heavily loaded with vegetables such as green beans and yard long beans; it was named the “vegetable-rich” diet. The second dietary pattern was loaded mainly with fruits; it was named the “fruit-rich” diet. The third dietary pattern was loaded with meat, poultry, and animal organs; it was named the “meat-rich” diet.

TABLE 2

TABLE 2

Table 3 shows the covariate-adjusted hazard ratios by category of dietary patterns for mortality from all causes and selected causes. A strong inverse association between the fruit-rich diet and all causes of mortality was observed (HR = 0.94 by modeling categories of fruit-rich diet as a continuous variable; 95% CI = 0.89–0.98), with the highest quartile showing a risk reduction of 20% relative to the lowest quartile. An inverse association was also observed for cardiovascular disease, stroke, and diabetes (HR = 0.89, 0.79 and 0.51, respectively). There was a modest positive association between the meat-rich diet and all causes of mortality. A positive association was also found between the meat-rich diet and diabetes (HR = 1.18 by modeling categories of meat-rich diet as a continuous variable; 95% CI = 0.98–1.42), with the highest quartile of meat consumption associated with a more than 72% increase in risk compared with the lowest quartile. No overall association of dietary pattern and cancer mortality was observed (Table 4), although the meat-rich diet was related to an elevated risk of colorectal cancer.

TABLE 3

TABLE 3

TABLE 4

TABLE 4

To evaluate the influence of chronic disease at baseline on these associations, we conducted further analyses stratified by the presence of diabetes, hypertension, coronary heart disease, stroke, and cancer at baseline (Table 5). Generally, the HRs of mortality for women with a chronic disease at baseline were consistent with that of all women in the study, whereas the HRs in the healthy women at baseline were attenuated. There were no interaction effects between the presence of chronic disease at baseline and dietary patterns for most cause-specific mortalities, with the exception of breast cancer.

TABLE 5

TABLE 5

TABLE 5

TABLE 5

To eliminate the possibility that the observed associations were skewed by dietary change resulting from recent diagnosis of chronic disease, we repeated the risk analysis using the Cox proportional hazard model and excluding participants who died within 1 year of the baseline survey (196 deaths) or whose disease was diagnosed within 2 years of the baseline survey. The associations remained unchanged by these exclusions (data not shown).

Back to Top | Article Outline

DISCUSSION

There has been increasing interest in the identification of dietary patterns as an alternative or complementary approach to single-nutrient analysis in relation to risk of disease or death. Dietary patterns are characterized on the basis of habitual food consumption, represent a combination of nutrients and foods, and may be a better predictor of health outcomes than any single nutrient.5

The number and description of dietary patterns identified to date has varied widely. One in particular, the Mediterranean diet, has been evaluated in many studies for its benefits to health.1 Examples of other dietary patterns that are often reported include the animal fat pattern identified by Akin et al19 and the “healthy” pattern identified by Pryer et al20 In our large, population-based, cohort study we identified 3 common dietary patterns found in Chinese women between 40 and 70 years of age, based on a comprehensive and validated FFQ. Our results were similar to those found in the Multiethnic Cohort Study.21 The fruit- and vegetable-rich diets in our study population were similar to the “prudent” patterns in the Nurses’ Health Study,22 and the meat-rich diet was similar to their “western” pattern.

Many epidemiologic studies have examined the association of dietary patterns with demographic or anthropometric characteristics23,24 or with chronic disease.25–27 Some studies have investigated the relationship of the Mediterranean dietary pattern and longevity in Greek, Danish, Australian, and Spanish populations. These studies have all been relatively small scale (including fewer than 400 subjects),28–30 or were conducted exclusively among elderly persons in Western populations.31 Few studies have shown an association of dietary patterns with cause-specific mortality.32

Our study found that a fruit-rich diet was associated with a reduction in all-cause mortality. This reduction in mortality was also evident with respect to death due to both cardiovascular disease and diabetes, although it was slightly more pronounced with respect to the latter. The meat-rich diet in our study showed a weak association with an increase in risk for all-cause mortality and mortality due to diabetes and colorectal cancer. Taking our fruit-rich and vegetable-rich diets together yields a dietary pattern similar to the Mediterranean diet, decreasing the risk of mortality by 6% with each quartile increase. This magnitude of risk reduction or increase associated with the dietary patterns agrees, in general, with findings from other reports on the Mediterranean-type diet, even reports that used different methods for deriving the dietary scores. Studies in Greece and Denmark found a 17%–20% reduction in smoking-adjusted risk of mortality with a 1-unit increase in a Mediterranean diet score.11–20 Similarly, a 13% reduction in age-, smoking-, and alcohol-adjusted risk of mortality in men (n = 3045) in the highest third of a Healthy Diet Index was noted in the Seven Countries Study.12 Conversely, Osler et al13 reported no association of their 4-point Healthy Food Index with all-cause mortality. In addition, several studies in European populations have produced conflicting results regarding the association between mortality and the Mediterranean diet.28

In our study, we found an inverse association between the fruit-rich diet and risk of mortality caused by type 2 diabetes (HR = 0.51) and a positive association between the meat-rich diet and mortality due to this condition (HR = 1.18). Although diet is widely believed to play an important role in the development of type 2 diabetes, its specific mechanisms have not been clearly defined. One possibility is that higher vitamin C intake resulting from high intake of fruits may play a role in the modulation of insulin action. Paolisso et al reported that plasma vitamin C level was associated with higher insulin action in both healthy and diabetic people.33 Another possible protective effect of a high-fruit diet is through the combined action of antioxidants in fruits.34 This combined action has been suggested as a possible reason for the controversial inconsistencies between supplementation trials and observational studies on the health effects of antioxidants. Factors other than dietary antioxidants may also explain the findings. It is possible that individuals with diets high in antioxidants have healthier lifestyles, in general, than other people.35 Hu et al reported a low-glycaemic index diet with a higher amount of fiber products reduces glycaemic and insulinaemic responses and lowers the risk of diabetes.36

The meat-rich diet in our study was associated with mortality from colorectal cancer, especially in women who have been healthy at the baseline survey. This finding is generally in agreement with findings from other studies of dietary patterns and cancer incidence, or of diet index and cancer mortality. Slattery et al37 identified a “Western” pattern (including red meats, processed meats, and fast food) associated with an increased risk of colon cancer in women (OR = 1.49) and a “prudent” pattern (including fruits and vegetables) associated with a reduced risk of colon cancer in women (OR = 0.73). Fung et al38 identified a similar “Western” pattern that was associated with an increased risk of colorectal cancer (RR = 1.46) and a “prudent” pattern of fruits, vegetables, legumes, fish, and poultry that was inversely associated with colon cancer (RR = 0.71) in a large prospective study of nurses in the United States. However, in a prospective study conducted by Terry et al,39 the “Western” dietary pattern was not associated with an increased risk of colorectal cancer. Another multicenter cohort study, the Dietary Patterns and Cancer project, found that the vegetable pattern was generally not associated with colorectal cancer in any cohort. However, a pattern consisting mostly of pork, processed meats, and potatoes was associated with an increased risk of colorectal cancer in a study of Swedish women.25 In our study, the meat-rich diet was associated with an increased risk of mortality from colorectal cancer, and the vegetable-rich diet was associated with a modestly-decreased risk of death from colorectal cancer.

Because mortality reflects both incidence and survival of chronic diseases, the question remains as to whether dietary patterns are important in the etiology or in the prognosis of these diseases. We examined dietary patterns in relation to mortality of specific chronic diseases stratified by the presence of selected chronic diseases at baseline. Except for colorectal cancer, all inverse associations for the fruit-rich diet and positive associations for the meat-rich diet held for all-cause deaths and several cause-specific deaths (such as cardiovascular disease, stroke, coronary heart disease, and diabetes) for women with those chronic diseases at baseline. An interaction between the presence of chronic disease at baseline and the meat-rich diet was found for breast cancer mortality as well, suggesting that fruit- and meat-rich diets are associated with life expectancy or prognosis in women with those diseases. The meat-rich diet may be associated with both the etiology and prognosis of colorectal cancer, because death due to this disease was highly associated with the meat-rich diet in healthy women at baseline survey.

In nutritional epidemiology, dietary patterns may be defined theoretically, in which case foods are grouped according to some a priori criteria of nutritional health. They may also be defined empirically, in which case foods are reduced to a few dietary patterns through statistical manipulation and then evaluated a posteriori. Theoretically derived dietary patterns generally use a dietary index to rank more or less healthy dietary behaviors.11–12 Such structures are built upon current nutritional knowledge or theory, include variables from current nutrition guidelines, recommendations, and specific dietary components, and provide an overall measure of dietary quality. Conflicts can arise, however, when guidelines or recommendations do not have scientific consensus. Also, these patterns often include different foods or different weightings of foods, resulting in indices that measure different definitions of “healthy” behavior. Principal component analysis and cluster analysis are 2 commonly used empirical methods for deriving dietary patterns. Principal component analysis reduces data into patterns based on intercorrelations between dietary items, whereas cluster analysis reduces data into patterns based on individual differences in mean intake. Patterns derived from both factor analysis24 and cluster analysis40 are comparable and are similarly associated with plasma lipids.41 Compared with cluster analysis, the major dietary patterns derived using principal component have reasonable reproducibility and validity.4 The challenge in principal component analysis is that the relationships do not hold true for individuals but rather for a dietary pattern, because factor scores are continuous variables and individuals have scores for each factor. Clusters are arguably easier to handle and interpret in the analysis, because they are mutually exclusive and continuous. However, their reproducibility and validity are not clear.5

Strengths of our study include its prospective nature, its large size, its reliance on a population sample framework, and its use of a validated, comprehensive FFQ. There are also some limitations for this study. Mortality is a complex end point and is strongly influenced by factors such as treatment, screening practices, and severity of disease. Unmeasured variables associated with diet may have confounded our observations or resulted in suboptimal measurement. The average follow-up time was 5.7 years, which limited the statistical power to detect associations because of small numbers for some cancer endpoints. A longer follow-up period would result in a more stable and powerful analysis.

In conclusion, we identified 3 main dietary patterns in a population of Chinese women and found that a fruit-rich diet reduced total mortality and mortality caused by cardiovascular disease and diabetes. Conversely, the meat-rich diet increased the risk of all-cause mortality, and especially mortality caused by diabetes and colorectal cancer.

Back to Top | Article Outline

ACKNOWLEDGMENTS

We thank the participants and staff members of the Shanghai Women's Health Study for their important contributions. We also thank Bethanie Hull for her assistance in manuscript preparation.

Back to Top | Article Outline

REFERENCES

1. Willett WC, Sacks F, Trichopoulou A, et al. Mediterranean diet pyramid: a cultural model for healthy eating. Am J Clin Nutr. 1995;61(suppl 6):S1402–A1406.
2. Vecchia CL, Franceschi S. Nutrition and gastric cancer. Can J Gastroenterol. 2000;14(suppl D):51D–54D.
3. Riboli E, Norat T. Cancer prevention and diet: opportunities in Europe. Public Health Nutr. 2001;4:475–484.
4. Ledikwe JH, Wright HS, Mitchell DC, et al. Dietary pattern of rural older adults are associated with weight and nutritional status. J Am Geriatr Soc. 2004;52:589–595.
5. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol. 2002;13:3–9.
6. Quatromoni PA, Copenhafer DL, Demissie S, et al. The internal validity of a dietary pattern analysis. The Framingham Nutrition Studies. J Epidemiol Community Health. 2002;56:381–388.
7. Balder HF. Common and country-specific dietary patterns in four European cohort studies. J Nutr. 2003;133:4246–4251.
8. Kant AK, Schatzkin A, Graubard BI, et al. A prospective study of diet quality and mortality in women. JAMA. 2000;283:2109–2115.
9. Kumagai S, Shibata H, Watanabe S, et al. Effect of food intake pattern on all-cause mortality in the community elderly: a 7-year longitudinal study. J Nutr Health Aging. 1999;3:57–61.
10. Michels KB, Wolk A. A prospective study of variety of healthy foods and mortality in women. Int J Epidemiol. 2002;31:847–854.
11. Osler M, Schroll M. Diet and mortality in a cohort of elderly people in North European community. Int J Epidemiol. 1997;26:155–159.
12. Huijbregts P, Feskens E, Raasnen L, et al. Dietary pattern and 20-y mortality in elderly men in Finland, Italy, and the Netherlands: longitudinal cohort study. Br Med J. 1997;81:13–17.
13. Osler M, Heitmann BL, Gerdes LU, et al. Dietary patterns and mortality in Danish men and women: a prospective observational study. Br J Nutr. 2001;85:219–225.
14. Willett WC. Diet and health: what should we eat? Science. 1994;264:532–537.
15. Zheng W, Chow WH, Yang G, et al. The Shanghai Women's Health Study: rationale, study design, and baseline characteristics. Am J Epidemiol. 2005;162:1123–1131.
16. Shu XO, Yang G, Jin F, et al. Validity and reproducibility of the food frequency questionnaire used in the Shanghai Women's Health Study. Eur J Clin Nutr. 2004;58:17–23.
17. Kleinbaum DG, Kupper LL, Muller KE. Variable reduction and factor analysis. In: Applicated Regression Analysis and Other Multivariable Methods. Boston, MA: PWS-Kent publishing company; 1988:595:–640.
18. Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc. 2000;32(suppl):S498–S504.
19. Akin JS, Guilkey DK, Popkin BM, et al. Cluster analysis of food consumption patterns of older Americans. J Am Diet Assoc. 1986;86:616–624.
20. Pryer JA, Cook A, Shetty P. Identification of groups who report similar patterns of diet among a representative national sample of British adults aged 65 years of age or more. Public Health Nutr. 2001;4:787–795.
21. Park SY, Murphy SP, Wilkens LR, et al. Dietary patterns using the Food Guide Pyramid groups are associated with sociodemographic and lifestyle factors: the multiethnic cohort study. J Nutr. 2005;135:843–849.
22. Trichopoulou A, for members of the EPIC-Elderly Prospective Study Group. Modified Mediterranean diet and survival: EPIC-elderly prospective cohort study. BMJ. 2005;330:991–995.
23. Schroder H, Marrugat J, Vila J, et al. Adherence to the traditional Mediterranean diet is inversely associated with body mass index and obesity in a Spanish population. J Nutr. 2004;134:3355–3361.
24. Newby PK, Muller D, Hallfrisch J, et al. Food patterns measured by factor analysis and anthropometric changes in adults. Am J Clin Nutr. 2004;80:504–513.
25. Dixon LB, Balder HF, Virtamen MJ, et al. Dietary patterns associated with colon and rectal cancer: results from the dietary patterns and cancer (DIETSCAN) project. Am J Clin Nutr. 2004;80:1003–1011.
26. Lopez-Garci E, Schulze MB, Fung TT, et al. Major dietary patterns are related to plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr. 2004;80:1029–1035.
27. Kim MK, Sasaki S, Sasazuki S, et al. Prospective study of three major dietary patterns and risk of gastric cancer in Japan. Int J Cancer. 2004;110:435–442.
28. Kouris-Blazos A, Gnardellis C, Wahlqvist ML, et al. Are the advantages of the Mediterranean diet transferable to other population? A cohort study in Melbourne, Australia. Br J Nutr. 1999;82:57–61.
29. Lasheras C, Fernandez S, Patterson AM. Mediterranean diet and age with respect to overall survival in institutionalized, nonsmoking elderly people. Am J Clin Nutr. 2000;71:987–992.
30. Trichopoulou A, Critselis E. Mediterranean diet and longevity. Eur J Cancer Prev. 2004;13:453–456.
31. Trichopoulou A, Kouris-Blazos A, Wahlqvist ML, et al. Diet and overall survival in the elderly. Br Med J. 1995;311:1457–1460.
32. Trichopoulos D, Lagiou P. Dietary patterns and mortality. Br J Nutr. 2001;85:133–134.
33. Paolisso G, D'Amore A, Balbi V, et al. Plasma vitamin C affects glucose homeostasis in healthy subjects and in non-insulin-dependent diabetics. Am J Physiol. 1994;266:E261–E268.
34. Ford ES, Mokdad AH. Fruit and vegetable consumption and diabetes mellitus incidence among U.S. adults. Prev Med. 2001;32:33–39.
35. Montonen J, Knekt P, Järvinen R, et al. Dietary antioxidant intake and risk of type 2 diabetes. Diabetes Care. 2004;27:362–366.
36. Hu FB, Van Dam RM, Liu S. Diet and risk of type II diabetes: the role of types of fat and carbohydrate. Diabetologia. 2001;44:805–817.
37. Slattery ML, Boucher KM, Caan BJ, et al. Eating patterns and risk of colon cancer. Am J Epidemiol. 1998;148:4–16.
38. Fung T, Hu FB, Fuchs C, et al. Major dietary patterns and the risk of colorectal cancer in women. Arch Intern Med. 2003;163:309–314.
39. Terry P, Hu FB, Hansen J, et al. Prospective study of major dietary patterns and colorectal cancer risk in women. Am J Epidemiol. 2001;154:1143–1149.
40. Newby PK, Muller D, Hallfrisch J, et al. Dietary patterns and changes in body mass index and waist circumference in adults. Am J Clin Nutr. 2003;77:1417–1425.
41. Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev. 2004;62:177–203.
© 2007 Lippincott Williams & Wilkins, Inc.