American Journal of Epidemiology Advance Access originally published online on January 12, 2007
American Journal of Epidemiology 2007 165(6):603-610; doi:10.1093/aje/kwk061
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ORIGINAL CONTRIBUTIONS |
Dietary Patterns and Diabetes Incidence in the Melbourne Collaborative Cohort Study
1 Cancer Epidemiology Centre, The Cancer Council Victoria, Melbourne, Victoria, Australia
2 School of Population Health, University of Melbourne, Melbourne, Victoria, Australia
3 Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia
4 St. Vincent's Hospital Melbourne, Fitzroy, Victoria, Australia
Correspondence to Allison Hodge, The Cancer Council Victoria, 1 Rathdowne Street, Carlton, VIC 3053, Australia (e-mail: allison.hodge{at}cancervic.org.au).
Received for publication April 28, 2006. Accepted for publication September 14, 2006.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The authors investigated the association of dietary patterns and type 2 diabetes in a 4-year prospective study of 36,787 adults in the Melbourne Collaborative Cohort Study (1990–1994). A total of 31,641 (86%) participants completed follow-up, and 365 cases were identified. Four factors with eigenvalues of greater than 2 were identified using the principal factor method with 124 foods/beverages, followed by orthogonal rotation. Variables with factor loadings having absolute values of 0.3 or greater were used in interpreting the factors. Odds ratios for diabetes incidence across quintiles of factor scores were computed by use of logistic regression, adjusting for age, energy intake, family history of diabetes, country of birth, and other factor scores. Factor 1, characterized by olive oil, salad vegetables, and legumes and by avoidance of sweet bakery items, margarine, and tea, was associated with country of birth but not with diabetes. Factor 2, characterized by salad and cooked vegetables, was inversely associated with diabetes. Factor 3, characterized by meats and fatty foods, was associated with increased diabetes risk. A range of fruits loaded strongly on factor 4, which showed little association with diabetes. Avoidance of a dietary pattern including meats and fatty foods, as well as adherence to a pattern including salad and cooked vegetables, is recommended.
diabetes mellitus, type 2; diet; factor analysis, statistical; food habits; prospective studies
Abbreviations: CI, confidence interval; OR, odds ratio
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Type 2 diabetes is increasing in prevalence (1), and dietary modification combined with physical activity has been shown to reduce incidence (2–4). The effects of specific dietary patterns, however, are unclear. A recent review identified 93 studies using factor or cluster analysis to examine eating patterns since 1980, reflecting the growing interest in this approach to nutritional epidemiology (5). The use of dietary patterns avoids focusing on single foods or nutrients that can be correlated with, or interact with, each other and assesses combinations of foods, which alone may have effects too small to be identified.
Four studies have examined dietary patterns and diabetes incidence (6–9). In US men, a "Western" diet factor, characterized by red meat, processed meat, French fries, high-fat dairy products, refined grains, sweets, and desserts, was associated with increased risk of diabetes, while a "prudent" diet, characterized by vegetables, fruit, fish, poultry, and whole grains, was associated with reduced risk (9). In US women, similar factors were identified, and a Western factor was associated with increased diabetes risk (6). Reduced rank regression was used to identify a single factor associated with metabolic risk factors that predicted reduced diabetes incidence in a German population. A high score for this factor reflected a high intake of fruit and low intakes of high-calorie soft drinks, beer, red meat, processed meat, poultry, legumes, and bread other than whole grain (7). In a Finnish study, two dietary patterns were identified: A prudent pattern was associated with reduced diabetes risk, and a "conservative" pattern was associated with increased risk. The prudent pattern included fruit and vegetables, while the conservative pattern included butter, potatoes, and whole milk. Red meat contributed to both factors, but more strongly to the conservative factor, while processed meat contributed to the conservative factor (8).
In the Melbourne Collaborative Cohort Study, the dietary glycemic index and white bread were positively associated with diabetes risk (10), and alcohol, particularly wine, was inversely associated (11). To extend our understanding of diet and diabetes, the aim of this study was to identify dietary patterns by use of factor analysis and to determine their association with risk of diabetes in the Melbourne Collaborative Cohort Study.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Subjects
The Melbourne Collaborative Cohort Study recruited 41,528 people (17,049 men), aged 27–75 years (99.3 percent were aged 40–69 years), between 1990 and 1994. The study included 5,425 migrants from Italy and 4,535 from Greece. The Cancer Council Victoria's Human Research Ethics Committee approved the study protocol. Subjects gave written consent to participate and for the investigators to obtain access to their medical records.
For this analysis, people with diabetes at baseline (self-reported (n = 1,549) or diabetic plasma glucose (n = 324)), those who reported having angina or suffering a heart attack prior to baseline, those who did not report diabetes at baseline but later reported a date of diabetes diagnosis before baseline, those with energy intakes in the top or bottom 1 percent of the sex-specific distributions, and those with missing values for relevant risk factors measured at baseline were excluded. These exclusions left 36,787 subjects.
Baseline glucose measurement
Plasma glucose was measured using a Kodak Ektachem analyzer (Rochester, New York). For the 68 percent who were fasting, plasma glucose values of 7.8 mmol/liter or above were considered diabetic and, for those who were not fasting, diabetes was defined as equal to or greater than 11.1 mmol/liter, according to the World Health Organization criteria of the time (12).
Dietary assessment
Dietary information was collected by use of a 121-item, self-administered, food frequency questionnaire, specifically developed for the Melbourne Collaborative Cohort Study (13). To calculate nutrient intakes, we used sex-specific standard portions, together with Australian food composition data (14).
Repeatability of food frequency questionnaire measurements
From July 1992 to June 1993, 275 subjects were invited to participate in a repeatability study that involved completing a second questionnaire 12 months after baseline. Selection was stratified by sex, country of birth (Australia, Italy, Greece), 10-year age group, and month of attendance. Intraclass correlations were calculated for each of the 121 items from the food frequency questionnaire expressed as daily equivalent frequencies. Data were not available to assess repeatability for alcoholic beverages or oils.
Measurement of nondietary risk factors
A structured interview schedule was used to obtain information on country of birth, smoking, alcohol consumption, physical activity, education, weight change over the last 5 years, and history of diabetes in first-degree relatives. Frequencies of walking, vigorous exercise, and less vigorous exercise over the last 6 months were coded as follows: 0 (none), 1.5 (one or two times per week), and 4 (three or more times per week). A total activity score was calculated as the frequency of walking plus less vigorous exercise plus two times the frequency of vigorous exercise.
Standard methods were used to measure height, weight, and waist and hip circumferences, from which body mass index (kg/m2) and the waist/hip ratio were calculated.
Follow-up and ascertainment of diabetes status
Incident cases of diabetes were identified from a self-administered questionnaire mailed to participants about 4 years after baseline. Participants were asked: "Has a doctor ever told you that you have had diabetes?" and, if yes, for the year of diagnosis. For all self-reported incident cases, except those who reported a diagnosis date before baseline and who were excluded, confirmation of diagnosis was sought from physicians nominated by participants. Physicians were asked to specify if the participant had diabetes and, if so, to indicate whether it was type 1 or type 2.
Statistical analysis
Factor analysis was performed on the 121 items from the food frequency questionnaire, plus olive and vegetable oils and alcohol from wine. Many studies include alcoholic beverages in their dietary questionnaires and, hence, in dietary pattern analysis (6, 7, 9). There is also evidence that moderate alcohol intake is associated with reduced risk of diabetes (15, 16) and may be associated with other dietary intake. In the study cohort, wine was the predominant source of alcohol and showed an inverse association with diabetes (for 
10 g/day vs. lifelong abstainer for women: odds ratio (OR) = 0.43, 95 percent confidence interval (CI): 0.26, 0.70; ptrend = 0.026; for 20 g/day vs. lifelong abstainer for men: OR = 0.61, 95 percent CI: 0.37, 1.02; ptrend = 0.018) (11). For the 121 items, intake was measured as daily equivalent frequency, while intakes of olive and vegetable oils were measured in milliliters/week. Alcohol from wine was measured as grams/day.
The principal factor method was used to extract factors, followed by orthogonal (varimax) rotation to assist in interpretation of the factors and to ensure that the factors were uncorrelated. Factors with eigenvalues of 2 or greater were retained. Variables with factor loadings having absolute values of 0.3 or greater were used in interpreting the factors (17). Scores were computed for rotated factors as the sum of products of observed variables multiplied by their factor loading, and they were analyzed as quintile groups. Factors were initially extracted separately for men and women, with similar results, so factors derived from the pooled data were used.
Linear regression was used to calculate how much of the variance in factor scores was associated with country of birth and other potential confounders. Because of the anticipated close association between country of birth and dietary patterns, it was important to determine whether country of birth could be considered as a potential confounder or as a surrogate for factor score. Correlations between factor scores and for factor scores with nutrient and food group intakes were calculated.
The relation between quintiles of the dietary factor scores and risk of diabetes was examined by use of logistic regression. Age, country of birth, physical activity score, weight change, education, family history of diabetes, and smoking status were considered as potential confounders. Forward stepwise logistic regression, with p = 0.2 for removal and p = 0.05 for entry, was used to identify which of these to include. The final models included the four dietary factor scores simultaneously, with age, country of birth, energy intake, and family history of diabetes, both with and without body mass index and waist/hip ratio, which were considered possible intermediates in the causal pathway. Tests for trend across quintiles were performed by use of the medians in each group. Interactions between sex and factor scores were tested prior to analyzing men and women together. Because the factors appeared similar to our previously defined food groups, we also computed odds ratios and 95 percent confidence intervals for quartiles of intake of vegetable, fruit, and meat groups with the same confounders, to determine whether factors showed different associations with diabetes. In addition, models including factor scores and quartiles of intake for meat and white bread, which were both associated with diabetes risk and loaded on factor 3, were computed to determine whether these items were responsible for the association of factor 3 and diabetes. The main factor score analyses were repeated with only the 303 confirmed cases of type 2 diabetes.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
A total of 31,641 participants (86 percent of those eligible) completed the follow-up question on diabetes. People who completed follow-up had similar levels of risk factors for type 2 diabetes compared with those who did not complete the questionnaire, although body mass index was slightly lower: body mass index (mean: 26.6 vs. 27.2 kg/m2), age (mean: 54.3 vs. 54.5 years), and fasting plasma glucose (mean: 5.5 vs. 5.5 mmol/liter). Eighty-seven percent of women and 85 percent of men completed follow-up. Greek-born participants (80 percent) were slightly less likely to complete follow-up than were those born elsewhere (86–87 percent).
A diagnosis of diabetes after baseline was reported by 459 participants. Of 399 people for whom a response was obtained from their physician, 303 (76 percent) were confirmed as type 2. Because the proportion confirmed was high, two people for whom the physician did not know type, or did not know the diabetes status, were considered to be cases, as were 60 people for whom no response was available. Participants whose physicians reported that they had type 1 diabetes (n = 11), had impaired glucose tolerance (n = 1), or did not have diabetes (n = 82) were classified as noncases, along with those who did not report diabetes at follow-up. After exclusions, 365 cases and 31,276 noncases were eligible for analysis.
Table 1 shows baseline characteristics of the participants by diabetes status at follow-up. People who developed diabetes tended to be older, more obese, more likely to be southern European, less active, less educated, more likely to have a family history of diabetes, and more likely to have gained weight in the last 5 years. Men were also more likely to be smokers.
|
Four factors with eigenvalues of 2.39 or greater and explaining 68 percent of the variance were identified. Rotated factor loadings for food items with a loading having absolute values of 0.2 or greater for any factor are given in table 2. Olive oil, cooked dried legumes, and some salad vegetables, as well as avoidance of sweet biscuits, cakes and pastries, margarine, and tea, characterized the first factor. Factor 2 was characterized by intake of a variety of salad and cooked vegetables. The third factor was predominantly a meat factor but also had loadings having absolute values of 0.3 or greater for savory pastries, fried eggs, fried fish, and fried potatoes. The last factor was characterized by frequent consumption of fruit. Correlations with energy intake were –0.2, 0.3, 0.6, and 0.2 for factors 1–4, respectively. The factor 2 score was strongly correlated with intake of vegetables (not potatoes, r = 0.83), the factor 3 score was strongly correlated with meat (r = 0.79) and fresh meat (r = 0.73), and the factor 4 score was correlated with fruit intake (r = 0.89).
|
The adjusted R2 from linear regression models with factor scores as the dependent variable indicated that country of birth explained between 5 percent (factor 3) and 35 percent (factor 1) of the variance in factors' scores; thus, although dietary patterns were associated with country of birth, they were not a substitute (table 3). Participants born in Australia and the United Kingdom scored higher on factor 2 (vegetables), while those born in Greece and Italy scored higher on factors 1 (olive oil, salad, legumes, less of sweet bakery items, tea, and margarine) and 4 (fruit), and those born in Greece scored highest for factor 3 (meat). Education was associated with factor 1, but this was largely due to confounding by country of birth. There was little association with other risk factors (adjusted R2: all p's < 0.04) (table 3).
|
Table 4 shows odds ratios for quintiles of factor scores. In model 1 (without body mass index and waist/hip ratio), factor 2 showed an inverse association, while factor 3 was positively associated with diabetes. There was no significant association with factor 1 or 4. The addition of body mass index and waist/hip ratio attenuated the associations, and no statistically significant trends remained. p values for interaction terms between sex and factor scores were all large except for factor 3, where p = 0.01 for model 1 and p = 0.02 for model 2. However, as the association between factor 3 and diabetes was positive in sex-specific analyses for women (for quintile 5 vs. quintile 1: OR = 3.44, 95 percent CI: 1.85, 6.41; ptrend = 0.001 in model 1) and men (for quintile 5 vs. quintile 1: OR = 1.97, 95 percent CI: 1.01, 3.85; ptrend = 0.04 in model 1), data were pooled.
|
When the analyses were repeated with only the confirmed cases of diabetes, the positive association between factor 3 and diabetes remained (for quintile 5 vs. quintile 1: OR = 2.95, 95 percent CI: 1.75, 4.99; ptrend = 0.001 in model 1), although the inverse association with factor 2 was no longer significant (for quintile 5 vs. quintile 1: OR = 0.77, 95 percent CI: 0.51, 1.18; ptrend = 0.102 in model 1).
There was little evidence for an association between intake of vegetables (for quartile 4 vs. quartile 1: OR = 0.93, 95 percent CI: 0.67, 1.28; ptrend = 0.55) or fruit (for quartile 4 vs. quartile 1: OR = 0.88, 95 percent CI: 0.66, 1.19; ptrend = 0.75) with diabetes, but the meat group showed a modest association (for quartile 4 vs. quartile 1: OR = 1.53, 95 percent CI: 1.10, 2.14; ptrend = 0.02). A positive association between white bread and type 2 diabetes has previously been reported for the study cohort (10). In models including meat or white bread along with factor scores, meat strengthened the association between factor 3 and diabetes (for quintile 5 vs. quintile 1: OR = 3.01, 95 percent CI: 1.70, 5.33; ptrend = 0.002), while bread attenuated the association (for quintile 5 vs. quintile 1: OR = 2.46, 95 percent CI: 1.53, 3.96; ptrend = 0.003). When white bread and meat were included simultaneously in the model, the odds ratio for the top quintile of factor 3 was similar to that for the original model (OR = 2.64, 95 percent CI: 1.46, 4.76; ptrend = 0.02).
A total of 242 people completed the reproducibility study. For the 83 food frequency questionnaire foods included in table 2, the intraclass correlations ranged from less than 0.1 for rice dishes, egg dishes, beef/veal schnitzel, legume soup, fried fish, and luncheon meats to 0.84 for tea, 0.76 for pudding, 0.61 for carrot, and 0.60 for margarine, with a median of 0.30. It was not possible to test the reproducibility of the factors, as there were insufficient people in the reproducibility study.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Four dietary factors were identified. The first factor, characterized by olive oil, salad vegetables, and legumes, with avoidance of sweet bakery items, margarine, and tea, was closely associated with country of birth but not with diabetes. The second factor, characterized by a variety of salad and cooked vegetables, was inversely associated with diabetes. The third factor, characterized by meats and fatty fried foods, was positively associated with diabetes. A range of fruits loaded strongly on factor 4, which showed no association with diabetes risk. Positive associations between meat or white bread and diabetes did not explain the association observed for factor 3.
The high response rate and the small differences between respondents and nonrespondents should minimize response bias. Compared with other studies of diabetes incidence (18–21), our study had the advantage of measuring blood glucose at baseline, so people with elevated plasma glucose levels could be excluded, although this would not have excluded all those who would be identified by an oral glucose tolerance test (22, 23).
At follow-up, type 2 diabetes was confirmed in 76 percent of self-reported cases. This compares favorably with the results from the Iowa Women's Health Study where, in a substudy, diabetes was confirmed in only 64 percent of self-reported cases (19). Excluding the unconfirmed cases did not change our conclusions regarding factor 3 and diabetes risk, although the association for factor 2 was no longer apparent. Some incident cases would have been missed because we did not screen participants at follow-up, but underascertainment of cases would affect the results only if it were associated with the exposure (24).
One advantage of our analyses over some other dietary factor analyses is that foods were not grouped before the factor analysis. This avoids making assumptions about what should be put together. For example, in a study of US men, there was a single item "fish," which loaded on a "healthy" diet pattern (9), but in our data fried and steamed fish loaded on different factors, having opposite associations with diabetes.
Random error in measuring intake and dietary change during follow-up is likely to have attenuated the associations. Our measurement of diet was based on a single food frequency questionnaire administered at baseline that may not have been representative of consumption over the longer term. In a subset of 242 participants, the food frequency questionnaire showed only fair to moderate agreement for foods when administered on two occasions 12 months apart.
Dietary factor analysis in different populations has identified factors considered healthy (fruit, vegetables, and whole grains) and less healthy (meat and fatty foods) (5). A meat and fatty food factor was identified in our study but, unlike other reports, fruit and vegetables fell into two separate factors. This may be because southern European migrants were included in the cohort to broaden the range of dietary exposures. Participants born in Australia and the United Kingdom scored higher on the vegetable factor (factor 2), while those born in Greece and Italy scored higher on the fruit factor (factor 4), suggesting that in our study people who ate fruit more frequently did not always eat vegetables more frequently and vice versa.
The factor reflecting intake of a variety of vegetables was inversely associated with diabetes. This is generally consistent with the inverse associations for the prudent pattern in Finland (8) and US men (9), although in both these studies there were foods other than vegetables in the healthier pattern. Vegetables as a group (excluding potatoes) were not associated with diabetes in the Melbourne Collaborative Cohort Study, but other items that loaded less strongly on factor 2, such as whole grain bread, yogurt, chicken dishes, steamed fish, and avoidance of white bread, may contribute to the inverse association. Unlike other studies (5), this study found that fruit was in a factor separate to other "healthy" foods, which showed no association with diabetes.
The association observed for the meat factor and diabetes is consistent with the findings of previous studies (6–9). Factor 3, on which meat items loaded heavily, was more strongly associated with diabetes than was the meat group alone. White bread, which was previously shown to be associated with diabetes risk in the cohort (10), also loaded quite strongly on factor 3. The association of factor 3 with diabetes was independent of meat and white bread. The fatty foods in factor 3 may also be important in regard to increasing diabetes risk. Fish, chicken, and potatoes, when cooked in ways other than frying, loaded on factors that were not associated with increased risk of diabetes. On the other hand, the specific combination of meats in factor 3 compared with the meat group (fresh meat includes beef/veal schnitzel, beef/veal roast, beef steak, beef rissole, beef dish, lamb roast/chops, lamb dish, and pork roast) may be more closely associated with diabetes.
The finding of associations between factor scores and diabetes, which are stronger than and independent of individual food groups, is important in demonstrating the potential for modifying diabetes risk through diet, even if the association is mediated to a large extent by obesity. While it is intuitive that combining several foods having weak associations with diabetes may result in a factor with a stronger association than that of any individual food, caution is required in translating our findings into a public health message. The level of factor loading required to include an item in a factor is arbitrary, and the intake of each food in the factor required to achieve an effect is not clear. Nonetheless, our results suggest that avoiding a dietary pattern consisting of meats and fried or fatty foods and including a variety of vegetables may reduce the risk of type 2 diabetes.
|
ACKNOWLEDGMENTS |
|---|
Cohort recruitment was funded by VicHealth and The Cancer Council Victoria. This study was funded by National Health and Medical Research Council grants 209057 and 126403 and was further supported by infrastructure provided by The Cancer Council Victoria.
The authors are grateful for the contributions of many people, including the original investigators and the diligent team who recruited the participants and continue working on follow-up.
Conflict of interest: none declared.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
- King H, Aubert RE, Herman WH. (1998) Global burden of diabetes, 1995 –2025. Prevalence, numerical estimates, and projections. Diabetes Care 21:1414–31.[Abstract]
- Diabetes Prevention Program Research. (2002) Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 346:393–403.
[Abstract/Free Full Text] - Pan XR, Li GW, Hu YH, et al. (1997) Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance. The Da Qing IGT and Diabetes Study. Diabetes Care 20:537–44.[Abstract]
- Tuomilehto J, Lindstrom J, Eriksson JG, et al. (2001) Prevention of type 2 diabetes mellitus by change in lifestyle among subjects with impaired glucose tolerance. N Eng J Med 344:1343–50.
[Abstract/Free Full Text] - Newby PK and Tucker KL. (2004) Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 62:177–203.[CrossRef][ISI][Medline]
- Fung TT, Schulze M, Manson JE, et al. (2004) Dietary patterns, meat intake, and the risk of type 2 diabetes in women. Arch Intern Med 164:2235–40.
[Abstract/Free Full Text] - Heidemann C, Hoffmann K, Spranger J, et al. (2005) A dietary pattern protective against type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC)–Potsdam Study cohort. Diabetologia 48:1126–34.[CrossRef][ISI][Medline]
- Montonen J, Knekt P, Harkanen T, et al. (2005) Dietary patterns and the incidence of type 2 diabetes. Am J Epidemiol 161:219–27.
[Abstract/Free Full Text] - van Dam RM, Rimm EB, Willett WC, et al. (2002) Dietary patterns and risk for type 2 diabetes mellitus in US men. Ann Intern Med 136:201–9.
[Abstract/Free Full Text] - Hodge AM, English DR, O'Dea K, et al. (2004) Glycemic index and dietary fiber and the risk of type 2 diabetes. Diabetes Care 27:2701–6.
[Abstract/Free Full Text] - Hodge A, English D, O'Dea K, et al. (2006) Alcohol intake, consumption pattern and beverage type, and the risk of type 2 diabetes. Diabet Med 23:690–7.[CrossRef][ISI][Medline]
- World Health Organization. (1999) Definition, diagnosis and classification of diabetes mellitus and its complications. Report of a WHO Consultation. Part 1: diagnosis and classification of diabetes mellitus. (Department of Noncommunicable Disease Surveillance, World Health Organization, Geneva, Switzerland).
- Ireland P, Jolley D, Giles G, et al. (1994) Development of the Melbourne FFQ: a food frequency questionnaire for use in an Australian prospective study involving and ethnically diverse cohort. Asia Pac J Clin Nutr 3:19–31.[Medline]
- Lewis J, Milligan G, Hunt A. (1995) NUTTAB95 nutrient data table for use in Australia. (Australian Government Publishing Service, Canberra, Australia).
- Carlsson S, Hammar N, Grill V. (2005) Alcohol consumption and type 2 diabetes meta-analysis of epidemiological studies indicates a U-shaped relationship. Diabetologia 48:1051–4.[CrossRef][ISI][Medline]
- Howard AA, Arnsten JH, Gourevitch MN. (2004) Effect of alcohol consumption on diabetes mellitus: a systematic review. Ann Intern Med 140:211–19.
[Abstract/Free Full Text] - Williams DE, Prevost AT, Whichelow MJ, et al. (2000) A cross-sectional study of dietary patterns with glucose intolerance and other features of the metabolic syndrome. Br J Nutr 83:257–66.[ISI][Medline]
- Liu S. (2002) Intake of refined carbohydrates and whole grain foods in relation to risk of type 2 diabetes mellitus and coronary heart disease. J Am Coll Nutr 21:298–306.
[Abstract/Free Full Text] - Meyer KA, Kushi LH, Jacobs DR Jr, et al. (2000) Carbohydrates, dietary fiber, and incident type 2 diabetes in older women. Am J Clin Nutr 71:921–30.
[Abstract/Free Full Text] - Salmerón J, Ascherio A, Rimm EB, et al. (1997) Dietary fiber, glycaemic load, and risk of NIDDM in men. Diabetes Care 20:545–50.[Abstract]
- Salmerón J, Manson JE, Stampfer MJ, et al. (1997) Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 277:472–7.[Abstract]
- Gabir MM, Hanson RL, Dabelea D, et al. (2000) The 1997 American Diabetes Association and 1999 World Health Organization criteria for hyperglycemia in the diagnosis and prediction of diabetes. Diabetes Care 23:1108–12.
[Abstract/Free Full Text] - Harris MI, Klein R, Welborn TA, et al. (1992) Onset of NIDDM occurs at least 4–7 yr before clinical diagnosis. Diabetes Care 15:815–19.[Abstract]
- Rothman KJ and Greenland S. (1998) Modern epidemiology. (Lippincott-Raven, Philadelphia, PA).
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M.-C. Yeh, S. B. Ickes, L. M. Lowenstein, K. Shuval, A. S. Ammerman, R. Farris, and D. L. Katz Understanding barriers and facilitators of fruit and vegetable consumption among a diverse multi-ethnic population in the USA Health Promot. Int., March 1, 2008; 23(1): 42 - 51. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. R Harriss, D. R English, J. Powles, G. G Giles, A. M Tonkin, A. M Hodge, L. Brazionis, and K. O'Dea Dietary patterns and cardiovascular mortality in the Melbourne Collaborative Cohort Study Am. J. Clinical Nutrition, July 1, 2007; 86(1): 221 - 229. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||