RESEARCH
Review
A Posteriori Dietary Patterns Are Related to Risk of Type 2 Diabetes: Findings from a Systematic Review and Meta-Analysis Claire T. McEvoy, PhD, RD*; Christopher R. Cardwell, PhD; Jayne V. Woodside, PhD; Ian S. Young, MD, MB, BCh, FRCP; Steven J. Hunter, MD, MB, BCh, FRCP; Michelle C. McKinley, PhD ARTICLE INFORMATION
ABSTRACT
Article history:
Our review and meta-analysis examined the association between a posterioriederived dietary patterns (DPs) and risk of type 2 diabetes mellitus. MEDLINE and EMBASE were searched for articles published up to July 2012 and data were extracted by two independent reviewers. Overall, 19 cross-sectional, 12 prospective cohort, and two nested case-control studies were eligible for inclusion. Results from cross-sectional studies reported an inconsistent association between DPs and measures of insulin resistance and/or glucose abnormalities, or prevalence of type 2 diabetes. A meta-analysis was carried out on nine prospective cohort studies that had examined DPs derived by principle component/factor analysis and incidence of type 2 diabetes risk (totaling 309,430 participants and 16,644 incident cases). Multivariate-adjusted odds ratios were combined using a random-effects meta-analysis. Two broad DPs (Healthy/Prudent and Unhealthy/Western) were identified based on food factor loadings published in original studies. Pooled results indicated a 15% lower type 2 diabetes risk for those in the highest category of Healthy/Prudent pattern compared with those in the lowest category (95% CI 0.80 to 0.91; P19 y
Yap and colleagues54
Malaysia
Chinese 30-65 y menþwomen Elderly 70-79 y menþwomen
Anderson United States and colleagues55
24-Hour diet recall
Factor
1. White rice and kimchib 2. Meat and alcoholb 3. High fat, sweets, coffeeb 4. Grains, vegetables, fishb
Fasting glucose
BMI, energy intake, smoking, alcohol, physical activity
179
91-Item food frequency questionnaire
Factor
1. Balancedb 2. Meat, rice, and noodlesf
Fasting glucose Glycated hemoglobin
Smoking, alcohol, physical activity
1,751
108-Item food frequency questionnaire
Cluster
1. Healthy foodsae 2. Breakfast cerealsbg 3. Meat and alcoholbg 4. Sweets & dessertsbg 5. Refine grainsbg 6. High-fat dairybd
Fasting glucose Fasting insulin Insulin resistance (HOMA) Oral glucose tolerance test
Study site, education, physical activity, smoking, energy intake, BMI
9,850
Significantly negatively associated with 1 outcome measurements of glucose tolerance. Not significantly associated with outcome measurement(s) of glucose tolerance. BMI¼body mass index. d Significantly positively associated with 1 outcome measurements of insulin resistance. e Significantly negatively associated with 1 outcome measurements of insulin resistance. f Significantly positively associated with 1 outcome measurements of glucose tolerance. g Not significantly associated with outcome measurement(s) of insulin resistance. h Significant association in men but not in women. i Significant association in women but not in men. j SI¼insulin sensitivity index. k HOMA¼homeostatic model assessment. l Significantly inversely related to T2DM risk. m Not significantly related to T2DM risk. n QUICKI¼Quantitative Insulin Sensitivity Check Index. a
b
--
c
RESEARCH
8
Table 2. Cross-sectional studies (N¼19) examining a posteriori dietary patterns and insulin resistance, glucose abnormalities, or prevalence of type 2 diabetes mellitus (T2DM) (continued)
2014 Volume -
Number -
RESEARCH patterns and glucose abnormalities, whereas nine studies reported no significant association.29,39-41,45,48,51,53,54 Eight studies reported significant positive associations between generally Unhealthy/Western patterns and glucose abnormalities,40,41,44,45,47,49,51,54 whereas six studies reported no significant association.29,30,39,48,53,55 One study47 reported a significant inverse partially adjusted correlation between FG level and an Unhealthy/Western pattern in a multiethnic cohort that included individuals with type 2 diabetes.
Cross-Sectional Studies Investigating a posteriori DPs and Type 2 Diabetes Prevalence Three studies investigated associations between a posteriori DPs and prevalence of diabetes determined by measuring FG concentration.46,47,52 One study in elderly Mediterranean adults found a significant inverse relationship between FAderived Healthy/Prudent patterns and prevalence of type 2 diabetes,46 but no significant association between Unhealthy/ Western patterns and type 2 diabetes prevalence. In the Trivandrum region of India, a Healthy/Prudent pattern was also significantly inversely associated with type 2 diabetes and prediabetes risk.52 In addition, a FA-derived Unhealthy/ Western pattern was positively associated with increased prevalence of type 2 diabetes in a multiethnic cohort; however, the association was nonsignificant after adjustment for energy intake.47
Prospective Cohort Studies and Type 2 Diabetes Risk Meta-Analysis of Prospective Studies Investigating a posteriori DPs and Type 2 Diabetes Risk. Nine prospective cohort studies were eligible for inclusion in the pooled analysis,22-24,61-66 as shown in Table 3. Of these studies, five were conducted in the United States, one in Australia, one in Finland, and two in China. Overall, 309,430 participants and 16,644 incident cases of type 2 diabetes were included in the analysis. All studies involved adults free of type 2 diabetes at baseline. The mean age of participants ranged from 24 to 84 years. The duration of follow-up ranged from 4 to 23 years. Baseline habitual dietary intake was assessed using a validated FFQ in eight studies, whereas a DH method was used in one study.63 Diabetes incidence was ascertained using a combination of self-report, alongside physician confirmation, and/or health plans and drug reimbursement registry, in seven studies24,61-66 and directly measured using FG concentration in two studies.22,23 For each of the included prospective studies, the predominant factor loading foods within the Healthy/Prudent and Unhealthy/Western patterns are detailed in Table 3. The association between the highest vs lowest categories of intake of the Healthy/Prudent DP and type 2 diabetes risk is shown in Figure 5. The combined results of all nine studies demonstrated a significant reduction in type 2 diabetes risk for individuals in the highest category of the Healthy/Prudent pattern compared with those in the lowest category of the Healthy/Prudent pattern (OR 0.85, 95% CI 0.80 to 0.91; P
Review
A Posteriori Dietary Patterns Are Related to Risk of Type 2 Diabetes: Findings from a Systematic Review and Meta-Analysis Claire T. McEvoy, PhD, RD*; Christopher R. Cardwell, PhD; Jayne V. Woodside, PhD; Ian S. Young, MD, MB, BCh, FRCP; Steven J. Hunter, MD, MB, BCh, FRCP; Michelle C. McKinley, PhD ARTICLE INFORMATION
ABSTRACT
Article history:
Our review and meta-analysis examined the association between a posterioriederived dietary patterns (DPs) and risk of type 2 diabetes mellitus. MEDLINE and EMBASE were searched for articles published up to July 2012 and data were extracted by two independent reviewers. Overall, 19 cross-sectional, 12 prospective cohort, and two nested case-control studies were eligible for inclusion. Results from cross-sectional studies reported an inconsistent association between DPs and measures of insulin resistance and/or glucose abnormalities, or prevalence of type 2 diabetes. A meta-analysis was carried out on nine prospective cohort studies that had examined DPs derived by principle component/factor analysis and incidence of type 2 diabetes risk (totaling 309,430 participants and 16,644 incident cases). Multivariate-adjusted odds ratios were combined using a random-effects meta-analysis. Two broad DPs (Healthy/Prudent and Unhealthy/Western) were identified based on food factor loadings published in original studies. Pooled results indicated a 15% lower type 2 diabetes risk for those in the highest category of Healthy/Prudent pattern compared with those in the lowest category (95% CI 0.80 to 0.91; P19 y
Yap and colleagues54
Malaysia
Chinese 30-65 y menþwomen Elderly 70-79 y menþwomen
Anderson United States and colleagues55
24-Hour diet recall
Factor
1. White rice and kimchib 2. Meat and alcoholb 3. High fat, sweets, coffeeb 4. Grains, vegetables, fishb
Fasting glucose
BMI, energy intake, smoking, alcohol, physical activity
179
91-Item food frequency questionnaire
Factor
1. Balancedb 2. Meat, rice, and noodlesf
Fasting glucose Glycated hemoglobin
Smoking, alcohol, physical activity
1,751
108-Item food frequency questionnaire
Cluster
1. Healthy foodsae 2. Breakfast cerealsbg 3. Meat and alcoholbg 4. Sweets & dessertsbg 5. Refine grainsbg 6. High-fat dairybd
Fasting glucose Fasting insulin Insulin resistance (HOMA) Oral glucose tolerance test
Study site, education, physical activity, smoking, energy intake, BMI
9,850
Significantly negatively associated with 1 outcome measurements of glucose tolerance. Not significantly associated with outcome measurement(s) of glucose tolerance. BMI¼body mass index. d Significantly positively associated with 1 outcome measurements of insulin resistance. e Significantly negatively associated with 1 outcome measurements of insulin resistance. f Significantly positively associated with 1 outcome measurements of glucose tolerance. g Not significantly associated with outcome measurement(s) of insulin resistance. h Significant association in men but not in women. i Significant association in women but not in men. j SI¼insulin sensitivity index. k HOMA¼homeostatic model assessment. l Significantly inversely related to T2DM risk. m Not significantly related to T2DM risk. n QUICKI¼Quantitative Insulin Sensitivity Check Index. a
b
--
c
RESEARCH
8
Table 2. Cross-sectional studies (N¼19) examining a posteriori dietary patterns and insulin resistance, glucose abnormalities, or prevalence of type 2 diabetes mellitus (T2DM) (continued)
2014 Volume -
Number -
RESEARCH patterns and glucose abnormalities, whereas nine studies reported no significant association.29,39-41,45,48,51,53,54 Eight studies reported significant positive associations between generally Unhealthy/Western patterns and glucose abnormalities,40,41,44,45,47,49,51,54 whereas six studies reported no significant association.29,30,39,48,53,55 One study47 reported a significant inverse partially adjusted correlation between FG level and an Unhealthy/Western pattern in a multiethnic cohort that included individuals with type 2 diabetes.
Cross-Sectional Studies Investigating a posteriori DPs and Type 2 Diabetes Prevalence Three studies investigated associations between a posteriori DPs and prevalence of diabetes determined by measuring FG concentration.46,47,52 One study in elderly Mediterranean adults found a significant inverse relationship between FAderived Healthy/Prudent patterns and prevalence of type 2 diabetes,46 but no significant association between Unhealthy/ Western patterns and type 2 diabetes prevalence. In the Trivandrum region of India, a Healthy/Prudent pattern was also significantly inversely associated with type 2 diabetes and prediabetes risk.52 In addition, a FA-derived Unhealthy/ Western pattern was positively associated with increased prevalence of type 2 diabetes in a multiethnic cohort; however, the association was nonsignificant after adjustment for energy intake.47
Prospective Cohort Studies and Type 2 Diabetes Risk Meta-Analysis of Prospective Studies Investigating a posteriori DPs and Type 2 Diabetes Risk. Nine prospective cohort studies were eligible for inclusion in the pooled analysis,22-24,61-66 as shown in Table 3. Of these studies, five were conducted in the United States, one in Australia, one in Finland, and two in China. Overall, 309,430 participants and 16,644 incident cases of type 2 diabetes were included in the analysis. All studies involved adults free of type 2 diabetes at baseline. The mean age of participants ranged from 24 to 84 years. The duration of follow-up ranged from 4 to 23 years. Baseline habitual dietary intake was assessed using a validated FFQ in eight studies, whereas a DH method was used in one study.63 Diabetes incidence was ascertained using a combination of self-report, alongside physician confirmation, and/or health plans and drug reimbursement registry, in seven studies24,61-66 and directly measured using FG concentration in two studies.22,23 For each of the included prospective studies, the predominant factor loading foods within the Healthy/Prudent and Unhealthy/Western patterns are detailed in Table 3. The association between the highest vs lowest categories of intake of the Healthy/Prudent DP and type 2 diabetes risk is shown in Figure 5. The combined results of all nine studies demonstrated a significant reduction in type 2 diabetes risk for individuals in the highest category of the Healthy/Prudent pattern compared with those in the lowest category of the Healthy/Prudent pattern (OR 0.85, 95% CI 0.80 to 0.91; P
