Abstract
Background
Dietary diversity may affect type 2 diabetes (T2D) but no studies have examined protein diversity by source. We examined five diversity scores and the 10-year risk of T2D and effect modification.
Methods
A prospective study of 10 363 incident T2D cases and a representative sub-cohort of 13 937 individuals sampled from a cohort of 340 234 participants in eight European countries (1993–2007). Five diversity scores were derived from self-reported diet data (gr/day): diversity of food groups (range: 0–5); and diversity within subtype of vegetables (0–4); meat/alternatives (0–6); animal-protein (0–8); and plant-protein sources (0–5). Country-specific hazard ratios (HRs) and 95% confidence intervals (CIs) were obtained by using Prentice-weighted Cox regression and combined by using mixed-effects models. Models were stratified by sex (male/female) and obesity status (body mass index ≥ 30 kg/m2; waist circumference ≥ 88 cm for females and ≥102 cm for males).
Results
Daily intake of five food groups (versus up to three) was linked to lower T2D incidence overall [HR 0.86 (95% CI 0.75, 0.98)], in females [0.86 (0.77, 0.96)], and in people without central obesity [0.79 (0.70, 0.89)]. Three or more subtypes of plant protein were inversely associated with T2D overall [0.78 (0.65, 0.98)], in females [0.75 (0.62, 0.90)] and people without central obesity [0.82 (0.68, 1.00)]. Additionally, consuming three subtypes of vegetables was inversely associated with T2D overall [0.90 (0.83, 0.98)] and in males [0.85 (0.73, 0.99)].
Conclusion
Diabetes prevention may benefit not only from a diet consisting of five different food groups, but also from a diet that is diverse in plant-protein sources, with specific benefits for female Europeans and those without central obesity.
Keywords: dietary diversity, protein diversity, type 2 diabetes, case–cohort study
Key Messages.
This study aimed to examine the prospective association of five diversity scores with the 10-year risk of T2D and assessed differences by sex and obesity status.
A varied diet rich in five food groups and diverse plant-protein sources may help in diabetes prevention, particularly among female Europeans and those without central obesity.
This study supports the 2019 EAT-Lancet report and other national dietary guidelines that advocate for the “diversity” of protein products.
Introduction
In 2021, diabetes affected 10.5% of the global population, contributing to 6.7 million deaths and USD 966 billion in health expenditures [1]. Type 2 diabetes (T2D) represents ∼90% of all diabetes diagnoses and the majority of diabetes-related burdens [1]. T2D prevention is therefore a healthcare and policy priority. Diet is a recognized modifiable factor for T2D prevention [2] and dietary diversity is a key aspect of healthy eating [3–5]. Unlike the frequency [6], quantity [7, 8], or quality of food intake [3–5], dietary diversity reflects the variety of food groups consumed, offering unique benefits for metabolic health. While frequency counts food items, diversity captures both variety and number. Even with similar food quantity, a limited range of foods may be consumed [9, 10]. Additionally, a healthy diet can lack diversity; e.g. French adults who were following US nutrition guidelines had the lowest diversity score [11]. A diverse diet is not necessarily lower in energy, trans fats, or salt [12]. Therefore, dietary diversity represents an additional facet of diet that is not fully captured by other measures. Our previous review [13] identified prospective studies of the EPIC-Norfolk cohort reporting a reduced risk of T2D from both greater dietary diversity across major food groups and greater diversity within vegetables [9, 10]. However, this literature is limited by the use of single populations, omission of known confounders, and a focus on only total dietary diversity [13]. We also found that this literature does not consider important effect modifiers of sex and weight status [13].
The diversity of protein-rich foods is crucial for health and managing T2D because consuming a range of amino acids supports the production of essential building blocks for cells, including pancreatic beta cells [14]. This diversity also facilitates vital biochemical functions such as tissue growth, repair, and the synthesis of hormones and enzymes necessary for insulin production and effective blood glucose regulation [14]. There is strong and consistent evidence that total protein consumption matters for T2D risk [15–18], but findings are mixed, depending on the protein source. A higher quantity of animal protein increases the risk of T2D whereas the association for the amount of plant protein is null [19]. Moreover, T2D risk varies by the subtype of animal protein. A higher quantity of red and processed meat increases the incidence of T2D whereas dairy products appear to be protective [15, 17]. Higher intakes of egg and white meat either reduce or do not alter T2D risk [15, 17]. Different protein-rich foods contain various bioactive compounds and the relative contribution and biological effects of protein intake may vary when a diet is highly diverse in plant- or animal-protein sources [20]. Yet, no studies have examined “diversity” in protein-rich foods based on dietary source. Beyond the quantity of protein-rich foods, protein “diversity” may offer complementary insights for T2D prevention. Understanding the role of diverse protein-rich food subtypes for developing T2D is timely given the 2019 EAT-Lancet report [21] and other national dietary guidelines [22] that promote a “diversity” of protein products and also ageing-related concerns about the maintenance of healthy muscle mass among older adults [23] and optimal bone health among vegetarians [24].
Both sex and adiposity can modify biological and socio-behavioral pathways linking dietary diversity to T2D, yet modifiers remain unexplored. Previous research indicates that differences in body composition and metabolic profiles contribute to sex-specific variations in the impact of diet on T2D [25]. Males typically have a higher percentage of lean body mass than females and may benefit more from a diverse diet because their higher muscle mass can enhance the metabolism of nutrients and improve insulin sensitivity more effectively. Obesity may also modify the association of dietary diversity and T2D through interactions involving inflammation, oxidative stress, hormonal factors [26], microbiome changes [27], and metabolic dysregulation [28, 29]. Obesity induces chronic low-grade inflammation and heightened oxidative stress [26], which suggests that the anti-inflammatory effects of greater dietary diversity could more effectively alleviate these adverse effects in individuals with obesity.
This multicountry, population-based study aimed to investigate the prospective association of multiple dietary diversity scores with incident T2D and test whether sex and obesity are possible modifiers. We hypothesized that (i) dietary diversity will be associated with incident T2D; and (ii) sex and adiposity will modify the association, with greater diversity overall and within groups more strongly associated with lower T2D incidence in males or people with obesity.
Methods
Study design and population
Data from the EPIC-InterAct case–cohort study provided a final sample size of 23 649 (including 10 363 incident cases of T2D) for complete-case analysis over a mean follow-up duration of 9.90 years (234 324 person-years). Details of data collection, sample selection, and country-specific sample sizes and dietary assessment methods are available in Supplementary Methods, Supplementary Figure S1, and Supplementary Tables S1 and S2.
Dietary diversity scores
Five diversity scores were derived from self-reported diet data (gr/day) (Table 1). Details on the methods and examples of the scoring approach are given in Supplementary Methods and Supplementary Tables S3 and S4. Scores had small to medium correlations (range: 0.18 to 0.35), except for meat/alternative diversity and animal-protein diversity (Pearson’s r = 0.80) (Supplementary Table S5).
Table 1.
Summary of dietary diversity scores evaluated in the EPIC-InterAct Study.
| Diversity measure | Diet diversity score (DDS) type | Diversity score component |
|---|---|---|
| Diversity across food groups (total dietary diversity) | DDS-total5 (0–5), diversity across five food groups | Meat and alternativea; dairy products; grains; fruits; vegetables |
| Diversity within a food group (vegetable diversity; meat and alternative diversity; protein diversity) | DDS-veg (0–4), diversity across four subtypes of vegetables | Fruiting vegetables; leafy vegetables; root vegetables; other vegetables |
| DDS-meat (0–6), diversity across six subtypes of meat/protein foods | Red meat and processed meat; organ meat; poultry; fish and seafood; eggs; legumes/beans, nuts, and seeds | |
| DDS-ProtA (0–8), diversity across eight subtypes of animal-based protein foods | Red meat; processed meat; organ meat; poultry; fish and seafood; eggs; milk and dairy products; cheese | |
| DDS-ProtP (0–5), diversity across subtypes of five plant-based protein foods | Legumes/beans; nuts and seeds; grains—rice and pasta; grains—bread; grains—other cerealsb |
Includes red meat, processed meat, organ meat, poultry, fish and seafood, eggs, legumes/beans, nuts, and seeds.
Includes flour, flakes, starches, semolina, breakfast cereals, salty biscuits, aperitif biscuits, crackers, dough, and pastry.
Statistical analysis
Means (standard deviations) and frequencies (proportions) described the characteristics of the sub-cohort (n = 13 937). Prentice-weighted Cox regression using age as a timescale was used to examine prospective associations of each dietary diversity score with T2D incidence. Different models were specified based on a causal directed acyclic graph (Supplementary Figure S2). Multivariable meta-analysis pooled country-specific hazard ratios (HRs) that had been adjusted for measured covariates in each respective country. A continuous variable for each diversity score was used to test for linear trend. Country-specific findings by using random-effects meta-analysis are shown on forest plots. Results are provided as HRs and 95% confidence intervals (CIs).
Effect modification was assessed by using Cox regression analyses of final adjusted models stratified by potential effect modifiers, such as sex, body mass index (BMI), and waist circumference (WC). A multivariable meta-analysis approach was employed in sex-stratified analyses, whereas a simple-effects Prentice-weighted Cox regression was used for the obesity-stratified model due to data constraints. Obesity models were mutually adjusted for BMI and WC. Multiplicative interaction between dietary diversity scores and sex or obesity (continuous measures of BMI and WC) in the most adjusted model tested for group differences by using the post-calculated omnibus test (testparm). A series of robustness checks extended final models to account for additional potential confounders and considerations (details in Supplementary Methods). Statistical analyses were performed by using Stata version 15.1.
Results
Diversity score distribution, country-specific average scores by sex and age, and overall characteristics in the sub-cohort sample are reported in Supplementary Results, Supplementary Figure S3, and Supplementary Tables S6–S8. The total diet diversity across food groups was socially patterned by using key socio-demographic, lifestyle, and health characteristics (Table 2). The mean daily consumption of foods and nutrients differed across levels of dietary diversity (Supplementary Table S9).
Table 2.
Baseline characteristics across total dietary diversity score in the EPIC-InterAct sample.
|
Dietary diversity score of an overall diet
g
| ||||
|---|---|---|---|---|
| Means (SD) or frequency (%) | ≤3 (n = 3486) | 4 (n = 5933) | 5 (n = 4518) | P-value h |
| Sociodemographic characteristics | ||||
| Age at recruitment (years) | 52.3 (9.2) | 51.6 (9.2) | 52.4 (8.9) | <0.001 |
| Sex [n (%) of females] | 2232 (64%) | 3747 (63%) | 2845 (63%) | 0.58 |
| Education [n (%)] | <0.001 | |||
| None | 285 (8) | 535 (9) | 340 (8) | |
| Primary school | 1262 (36) | 1855 (31) | 1392 (31) | |
| Technical/professional school | 821 (24) | 1300 (22) | 1004 (22) | |
| Secondary school | 480 (14) | 966 (16) | 704 (16) | |
| University degree | 607 (17) | 1231 (21) | 1043 (23) | |
| Not specified | 31 (1) | 46 (1) | 35 (1) | |
| Marital statusa [n (%)] | 0.61 | |||
| Single | 227 (10) | 341 (9) | 244 (8) | |
| Married/living together | 1746 (77) | 2936 (78) | 2257 (78) | |
| Divorced/separated | 194 (9) | 304 (8) | 239 (8) | |
| Widowed | 98 (4) | 169 (5) | 136 (5) | |
| Employment statusb [n (%)] | 0.002 | |||
| Employed | 967 (67) | 1628 (68) | 1250 (69) | |
| Housewife | 105 (7) | 170 (7) | 80 (4) | |
| Retired | 249 (17) | 436 (18) | 358 (17) | |
| Unemployed | 86 (6) | 116 (5) | 99 (5) | |
| Student | 11 (1) | 15 (1) | 19 (1) | |
| Other | 21 (1) | 35 (1) | 13 (1) | |
| History of pregnancye [n (%)] | 1899 (89) | 3230 (89) | 2458 (89) | 0.65 |
| Menopausal status [n (%)] | 0.01 | |||
| Premenopausal | 726 (33) | 1340 (36) | 939 (33) | |
| Postmenopausal | 1085 (49) | 1646 (44) | 1333 (47) | |
| Perimenopausal | 342(15) | 629 (17) | 460 (16) | |
| Surgical menopausef | 79 (4) | 132 (4) | 113 (4) | |
| Lifestyle characteristics | ||||
| Physical activity [n (%)] | <0.001 | |||
| Inactive | 924 (27) | 1460 (25) | 937 (21) | |
| Moderately inactive | 1223 (35) | 1976 (33) | 1502 (33) | |
| Moderately active | 763 (22) | 1334 (22) | 1055 (23) | |
| Active | 576 (17) | 1163 (20) | 1024 (23) | |
| Smoking status [n (%)] | <0.001 | |||
| Never | 1525 (44) | 2758 (46) | 2385 (53) | |
| Former | 886 (25) | 1601 (27) | 1271 (28) | |
| Current | 1059 (30) | 1547 (26) | 841 (19) | |
| Unknown | 16 (0) | 27 (0) | 21 (0) | |
| Alcohol-drinking history [n (%)] | 0.01 | |||
| Never | 308 (9) | 442 (7) | 332 (7) | |
| Former | 178 (5) | 347 (6) | 266 (6) | |
| Only at recruitment | 161 (5) | 251 (4) | 181 (4) | |
| Lifetime | 2164 (62) | 3598 (61) | 2746 (61) | |
| Unknown | 675 (19) | 1295 (22) | 993 (22) | |
| History of diseases | ||||
| Myocardial infarctione [n (%)] | 56 (2) | 64 (1) | 47 (1) | 0.07 |
| Strokee [n (%)] | 37 (1) | 38 (1) | 25 (1) | 0.02 |
| Hypertensione [n (%)] | 648 (19) | 1052 (18) | 821 (18) | 0.005 |
| Dyslipidemiae [n (%)] | 505 (17) | 855 (17) | 640 (17) | 0.52 |
| Cancer n (%) | 96 (3) | 185 (3) | 141 (3) | 0.56 |
| Family history of diabetesc [n (%)] | 354 (19) | 559 (18) | 499 (20) | 0.30 |
| Anthropometric characteristics | ||||
| Height (cm) | 165.24 (9.07) | 165.94 (9.29) | 166.31 (9.39) | <0.001 |
| Weight (kg) | 71.08 (13.20) | 71.73 (13.41) | 72.41 (13.23) | <0.001 |
| Waist circumference (WC)d (cm) | 86.14 (12.68) | 86.19 (12.58) | 86.33 (12.40) | 0.76 |
| Central obesity (WC ≥ 88 cm for females and WC ≥ 102 for males) | 1113 (32) | 1761 (30) | 1307 (29) | 0.01 |
| BMI (kg/m2) | 25.98 (4.15) | 26.01 (4.20) | 26.15 (4.18) | 0.14 |
| Overweight, BMI = 25–29.9 kg/m2 [n (%)] | 1369 (39) | 2347 (40) | 1860 (41) | |
| Obesity, BMI ≥ 30 kg/m2 [n (%)] | 550 (16) | 934 (16) | 730 (16) | 0.23 |
EPIC-InterAct sample size is n = 23 649.
Excludes Spain and Denmark, for which information on marital status was not collected.
Excludes France, Italy, Spain, Denmark, and Utrecht (Netherlands), for which information on employment status was not collected.
Excludes Italy, Spain, Oxford (UK), and Heidelberg (Germany), for which information on family history of diabetes was not collected.
Excludes Umeå (Sweden), for which information on WC was not collected.
Missing data: history of pregnancy (n = 303), myocardial infarction (n = 178), stroke (n = 1104), hypertension (n = 43), hyperlipidemia (n = 2326).
Bilateral ovariectomy.
Number of major food groups (0–5) consumed more than the median cut points.
Calculated by analysis of variance and chi-squared test for continuous and categorical variables, respectively.
Across countries, consuming four to five subtypes of plant-protein sources was associated with a lower incidence of T2D [HR 0.78 (95% CI 0.65, 0.93)] compared with no consumption (Table 3). There was also a 4% lower incidence of T2D with every new subtype of plant-based protein consumed [0.96 (0.93, 1.00)] (Supplementary Table S10). Sex and obesity also modified the association of plant-protein diversity with T2D (Tables 4 and 5). There was a consistent inverse association of plant-protein diversity with incident T2D in females and no clear pattern in males. Females consuming three subtypes of plant protein had a 25% lower incidence of T2D [0.75 (0.62, 0.90)] compared with no consumption. Notably, sex differences were not detected statistically based on CIs of one group overlapping with HRs of the other group and an omnibus test (P = 0.35). Individual P-values for interactions were all nonsignificant. Individuals both with and without obesity showed an inverse association between plant-protein diversity and incident T2D, but only participants without central obesity reporting five subtypes had an 18% lower incidence [0.82 (0.68, 1.00)] compared with no intake. Associations were different between the groups with and without obesity (P = 0.04). Country-specific findings are shown in Supplementary Figure S4.
Table 3.
Adjusted HRs (95% CIs) of incident T2D for multiple diversity scores in the EPIC-InterAct sample.
| Diet diversity score a | Number of groups | Cases/total | Model 1 HR (95% CI) | Model 2 HR (95% CI) | Model 3 HR (95% CI) | Model 4 HR (95% CI) |
|---|---|---|---|---|---|---|
| Diversity of food groups | 0–3 | 2827/6144 | 1.00 | 1.00 | 1.00 | 1.00 |
| 4 | 4377/10 030 | 0.95 (0.82, 1.10) | 1.00 (0.86, 1.17) | 1.01 (0.86, 1.18) | 0.95 (0.81, 1.11) | |
| 5 | 3159/7475 | 0.84 (0.75, 0.95) | 0.91 (0.80, 1.04) | 0.95 (0.82, 1.09) | 0.86 (0.75, 0.98) | |
| Vegetable diversity | 0–1 | 3803/8670 | 1.00 | 1.00 | 1.00 | 1.00 |
| 2 | 2788/6335 | 0.98 (0.90, 1.06) | 1.00 (0.93, 1.08) | 1.04 (0.95, 1.13) | 0.98 (0.84, 1.16) | |
| 3 | 2540/5724 | 0.96 (0.84, 1.11) | 1.00 (0.88, 1.14) | 1.04 (0.90, 1.19) | 1.00 (0.81, 1.23) | |
| 4 | 1232/2920 | 0.88 (0.78, 0.99) | 0.92 (0.83, 1.02) | 0.98 (0.89, 1.07) | 0.90 (0.83, 0.98) | |
| Meat diversity | 0 | 386/966 | 1.00 | 1.00 | 1.00 | 1.00 |
| 1 | 1316/3257 | 0.90 (0.74, 1.09) | 0.91 (0.75, 1.11) | 0.94 (0.77, 1.14) | 0.88 (0.68, 1.14) | |
| 2 | 2354/5442 | 1.11 (0.95, 1.29) | 1.13 (0.96, 1.32) | 1.17 (0.99, 1.39) | 1.10 (0.86, 1.40) | |
| 3 | 2736/6021 | 1.20 (0.99, 1.46) | 1.23 (1.01, 1.50) | 1.30 (1.06, 1.59) | 1.05 (0.81, 1.37) | |
| 4 | 2109/4806 | 1.16 (0.93, 1.44) | 1.17 (0.93, 1.47) | 1.24 (0.97, 1.57) | 0.98 (0.71, 1.35) | |
| 5–6 | 1462/3157 | 1.33 (1.11, 1.59) | 1.38 (1.15, 1.65) | 1.51 (1.25, 1.83) | 1.13 (0.91, 1.39) | |
| Animal-protein diversity | 0–1 | 779/1913 | 1.00 | 1.00 | 1.00 | 1.00 |
| 2 | 1400/3274 | 1.07 (0.90, 1.27) | 1.08 (0.90, 1.30) | 1.10 (0.91, 1.33) | 0.99 (0.74, 1.33) | |
| 3 | 2017/4664 | 1.07 (0.94, 1.23) | 1.09 (0.93, 1.27) | 1.14 (0.95, 1.37) | 0.99 (0.71, 1.39) | |
| 4 | 2277/5244 | 1.16 (0.99, 1.36) | 1.17 (0.99, 1.38) | 1.23 (1.04, 1.45) | 0.98 (0.73, 1.31) | |
| 5 | 1951/4428 | 1.21(1.01, 1.44) | 1.25 (1.04, 1.50) | 1.31 (1.08, 1.59) | 0.96 (0.72, 1.29) | |
| 6–8 | 1939/4126 | 1.41 (1.20, 1.66) | 1.45 (1.22, 1.73) | 1.55 (1.29, 1.85) | 1.08 (0.80, 1.48) | |
| Plant-protein diversity | 0 | 851/1732 | 1.00 | 1.00 | 1.00 | 1.00 |
| 1 | 2339/4992 | 0.92 (0.82, 1.03) | 0.93 (0.81, 1.07) | 0.94 (0.79, 1.11) | 0.92 (0.73, 1.16) | |
| 2 | 3087/6851 | 0.85 (0.73, 0.98) | 0.88 (0.76, 1.02) | 0.90 (0.77, 1.06) | 0.94 (0.82, 1.07) | |
| 3 | 2462/5798 | 0.80 (0.71, 0.91) | 0.86 (0.74, 1.01) | 0.90 (0.76, 1.08) | 0.90 (0.69, 1.17) | |
| 4–5 | 1624/4276 | 0.68 (0.56, 0.82) | 0.73 (0.61, 0.88) | 0.79 (0.64, 0.97) | 0.78 (0.65, 0.93) |
HRs and 95% CIs were estimated by country, adjusted for age [2], sex, within-country regions, and family history of diabetes in Model 1; Model 2, further adjusted for education level, employment status, and marital status; Model 3, further adjusted for smoking status, alcohol intake, and physical activity; and Model 4, further adjusted for energy intake and BMI. Final sample size is 23 649. Country-specific estimates were pooled with a random-effects meta-analysis (Supplementary Figure S4). Bold, statistically significant (p<0.05).
Dietary diversity scores were constructed based on the number of unique food sub/groups in a diet or the number of unique food subtypes within a food group.
Table 4.
Adjusted HRs (95% CIs) of incident T2D for multiple diversity scores by sex in the EPIC-InterAct sample.
| Males (n = 9864) |
Females (n = 13 785) |
|||||
|---|---|---|---|---|---|---|
| Diet diversity score a | Number of groups | Cases/total | HR (95% CI) | Cases/total | HR (95% CI) | P for omnibus test b |
| Diversity of food groups | 0–3 | 1456/2629 | 1.00 | 1371/3515 | 1.00 | |
| 4 | 2109/4151 | 0.89 (0.67, 1.18) | 2268/5879 | 0.99 (0.90, 1.09) | ||
| 5 | 1512/3084 | 0.84 (0.68, 1.04) | 1647/4391 | 0.86 (0.77, 0.96) | 0.58 | |
| Vegetable diversity | 0–1 | 1957/3730 | 1.00 | 1846/4940 | 1.00 | |
| 2 | 1368/2609 | 0.95 (0.76, 1.18) | 1420/3726 | 0.95 (0.76, 1.20) | ||
| 3 | 1179/2353 | 0.85 (0.73, 0.99) | 1361/3371 | 1.03 (0.76, 1.41) | ||
| 4 | 573/1172 | 0.84 (0.71, 1.00) | 659/1748 | 0.93 (0.77, 1.11) | 0.40 | |
| Meat diversity | 0 | 183/375 | 1.00 | 203/591 | 1.00 | |
| 1 | 650/1358 | 0.68 (0.31, 1.50) | 666/1899 | 0.89 (0.56, 1.42) | ||
| 2 | 1176/2279 | 0.80 (0.31, 2.06) | 1178/3163 | 1.10 (0.78, 1.56) | ||
| 3 | 1341/2552 | 0.80 (0.36, 1.76) | 1395/3469 | 1.12 (0.70, 1.81) | ||
| 4 | 1028/2.001 | 0.75 (0.33, 1.73) | 1081/2805 | 1.01 (0.64, 1.60) | ||
| 5–6 | 699/1299 | 1.04 (0.62, 1.77) | 763/1858 | 1.08 (0.74, 1.58) | 0.49 | |
| Animal-protein diversity | 0–1 | 388/759 | 1.00 | 391/1154 | 1.00 | |
| 2 | 728/1417 | 0.93 (0.60, 1.44) | 672/1857 | 1.07 (0.76, 1.49) | ||
| 3 | 997/1967 | 0.97 (0.63, 1.50) | 1020/2697 | 1.13 (0.76, 1.67) | ||
| 4 | 1120/2208 | 0.93 (0.55, 1.56) | 1157/3036 | 1.02 (0.78, 1.32) | ||
| 5 | 954/1852 | 1.05 (0.62, 1.77) | 997/2576 | 0.99 (0.77, 1.27) | ||
| 6–8 | 890/1661 | 1.11 (0.62, 1.99) | 1049/2465 | 1.24 (0.92, 1.67) | 0.71 | |
| Plant-protein diversity | 0 | 377/656 | 1.00 | 474/1076 | 1.00 | |
| 1 | 1144/2064 | 1.06 (0.85, 1.33) | 1195/2928 | 0.74 (0.50, 1.10) | ||
| 2 | 1581/3017 | 1.01 (0.82, 1.25) | 1506/3834 | 0.85 (0.64, 1.12) | ||
| 3 | 1240/2471 | 1.14 (0.86, 1.52) | 1222/3327 | 0.75 (0.62, 0.90) | ||
| 4–5 | 735/1656 | 0.73 (0.51, 1.05) | 889/2620 | 0.80 (0.58, 1.11) | 0.35 | |
HRs and 95% CIs were obtained by using Prentice-weighted Cox regression with age as the underlying timescale, first by country and then combined by multivariable meta-analysis: n cases = 10 363 (5077 males and 5286 females). HR estimates were adjusted for age [2], region, education level, smoking status, alcohol intake, physical activity, energy intake, and BMI (n = 23 649). Bold, statistically significant (p<0.05).
Dietary diversity scores were constructed based on the number of unique food sub/groups in a diet or the number of unique food subtypes within a food group.
Statistical test for group differences used multiplicative interaction between dietary diversity scores and sex in the most adjusted model, with the post-calculated omnibus test (testparm).
Table 5.
Adjusted HRs (95% CIs) of incident T2D for multiple diversity scores by central obesity status in the EPIC-InterAct sample.
|
Normal WC
b
(n = 13 367) |
Central obesity
b
(n = 8448) |
|||||
|---|---|---|---|---|---|---|
| Diet diversity score a | Number of groups | Cases/total | HR (95% CI) | Cases/total | HR (95% CI) | P for omnibus test c |
| Diversity of food groups | 0–3 | 1096/3377 | 1.00 | 1464/2185 | 1.00 | |
| 4 | 1624/5653 | 0.83 (0.75, 0.93) | 2439/3611 | 1.04 (0.92, 1.17) | ||
| 5 | 1233/4337 | 0.79 (0.70, 0.89) | 1708/2652 | 0.87 (0.76, 1.00) | 0.70 | |
| Vegetable diversity | 0–1 | 1438/4768 | 1.00 | 1926/2924 | 1.00 | |
| 2 | 1058/3621 | 0.94 (0.85, 1.05) | 1538/2276 | 1.04 (0.92, 1.18) | ||
| 3 | 974/3281 | 0.98 (0.88, 1.10) | 1442/2142 | 1.02 (0.90, 1.16) | ||
| 4 | 483/1697 | 0.95 (0.82, 1.10) | 705/1106 | 0.89 (0.76, 1.05) | 0.99 | |
| Meat diversity | 0 | 133/547 | 1.00 | 201/307 | 1.00 | |
| 1 | 469/1796 | 1.07 (0.84, 1.36) | 655/1006 | 0.84 (0.63, 1.11) | ||
| 2 | 904/3074 | 1.32 (1.04, 1.66) | 1211/1837 | 0.89 (0.68, 1.16) | ||
| 3 | 1096/3421 | 1.39 (1.10, 1.75) | 1476/2213 | 0.86 (0.66, 1.12) | ||
| 4 | 838/2777 | 1.31 (1.04, 1.67) | 1184/1811 | 0.83 (0.63, 1.09) | ||
| 5–6 | 513/1752 | 1.29 (1.00, 1.67) | 884/1274 | 0.90 (0.67, 1.20) | 0.12 | |
| Animal-protein diversity | 0–1 | 271/1015 | 1.00 | 329/502 | 1.00 | |
| 2 | 498/1758 | 1.04 (0.86, 1.27) | 707/1074 | 0.96 (0.76, 1.22) | ||
| 3 | 789/2632 | 1.12 (0.93, 1.35) | 1059/1635 | 0.91 (0.73, 1.14) | ||
| 4 | 881/3030 | 1.05 (0.88, 1.27) | 1261/1914 | 0.92 (0.73, 1.15) | ||
| 5 | 806/2613 | 1.12 (0.93, 1.36) | 1086/1651 | 0.84 (0.67, 1.07) | ||
| 6–8 | 708/2319 | 1.19 (0.97, 1.46) | 1169/1672 | 0.95 (0.74, 1.21) | 0.01 | |
| Plant-protein diversity | 0 | 342/994 | 1.00 | 499/713 | 1.00 | |
| 1 | 911/2846 | 0.94 (0.79, 1.11) | 1321/1930 | 1.00 (0.82, 1.22) | ||
| 2 | 1241/3947 | 0.93 (0.79, 1.11) | 1601/2406 | 0.95 (0.78, 1.15) | ||
| 3 | 914/3262 | 0.92 (0.77, 1.09) | 1306/1970 | 0.95 (0.78, 1.16) | ||
| 4–5 | 545/2318 | 0.82 (0.68, 1.00) | 884/1429 | 0.85 (0.69, 1.06) | 0.04 | |
HRs and 95% CIs were obtained by using Prentice-weighted Cox regression with age as the underlying timescale (total cases: n = 9564), adjusted for age [2], sex, region, education level, smoking status, alcohol intake, physical activity, energy intake, and BMI (n = 21 815). Bold, statistically significant (p<0.05).
Dietary diversity scores were constructed based on the number of unique food sub/groups in a diet or the number of unique food subtypes within a food group.
Normal WC, <88 cm for women and <102 for men; central obesity, WC of ≥88 cm for females and ≥102 cm for males.
Statistical test for group differences used multiplicative interaction between dietary diversity scores and central obesity in the most adjusted model, with the post-calculated omnibus test (testparm).
For total dietary diversity and vegetable diversity, the greatest diversity level was associated with lower T2D incidence [0.86 (0.75, 0.98) and 0.90 (0.83, 0.98), respectively] compared with the lowest scores (Table 3). Each additional major food group in the daily diet was associated with a lower T2D incidence of 9%; however, no linear association between vegetable diversity and T2D incidence was observed (Supplementary Table S10). Associations were also modified by sex and obesity, although overall tests of group difference were nonsignificant. Five food groups reduced T2D incidence in females [0.86 (0.77, 0.96)], with a similar pattern in males. Three or more subtypes of vegetables reduced T2D incidence only in males by 15%– 16% (Table 4). Similarly, four and five major food groups reduced T2D incidence by 17% and 21% in those without central obesity, but associations for vegetable diversity were not modified by WC (Table 5). No clear associations were found for the diversity of meat/alternative subtypes and animal-protein subtypes. However, increasing meat and alternative diversity was associated with incident T2D in those without central obesity (HR range: 1.29–1.39).
Robustness checks showed that adjusted HRs were largely unchanged for total, vegetable, and plant-protein diversity, although some associations lost significance after adjustment for quantity, reproductive history, or baseline HbA1c (Supplementary Tables S11–S15). Associations were attenuated when considering other diversity scores for plant-protein diversity in the overall sample and the subpopulation without central obesity, for total dietary diversity in females, and for vegetable diversity in males.
General obesity results and sensitivity analyses are provided in Supplementary Tables S16–S18.
Discussion
This multi-country, prospective study examined multiple dietary diversity scores and incident T2D in Europeans overall, by sex, and by obesity status. Overall, consuming four or five different subtypes of plant-protein sources was consistently linked to a lower risk of developing T2D compared with no intake. Consuming five major food groups and four subtypes of vegetables was also associated with lower T2D risk. Results also revealed effect modification by sex for total dietary diversity, vegetable diversity, and plant-protein diversity. The associations of plant-protein diversity and T2D risk were also modified by central obesity status. Findings for total dietary diversity were robust to multiple confounders; however, some results for within-food-group diversity were sensitive to model specification.
No other study has considered separate measures for animal-protein diversity and plant-protein diversity. One previous study assessed sources of protein but combined animal and plant protein into a single diversity score [30]. This study demonstrated that consuming four or five subtypes of plant-protein diversity reduced the risk of T2D by 22% in Europeans across multiple countries, even after adding additional confounders except other diversity scores. Meta-analyses of the quantity of plant protein revealed that the link to T2D risk was either restricted to females [16] or was a U-shaped pattern [18]; however, a previous InterAct study found no association for both sexes [19]. Our results differ because consuming a “variety” of plant-based proteins might ensure the intake of necessary amino acids and correlate with improved nutritional status and microbial diversity [31]—all of which could influence diabetes risk [30, 32]. Thus, our work may help to better understand the unique role of plant-protein “diversity” for diabetes, as diets that are high in plant-protein quantity may still lack diverse foods with key bioactive compounds for metabolic and/or microbial health.
Results of total dietary diversity add a new multicountry perspective to the mixed evidence on dietary diversity and risk of T2D [13]. Of the six medium-quality longitudinal studies on this topic, four report no association between the diversity of individual food items or food subgroups with T2D [13]. Nevertheless, previous research using the EPIC-Norfolk cohort (part of InterAct) found that total dietary diversity was linked to a larger risk reduction (30%) in British adults [9]. This study extends knowledge by showing that this prospective association is replicated among Northern (e.g. Sweden, Denmark) and Southern (e.g. Spain) Europeans. Other cross-sectional evidence also suggests that a diet diverse in five main food groups improves glucose homeostasis [33] and diabetes-specific mortality [34]. Thus, there may be an optimal number of food groups for a diverse diet to support diabetes prevention.
Results for vegetable diversity corroborate existing research [9, 10]. In this study, daily intake of four subtypes of vegetables lowered the risk of T2D by 10%, which is smaller than the 23%–33% lower risk previously reported for diets with 12 different vegetables or 4 subtypes of vegetables in British adults [9, 10]. Differences in observed effect sizes may be due to averaging across multiple populations and/or scoring approaches and thresholds for counting vegetable subtypes toward a diversity score (g/d versus number/week) [9, 10]. The hazard ratio was mildly attenuated after mutually adjusting other diversity scores, but were robust to adding quantity and quality; both confounders were omitted in earlier work reporting no association between vegetable diversity and T2D [35].
Some effect modification was found. Greater vegetable diversity appeared to lower T2D risk in males while greater plant-protein diversity lowered T2D risk in females. Unlike females, males are more susceptible to oxidative stress and have lower antioxidant capacity, as they are not protected by estrogen [36]. Males may therefore gain more benefit than females from a diet that is diverse in vegetables that are high in antioxidants. However, some caution is needed, as results for vegetable diversity in males were altered in both directions, depending on the additional confounders; in particular, the association was not independent of other dietary diversity scores. Only females showed a reduced T2D risk with five food groups and three plant-protein subtypes, even after accounting for quantity of intake, diet quality, and other dietary diversity scores. Thus, female Europeans could be encouraged to consume all five food groups along with three subtypes of plant protein.
Contrary to our hypothesis, the association of total dietary diversity and plant-protein diversity with T2D was observed in those without central obesity, although the magnitudes and directions of association were similar in people with central obesity. Due to fewer people with central obesity reporting the highest dietary diversity score, CIs were wide. We noted that plant-protein diversity was more strongly associated with T2D in Europeans without central obesity after adjusting for quantity, quality, and fat ratio, but was not independent of other dietary diversity scores. Participants with and without obesity may consume food with different processing and preparation methods that were not accounted for in this study. Moreover, the incidence of T2D may differ for metabolically healthy and unhealthy obesity, and future work should consider obesity phenotypes. The role of plant-protein diversity for diabetes prevention for subpopulations warrants further investigation.
The biological mechanism linking dietary diversity to T2D requires further research. Experimental studies show that greater dietary diversity can improve the diversity of gut microbiota and protect against metabolic disorders [37]. A diverse diet is known to ensure various essential nutrients necessary for optimal physiological functioning [38, 39]. Different plant-protein sources are also rich in dietary fiber, often have varying amino acid profiles, and contain a wide range of phytochemicals and antioxidants, each offering unique health benefits [40]. By incorporating a diverse array of plant-protein sources, we may increase the likelihood of obtaining a more complete set of essential amino acids and expose the body to a broad spectrum of these essential nutrients, thereby promoting overall health [40].
Overall, this study corroborates the American [22] and Australian [41] Dietary Guideline recommendations to consume a diversity of “five” food groups and provide a basis for future updates to other dietary guidelines that recommend a diet with only three food groups (e.g. Canada’s Food Guide 2023). These novel results support dietary interventions that consider diversity in protein-rich foods based on their food sources. Importantly, the 2019 EAT-Lancet Commission has recommended the adoption of a “universal healthy reference diet” including greater “diversity” of plant-based foods to improve human health and planetary sustainability [21]. However, this recommendation focuses on amount rather than diversity, as we found no citation to empirical evidence supporting this statement on diversity independently of quantity. Our study is therefore the first to generate empirical support for this recommendation. Although greater diversity of plant-protein subtypes may also ensure a higher quantity of intake, we showed that greater variety of plant-protein subtypes is inversely associated with T2D risk, independently of both quantity and diet quality. We urge greater clarity in future reports regarding quantity versus variety of items for “healthy eating” recommendations.
Multiple strengths of this study are noted. First, the prospective design reduces the concern for recall bias in the assessment of dietary intakes and is more amenable to causal inference. Second, the large sample size and high number of incident cases improve the precision of association estimates. Third, external verification of incident cases minimizes outcome misclassification. Additionally, a wide range of co-variables helped to mitigate bias due to confounding. In particular, total energy intake (kcal) addressed confounding of dietary factors that were not included in the diversity scores, such as sugar-sweetened beverages that are a major source of added sugars and excess caloric intake. Fourth, the multicountry dataset enhances the external validity of the findings. Fifth, we used a Food and Agriculture Organization-based food-group classification, ensuring comparability with previous research. Sixth, we assessed diversity both between and within food groups by using counts so that the findings can inform food-based nutrition education and diabetes-prevention messaging. Finally, our country- and sex-specific median cut points helped to enhance precision and cultural relevance by tailoring diversity scores to our population’s behaviors.
Limitations of this work include exposure-related bias due to data collection, reporting, and scoring. In particular, the approach of using median cut points for diversity scores limits generalizability and cross-study comparisons, as the cut points do not provide precise values for targeted dietary interventions. Food-frequency questionnaires capture the relative habitual intakes, and not absolute intake. Therefore, the inference based on absolute dietary intakes is limited in this study, while ranking individuals in terms of dietary diversity was considered adequate. Misclassification bias in an unknown direction may happen due to the single baseline diet assessment that ignored intra-individual variation over time. Reporting bias may stem from social patterning in which socioeconomic status influences diet reporting, although models included multiple socioeconomic status variables. Misclassification in scoring diversity may have occurred, as we used a quantity-based approach instead of the scoring approach using frequency for consumption thresholds (e.g. 1/week, 2/week, etc.). Nevertheless, our score counted food groups commonly used in the literature. Aggregated dietary data with no detailed information (e.g. food subtypes, cooking method, etc.) prevented our assessment of diversity within certain food groups (e.g. dairy) and stratification of total dietary diversity based on food healthfulness. Additionally, the high correlation between diversity scores for animal protein and meat/alternatives indicated strong overlap and redundancy of the exposure, making it difficult to discern the construct validity of meat/alternatives diversity. Nevertheless, the inclusion of a meat/alternatives diversity score that incorporates nonmeat protein items was done for comparability with prior research. Furthermore, potential misclassification of participants with undiagnosed T2D as nondiabetic (false negatives) might also underestimate the risk. However, reverse causation is unlikely, as prevalent cases of T2D were excluded at the cohort baseline and a sensitivity analysis excluded participants with undiagnosed T2D (HbA1c ≥ 6.5%). Finally, results may not be generalizable to non-European populations with a different diet, body composition, and T2D incidence.
Conclusion
Results highlight the importance of a diet with five food groups and the unique role of four or more plant-protein subtypes for diabetes prevention in Europe. Specific subpopulations, such as females and individuals without central obesity, may benefit most from greater overall and within-group diversity. This research offers cross-national evidence to support the dietary guideline recommendations of eating foods from five different food groups and incorporating a variety of vegetables and also plant-based proteins in the diet.
Ethics approval
All participants provided informed consent, and the EPIC-InterAct Study was approved by the local ethics committee in the participating countries and the Internal Review Board of the International Agency for Research on Cancer. The Behavioral Research Ethics Board of the University of British Columbia also approved the secondary data analysis of EPIC-InterAct data for the current study.
Supplementary Material
Acknowledgements
H.M. is funded by a Canadian Institutes of Health Research (CIHR) Vanier Scholarship (2021–24) and a UBC Four-Year Doctoral Fellowship (2020–23). A.I.C. is funded by a Michael Smith Foundation for Health Research (MSFHR) Scholar Award (2021–26). R.A.M. is funded by a MSFHR Scholar Award (grant #17644; 2018–2023). M.J. is funded by a Canada Research Chair (Tier 2). D.B.I. is supported by the Independent Research Fund Denmark (grant 1057-00016B). We thank all EPIC participants and staff for their contribution to the study. We also thank Nicola Kerrison (MRC Epidemiology Unit, Cambridge) for managing the data for the InterAct Project. Funding for the InterAct Project was provided by the EU FP6 program (grant number LSHM_CT_2006_037197). In addition, InterAct investigators acknowledge funding from the following agencies: IS, JWJB, and YTvdS. Verification of diabetes cases was additionally funded by NL Agency grant IGE05012 and an Incentive Grant from the Board of the UMC Utrecht (Netherlands); HBBdM, AMWS, and DLvdA: Dutch Ministry of Public Health, Welfare and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), Statistics Netherlands (Netherlands); FLC: Cancer Research UK; PWF: Swedish Research Council, Novo nordisk, Swedish Heart Lung Foundation, Swedish Diabetes Association; JH, CD and AT: Danish Cancer Society; RK: Deutsche Krebshilfe; SP: Associazione Italiana per la Ricerca sul Cancro; JRQ: Asturias Regional Government; MT: Health Research Fund (FIS) of the Spanish Ministry of Health; the CIBER en Epidemiología y Salud Pública (CIBERESP), Spain; Murcia Regional Government; RT: AIRE-ONLUS Ragusa, AVIS-Ragusa, Sicilian Regional Government.
Contributor Information
Hadis Mozaffari, Faculty of Land and Food Systems, University of British Columbia, Vancouver, Canada.
Fumiaki Imamura, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK.
Rachel A Murphy, School of Population and Public Health, University of British Columbia, Vancouver, BC, Canada; Cancer Control Research, BC Cancer Agency, Vancouver, BC, Canada.
Mahsa Jessri, Faculty of Land and Food Systems, University of British Columbia, Vancouver, Canada.
Stephen J Sharp, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK.
Nita G Forouhi, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK.
Nicholas J Wareham, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK.
Daniel B Ibsen, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Cambridge, UK; Department of Public Health, Aarhus University, Aarhus, Denmark; Steno Diabetes Center Aarhus, Aarhus University Hospital, Aarhus, Denmark; Department of Nutrition, Exercise and Sports, University of Copenhagen, Frederiksberg, Denmark.
Christina C Dahm, Department of Public Health, Aarhus University, Aarhus, Denmark.
José María Huerta, Department of Epidemiology, Murcia Regional Health Council-IMIB, Murcia, Spain; Centro de investigación Biomédica en Red de Epidemiología y Salud Pública (CIBERESP), Madrid, Spain.
Esther Molina-Montes, Centro de investigación Biomédica en Red de Epidemiología y Salud Pública (CIBERESP), Madrid, Spain; Department of Nutrition and Food Science, University of Granada, Granada, Spain; “José Mataix Verdú” Institute of Nutrition and Food Technology (INYTA), University of Granada, Granada, Spain; Instituto de Investigacion Biosanitaria de Granada (ibs.Granada), University of Granada, Granada, Spain.
Daniela V Nickel, Department of Molecular Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany; German Center for Diabetes Research (DZD), Neuherberg, Germany; Institute of Nutritional Science, University of Potsdam, Nuthetal, Germany.
Olov Rolandsson, Department of Public Health and Clinical Medicine, Family Medicine, Umeå University, 90187, Umeå, Sweden.
Carlotta Sacerdote, Unit of Cancer Epidemiology, Città della Salute e della Scienza University-Hospital, Turin, Italy.
Matthias B Schulze, Department of Molecular Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany; German Center for Diabetes Research (DZD), Neuherberg, Germany; Institute of Nutritional Science, University of Potsdam, Nuthetal, Germany.
Jon Ander Gonzalez-Martin, Osakidetza Basque Health Service, Cruces University Hospital, Preventive Medicine Deparment, Barakaldo, Spain.
Marcela Guevara, Centro de investigación Biomédica en Red de Epidemiología y Salud Pública (CIBERESP), Madrid, Spain; Instituto de Salud Pública y Laboral de Navarra, Gobierno de Navarra, Pamplona, Spain; Navarra Institute for Health Research (IdiSNA), University Hospital of Navarra (HUN), Pamplona, Spain.
Peter M Nilsson, Department of Clinical Sciences and Medicine, Lund University, University Hospital Malmö, Malmö, Sweden.
Salvatore Panico, Dipartimento di Medicina Clinica e Chirurgia, Federico II University, Naples, Italy.
Anna Winkvist, Sustainable Health, Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden.
Annalijn I Conklin, Faculty of Land and Food Systems, University of British Columbia, Vancouver, Canada; Collaboration for Outcomes Research and Evaluation, Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, Canada; Centre for Advancing Health Outcomes, Providence Healthcare Research Institute, St Paul’s Hospital, Vancouver, Canada.
Author contributions
A.I.C. and H.M. conceptualized the study. A.I.C. submitted the data request and proposal to EPIC-InterAct scientific committee. H.M., F.I., R.M., M.J., and A.I.C. designed the analysis. H.M. performed the data analysis. F.I., S.J.S., and A.I.C. contributed to data analysis. N.J.W., S.J.S., and N.G.F. coordinated the EPIC-InterAct Study and N.J.W. is the chief investigator. H.M. and A.I.C. drafted the original manuscript. All authors contributed to interpretation of the data and all authors revised the article critically for important intellectual content and approved the final version for publication. A.I.C. serves as the guarantor for the study and affirms responsibility for the overall integrity of the work, including the study design, data collection, analysis, and interpretation.
Supplementary data
Supplementary data is available at IJE online.
Conflict of interest: None declared.
Funding
No specific funding was received for this secondary data analysis study. Details of salary support and EPIC-InterAct funding are given in the acknowledgement section.
Data availability
EPIC-InterAct Study data described in the manuscript, code book, and analytic code cannot be deposited publicly, as these collaborative data originate from multiple research institutions across eight European countries with different legal frameworks. The authors confirm that researchers seeking the analysis dataset for this work can submit a data request to the EPIC-InterAct Study central contact point by emailing interact@mrc-epid.cam.ac.uk.
Use of artificial intelligence (AI) tools
None.
References
- 1. Sun H, Saeedi P, Karuranga S et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract 2022;183:109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Salas-Salvadó J, Martinez-Gonzalez M, Bulló M, Ros E. The role of diet in the prevention of type 2 diabetes. Nutr Metab Cardiovasc Dis 2011; 21(Suppl 2):B32–48. [DOI] [PubMed] [Google Scholar]
- 3. Thiele S, Mensink GB, Beitz R. Determinants of diet quality. Public Health Nutr 2004;7:29–37. [DOI] [PubMed] [Google Scholar]
- 4. Drescher LS, Thiele S, Mensink GB. A new index to measure healthy food diversity better reflects a healthy diet than traditional measures. J Nutr 2007;137:647–51. [DOI] [PubMed] [Google Scholar]
- 5. Krebs-Smith SM, Smiciklas-Wright H, Guthrie HA, Krebs-Smith J. The effects of variety in food choices on dietary quality. J Am Diet Assoc 1987;87:897–903. [PubMed] [Google Scholar]
- 6. Magklis E, Howe LD, Johnson L. Eating style and the frequency, size and timing of eating occasions: a cross-sectional analysis using 7-day weighed dietary records. Sci Rep 2019;9:15133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Manig C, Moneta A, More or better? Measuring quality versus quantity in food consumption. Papers Econ Evol 2009; Discussion Paper No. 2009–10.
- 8. Marriott BM, Effects of food quality, quantity, and variety on intake. In: BM Marriott, ed. Not Eating Enough: Overcoming Underconsumption of Military Operational Rations. Washington, DC: National Academies Press, 1995, 161–179. [PubMed] [Google Scholar]
- 9. Conklin AI, Monsivais P, Khaw K-T, Wareham NJ, Forouhi NG. Dietary diversity, diet cost, and incidence of type 2 diabetes in the United Kingdom: a prospective cohort study. PLoS Med 2016;13:e1002085. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Cooper AJ, Sharp SJ, Lentjes MA et al. A prospective study of the association between quantity and variety of fruit and vegetable intake and incident type 2 diabetes. Diabetes Care 2012;35:1293–300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Drewnowski A, Henderson SA, Shore A, Fischler C, Preziosi P, Hercberg S. Diet quality and dietary diversity in France: implications for the French paradox. J Am Diet Assoc 1996;96:663–9. [DOI] [PubMed] [Google Scholar]
- 12. Kennedy E, Ohls J, Carlson S, Fleming K. The healthy eating index: design and applications. J Am Diet Assoc 1995;95:1103–8. [DOI] [PubMed] [Google Scholar]
- 13. Mozaffari H, Hosseini Z, Lafrenière J, Conklin AI. The role of dietary diversity in preventing metabolic-related outcomes: findings from a systematic review. Obes Rev 2021;22:e13117. [DOI] [PubMed] [Google Scholar]
- 14. Rose AJ. Amino Acid Nutrition and Metabolism in Health and Disease. Basel: MDPI, 2019:2623. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Fan M, Li Y, Wang C et al. Dietary protein consumption and the risk of type 2 diabetes: a dose-response meta-analysis of prospective studies. Nutrients 2019;11:2783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Shang X, Scott D, Hodge AM et al. Dietary protein intake and risk of type 2 diabetes: results from the Melbourne Collaborative Cohort Study and a meta-analysis of prospective studies. Am J Clin Nutr 2016;104:1352–65. [DOI] [PubMed] [Google Scholar]
- 17. Tian S, Xu Q, Jiang R, Han T, Sun C, Na L. Dietary protein consumption and the risk of type 2 diabetes: a systematic review and meta-analysis of cohort studies. Nutrients 2017;9:982. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Zhao L-G, Zhang Q-L, Liu X-L, Wu H, Zheng J-L, Xiang Y-B. Dietary protein intake and risk of type 2 diabetes: a dose–response meta-analysis of prospective studies. Eur J Nutr 2019;58:1351–67. [DOI] [PubMed] [Google Scholar]
- 19. Van Nielen M, Feskens EJ, Mensink M et al. ; InterAct Consortium. Dietary protein intake and incidence of type 2 diabetes in Europe: the EPIC-InterAct Case-Cohort Study. Diabetes Care 2014;37:1854–62. [DOI] [PubMed] [Google Scholar]
- 20. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3–9. [DOI] [PubMed] [Google Scholar]
- 21. Willett W, Rockström J, Loken B et al. Food in the Anthropocene: the EAT–Lancet Commission on healthy diets from sustainable food systems. Lancet 2019;393:447–92. [DOI] [PubMed] [Google Scholar]
- 22. U.S. Department of Agriculture and U.S. Department of Health and Human Services. Dietary Guidelines for Americans, 2020–2025. 9th ed. 2020. https://www.dietaryguidelines.gov/ (7 May 2020, date last accessed).
- 23. Ouyang Y, Huang F, Zhang X et al. Association of dietary protein intake with muscle mass in elderly Chinese: a cross-sectional study. Nutrients 2022;14:5130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Craig WJ. Nutrition concerns and health effects of vegetarian diets. Nutr Clin Pract 2010;25:613–20. [DOI] [PubMed] [Google Scholar]
- 25. Chen Y, Kim M, Paye S, Benayoun BA. Sex as a biological variable in nutrition research: from human studies to animal models. Annu Rev Nutr 2022;42:227–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Day C, Bailey CJ. Obesity in the pathogenesis of type 2 diabetes. Diabetes Vasc Dis 2011;11:55–61. [Google Scholar]
- 27. Musso G, Gambino R, Cassader M. Obesity, diabetes, and gut microbiota: the hygiene hypothesis expanded? Diabetes Care 2010;33:2277–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Rosen ED, Kaestner KH, Natarajan R et al. Epigenetics and epigenomics: implications for diabetes and obesity. Diabetes 2018;67:1923–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Van Dijk SJ, Tellam RL, Morrison JL, Muhlhausler BS, Molloy PL. Recent developments on the role of epigenetics in obesity and metabolic disease. Clin Epigenet 2015;7:66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Zhou C, Liu C, Zhang Z et al. Variety and quantity of dietary protein intake from different sources and risk of new-onset diabetes: a nationwide cohort study in China. BMC Med 2022;20:6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Lee-Sarwar KA, Ramirez L. Diversifying Your Diet Portfolio: Potential Impacts of Dietary Diversity on the Gut Microbiome and Human Health. Oxford: Oxford University Press; 2022, 844–845. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Lavigne C, Marette A, Jacques H. Cod and soy proteins compared with casein improve glucose tolerance and insulin sensitivity in rats. Am J Physiol Endocrinol Metab 2000; 278: E491–E500. [DOI] [PubMed] [Google Scholar]
- 33. Azadbakht L, Mirmiran P, Azizi F. Dietary diversity score is favorably associated with the metabolic syndrome in Tehranian adults. Int J Obes (Lond) 2005;29:1361–7. [DOI] [PubMed] [Google Scholar]
- 34. Lee M-S, Huang Y-C, Su H-H, Lee M-Z, Wahlqvist ML. A simple food quality index predicts mortality in elderly Taiwanese. J Nutr Health Aging 2011;15:815–21. [DOI] [PubMed] [Google Scholar]
- 35. Azadbakht L, Mirmiran P, Esmaillzadeh A, Azizi F. Dietary diversity score and cardiovascular risk factors in Tehranian adults. Public Health Nutr 2006;9:728–36. [DOI] [PubMed] [Google Scholar]
- 36. Kander MC, Cui Y, Liu Z. Gender difference in oxidative stress: a new look at the mechanisms for cardiovascular diseases. J Cell Mol Med 2017;21:1024–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Lee-Sarwar KA, Ramirez L. Diversifying your diet portfolio: potential impacts of dietary diversity on the gut microbiome and human health. Am J Clin Nutr 2022;116:844–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Verger EO, Le Port A, Borderon A et al. Dietary diversity indicators and their associations with dietary adequacy and health outcomes: a systematic scoping review. Adv Nutr 2021;12:1659–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Louie JCY, Jones M, Barclay AW, Brand-Miller JC. Dietary glycaemic index and glycaemic load among Australian adults: results from the 2011–2012 Australian Health Survey. Sci Rep 2017;7:4699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Sá AGA, Moreno YMF, Carciofi BAM. Plant proteins as high-quality nutritional source for human diet. Trends Food Sci Technol 2020;97:170–84. [Google Scholar]
- 41. National Health and Medical Research Council. Australian Guide to Healthy Eating. 2017. https://www.eatforhealth.gov.au/guidelines/australian-guide-healthy-eating/ (7 May 2020, date last accessed).
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
EPIC-InterAct Study data described in the manuscript, code book, and analytic code cannot be deposited publicly, as these collaborative data originate from multiple research institutions across eight European countries with different legal frameworks. The authors confirm that researchers seeking the analysis dataset for this work can submit a data request to the EPIC-InterAct Study central contact point by emailing interact@mrc-epid.cam.ac.uk.