If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
School of Public Health, Xi'an Jiaotong University Health Science Center, Xi'an, Shaanxi, ChinaNutritional Epidemiology, Department of Clinical Sciences Malmö, Lund University, Malmö, Sweden
Department of Epidemiology, School of Public Health and Tropical Medicine, Tulane University, New Orleans, LA, USADepartment of Nutrition, Harvard T.H. Chan School of Public Health, Boston, MA, USA
In 2019, the EAT-Lancet Commission proposed a global sustainable and healthy diet.
•
Higher adherence to the EAT-Lancet diet was associated with decreased risk of type 2 diabetes.
•
The association was independent of the genetic susceptibility to type 2 diabetes.
•
Our results can offer important support for public policy and guidelines on the benefits of healthy and sustainable diets.
Abstract
Background and aims
In 2019, the EAT-Lancet Commission proposed a mainly plant-based diet that nurtures human health and supports environmental sustainability. However, its association with type 2 diabetes (T2D) has not been widely studied, and it remains unclear whether genetic susceptibility for T2D can modify this association. The aim was therefore to investigate the association between the EAT-Lancet diet and risk of T2D and assess whether the association differs by the genetic predisposition to T2D.
Methods
A total of 24,494 participants from the Malmö Diet and Cancer study were analyzed. Dietary intake was assessed using a modified diet history methodology, and an EAT-Lancet diet index (range from 0 to 42 points) was constructed based on the EAT-Lancet reference diet. National and local registers were used to identify T2D cases during follow-up. Cox proportional hazards regression model was applied to estimate the association between the EAT-Lancet diet index and risk of T2D. Genetic predisposition to T2D was captured based on 116 single nucleotide polymorphisms.
Results
During a median of 24.3 years of follow-up, 4197 (17.1 %) T2D cases were documented. Compared with those with the lowest adherence to the EAT-Lancet diet (≤13 points), participants who had the highest adherence (≥23 points) showed an 18 % (95 % CI: 4 %–30 %) lower risk of T2D (P for trend <0.01). There was no significant multiplicative interaction between genetic predisposition to T2D and the EAT-Lancet diet index (P = 0.59). Also, no significant additive interaction between the genetic risk and the EAT-Lancet diet was seen (P = 0.44). The highest risk was observed among the 22.9 % of the individuals with high genetic risk and low EAT-Lancet diet score (HR = 1.79; 95 % CI: 1.63, 1.96).
Conclusions
Our findings indicate that high adherence to the EAT-Lancet diet was associated with decreased risk of incident T2D among people with different genetic risks.
Globally, the number of adults (20–79 years) living with diabetes is 537 million in 2021, and this figure is projected to rise to 783 million by 2045 [
]. Type 2 diabetes (T2D) is the most common type of diabetes, accounting for >90 % of diabetes cases. People with T2D are at increased risk of cardiovascular disease (CVD) and mortality [
]. Thus, the prevention of T2D is an urgent public health priority. Although the etiology of T2D is not completely understood, both genetic and environmental factors contribute to the risk of T2D [
]. Because analyses of single nutrients or foods cannot account for complicated interactions, dietary pattern analysis has emerged as a high priority to examine the association between diet and T2D [
]. In particular, a sustainable and mainly plant-based dietary pattern that promotes human and planetary health was proposed by the EAT-Lancet Commission in 2019 [
] and has been gaining increased attention. This dietary pattern, also referred to as the EAT-Lancet reference diet, emphasizes the intake of vegetables, fruits, legumes, whole grains, and nuts, and limits the intake of animal source foods, added sugar, and saturated fat [
]. Compared with previous plant-based diets defined by three diet quality indexes (an overall plant-based diet index, a healthy plant-based diet index, and an unhealthy plant-based diet index) [
], the EAT-Lancet diet index per se is a healthy plant-based diet index (i.e., high-quality diet). So far, to the best of our knowledge, only three previous studies have focused on the association between the EAT-Lancet diet and the risk of T2D [
]. Moreover, the scoring method of the EAT-Lancet diet index applied in the three studies encompassed binary cut-offs in intake of foods, limiting the possibility to capture variations in intake [
]. Due to its gradual scoring criteria, we expect this new EAT-Lancet diet index to provide more detailed information on the adherence to the diet proposed by the EAT-Lancet Commission [
Furthermore, although two studies have indicated that genetic susceptibility significantly modified the association between dietary patterns and risk of T2D [
], other studies have consistently indicated that better adherence to a healthy diet or lifestyle was associated with a lower risk of T2D regardless of genetic susceptibility [
]. To date, no studies have assessed whether the association between the EAT-Lancet diet and risk of T2D is modified by genetic susceptibility to T2D.
In this prospective cohort study, we examined the association between the EAT-Lancet diet index and the risk of T2D in a large population-based cohort of Swedish adults and hypothesized that there was an inverse association. In addition, we assessed whether the association between the EAT-Lancet diet index and risk of incident T2D differed according to the genetic predisposition to T2D (defined by a polygenic risk score, PRS) by testing gene-diet interactions, and hypothesized that the effect of the EAT-Lancet diet index on T2D would differ across people with different genetic susceptibility.
2. Methods
2.1 Study population
The study participants were from the Malmö Diet and Cancer (MDC) study, a prospective population-based cohort study. The design of the MDC study has been described previously [
]. In short, recruitment was carried out between 1991 and 1996, enrolling adults born between 1923 and 1950 residing in Malmö, Sweden. In total, 30,446 individuals participated in the baseline examination, and 28,098 individuals provided dietary, questionnaire, and anthropometric data. All participants signed written informed consent. The study protocol was approved by the Ethical Committee at the Medical Faculty at Lund University (approval number: LU 51/90).
For the current study, we excluded participants with baseline examination during 1991 because of missing data on legumes separately (n = 2128). Also, we excluded those who had prevalent diabetes at baseline (n = 1182, based on self-reported diabetes diagnosis, self-reported diabetes medication, or information from registries) and had missing covariates (n = 294), including body mass index (BMI) (n = 35), leisure-time physical activity (n = 133), smoking (n = 7), education (n = 48), and information on family history of diabetes (n = 71). Finally, a total of 24,494 participants were included in the diet-T2D analysis. For the gene-diet interaction analysis, we further excluded non-European individuals (n = 1053) based on genetic data, leaving 23,441 participants (Fig. 1).
]. The method consisted of a 7-day food diary, a 168-item food frequency questionnaire (FFQ), and an in-person interview. In the 7-day food diary, information on cooked/main meals, cold beverages, and dietary supplements was obtained. In the FFQ, the frequency and portion size information of foods habitually consumed in the past year (not overlapping with the 7-day food diary) was collected. During the in-person interview, information on cooking habits and habitual portion sizes of foods reported in the food diary was inquired about. Individual food items were converted into food intake amounts (g/day) by combining the information from the food diary/interview and the FFQ. Daily total energy intake was estimated according to the Swedish Food Database PC KOST2–93 of the Swedish National Food Agency. The validity and reproducibility of the dietary assessment method indicated that the method is reliable and valid for assessing the MDC study participants' habitual dietary intake [
An EAT-Lancet diet index was calculated to estimate the adherence to the EAT-Lancet Commission recommendations on healthy diets from sustainable food systems [
]. Briefly, the EAT-Lancet diet index contains 7 emphasized components (whole grains, vegetables, fruits, fish, legumes, nuts, and unsaturated oils) and 7 limited components (potatoes, dairy, beef and lamb, pork, poultry, eggs, and added sugar). For emphasized components, a score of 0 was assigned to the lowest adherence and a score of 3 to the highest; for limited components, the scoring pattern was inversed, with 3 being assigned to the lowest adherence and 0 to the highest. The boundaries of the different scores (0, 1, 2, or 3) to the individual components of EAT-Lancet attributed were based on the target intake levels and the reference intervals, as described in our previous study [
]. Then, the 14 component scores were summed to construct the EAT-Lancet diet index ranging from 0 (worst) to 42 (best), with higher scores indicating higher diet quality. Supplementary Table 1 displays the numbers of participants in each category and the cutoff points of individual components of the EAT-Lancet diet index. Supplementary Fig. 1 shows the distribution of the EAT-Lancet diet index in the study sample. To avoid too small extreme groups that may include many outliers with unreliable dietary data and ensure adequate numbers of participants in each group, we divided the participants into five groups (≤13, 14–16, 17–19, 20–22, and ≥23 points).
2.3 Ascertainment of incident T2D cases
Diabetes cases were retrieved by linking the Swedish personal identification number with eight national and local registers as well as re-examination screenings of the study participants, which have been previously described in detail [
], the Malmö HbA1c register, the regional Diabetes 2000 register of the Scania region, the Swedish inpatient register, the Swedish outpatient register, the Swedish Prescribed Drug Register, the Swedish Cause of Death Register, and the ANDIS (All New Diabetics in Scania) study [
The role of circulating galectin-1 in type 2 diabetes and chronic kidney disease: evidence from cross-sectional, longitudinal and mendelian randomisation analyses.
]. Diagnosis of diabetes was retrieved using the ICD10 codes E10-E14 and O244-O249. In the Swedish Prescribed Drug Register, a filled prescription of glucose-lowering medications (ATC code A10) or insulin was required for diagnosis of diabetes. Cases with incident type 1 diabetes (n = 167), latent autoimmune diabetes in adults (n = 15), secondary diabetes (n = 4), and other diabetes conditions (n = 11) were considered non-T2D cases. Follow-up was censored at the date of the first incident diabetes, death, emigration, or 31 December 2020, whichever occurred first.
2.4 PRS
DNA was extracted from blood samples and the MDC study participants were genotyped using the Illumina GSA v1 genotyping array. Details on imputation procedures and quality control have been previously described elsewhere [
Single nucleotide polymorphisms in close proximity to the fibroblast growth factor 21 (FGF21) gene found to be associated with sugar intake in a swedish population.
]. We constructed a weighted PRS for T2D using 116 single nucleotide polymorphisms (SNPs), selected from a previous genome-wide association study and which passed quality control [
] (Supplementary Table 2). Each SNP was coded as 0, 1, or 2 according to the number of risk alleles. The weighted PRS was calculated using the following equation: PRS = (β1 × SNP1 + β2 × SNP2 + … + β116 × SNP116) × (116/sum of the β coefficients), where SNPn is the risk allele number of each SNP, the β coefficients were from the genome-wide association study [
]. The PRS ranged from 89.3 to 134.7, with higher scores representing a higher genetic predisposition to T2D. Participants were then categorized into low (quintile 1), medium (quintiles 2–4), and high (quintile 5) genetic risk groups according to the PRS quintiles.
2.5 Assessment of covariates
Information on age and sex was obtained via the Swedish personal identification number. Sociodemographic and lifestyle covariates were collected using a questionnaire at baseline and included education (elementary, primary and secondary, upper secondary, further education without a degree, and university degree), leisure-time physical activity, alcohol consumption, and smoking status (current, former, and never). The leisure-time physical activity was assessed using a modified Minnesota Leisure Time Physical Activity questionnaire [
], and was expressed in metabolic equivalent task hours per week (MET-hour/week). Participants were then categorized into five groups: <7.5, ≥7.5–15, ≥15–25, ≥25–50, and ≥50 MET-hour/week. Alcohol consumption was divided into 6 groups, including zero-consumers (i.e., no consumption in the last year from the questionnaire and the 7-day diary) and sex-specific quintiles based on the intake from the 7-day diary (i.e., <0–0.9, ≥0.9–4.3, ≥4.3–8.1, ≥8.1–14.0, and ≥14.0 g/day for women and <0–3.4, ≥3.4–9.1, ≥9.1–15.7, ≥15.7–25.7, and ≥25.7 g/day for men). Information on the use of lipid-lowering and antihypertensive drugs, family history of diabetes, and registers-sourced personal history of CVD and cancer were also collected at baseline. Blood pressure was measured using a mercury column sphygmomanometer after the participant had been in a supine position for 10 min. Hypertension was defined as systolic blood pressure of 140 mmHg or higher, diastolic blood pressure of 90 mmHg or higher, or use of antihypertensive drugs. Weight and height were measured by trained staff and BMI was calculated as body weight in kilograms divided by height in meters squared (kg/m2). Due to the shortening time of the dietary interview from 60 min (March 1991 to August 1994) to 45 min (from September 1994), the dietary assessment version (method) was, thus, introduced as a new binary variable. The variable “season” represents the season of baseline diet measurements.
2.6 Statistical analysis
The normal distribution of the continuous variables was assessed by using Q-Q plots. For baseline characteristics, continuous variables are presented as means with standard deviation (SD) or medians (interquartile ranges), whereas categorical variables are presented as percentages. As per STROBE guidelines [
], we did not use P values to evaluate participant characteristics.
Cox proportional hazards regression models were applied to estimate hazard ratios (HRs) and 95 % confidence intervals (CIs) for the association of the EAT-Lancet diet index with T2D risk. The EAT-Lancet diet index was entered into the models as five groups (≤13, 14–16, 17–19, 20–22, and ≥23 points), with the lowest group as the reference. The proportional hazards assumption was checked by Schoenfeld residuals, and the results indicated that the assumptions had not been violated. Four multivariable models were constructed. Model 1 adjusted for age, sex, dietary assessment version (method), season, and log-transformed total energy intake. Model 2 further adjusted for leisure-time physical activity, alcohol consumption, smoking status, and education; Model 3 further adjusted for family history of diabetes, use of lipid-lowering drugs, hypertension at baseline, as well as history of CVD and cancer from registers. Because BMI can be a potential mediator for the association, we adjusted for it separately in model 4. The factors included in the multivariable models were selected based on prior knowledge. Linear trends were tested for significance by including the categories of the EAT-Lancet diet index (≤13: 1, 14–16: 2, 17–19: 3, 20–22: 4, and ≥23: 5) as an ordinal variable in the model. Furthermore, the dose-response association between the EAT-Lancet diet index and risk of T2D was confirmed by the restricted cubic spline function, with three knots placed at the 25th, 50th, and 75th percentiles of exposure [
]. From a public health perspective, the population attributable risk (PAR) was computed to estimate the proportion of T2D cases in this cohort that theoretically would be prevented if all individuals had the highest adherence to the EAT-Lancet diet index (≥23 points). The PAR calculation formula was PAR = Pc × (1–1/HR) × 100 % [
], where Pc denotes the prevalence of not adhering to the diet index (<23 points) among T2D cases (93.4 %) and HR represents the fully adjusted HR of T2D in relation to the diet index. The HR (95 % CI) used in this formula was 1.16 (1.02, 1.31), which was the effect size of not following the diet (<23 points) compared to following (≥23 points). Moreover, to assess the influence of individual components on the association between the overall EAT-Lancet diet score and T2D, we iteratively recomputed the scores, without each component and evaluated the association between the recomputed score and risk of T2D, with adjustment for the excluded component and variables in model 4.
We performed a series of sensitivity analyses to assess the robustness of our primary results. First, we excluded participants with prevalent CVD (including coronary events and stroke) (n = 696) and cancer (n = 1495) at baseline. Second, to minimize measurement errors in dietary intake, we excluded those with under/over-energy reporters (n = 4358) defined by the method by Black and Goldberg [
] and those with substantial diet change before the baseline (n = 4193) assessed by a question “Have you substantially changed your eating habits because of illness or for some other reason?”. Third, we excluded those with incident non-T2D instead of censoring them, including incident type 1 diabetes (n = 167), latent autoimmune diabetes in adults (n = 15), secondary diabetes (n = 4), and other diabetes conditions (n = 11). Fourth, we excluded T2D cases that occurred within the first two (n = 278) or four years (n = 513) of follow-up to assess whether the results might have been influenced by reverse causation bias. Fifth, we censored participants at 5, 10, 15, 20, and 25 years of follow-up. Finally, we examined potential modification effects by sex.
We further performed a stratified analysis by T2D-PRS (low, intermediate, and high) to assess the association of the EAT-Lancet diet index with the risk of T2D across participants with different genetic susceptibility to T2D. To investigate whether genetic susceptibility modified the association of the EAT-Lancet diet index with T2D, we used the likelihood ratio test to test for the statistical significance of the multiplicative interaction term PRS*EAT-lancet diet score by including them as continuous variables [
]. Furthermore, we performed joint analyses of PRS (< or ≥median) and EAT-Lancet diet index (< or ≥median) to examine how their combination was associated with risk of T2D. The relative excess risk due to interaction (RERI) and corresponding 95 % CI was used to quantify interaction on the additive scale [
]. Because additive interaction is usually focused on dichotomous determinants, we split the PRS and the EAT-Lancet diet score at the medians in the joint analyses.
All statistical analyses were performed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA) and R software, version 4.1.3 (R Foundation for Statistical Computing). A two-sided P value of <0.05 was considered statistically significant.
3. Results
Among the 24,494 participants, the mean (SD) age was 58.1 (7.7) years at baseline, and 9418 (38.5 %) were men. During a total of 496,849 person-years, 4197 (17.1 %) T2D cases were ascertained. The median (interquartile range) follow-up duration was 24.3 (14.8, 26.3) years, and the maximum was 29.7 years.
Table 1 shows the baseline characteristics of participants, overall and according to categories of the EAT-Lancet diet index. Overall, participants with higher EAT-Lancet diet scores were less likely to be men, tended to have higher education and engage in more leisure-time physical activity, were less likely to be current smokers, were more likely to be former and never smokers, and had a lower total energy intake. Furthermore, among the participants, 0.3 % reported no consumption of meat, poultry, or fish.
Table 1Baseline characteristics of the study participants, overall and according to categories of the EAT-Lancet diet index (n = 24,494)
Table 2 presents the association between the EAT-Lancet diet index and the risk of T2D. In the model adjusted for age, sex, dietary assessment version, season, and total energy intake, when comparing the highest with the lowest groups of adherence, the multivariable HR (95 % CI) for incident T2D was 0.70 (0.60, 0.82), P for trend <0.0001. These estimates were attenuated but remained statistically significant after further adjustment for lifestyle factors, personal and family medical history, and BMI (models 2 and 3). In the fully adjusted model (model 4), the multivariable HR (95 % CI) for T2D among participants with the highest vs. those with the lowest EAT-Lancet diet index was 0.82 (0.70, 0.96), P for trend <0.01. The dose-response analysis shows that the EAT-Lancet diet index had a linear association with T2D risk (P for overall association =0.01 and P for non-linear =0.70) (Fig. 2). The PAR of incident T2D for not adhering to the EAT-Lancet diet index (<23 points) was 12.9 % (95 % CI: 1.83 %, 22.1 %).
Table 2Association between EAT-Lancet diet index and risk of type 2 diabetes in the Malmö Diet and Cancer Study (n = 24,494)
Values are given as hazard ratios and 95 % confidence intervals within parentheses.
.
EAT-Lancet diet index categories
P for trend
≤13
14–16
17–19
20–22
≥23
Number of participants
2379
5846
8727
5566
1976
–
Number of cases
447
1107
1448
919
276
–
Person-years
45,483
114,875
177,674
115,826
42,990
–
Incidence per 1000 person-years
9.83
9.64
8.15
7.93
6.42
–
Model 1
1.00 (reference)
1.01 (0.90, 1.13)
0.86 (0.77, 0.96)
0.86 (0.76, 0.96)
0.70 (0.60, 0.82)
<0.0001
Model 2
1.00 (reference)
1.09 (0.98, 1.22)
0.98 (0.88, 1.10)
1.00 (0.89, 1.13)
0.84 (0.72, 0.99)
0.01
Model 3
1.00 (reference)
1.07 (0.96, 1.20)
0.96 (0.86, 1.07)
0.97 (0.86, 1.10)
0.84 (0.72, 0.98)
<0.01
Model 4
1.00 (reference)
1.04 (0.93, 1.16)
0.90 (0.81, 1.01)
0.94 (0.83, 1.06)
0.82 (0.70, 0.96)
<0.01
Model 1: adjusted for age, sex, dietary assessment version (method), season, and total energy intake.
Model 2: adjusted for variables in model 1 plus leisure-time physical activity, alcohol consumption, smoking status, and educational level.
Model 3: adjusted for variables in model 2 plus family history of diabetes, lipid-lowering medication, hypertension at baseline, history of cardiovascular disease and cancer.
Model 4: adjusted for variables in model 3 plus body mass index.
a Values are given as hazard ratios and 95 % confidence intervals within parentheses.
Fig. 2The dose-response association between the EAT-Lancet diet index and risk of type 2 diabetes using restricted cubic spline, with three knots placed at the 25th, 50th, and 75th percentiles of exposure. Adjusted for age, sex, dietary assessment version (method), season, total energy intake, leisure-time physical activity, alcohol consumption, smoking status, education, family history of diabetes, use of lipid-lowering drugs, hypertension at baseline, history of cardiovascular disease and cancer as well as body mass index. HR, hazard ratio. The solid lines are point estimates, and the dashed lines are 95 % confidence intervals.
Table 3 displays the associations between iteratively recomputing the EAT-Lancet diet scores and the risk of T2D. Similar results were observed when iteratively recomputing the scores.
Table 3Associations between the recomputed EAT-Lancet diet score and risk of type 2 diabetes in the Malmö Diet and Cancer Study (n = 24,494).
Hazard ratios (95 % confidence intervals) for per 3-point increase (about one standard deviation).
P value
Whole grains
0.96 (0.93, 0.99)
0.01
Potatoes
0.95 (0.92, 0.98)
<0.001
Vegetables
0.96 (0.93, 0.99)
<0.01
Fruits
0.96 (0.93, 0.99)
0.02
Dairy
0.95 (0.92, 0.98)
<0.001
Beef and lamb
0.97 (0.94, 1.00)
0.03
Pork
0.96 (0.93, 0.99)
<0.01
Poultry
0.96 (0.93, 0.99)
<0.01
Eggs
0.97 (0.94, 1.00)
0.03
Fish
0.95 (0.92, 0.98)
<0.001
Legumes
0.95 (0.92, 0.98)
<0.01
Nuts
0.95 (0.92, 0.98)
<0.001
Unsaturated oils
0.94 (0.91, 0.97)
<0.001
Added sugar
0.95 (0.92, 0.98)
<0.01
Multivariable Cox models adjusted for age, sex, dietary assessment version (method), season, total energy intake, leisure-time physical activity, alcohol consumption, smoking status, educational level, family history of diabetes, lipid-lowering medication, hypertension at baseline, history of cardiovascular disease and cancer, body mass index, and the excluded component.
a Hazard ratios (95 % confidence intervals) for per 3-point increase (about one standard deviation).
The associations between the EAT-Lancet diet index and risk of T2D did not materially change in the various sensitivity analyses with further exclusions, including excluding participants with prevalent CVD and cancer (Supplementary Table 3), excluding those with energy intake misreporting and those with self-reported substantial diet change in the past (Supplementary Table 4), excluding those with incident non-T2D instead of censoring them (Supplementary Table 5), excluding T2D cases ascertained within the first two or four years of follow-up (Supplementary Table 6), and censoring participants at different follow-up time points (Supplementary Table 7). Moreover, the association did not differ between men and women (Supplementary Table 8).
Participants with high PRS had a significantly increased risk of incident T2D compared to those with low PRS, with an age- and sex-adjusted HR (95 % CI) of 2.39 (2.15, 2.66) (Supplementary Table 9). Fig. 3 presents the association between the EAT-Lancet diet index and incident T2D according to T2D-PRS. Inverse but not statistically significant association was observed in low and high genetic risk categories, while a significant inverse association was found in medium genetic risk subgroup (P for interaction = 0.59). Fig. 4 shows the joint association of the EAT-Lancet diet index and genetic susceptibility on the risk of T2D. Compared with those with low genetic risk and high adherence to the EAT-Lancet diet index (27.2 %), those with high genetic risk and low adherence to the EAT-Lancet diet index (22.9 %) had the highest risk of T2D, with HR (95 % CI) of 1.79 (1.63, 1.96). The additive interaction was not statistically significant (RERI = −0.01; 95 % CI: −0.19, 0.16; P = 0.44).
Fig. 3Association between the EAT-Lancet diet index and risk of type 2 diabetes according to the polygenic risk score. Adjusted for age, sex, dietary assessment version (method), season, total energy intake, leisure-time physical activity, alcohol consumption, smoking status, education, family history of diabetes, use of lipid-lowering drugs, hypertension at baseline, history of cardiovascular disease and cancer as well as body mass index. HR, hazard ratio (per one SD EAT-Lancet diet index); CI, confidence interval.
Fig. 4Joint association of the EAT-Lancet diet index and genetic susceptibility with risk of type 2 diabetes. Adjusted for age, sex, dietary assessment version (method), season, total energy intake, leisure-time physical activity, alcohol consumption, smoking status, education, family history of diabetes, use of lipid-lowering drugs, hypertension at baseline, history of cardiovascular disease and cancer as well as body mass index. HR, hazard ratio; CI, confidence interval; RERI, relative excess risk due to interaction.
In this large population-based prospective cohort study with a median follow-up of 24 years, we found that a higher adherence to the EAT-Lancet diet index was associated with a lower risk of T2D in a dose-response manner. Such association was independent of the genetic susceptibility to T2D and was robust in several sensitivity analyses. Assuming causality, our data demonstrated that if all participants adhered to the EAT-Lancet diet index of ≥23 points, 12.9 % of the T2D cases in this population would be prevented.
Three previous studies had investigated the association between the EAT-Lancet diet (different from our index) and the risk of T2D [
]. Results from the UK in the European Prospective Investigation into Cancer and Nutrition Oxford study indicated that greater adherence to the EAT-Lancet diet was associated with a lower risk of diabetes [
]. In addition, data from the UK Biobank showed an inverse association between the EAT-Lancet diet and the risk of T2D; however, this inverse association disappeared after further adjustment for BMI [
]. In the three studies, the authors used the binary scoring criterion to construct an EAT-Lancet diet score, which does not adequately allow for the degrees of adherence to the EAT-Lancet reference diet. Thus, there is concern that the binary criterion EAT-Lancet diet score may not optimally evaluate the EAT-Lancet diet [
]. In the current study, we used a newly developed EAT-Lancet diet index with the gradual scoring criterion, which may better reflect adherence to the reference diet proposed by the EAT-Lancet Commission [
]. Our results indicated that the EAT-Lancet diet index was significantly and linearly associated with the risk of T2D. The findings add new evidence to help understand the health effect of the EAT-Lancet reference diet on incident T2D.
Our findings were also supported by a meta-analysis of 9 cohort studies, in which plant-based dietary patterns had a significant inverse linear dose-response association with the risk of T2D [
]. In addition, recent findings from the Women's Health Initiative indicated that greater adherence to the plant-predominant Portfolio, Dietary Approaches to Stop Hypertension, and Mediterranean diets was prospectively associated with lower risk of T2D in postmenopausal women [
]. A recent umbrella review also indicated that healthy plant-based diets such as Mediterranean diet, Dietary Approaches to Stop Hypertension, and Healthy Eating Index were inversely associated with risk of T2D [
]. Moreover, recent findings from the Women's Health Initiative and the UK Biobank showed that substituting plant protein for animal protein was associated with decreased risk of T2D [
]. Furthermore, assuming a causal relationship between the EAT-Lancet diet index and T2D, our study showed that 12.9 % of T2D cases could have been prevented if all participants with the lower adherence (<23 points) shifted to the highest adherence (≥23 points, 8.1 % of individuals met this criterion) to the EAT-Lancet diet index. Of note, even if it was the highest group, it was still far from the maximum score/adherence of 42 points, suggesting that more strictly adhering to the EAT-Lancet diet may have more benefits.
Because single food-disease risk associations do not account for complex interactions between dietary components [
], we did not analyze the associations between individual components of the EAT-Lancet diet index and the risk of T2D. Instead, we iteratively recomputed the scores of the EAT-Lancet diet index by excluding each component and evaluating the association between the recomputed score and T2D. Such an approach can provide a deeper insight into the influence of individual components on the association between the overall EAT-Lancet diet score and T2D. Our findings indicated that the inverse association between the overall EAT-Lancet diet score and risk of T2D was only slightly influenced by the single components.
Consistent with previous studies that documented no significant interaction between genetic susceptibility and healthy dietary pattern or lifestyle on the risk of incident T2D [
], we also showed that high adherence to the EAT-Lancet diet index was associated with a lower risk of T2D across people with different genetic susceptibility to T2D. From a public health standpoint, our current study highlights that the EAT-Lancet diet is healthy for the population regardless of genetic predisposition. The absence of interaction may be mainly due to the low percentage of the genetic susceptibility explained by the SNPs included. Although the PRS can be used to identify individuals at increased T2D risk, it consists of 116 SNPs that cover different metabolic pathways or have no known function. However, for T2D most of the SNPs have been involved in insulin secretion or resistance pathways, so the PRS may have a functional basis. Furthermore, a recent study used pathway-specific PRS (by impaired insulin secretion and increased insulin resistance separately) and also did not show a significant interaction between diet quality and T2D [
]. This finding suggests that the interaction between healthy dietary patterns and T2D-PRS might likely be very small and undetectable by conventional approaches or confounded by other factors [
]. Therefore, future studies need to investigate the association based on PRS constructed by distinct panels of SNPs and should examine different T2D subtypes separately. Furthermore, our results indicated that participants with high PRS and low EAT-Lancet index had the highest risk for T2D, suggesting that the high genetic risk of T2D can be exaggerated by an unhealthy dietary pattern.
To our knowledge, this is the first study that examined the association between the EAT-Lancet diet, genetic susceptibility, and risk of T2D. Strengths of the current study include its large sample size, prospective study design, high follow-up rates (>99 %), and detailed data on dietary intake and covariates. In addition, we used multiple reliable registers to ascertain cases of T2D minimizing potential misclassification bias of the outcome.
Our study also has several limitations. First, measurement errors are inevitable in collecting dietary data. However, such bias is more likely to be non-differential and will, thus, bias the association toward null. Second, dietary data were collected once at baseline and not updated during the follow-up period. The baseline exposure might not capture changes in dietary intake over time. However, our results were similar when censoring participants according to follow-up time (0–5, 0–10, 0–15, 0–20, and 0–25 years). Also, some evidence indicates stability of dietary patterns over time [
]. Nevertheless, we acknowledge that dietary changes since baseline would lead to exposure misclassification during follow-up. However, such bias often leads to random measurement errors in estimating diet intake and can therefore underestimate the true association [
]. Third, although we have adjusted for a wide range of covariates, we cannot completely rule out the residual confounding and cannot establish causality. For example, the study only collected information on lipid-lowering medications rather than hypercholesterolemia and hypertriglyceridemia. Finally, our study participants are adults living in Sweden, so the findings might not be generalizable to other populations, especially in low-income and middle-income countries.
5. Conclusions
In conclusion, our study indicates that adherence to the EAT-Lancet diet index, reflecting the EAT-Lancet reference diet, was associated with a lower risk of T2D across all levels of genetic risk. Our results further support the EAT-Lancet Commission recommendations to adhere to a sustainable diet.
Funding
This work was supported by the Swedish Research Council (2020-01412), Heart and Lung Foundation (20190555 and 20200482), and Crafoord Foundation (20210674).
The authors would like to thank all the Malmö Diet and Cancer Study participants and staff for their contribution to the study. In addition, we thank diabetes registers including the Swedish National Diabetes Register.
The role of circulating galectin-1 in type 2 diabetes and chronic kidney disease: evidence from cross-sectional, longitudinal and mendelian randomisation analyses.
Single nucleotide polymorphisms in close proximity to the fibroblast growth factor 21 (FGF21) gene found to be associated with sugar intake in a swedish population.
00004-5&id=ga1.jpg){kind=link}
00004-5&id=gr1.jpg){kind=link}
00004-5&id=gr2.jpg){kind=link}
00004-5&id=gr3.jpg){kind=link}
00004-5&id=gr4.jpg){kind=link}