Highlights
-
•
Randomized trial of a 12-week, online, plant-based intervention in T2D patients.
-
•
Plant-based to total protein ratio increased by 35% (95% CI 21, 49%).
-
•
HbA1c decreased by a nonsignificant 3 mmol/mol (95% CI −7, 2).
-
•
Weight by −2.0 kg (95% CI −3.6, −0.5) and LDL-c by −0.3 mmol/L (−0.6, −0.0).
-
•
The resulting estimated reduction in relative CVD risk was 16% (95% CI −29, 1%).
ABSTRACT
Aim
Proof-of-concept study evaluating the effect of a 12-week, predominantly plant-based online dietary intervention compared to usual diet on change in glycated hemoglobin (HbA1c) and estimated cardiovascular disease (CVD) risk in people with type 2 diabetes (T2D).
Methods
People with T2D on stable medication were randomized to usual diet or a 12-week, predominantly plant-based online dietary intervention. Coprimary outcomes were change in HbA1c and estimated relative CVD risk calculated from change in HbA1c, low-density lipoprotein cholesterol (LDL-c), and systolic blood pressure (SBP).
Results
In total, 49 people were randomized, with 48 completing the trial. Mean age was 64 ± 9 years, and 50% were male. There was a nonsignificant 3 (95% CI −7, 2) mmol/mol reduction in HbA1c and a 16% (95% CI −29, 1%) estimated relative CVD risk reduction in the plant-based compared to the usual diet group. Plant-based to total protein intake increased by 35% (95% CI 21, 49%), saturated fat intake decreased by 4 energy-percent (95% CI −6, −1), weight decreased by 2.0 kg (95% CI −3.6; −0.5) and LDL-c by 0.3 mmol/L (95% CI −0.6; −0.0) in the plant-based compared to the usual diet group. There was no effect on SBP (−3 mmHg) (95% CI −10, 4). Medication use remained stable. No people experienced clinically important hypoglycemia. Mild gastrointestinal complaints in the plant-based group were generally well tolerated.
Conclusions
An online plant-based dietary intervention did not significantly reduce HbA1c and resulted in a trend toward a 16% estimated reduction in relative CVD risk. Several secondary outcomes, including dietary components, weight, and LDL-c improved.
Trial Registration
This trial was registered at clinicaltrials.gov under NCT05777746.
Graphical abstract
Abbreviations list
-
1.
LDL-c = low-density lipoprotein cholesterol
-
2.
HbA1c = glycated hemoglobin
-
3.
SBP = systolic blood pressure
-
4.
CVD = cardiovascular disease
-
5.
HR = hazard ratio
-
6.
T2D = type 2 diabetes
-
7.
SGLT-2 inhibitor = sodium/glucose cotransporter 2 inhibitor
-
8.
GLP1-agonist = glucagon-like peptide-1 receptor agonist
Background
People with type 2 diabetes (T2D) are at a two- to fourfold higher risk of cardiovascular disease (CVD) than people without T2D. To reduce CVD risk, guidelines recommend adequate cardiovascular risk management, including lowering glycated hemoglobin (HbA1c), low-density lipoprotein-cholesterol (LDL-c), and systolic blood pressure (SBP). Lifestyle recommendations, including dietary advice, are the first step to reduce CVD risk. The 2023 European Society of Cardiology (ESC) Guidelines for CVD management in people with T2D suggest a Mediterranean diet to reduce cardiovascular risk. Additionally, the 2021 ESC CVD Prevention Guidelines recommend adopting a more plant-based (ie, plant-forward) diet. , Dietary guidelines by the European Association for the Study of Diabetes and American Diabetes Association suggest the Mediterranean, vegetarian and Nordic dietary patterns to improve cardiovascular risk profiles in people with T2D. Similarly, the American Diabetes Association suggests dietary patterns including the Mediterranean, Dietary Approaches to Stop Hypertension (DASH), vegetarian, and plant-based diet.
Observational studies in the general population have found that a more plant-based (ie, plant-forward) diet is associated with a lower risk of CVD. Previous trials in people with T2D indicate that predominantly plant-based diets, ranging from lacto-ovo vegetarian to whole-food vegan diets, might lower HbA1c, LDL-c, and potentially blood pressure (BP). ,, However, these trials required people to attend multiple face-to-face sessions, such as group meetings and cooking sessions. , Organizing face-to-face sessions is resource-intensive, and attending them is time-consuming. Online sessions may be more easily integrated into people’ daily lives, potentially lower participation barriers, and offer greater scalability at low costs. Despite these potential benefits, the evidence for online plant-based dietary interventions is limited in general and no randomized trials with such interventions have been conducted in people with T2D. , Therefore, this proof-of-concept trial aimed to evaluate the effect of a 12-week, online, predominantly plant-based dietary intervention compared to a usual diet on change in HbA1c and change in estimated relative CVD risk in people with T2D.
Methods
Trial design
This was an investigator-blinded, parallel, randomized controlled trial. The trial was funded by a Regio Deal Foodvalley (grant no. 162135), a government-funded program aimed at, amongst others, promoting a healthy diet. Funders were not involved in the design of the study or interpretation of the results. Randomization was performed using computer-generated randomized block randomization. People were randomized in a 1:1 ratio to a 12-week online plant-based dietary intervention or usual diet (Graphical Abstract [in the online version]). Randomization and data collection were performed at the University Medical Center Utrecht. The primary outcome was measured at the 12-week follow-up, as HbA1c reflects the mean serum glucose level over the past 8 to 12 weeks (Supplementary Figure 1). Therefore, at least 8 to 12 weeks of dietary change are needed to translate into a change in HbA1c. All investigators collecting data through physical examination, such as research nurses and lab technicians, were blinded to the allocation of the people. People were instructed not to disclose information regarding allocation to the research nurses who performed the blinded measurements. Randomization was performed by a trial physician who oversaw trial logistics and only collected data on adverse events. During the statistical analyses, treatment allocation was concealed until the analysis had been formalized by the principal investigator. The trial design was reviewed and approved by the Medical Ethics Review Committee NedMec (Utrecht, The Netherlands, protocol number NL80378.041.22). The trial was registered at Clinicaltrials.gov (identifier NCT05777746) and conducted in line with the Good Clinical Practice Standards. The trial and its outcomes were reported in line with the Consolidated Standards of Reporting Trials statement for trials with the extension for surrogate outcomes.
The online plant-based intervention
The online plant-based dietary intervention consisted of three elements: (1) an online program providing weekly information on plant-based diets and dietary change via email, video, and recipes and grocery lists; (2) three online appointments with a dietician; and (3) weekly online peer support groups (Table SI). The intervention was structured around weekly themes (Table SII). Each Monday, patients received an email introducing that week’s theme and directing them to the corresponding video with a small assignment (Figures S3-S5). Patients then received a daily email with information and/or motivational content (Figure S3). Patients could also access all material via the starting page (Figure S2). Recipes were available for breakfast, lunch, and dinner, and included some healthy snacks that could be enjoyed throughout the day (Figure S6). Ingredients for these recipes were accumulated on weekly grocery lists.
Dietary recommendations
The dietary intervention aimed for people to adopt a predominantly plant-based diet, defined as a plant-forward dietary pattern emphasizing plant-based foods while reducing (but not necessarily eliminating) animal-based foods. The dietary intervention aimed to increase plant-based whole foods such as vegetables (300 g/d), fruit (200 g/d), legumes (75 g/d), nuts (15 g/d), plant-based oils (40 g/d of unsaturated oils), and whole grains (230 g/d) (Supplementary Table II). Regarding animal-based protein sources, people were instructed to limit red meat, poultry, fish (especially breaded fish), and egg intake, aiming for zero grams per day (Supplementary Table II). Limiting high-sugar and high-fat dairy was recommended, as well as using fortified plant-based dairy alternatives (Supplementary Table II). Additionally, people were instructed to limit the intake of foods high in sugar and/or refined grains. People in the online plant-based intervention group were instructed to take vitamin B 12 supplements to maintain a stable serum vitamin B 12 level. People in the usual diet group continued with their usual dietary pattern during the intervention period. At the end of the study, after week 24, people in the usual diet group were offered to participate in a light version of the online plant-based dietary intervention (Supplementary Figure 1). People in both groups were asked to keep physical activity levels stable throughout the study.
Trial population and informed consent
The target population for this trial was people with T2D, diagnosed and recorded by their treating physician, who had been on a stable dose of glucose-lowering medication (insulin and/or oral medication) and, if applicable, lipid and BP-lowering medication for at least 2 months before screening (Table S3). Diagnosis was confirmed by retrieving patients’ medical records. People using insulin were only excluded in case of hypoglycemic unawareness. A complete overview of in- and exclusion criteria is provided in Supplementary Table I. People were recruited between October 2022 and September 2024 at the University Medical Center Utrecht, and at two general hospitals (Ziekenhuis Gelderse Vallei, Ede, and the Diakonessenhuis, Utrecht; The Netherlands). Patient recruitment proved challenging. In total, 329 information leaflets were distributed to potentially eligible people, and the study’s online social media campaign generated another 910 information requests. This resulted in a total of 982 people receiving the complete study information (ie, patient information form). Of these 982 people, after contacting the study team, 59 people attended a screening visit, and 49 people were eligible for participation and randomized (Supplementary Figure 7). Written informed consent was provided by all people before any study procedures were performed.
Outcomes
The study’s coprimary outcomes were HbA1c (mmol/mol) and change in estimated relative CVD risk (%), calculated from the changes in LDL-c, SBP and HbA1c. The change in estimated relative CVD risk was calculated using the average treatment effects of glucose, BP and LDL-cholesterol-lowering drugs from large clinical trials as described in the Supplementary Methods (Supplementary Equation 1).
Secondary outcomes
Cardiovascular risk factors
BP was measured at baseline three times at the left and three times at the right arm simultaneously to determine the arm with the highest BP. At subsequent visits, BP was measured at this arm. Three measurements were taken at 30-second intervals with people in a sitting position. Measurements were performed unattended and started 5 minutes after the research nurses had left the room. Blood samples were collected in the morning after at least 9 h of fasting. HbA1c was measured using high-performance liquid chromatography (ADAMS HA-8180, Arkray Inc), total and HDL-cholesterol and apolipoprotein-B were measured using the Siemens Atellica Chemistry Analyser (Siemens Healthineers, Germany), LDL-c was calculated using the Friedewald formula in people with serum triglycerides up to 8 mmol/L and measured directly in people with triglycerides >8 mmol/L. Non-HDL-c was calculated by subtracting HDL-c from total cholesterol. High-sensitivity CRP was measured using immunoturbidimetric (Atellica, Siemens Healthineers, Germany). Neutrophil and lymphocyte count were measured using optical light scatter (CELL-DYN Sapphire, Abbott Diagnostics), and the neutrophil-to-lymphocyte ratio was subsequently calculated. Weight and length were measured with people wearing no shoes and only light clothing.
Medication
People were asked about any changes to medication at the 12-week visit and changes made to medication prescriptions by healthcare providers were also monitored using the National Exchange Point for Healthcare (“Landelijk Schakelpunt”), which provides a comprehensive overview of medication prescriptions from other healthcare providers and pharmacies. Glucose-lowering medication use was quantified using the updated medication effect score, which quantifies the potency of glucose-lowering medication (Table SV). Lipid-lowering therapy use was categorized into high, moderate and low intensity using the LDL-c-lowering potency of prescribed statins and, if applicable, the potency of ezetimibe or PSCK9-inhibition. ,, BP-lowering medication was expressed as (1) mean dose and (2) number of prescribed medications.
Adherence
Dietary intake was measured using an app-based 3-day dietary record, given that dietary records have a higher correlation with biomarkers of recent food intake than food frequency questionnaires and are therefore better suited for measuring short-term changes in dietary intake. Dietary record was conducted using the Traqq app (Wageningen University and Research, Wageningen, the Netherlands). People received invitations to report dietary intake every 2 hours and could complete these by the end of that day. Data was collected on three consecutive days, including week and weekend days. Days were considered incomplete if people missed all timeslots in either the morning, afternoon, or evening. People with incomplete days were planned for new collection dates up to three times. Traqq has been validated against food frequency questionnaire data and serum biomarkers (eg, serum carotenoids such as β-carotene and lutein and n −3 polyunsaturated fatty acids). Validation indicated a good ranking ability for frequently consumed foods and for most nutrients, with more variation for less frequently consumed foods. Adequate intake of vitamin B 12 was assured by measuring serum vitamin B 12 at the 12-week visit.
Adverse events
The occurrence of (severe) adverse events was monitored throughout the study and was classified as definitely, potentially, or not related to the intervention. People were asked about the occurrence of hypoglycemia at each study visit. Clinically important hypoglycemia was defined as glucose <3.0 mmol/L, and mild hypoglycemia was defined as glucose ≥3.0 but <4.0 mmol/L, measured in serum, capillary blood, or with continuous glucose monitoring. Episodes during which symptoms of hypoglycemia occurred, but during which glucose was not measured, were classified as probable symptomatic hypoglycemia. Serum vitamin B 12 and iron were monitored. Vitamin B 12 deficiency was defined as serum vitamin B 12 <130 pmol/L. Iron deficiency was defined as a serum <25 µg/L for males or <20 µg/L for females.
Statistical methods
Sample size
The initial power calculation resulted in a sample size of 140 people, accounting for 15% dropout, and based on a 6 mmol/mol reduction in HbA1c and a standard deviation (SD) of 11 mmol/mol in the online plant-based intervention group. During the trial, recruitment progressed slower than anticipated, and two additional papers investigating the effect of a face-to-face plant-based dietary intervention were published. , After maximum recruitment effort, we managed to include 48 people. Based on updated evidence, a 4 to 6 mmol/mol reduction was expected for the online plant-based dietary intervention relative to the usual diet, and a 5 mmol/mol reduction is generally considered clinically relevant. ,, A posthoc power analysis showed an SD of HbA1c change in the online plant-based intervention group of 6.2 mmol/mol. For the current number of people and using a 6.2 mmol/mol SD, the study had a power of 71% to detect a 4 mmol/mol difference, 86% power for a 5 mmol/mol difference, and a 95% power to detect a 6 mmol/mol difference between the online plant-based dietary intervention and usual diet group.
Statistical tests
In the baseline table, continuous variables are presented as mean and SD or as median and interquartile range (IQR), depending on the underlying distribution. Categorical variables are presented as absolute numbers with corresponding percentages.
Predefined hierarchical testing of the coprimary outcomes was used. HbA1c was tested first for significance with a two-sided alpha of 0.05. Only when change in HbA1c was significant, the significance of change in estimated CVD risk could be tested with a two-sided alpha of 0.05. The analyses were performed blinded. Treatment allocation was unblinded after formalization of the analyses and approval by the principal investigator. Distributions were assessed visually using histograms and Q–Q plots. For normally distributed data, an unpaired t -test was used for between-group comparisons and a paired t -test for within-group comparisons. Differences were expressed using mean (95% confidence interval [CI]). For skewed outcomes, a Wilcoxon signed-rank test was used for within-group comparison, and a Mann–Whitney U test was used for between-group comparison. Differences were expressed using the median (IQR) for within-group changes. The between-group differences in median change of skewed variables were calculated using the Hodges-Lehmann Estimator using median with 95% CI for location of the population median. Differences in proportion with 95% CI were calculated using Fisher’s exact test and the exact 2 × 2 package in R.
A per-protocol analysis was conducted with people who adhered to their assigned allocation. For people in the online plant-based intervention group this was defined as completion of the online plant-based dietary intervention and at least a 10 percentage point increase in percentage of plant-based to total reported protein intake, thereby reflecting a relevant dietary change in line with current recommendations for a healthy protein transition. People in the usual diet group were included who reported no meaningful increase in plant-based protein intake (<5 percentage point increase).
As a sensitivity analysis, an ANCOVA linear regression model was used to assess the effect of baseline differences and changes in medication on the primary and secondary outcomes. , Three regression models were used: model 1 adjusted for baseline level of the risk factor (outcome), model 2 additionally adjusted for baseline medication use, and model 3 additionally adjusted for change in medication use.
Results
Trial population
Forty-nine people were randomized, of whom one patient withdrew after being randomized to the usual diet group. This resulted in 48 people being included in the present study: 25 in the online plant-based intervention and 23 in the usual diet group. Overall, people were aged 64 ± 9 years, and 50% were male ( Table 1 ). Most (63%) people attended tertiary education (university or equivalent), and 3% was currently smoking. Median diabetes duration was 14 [IQR: 5-17] years, and 35% of people had a history of CVD. At baseline, mean body mass index was 28.7 ± 4.6 kg/m 2, HbA1c was 57 ± 13 mmol/mol, LDL-c 2.3 ± 1 mmol/L, and SBP 132 ± 17 mmHg ( Table 1 ). In addition to using glucose-lowering medication, 27 people (56%) used BP-lowering and 35 (73%) used lipid-lowering medication (Table SVII).
Table 1
Baseline characteristics of the trial population stratified by dietary allocation.
| Overall ( n = 48) | Usual diet ( n = 23) | Plant-based dietary intervention ( n = 25) | |
|---|---|---|---|
| Age (y) | 64 (9) | 66 (8) | 63 (9) |
| Sex (male) | 24 (50) | 13 (57) | 11 (44) |
| Level of education | |||
| Primary school | 2 (4) | 1 (4) | 1 (4) |
| Practical secondary | 9 (19) | 5 (23) | 4 (16) |
| Theoretical secondary or vocational education | 7 (15) | 2 (9) | 5 (20) |
| Tertiary education | 30 (63) | 15 (63) | 15 (60) |
| Employment status | |||
| Working (employed, self-employed) | 26 (54) | 10 (44) | 16 (64) |
| Retired | 17 (35) | 10 (44) | 7 (28) |
| Unemployed | 1 (2) | 1 (4) | 0 (0) |
| Employment disability | 4 (8) | 2 (9) | 2 (8) |
| Smoking status | |||
| Current | 1 (2) | 0 (0) | 1 (4) |
| Former | 28 (58) | 11 (47.8) | 17 (68) |
| Never | 19 (40) | 12 (52.2) | 7 (28) |
| Pack-years | 9 [5, 48] | 5 [4, 17] | 22 [8, 53] |
| History of cardiovascular disease | 17 (35) | 9 (39) | 8 (32) |
| Coronary artery disease | 9 (19) | 5 (22) | 4 (16) |
| Cerebrovascular disease | 8 (17) | 5 (22) | 3 (12) |
| Peripheral artery disease | 2 (4) | 0 (0) | 2 (8) |
| Abdominal aortic aneurysm | 1 (2.1) | 0 (0) | 1 (4) |
| Diabetes duration (y) | 14 [5, 17] | 13 [6, 18] | 14 [4, 17] |
| Potency of glucose-lowering medication (Medication Effect Score) | 1.3 ± 0.9 | 1.1 (0.8) | 1.5 (1.0) |
| Blood pressure-lowering medication (%) | 27 (56) | 12 (52) | 15 (60) |
| Lipid-lowering medication (%) | 35 (73) | 16 (70) | 19 (76) |
| Body mass index (kg/m 2) | 28.7 (4.6) | 28.7 (3.8) | 28.7 (5.4) |
| Body mass index ≥30 kg/m 2 | 17 (35) | 7 (30) | 10 (40) |
| Glycated hemoglobin (mmol/mol) (%) | 57 (13) | 55 (11) | 59 (14) |
| 7.4 (1.2) | 7.2 (1.0) | 7.5 (1.3) | |
| Systolic blood pressure (mmHg) | 132 (17) | 133 (16) | 132 (17) |
| LDL-c (mmol/L) (mg/dL) | 2.3 (1.0) | 2.3 (1.0) | 2.3 (1.0) |
| 89 (38) | 89 (38) | 89 (38) | |
| eGFR (mL/min/1.73 m 2) | 102 [82, 150] | 94 [81, 123] | 113 [85, 154] |
| Ferritin (µg/L) | 65 [29, 116] | 107 [54, 153] | 37 [23, 78] |
| Vitamin B 12 (pmol/L) | 361 (154) | 394 (191) | 331 (106) |
Pack-years is the number of pack-years in current and previous smokers. Coronary artery disease was defined as a history of myocardial infarction or stenosis of the coronary arteries. Cerebrovascular disease was defined as a history of stroke, subarachnoid hemorrhage, retinal infarction, or carotid artery stenosis. Peripheral artery disease encompassed renal artery stenosis, intermittent claudication, critical limb ischemia, renal insufficiency, and atherosclerotic vascular disease at other locations. The Medication Effect Score reflects the total glucose-lowering potency of glucose-lowering medication combined (Alexopoulos et al.) (a more detailed description is provided in Table SV). A more detailed overview of medication intake is provided in Table SVII.
eGFR , estimated glomerular filtration rate; LDL-c , low-density-lipoprotein cholesterol.
Change in diet
Plant-based to total protein intake increased by 35 percentage points in the plant-based group compared to the usual diet group (95% CI 21, 49%). 83% of the online plant-based group increased plant-based protein intake by at least 10 percentage points and 29% in the usual diet group (between-group difference 54%, 95% CI 25, 72, Table 2 ). Saturated fat intake was reduced by 4 energy-percent (95% CI −6, −1) and fiber intake increased by 10 g/d (95% CI 2, 18) in the online plant-based compared to the usual diet group ( Table 2 ). Among food groups, intake of nuts (+26 g/d, 95% CI 4, 48) and legumes (+104 g/d, 95% CI 2, 206), while intake of animal-based foods such as meat (−71 g/d, 95% CI −125, −16) decreased (Table SIV). The online plant-based intervention group did not report a change in dietary energy intake compared to the usual diet group (−86 kcal/d, 95% CI −349, 177).
Table 2
Change in calculated macronutrient intake derived from self-reported dietary records.
| Usual diet ( n = 21) | Plant-based intervention ( N = 23) | ||||||
|---|---|---|---|---|---|---|---|
| Baseline | Wk 12 | Change within-group | Baseline | Wk 12 | Change within-group | Between-group difference in change (95% CI) | |
| Energy (kcal/d) | 1,584 (434) | 1,671 (516) | 87 (−99, 273) | 1,450 (542) | 1,450 (637) | 0 (−197, 197) | −86 (−349, 177) |
| Protein | |||||||
| – Total (E%/d) | 18 (4) | 18 (3) | 0 (−2, 2) | 18 (4) | 16 (3) | −2 (−4, 0) | −2 (−4, 0) |
| – Plant-based (g/d) | 29 (12) | 30 (14) | 0 (−6, 6) | 25 (13) | 42 (24) | 17 (8, 26) | 17 (7, 27) |
| – Animal-based (g/d) | 40 (19) | 44 (21) | 4 (−6, 14) | 37 (19) | 14 (17) | −23 (−32, −14) | −27 (−40, −14) |
| – Plant-based to total protein (%) * | 45 (15) | 44 (16) | −2 (−10: 6) | 45 (16) | 78 (25) | 33 (22, 44) | 35 (21, 49) |
| – Proportion of people with >10% increase in plant-based protein | 29% | 83% | 54% (25, 72) | ||||
| Fat | |||||||
| – Total (E%/d) | 35 (5) | 34.5 (7) | 0 (−4, 4) | 40 (11) | 37 (9) | −3 (−8, 2) | −3 (−9, 3) |
| – Saturated (g/d) | 23 (7) | 23.8 (7) | 1 (−2, 4) | 22 (12) | 16 (10) | −6 (−10, −2) | −7 (−12, −2) |
| – Saturated (E%) | 13 (3) | 13 (4) | 0 (−2, 2) | 13 (4) | 9 (4) | −4 (−6: −2) | −4 (−6, −1) |
| Carbohydrates | |||||||
| – Total (E%/d) | 43 (7) | 43 (9) | 0 (−3, 3) | 39 (11) | 43 (9) | 4 (−1, 9) | 5 (−1, 11) |
| – Mono- and disaccharides (g/d) | 69 (27) | 71 (38) | 1 (−11, 13) | 53 (31) | 56 (28) | 3 (−8, 14) | 1 (−15, 17) |
| – Fiber (g/d) | 19 (8) | 19 (8) | 0 (−4, 4) | 17 (7) | 27 (19) | 10 (3, 17) | 10 (2, 18) |
| – Fiber (g/mJ) | 2.9 (0.9) | 2.8 (1.1) | −0.1 (−0.7, 0.5) | 3.0 (1.0) | 4.5 (2.0) | 1.6 (0.8, 2.4) | 1.7 (0.7, 2.6) |
| Alcohol | |||||||
| – Users (%) | 6 (29%) | 8 (38%) | +9% | 7 (30%) | 6 (26%) | −4% | −14% |
| Ethanol intake (g/d) | 0 [0, 5] | 0 [0, 10] | 0 [0, 1] | 0 [0, 3] | 0 [0, 2] | 0 [−1, 0] | (−52, 25%) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
