A web-based low carbohydrate diet intervention significantly improves glycaemic control in adults with type 2 diabetes: results of the T2Diet Study randomised controlled trial

Dietary modification is essential for achieving and maintaining glycaemic targets in people with type 2 diabetes mellitus (T2DM). Although no single macronutrient distribution is universally recommended, carbohydrate intake has the greatest impact on glycaemic control. Low carbohydrate diets (LCDs; 10–<26% total energy from carbohydrate) are supported by a large evidence base showing improved HbA1c, body weight, and medication reductions, with better adherence than very low-carb/ketogenic diets in some reviews. LCDs emphasize high intake of non-starchy vegetables and nutrient-dense foods while limiting added sugars, refined grains, and processed foods, aligning with diabetes nutrition guidelines. Despite the need for ongoing self-management education and support, many people with T2DM face access barriers (workforce capacity, rural/remote location, inadequate reimbursement, costs, and COVID-19 disruptions). Web-based interventions could help overcome these barriers, yet no RCT had evaluated a web-based LCD programme. This study aimed to determine whether standard care plus a web-based healthy LCD education programme improves HbA1c at 16 weeks compared with standard care alone, with secondary aims including changes in anti-glycaemic medication, weight, BMI, dietary intake, and self-efficacy.

The paper summarizes extensive prior evidence indicating that carbohydrate reduction is an effective strategy for improving glycaemic control and clinical outcomes in T2DM. Definitions of carbohydrate levels are provided (high >45% energy; moderate 26–45%; low 10–<26%; ketogenic <10%). Systematic reviews and meta-analyses report that both ketogenic and LCD approaches reduce HbA1c, weight, and medication requirements, with LCDs often demonstrating better adherence. Concerns about nutritional quality in LCDs (e.g., potential reductions in fibre and increases in saturated fat) have been raised in earlier reviews, though LCDs can be designed to emphasize non-starchy vegetables, dietary fibre, and nutrient-dense foods. International diabetes guidelines recognize LCDs as a suitable option for people with T2DM. There is limited empirical evidence on web-based dietary interventions for T2DM, and prior web-based studies seldom assessed medication changes rigorously, highlighting a gap this RCT addresses.

Design: Two-arm parallel-group randomized controlled trial conducted remotely during the COVID-19 pandemic across Australia (metropolitan, rural, regional, and remote). Ethics approval: Deakin University Human Research Ethics Committee (2020-349). Trial registration: ACTRN12621000096853. Participants: Recruited via community organizations, social media, networks, and paid Facebook ads (n=10). Inclusion: adults 40–89 years, T2DM not on insulin, self-reported HbA1c ≥7.0% within 6 months, English-speaking, internet access, active email, Australia-based. Exclusion: other diabetes types; vegan/vegetarian; bariatric surgery; diagnosed cardiovascular or renal disease; conditions limiting participation; pregnancy/lactation; current/recent (≤3 months) weight-loss programme; concurrent clinical studies; risk of disordered eating (EAT-26); baseline HbA1c ≤5.6%. Written informed consent obtained. Randomisation and blinding: 1:1 allocation to intervention (standard care + web-based T2Diet LCD programme) or control (standard care only). Computer-generated sequence with random block sizes (2,4), stratified by gender and age (40–60, 61+). Allocation concealed off-site by a statistician. Researchers and participants were blinded until assignment. Primary outcome assessment (laboratory HbA1c) was blinded; secondary outcomes were self-reported. Data cleaning and statistical analyses were conducted blinded. Intervention: Automated 16-week web-based healthy LCD education programme (T2Diet). Recommendations: 50–100 g carbohydrate/day; ad libitum nutrient-dense lower-carb foods; emphasize high intake of non-starchy vegetables and dietary fibre; reduce starches, sugars, and discretionary foods. No specific prescriptions for protein/fat amounts, with emphasis on nutrient-dense sources. Approximately 3 days post-login, participants were contacted to discuss potential adverse effects and medication precautions, instructed to report adverse effects, continue standard care, and inform their GP. Participants were advised to log in at least weekly; programme use was otherwise self-directed. Control: Continued standard care only. Both groups completed end-of-study measures at 16 weeks. Outcomes and measures: Primary outcome: HbA1c measured by an external pathology lab (mail-in test kits). Secondary outcomes: anti-glycaemic medication intensity via Medication Effect Score (MES); dietary intake via 24-hour recall analyzed with FoodWorks; diabetes self-efficacy via DMSES (Australian version); weight and BMI from self-reported height/weight. Comorbidities checklist as exploratory outcome. Demographics and height at baseline; all other measures at baseline and 16 weeks. Sample size and statistics: Target n=100 (50/group) provided 80% power (α=0.05, two-sided) to detect a between-group HbA1c difference of 0.5% at 16 weeks, assuming SD 0.9, pre-post correlation 0.5, and 20% dropout. Intervention effects estimated using GEE models (Gaussian, identity link) with HbA1c as dependent variable; independent variables: group, time (baseline, week 16), and their interaction, adjusted for age and gender. Multiple imputation by chained equations (m=50) for missing data (auxiliary: group, age, gender, time-updating BMI, prior HbA1c). Similar analyses for continuous secondary outcomes. Two-sided tests with p<0.05 considered significant. Cohen’s d reported (0.2 small, 0.5 medium, 0.8 large). Analyses in Stata 17. Timelines: Recruitment Feb 1–Oct 10, 2021; follow-up completed Feb 18, 2022.

- Enrollment and analysis: 98 randomized (49 intervention; 49 control). Outcome analysis included 87 participants (40 intervention [82%]; 47 control [96%]). Seven missing primary outcomes imputed. - Baseline: Mean age 60.5±9.5 years; mean HbA1c 7.7%±1.2; mean weight 100.07±21.74 kg; mean BMI 34.73±7.25 kg/m²; 85% (n=74) on anti-glycaemic medications. - Primary outcome (HbA1c): Between-group differential change at 16 weeks −0.65% (95% CI −0.99 to −0.30; p<0.0001) favouring intervention. Within-group change: intervention −0.94% (SD 0.86) from 7.64±1.24 to 6.79±1.14; control −0.26% (SD 0.72) from 7.89±1.30 to 7.57±1.28. Effect size d=−0.86 (95% CI −1.32 to −0.40). - Anti-glycaemic medication (MES): Between-group differential change −0.40 (95% CI −0.62 to −0.19; p<0.0001). Intervention within-group −0.09 (SD 0.56) vs control +0.34 (SD 0.49). Effect size d=−0.83 (95% CI −1.28 to −0.38). Twenty-five percent of intervention participants had ≥20% medication reduction; medication needs increased in control. - Weight and BMI: Weight between-group difference −3.26 kg (95% CI −4.81 to −1.71; p<0.0001); BMI −1.11 kg/m² (95% CI −1.63 to −0.59; p<0.0001). Effect sizes: weight d=−1.00 (95% CI −1.46 to −0.54); BMI d=−1.01 (95% CI −1.47 to −0.55). Thirty-eight percent of intervention participants achieved ≥5% weight loss vs 9% control. - Dietary intake: Significant between-group differential change in carbohydrate intake −81.79 g (95% CI −127.16 to −36.42; p<0.0005). Intervention group energy percent from carbs decreased to 24.7% at 16 weeks (vs 39% control). No significant between-group differences for total energy, protein, saturated fat, or fibre. Significant increases in total fat (+21.72 g; p=0.040), monounsaturated fat (+10.33 g; p=0.034), and polyunsaturated fat (+8.36 g; p=0.003) in the intervention group. - Self-efficacy: DMSES increased by 6.95±24.55 in intervention and decreased by −1.11±26.42 in control; between-group differential change 8.18 (95% CI −2.55 to 18.90; p=0.14), not significant. - Adherence and safety: 80% of intervention participants adhered as advised. Minor adverse events: one report of brief constipation at LCD initiation and one instance of mild hypoglycaemia; both self-managed.

The web-based T2Diet LCD programme, when added to standard care, produced clinically meaningful improvements in HbA1c, weight, BMI, and reduced anti-glycaemic medication requirements compared with standard care alone over 16 weeks. These findings align with prior face-to-face LCD RCTs and meta-analyses, demonstrating that similar benefits can be achieved via remote, web-based delivery that may improve access to dietary education and support for people with T2DM. The HbA1c reduction (between-group −0.65%; within-group nearly −1% in the intervention) is associated with substantial reductions in risks of diabetes-related complications. The observed medication reductions underscore improved glycaemic control and are notable given that medication changes are rarely assessed robustly in web-based dietary interventions. Despite ad libitum guidance and energy intakes roughly aligned with recommendations, the intervention group lost more weight, potentially due to improved insulin dynamics and metabolic function. Nutritional quality concerns often raised about LCDs (reduced fibre, increased saturated fat) were not observed here; fibre and saturated fat did not differ between groups, while monounsaturated and polyunsaturated fats increased in the intervention group. Changes in self-efficacy did not appear to mediate outcomes, suggesting other determinants (knowledge, motivation, intentions) may have driven behaviour change. The results highlight the potential for scalable, remote dietary education integrated with standard care to enhance T2DM management.

An automated, web-based healthy low carbohydrate diet education programme, used alongside standard care, significantly improved glycaemic control and reduced body weight, BMI, and anti-glycaemic medication requirements in adults with T2DM over 16 weeks. Future research should evaluate larger-scale implementation, longer-term follow-up to assess sustainability and cardiometabolic outcomes, and integration through primary care pathways where routine T2DM monitoring occurs.

- Lack of additional biomarkers related to cardiometabolic risk, physical activity, psychological wellbeing, hunger, and satiety. - No longer-term follow-up; sustainability beyond 16 weeks not assessed. - Anthropometric outcomes (weight, height) were self-reported, introducing potential measurement error. - Dietary intake based on self-reported 24-hour recall, subject to reporting bias and measurement error. - Some outcomes (secondary) were assessed via self-report for both groups.