TL;DR: A six-month, fully remote three-arm trial in 161 U.S. adults with moderately controlled type 2 diabetes found that personalized Mediterranean-diet counseling with algorithm-driven meal feedback did not reduce major glucose swings more than usual care or standardized counseling.
The 2783-LB abstract published in Diabetes on June 5, 2026 reports the Diabetes Telemedicine Mediterranean Diet (DiaTeleMed) Study as a fully remote, 6-month, 3-arm randomized controlled trial. The study question was whether personalized dietary management improves glycemic variability in individuals with type 2 diabetes. The tested approaches were usual care, standardized Mediterranean diet-based counseling, and the same counseling plus real-time feedback on predicted postprandial glycemic response. The primary variability measure was mean amplitude of glycemic excursions, known as MAGE, assessed with continuous glucose monitoring. The reported conclusion was that the precision-nutrition approach did not improve MAGE or most variability indices beyond usual care or Mediterranean diet-based counseling in adults with moderately controlled disease. The Synapse summary of the same June 5, 2026 abstract presents the same design and the same conclusion in structured form.
Why researchers tested post-meal glucose responses
The investigators stated that dietary recommendations for type 2 diabetes aim to minimize postprandial glycemic response, abbreviated as PPGR. They also stated that the optimal strategy for minimizing that response remains unclear. They further stated that the impact of precision nutrition on glycemic control is also not well established. Those three statements define the rationale reported in the 2783-LB abstract published in Diabetes on June 5, 2026.
The trial therefore compared a standardized dietary counseling model with a personalized model that added prediction and feedback. The comparison was designed to separate the effect of Mediterranean diet-based diabetes self-management support from the additional effect of algorithm-driven PPGR information. The control condition was usual care control, abbreviated as UCC. The follow-up period was six months. The monitoring points for glucose variability were baseline, 3 months, and 6 months.
Who was eligible and who enrolled
Eligibility was defined as U.S. adults aged 21 to 80 years with moderately controlled type 2 diabetes. Moderately controlled was defined in the methods as HbA1c 6.5 to 8.0%. Those eligibility boundaries are stated in both the Diabetes abstract record and the Synapse-structured record.
Enrollment was reported as 161 participants in total. Allocation was reported as 59 participants in UCC, 61 participants in the Standardized intervention, and 41 participants in the Personalized intervention. The mean age was reported as 60 years with a standard deviation of 12 years. The proportion reported as female was 62.7%. The proportion reported as White was 46.6%. The proportion reported as non-Hispanic was 80.7%. Mean HbA1c at entry was reported as 6.77% with a standard deviation of 0.6%. Those baseline characteristics indicate a middle-aged to older adult sample with HbA1c near the lower portion of the eligible range.
What each of the three arms received
Participants were randomized to one of three conditions. The first condition was usual care control. The second condition was described as a Standardized intervention. The third condition was described as a Personalized intervention. Randomization into three parallel groups is the structure emphasized in the Synapse summary of the same June 5, 2026 abstract.
The Standardized arm received 14 behavioral counseling sessions focused on Mediterranean diet-based diabetes self-management. The number 14 and the Mediterranean diet-based diabetes self-management focus are stated explicitly in the methods. The Personalized arm received all Standardized components plus real-time, algorithm-driven feedback on predicted PPGR. That additive design means the difference between the two active arms was the real-time prediction and feedback layer, not the number or theme of counseling sessions. The UCC arm served as the comparator without the 14-session program described for the active arms.
The remote character of the program is stated as fully remote. The duration is stated as six months. The counseling program therefore represents a telemedicine-delivered behavioral nutrition intervention, while the personalized version adds a precision-nutrition informatics component.
How glucose variability was measured
Participants wore a continuous glucose monitor at baseline, 3 months, and 6 months. The purpose was to assess glycemic variability metrics, abbreviated as GV metrics. The named primary variability metric was mean amplitude of glycemic excursions, abbreviated as MAGE. Other GV metrics were also assessed, although only one additional metric is named with numerical results in the abstract.
The analytical approach was reported as linear mixed models used to compare changes in GV between arms. That method was applied to changes measured at the follow-up monitoring points relative to baseline. The outcome of interest was change in variability, rather than a single cross-sectional glucose value. The 3-month and 6-month wear periods provided the two post-randomization assessment windows.
What happened to the main measure of glucose swings
The change in MAGE did not differ significantly between UCC and Standardized, with a reported p-value of 0.24. The change in MAGE did not differ significantly between UCC and Personalized, with a reported p-value of 0.13. The change in MAGE did not differ significantly between Standardized and Personalized, with a reported p-value of 0.70. Those three pairwise comparisons cover every contrast among the three randomized groups.
The abstract reports those MAGE results as not significantly different. No pairwise MAGE comparison met the conventional threshold for statistical significance as presented. The Synapse-structured record lists the same three p-values and describes the finding as no significant differences in MAGE among the three arms. The implication stated by the investigators is that neither the standardized nor the personalized telemedicine nutrition program improved this excursion measure beyond the comparator over the observation period.
What happened to other variability measures
Changes in other GV metrics were also reported as not significantly different among the arms at 3 and 6 months. That statement applies broadly to the additional variability indices collected by continuous monitoring. It covers both the 3-month assessment and the 6-month assessment. It also covers comparisons among arms, rather than within-arm changes from baseline.
One exception is reported for the coefficient of variation, abbreviated as CV, at 6 months. At that single time point, the Personalized arm showed a greater reduction in CV compared with UCC. The reported mean difference was minus 1.84 percentage points. The reported 95% confidence interval was minus 3.57 to minus 0.11. The reported p-value was 0.04. That result is presented as a between-group difference in change favoring the Personalized arm over usual care at month six.
The investigators nevertheless concluded that the precision-nutrition intervention did not improve most GV indices beyond usual care or Mediterranean diet-based counseling. That wording preserves the distinction between one CV finding at six months and the broader pattern of no significant differences. Readers seeking precision-nutrition effects should therefore focus on the full variability profile, not the single CV contrast alone.
How to interpret the single CV difference
The CV result describes relative variability around mean glucose, expressed in percentage points in the abstract. The mean difference of minus 1.84% represents a larger reduction in the Personalized arm than in UCC at six months. The confidence interval extends from minus 3.57 to minus 0.11, which does not include zero in the reported analysis. The p-value of 0.04 is presented as statistically significant in that pairwise comparison.
That finding does not alter the reported MAGE results. The MAGE comparisons remained nonsignificant for UCC versus Standardized, UCC versus Personalized, and Standardized versus Personalized. It also does not alter the reported statement that changes in other GV metrics were not significantly different among arms at 3 and 6 months as a general pattern. The abstract positions the CV observation as an isolated six-month contrast within an otherwise nonsignificant variability profile.
Who conducted and supported the work
The author list begins with Lauren Thomas Berube and includes Chan Wang, Margaret Curran, Mary Lou Pompeii, Charlie Thao, Emily P. Ribeiro, Souptik Barua, Lu Hu, Huilin Li, David E. St-Jules, Eran Segal, Michael Bergman, and Collin J. Popp. The funding statement identifies the National Institutes of Health under grant number R01NR018916. Those attribution and support details appear in the Diabetes abstract record.
Disclosure information is also listed in the abstract. Most listed authors are reported as having no disclosures. Michael Bergman is reported with an ended advisory-panel relationship with Novo Nordisk and a current advisory-panel relationship with Abbott. Collin Popp is reported as a current consultant for Longevix. Those relationships are disclosed without additional explanation in the abstract record.
What this means for nutrition care delivered remotely
For adults whose characteristics match the enrolled sample, the trial does not support adding real-time predicted-PPGR feedback to Mediterranean diet-based telemedicine counseling for the purpose of reducing MAGE. It also does not show that 14 sessions of standardized Mediterranean diet-based diabetes self-management counseling reduced MAGE or most other variability measures more than usual care in this moderately controlled population. The setting was fully remote, the counseling dose was 14 sessions over six months, and the glucose assessment used continuous monitoring at three defined points.
Consumers and clinicians should not read the single CV reduction as evidence of broad glycemic benefit. The investigators explicitly limited their conclusion to no improvement in MAGE or most GV indices beyond the comparators. Any future use of algorithm-driven meal feedback would need additional evidence showing consistent effects across variability measures, time points, and populations beyond the 161 participants described here. Until such evidence is reported, Mediterranean diet-based counseling remains a counseling framework tested in this trial, not a variability-lowering intervention proven superior to usual care in this sample.