- βAI-personalized diets reduced daily time above 140 mg/dL by 2.42 hours compared to the Mediterranean diet
- β61% of participants achieved diabetes remission (HbA1c below 6.5%) after 6 months on the personalized diet
- βTriglycerides dropped by 49 mg/dL, offering significant cardiovascular benefit beyond blood sugar control
| Country | Israel |
| Journal | BMC medicine |
| Year | 2022 |
| Authors | Rein, Ben-Yacov, Godneva |
| PMID | 35135549 |
A Study from Israel Shows AI-Powered Personalized Nutrition May Outperform Standard Dietary Guidelines for New T2D Patients
A study from Israel, conducted by researchers at the prestigious Weizmann Institute of Science, offers a fascinating glimpse into the future of diabetes management β one where your diet is designed not by general guidelines, but by an algorithm that learns how your body responds to food. For the roughly 1.5 million Americans newly diagnosed with type 2 diabetes (T2D) each year, this research could signal a major shift in how doctors approach dietary treatment.
What the Researchers Did
Scientists enrolled 23 adults recently diagnosed with T2D (average age 53.5 years) in a randomized crossover trial. Each participant tried two different 2-week diets in sequence: a Personalized Postprandial-Targeting (PPT) diet β built by a machine learning algorithm using each person's clinical data and gut microbiome profile β and the widely recommended Mediterranean-style (MED) diet. All participants wore continuous glucose monitors (CGM) throughout the trial and logged meals via smartphone. Sixteen participants then continued the PPT diet for an additional 6 months.
Key Findings: The Numbers
The AI-personalized diet produced significantly better blood sugar outcomes than the Mediterranean diet:
- Average post-meal glucose spike was 19.8 mg/dL lower per hour on the PPT diet (p<0.001)
- Average daily glucose was 7.8 mg/dL lower on the PPT diet (p<0.001)
- Participants spent 2.42 fewer hours per day with glucose above 140 mg/dL β a key threshold American diabetes guidelines flag as a post-meal danger zone
- Blood fructosamine (a short-term glucose marker similar to A1C) dropped significantly more on the PPT diet
After 6 months on the PPT diet, the improvements deepened:
- HbA1c dropped by 0.39% β meaningful progress toward the American Diabetes Association's (ADA) target of below 7.0% for most T2D patients
- Fasting glucose fell by 16.4 mg/dL (from a baseline already in the diabetic range)
- Triglycerides dropped by 49 mg/dL β a heart health bonus, since the ADA recommends triglycerides below 150 mg/dL
- Remarkably, 61% of participants achieved diabetes remission, defined as HbA1c below 6.5% without medication
The Gut Microbiome Connection
Notably, some of the clinical improvements were linked to individual changes in gut microbiome composition, reinforcing why a one-size-fits-all diet may not work for everyone β even a diet as well-studied as the Mediterranean plan.
Why This Matters for US Patients
In the United States, the ADA currently recommends Mediterranean, low-carb, and DASH-style diets as top dietary approaches for T2D management. This Israeli research doesn't invalidate those recommendations β but it does suggest that personalization may dramatically improve outcomes, especially for the newly diagnosed. The 61% remission rate in this pilot trial is striking compared to typical remission rates seen in standard dietary interventions. Additionally, CGM technology β central to this study β is becoming increasingly accessible to American patients through programs like MDS Diabetes, making this approach more realistic than ever. While this was a small pilot trial and larger studies are needed, the science of matching your diet to your biology is moving fast, and US patients should ask their providers about personalized nutrition options.
Study Citation
Rein, Ben-Yacov, Godneva, et al. "Effects of personalized diets by prediction of glycemic responses on glycemic control and metabolic health in newly diagnosed T2DM: a randomized dietary intervention pilot trial." BMC Medicine (2022). PMID: 35135549. DOI: 10.1186/s12916-022-02254-y. ClinicalTrials.gov: NCT01892956.
