Data from wearable devices and food logs may help researchers understand how daily habits—and their timing—relate to metabolic health. A small observational study co-led by Ahmed Metwally, PhD, of Google’s Metabolic Health AI research program explored associations between meal timing, sleep patterns, physical activity, and glucose regulation, with the goal of informing personalized approaches to diabetes prevention and management. 

Researchers studied 36 adults from the San Francisco Bay Area: 16 with normoglycemia, 19 with prediabetes, and 1 with type 2 diabetes. They combined food logs, wearable measurements of sleep and physical activity, and continuous glucose monitoring, collecting data on 2,307 meals, 1,809 nights, and 2,447 days of activity. Participants also underwent oral glucose tolerance, insulin suppression, and isoglycemic intravenous glucose infusion tests to characterize their metabolic function. The researchers examined lifestyle–glucose associations and built machine learning models incorporating lifestyle and demographic features to predict metabolic characteristics, including muscle and adipose insulin resistance and incretin dysfunction. 

A greater proportion of daily calories consumed between 5 PM and 9 PM was associated with higher glycated hemoglobin levels, lower incretin function, and muscle insulin resistance. Greater variability in sleep efficiency was associated with higher nighttime and next-day mean glucose levels, while greater variability in bedtime was associated with higher next-day maximum glucose levels. More steps between 8 AM and 11 AM were associated with lower glucose levels within the next day among participants with muscle insulin resistance; among those with muscle insulin sensitivity, more steps between 2 PM and 5 PM were associated with lower glucose levels within the next 48 hours. 

The findings, reported in a preprint study that has not undergone peer review, require further investigation to determine whether changing these behaviors would improve glucose regulation. 

In part four of his conversation with AACE Endocrine AI Editor-in-Chief Johnson Thomas, MD, FSCE, FEAA, Dr. Metwally discusses what stood out in the study, why larger studies are needed, and how his research has influenced his own lifestyle priorities. 

(The following interview transcript has been edited for length and clarity.) 

Dr. Thomas: As part of the study, you found that when people ate appeared to matter more than what they ate, and the timing of exercise also seemed to affect glucose differently depending on the individual’s metabolic phenotype. Can you expand on the findings? 

Dr. Metwally: This was one of the studies from the same cohort that we used for metabolic phenotyping. In that cohort, we had gold standard metabolic testing, continuous glucose monitoring (CGM) at home, and wearables for the same participants, so we could capture their activity, activity intensity, and sleep time. And we also asked participants to log whatever they had eaten and when they had eaten it, so we could correlate that with the CGM traces and with their actual lifestyle. 

Our main question was about timing. What we found is that when you do things matters, as you suggested. Let’s say you ate a burger in the middle of the day, right before going to an exercise session. That is different from eating it right before going to bed. It’s completely different. 

When you eat and the order in which you eat different foods also matter. Let’s say you ate a certain amount of carbohydrates, but you had fat or protein before that. With the same amount of carbohydrates, the glycemic response is different. In many cases, it gets attenuated if you have fat or fiber before eating your carbohydrates. So, the timing and the order were things that we found very interesting. 

One thing that was also interesting in the study is that it’s not generalizable. You cannot say that timing affects everyone. Timing affects people with a specific subphenotype. So, if you have insulin resistance, you probably get that effect from timing. If you’re insulin sensitive, you don’t get that effect. If you have β-cell dysfunction, you get that effect. If you don’t, you don’t get that effect. All of these people are still normoglycemic, by the way, but because they have underlying dysfunction associated with a specific subphenotype, those are things that you want to optimize for. 

That was the whole analysis of it. It’s still a small study. But again, we can see signals, and those signals are statistically significant but still small. What we usually do with these small studies is test hypotheses: Do we see something called a sign of life? Is this something that’s worth investigating further, or is the signal not there and it’s all just noise? 

And we have seen enough signals. What we are trying to do now is scale these studies. Now that we have seen those signals, we want to do bigger studies and check whether the findings are still the same. Do they generalize to people from different ethnicities, races, and age groups? Or are they just specific to a particular group of people doing those things? 

I think it’s going to be great because it can make the claims more personalized and more accurate. If you are from the group of people that was tested, you should expect to see those results replicated in you. 

Dr. Thomas: On a personal note, did anything you learned by doing all this work lead you to change your own eating, exercise, or sleep habits? 

Dr. Metwally: The one thing I know I should do but haven’t done is improve my sleep. What I did change was the way I exercise. I’m a swimmer, but I’m trying to optimize resistance training. 

If you ask me why diabetes is what I’ve focused on, it’s because both of my grandparents have type 2 diabetes. But with my grandma specifically, I’ve seen every single complication that comes from type 2 diabetes. The problem with diabetes is the complications that come from it. 

And the only way to prevent it, at least to me and where I’m at, is to prevent insulin resistance. And that’s based on lifestyle factors. For insulin resistance, it's mostly strength training and weightlifting to add more muscle. So, after learning all of these different things about metabolic phenotypes—which lifestyle factors are important and which interventions and exercises should actually work for you—that’s what I’m trying to prioritize nowadays.