Adults with overweight or obesity taking the new generation of weight loss drugs semaglutide and tirzepatide tend to eat significantly less, leaving them vulnerable to nutritional deficiencies, according to one of the first real-world studies to examine dietary behaviour in people using glucagon-like peptide-1 receptor agonists and dual GLP-1/glucose-dependent insulinotropic polypeptide (GIP) dual agonists RAs (collectively referred to as GLP-1 RAs) with the help of an AI-powered nutritional tracking app.
The study, being presented at this year's European Congress on Obesity (ECO) in Istanbul, Turkey (12-15 May), is by Dr Valentina Vinelli from the metabolic health research team, led by Dr Rebecca De Lorenzo and Professor Patrizia Rovere Querini at IRCCS San Raffaele Hospital and Vita-Salute San Raffaele University in Milan, Italy, in collaboration with the Robin Health team.
"While GLP-1 RAs produce impressive weight loss, the question of whether meaningful muscle loss accompanies fat loss with these therapies is still hotly debated. What remains universally agreed, however, is that protecting muscle mass should be a central goal of any weight-loss programme," said lead author Dr Vinelli.
"Muscle health depends on sufficient protein intake and regular physical activity, especially resistance exercise. For people using GLP-1 RAs, reduced appetite can make it harder to meet protein needs, making what you eat matter more than ever, especially given that protein requirements tend to be higher in people with obesity and during active weight loss. Protein requirements should always be personalised, expressed as grams per kilogram of adjusted body weight, rather than treated as a one-size-fits-all target."
GLP1-RAs such as semaglutide (sold under the brand names Ozempic/Wegovy) and tirzepatide (Mounjaro/Zepbound) mimic hormones the body already produces, GLP-1 and GLP-1/GIP respectively, that are released in response to eating, increasing a feeling of fullness and reducing food cravings.
GLP-1 RAs have been shown to reduce calorie intake by between 16–39%, making them a powerful tool for the treatment of obesity. However, little research has examined their impact on diet quality or the intake of key macro- and micronutrients, such as protein, fibre, vitamins, iron and calcium, needed to maintain good health.
Inadequate micronutrient intake can put people at risk of a range of conditions, from hair loss, fatigue, and slow wound healing to an impaired immune system and osteoporosis.
"While GLP-1 RAs medications are transforming obesity care, we still know very little about how they affect eating patterns and nutrient intake," explained Dr Vinelli. "What's more, many people taking GLP-1 RAs receive little or no systematic guidance on diet quality, protein intake, or micronutrient adequacy."
Traditional dietary assessment tools such as food questionnaires and diaries are limited by recall bias. Recent advances in AI and digital health offer promising solutions, allowing users of mobile apps to record and analyse their eating habits in real time.
In response to these trends, researchers analysed 5,741 days of data from 332 adults with overweight or obesity (who gave their consent) using a mobile dietary tracking app between July 2025 and February 2026.
They compared total energy intake, macronutrient intake, weight-adjusted protein adequacy, meal patterns, and weight loss between 116 GLP-1 RAs users (using either semaglutide or tirzepatide) and 216 non-users (average BMI 31.3 vs 30.6 kg/m², 52.5 vs 50.0 years, 72% vs 63% female, respectively). Of the 116 GLP-1 RA users in the study, 87 logged at least one injection in the app. Of those, 62 were using tirzepatide and 25 were using semaglutide. The remaining 29 were identified as GLP-1 users through their profile but did not log individual injections. It is important to note that the dose data reflects what users chose to log in the app.
Of the 25 semaglutide users who logged their injections, 10 (40%) recorded a maximum dose of 0.25 mg (the starting dose), 4 (16%) recorded 0.5 mg, 6 (24%) recorded 1.0 mg, 2 (8%) recorded 1.7 mg, and 3 (12%) recorded 2.4 mg (the maintenance dose). Of the 62 tirzepatide users who logged their injections, 12 (19%) recorded a maximum dose of 2.5 mg (the starting dose), 21 (34%) recorded 5.0 mg, 10 (16%) recorded 7.5 mg, 17 (27%) recorded 10.0 mg, and 2 (3%) recorded 12.5 mg.
The Robin Health mobile app includes image recognition (from a photo of a meal the app analyses ingredients and calculates calories, carbohydrates, proteins, and fats), barcode scanning, as well as text and voice logging, providing real-time macronutrient intake estimates [1].
The analysis found that participants using GLP-1 RAs medications reported significantly lower total energy intake compared with non-users (average 1,102 vs 1,281 kcal/day) as well as significantly lower intake of macronutrients, including protein (53.8 vs 62.0 g/day), carbohydrates (128 vs 143 g/day), and fat (39.7 vs 45.7 g/day).
Despite lower absolute intake, macronutrient distribution (the proportion of daily calories derived from macronutrients) was nearly identical (protein 20% vs 19%; carbohydrate 47% vs 46%; fat 33% vs 34%), indicating a proportional reduction in food intake rather than changes in diet quality.
However, weight-adjusted daily protein intake was critically low in GLP-1 RAs users (0.6 g/kg/day), with the majority (88%) falling below the Italian national recommendation for protein intake (of 0.9 g/kg/day). In non-users protein consumption was also below this recommendation at 0.7 g/kg/day, with 69% below the national recommendation, showing that inadequate protein consumption is a concern regardless of the weight loss strategy employed, but more of a concern in those using GLP-1-RAs. According to the most recent Italian national food consumption survey (IV SCAI, 2017 to 2020, published in Nutrients 2025), the average Italian adult consumes 1.1 to 1.2 g/kg/day of protein, well above the LARN recommendation of 0.9 g/kg/day. Only about 7% of the general Italian population falls below this threshold.
Moreover, GLP-1 RAs users were much more likely to skip breakfast (31% vs 16% of days), lunch (31% vs 18%) and dinner (40% vs 30%) than non-users—reducing opportunities for adequate protein distribution across the day.
As expected, among the subset of 177 participants (76 GLP-1 RAs users and 101 non-users) with longitudinal weight data, GLP-1 RAs users experienced significantly greater weight loss than non-users (average −2.2 kg, −2.5% vs −0.2 kg, −0.2%), with nearly a quarter (24%,18 participants) achieving clinically meaningful weight loss (at least 5% body weight loss) compared with just 3% (3 participants) of non-users.
The authors explain that the modest weight loss seen in the GLP-1 group reflects the real-world conditions of the study. The study did not independently verify or control drug adherence; medication use was self-reported through the app, and it is possible that some users continued taking their medication without logging it. Most users tracked their weight for only a few weeks (median duration between first and last weight measurements across the group was 22 days), and the majority of those who did log their injections appeared to still be in the dose titration phase rather than on full maintenance doses. However, among the 13 users who tracked weight for at least 3 months, the average weight lost was 7.3%, on course towards the 10% plus weight loss seen at 12 months in randomised clinical trials.
The authors say: "Real-world GLP-1 RA use in adults with overweight or obesity is associated with widespread protein inadequacy and increased meal skipping, underscoring the urgent need for proactive nutritional monitoring and personalised dietary guidance to be integrated alongside obesity treatment to support long-term health beyond weight loss," said Dr Vinelli.
"In our future work, we will look beyond protein to explore how GLP-1 RAs affect the intake of other key nutrients and examine how real-time mobile apps could help people stay on track with healthy eating and lifestyle habits throughout their journey."
The authors acknowledge several limitations to the study, including the small sample size and its descriptive nature, which precludes causal inference and warrants confirmation in larger studies. In addition, dietary data were self-reported via app-based logging and may underestimate true intake, although AI-assisted multimodal capture likely reduces recall bias compared to traditional dietary assessment methods.
