Exploring Dietary Intake in Adults with Type 2 Diabetes Using GLP-1 Receptor Agonists: A Cross-Sectional Analysis

This is a cross-sectional study enrolling patients with T2D who attended visits at the Diabetes and Metabolic Diseases Clinic, "Città della Salute e della Scienza" Hospital of Turin in Italy from February to June 2025. Patients were enrolled if they met the following inclusion criteria: age between 40 and 75 years, diagnosis of T2D, and treatment with either GLP-1 receptor agonists or oral hypoglycemic agents initiated at least 1 month prior to the enrollment. Patients were excluded if they were receiving insulin therapy, either alone or in combination with other agents, or if they had any medical conditions that could affect food intake. Additionally, individuals who were pregnant or lactating, as well as those unable to provide informed consent, were excluded from the study. GLP-1 receptor agonists were prescribed within the therapeutic ranges commonly used in clinical practice for patients with T2D, with the following ranges: liraglutide 1.2–1.8 mg/day, dulaglutide 0.75–1.5 mg/week, and semaglutide 0.25–1 mg/week.

During the follow-up visit at the center, all participants provided socio-demographic information and underwent clinical and dietary assessments, along with anthropometric measurements.

Socio-demographic data included participants' age, sex, educational level, economic status, family composition, and smoking habits.

Dietary habits were collected using a medium-length food frequency questionnaire (FFQ), which has been validated for use in the Italian population [12]. This semi-quantitative tool evaluated habitual consumption patterns over the preceding year, focusing on 36 food items commonly consumed in Italy. Participants reported both the frequency and approximate quantity of intake for each item, selecting portion sizes from predefined options illustrated with reference images or standard household measures (e.g., cups, spoons, slices). To improve accuracy in estimating portion sizes, standardized food photographs were provided. The food items were organized into categories such as beverages (including coffee, alcoholic drinks, and soft drinks), dairy products, meat, fish, and eggs, cereals, fruits and vegetables (including legumes), fats and dressings, and miscellaneous items like sweets, fried foods, and fast food. Total energy (measured in kilocalories) and macronutrient intake (expressed both in grams and percentage of total calories) were calculated using WinFood® Pro 3.37.x software (Medimatica, Colonnella, Teramo, Italy, 2023).

The MDS (Mediterranean Diet Score) was calculated to assess participants' adherence to the Mediterranean diet, as it is a validated index in Italian and European populations with T2D and specifically adapted to the Mediterranean dietary context. The score was constructed following the method originally proposed by Trichopoulou et al. [13], assigning one point for each typical Mediterranean diet component. Specifically, nine food groups were considered: high consumption of vegetables, fruits and nuts, legumes, cereals, fish, and a high ratio of monounsaturated to saturated fats; low consumption of meat and meat products; and moderate alcohol intake. For each component, participants with consumption above the median (or below, for "unhealthy" components) were assigned one point. The total score ranges from 0 to 9 with higher values indicating greater adherence [13].

Physical activity levels were assessed using the International Physical Activity Questionnaire (IPAQ) short form [14]. This validated tool measures the frequency and duration of walking, moderate, and vigorous physical activities performed during the previous seven days, allowing classification of participants into different activity levels (low, moderate, high).

Body weight was recorded with participants wearing light clothing and no footwear, using a mechanical column scale (SECA model 711, Hamburg, Germany) with a precision of 0.1 kg. Height was measured to the nearest 0.1 cm using a SECA 220 stadiometer (Hamburg, Germany). Waist and neck circumferences were taken with a plastic measuring tape, positioned at the umbilicus level [15] and on the cricoid cartilage [16], respectively.

Eventually, the presence of other co-morbidities and/or diabetes complications, disease duration as well as biochemical parameters including HbA1c, fasting blood glucose and lipid profile were extracted from patients' medical records.

Data analyses were performed using IBM SPSS (version 28, Chicago, IL, USA). Continuous variables were reported as means mean ± standard deviation (SD), unless otherwise specified, whilst categorical variables were expressed as frequencies and percentages. The Kolmogorov–Smirnov test and the Shapiro–Wilk test were used to examine whether variables were normally distributed. Unpaired t-test or Whitney U test was used for continuous and non-continuous variables to determine differences in anthropometric, biochemical and nutritional variables between patients taking GLP-1RAs and those undertaking other medications. The chi-square or Fisher's exact tests were applied for categorical variables. The analysis of variance (ANOVA) or the Kruskal-Walli's test were used to compare variables between more than two groups, and post hoc comparisons were made where appropriate. Finally, Spearman's linear correlation was applied to evaluate correlations between adherence to Mediterranean Diet, dietary intake, age and anthropometric variables. Significance was defined as p < 0.05.

Although no a priori sample size calculation was performed, a post hoc power analysis was conducted using protein intake (expressed as percentage of total energy) as the benchmark variable, given its role on nutritional status. Using 15% of total energy from protein as the reference value recommended by dietary guidelines, the analysis indicated a statistical power of 81% (α = 0.05, two-tailed t-test), supporting the adequacy of the sample size for this outcome.

All participants provided written informed consent prior to their inclusion in this study. The research protocol was reviewed and approved by the local Ethics Committee of Città della Salute e della Scienza di Torino hospital (protocol number CS/662 of 19 October 2015). All procedures were conducted in accordance with the ethical standards of the Declaration of Helsinki.

Energy and macronutrients intake retrieved from the semi-quantitative FFQ were calculated in both absolute and percentage values for both GLP-1RAs and other medications users and shown in Table 2. Results were similar in both groups, except for monounsaturated fatty acids (MUFA) resulting higher for GLP-1 users (p = 0.03) and characterized by low intakes of carbohydrates (44% of energy intake -EI-), total (~16 g) and adjusted dietary fibers per 1000 kcal/day (~11 g/1000 kcal) as well as by a high intake of fats (39–40% of EI), with saturated fatty acids (SFAs) below 10% of EI. The macronutrient distribution observed in both groups was not in line with the recommendations proposed by the European Association for the Study of Diabetes (EASD) dietary guidelines, as shown in Table 2. Similar data were observed in both absolute and percentage data when the analysis was separated by sex (Table S1).

In Figure 1, the percentage of patients meeting or not the dietary guidelines are presented, highlighting that the intake of carbohydrate was reduced in over 50% of patients, whereas fats were elevated in about 70% of them, unrelated to drugs treatment. Also, we found that none of the patients in GLP-1RAs group, but just one in the other group met fiber recommendations, whereas high intake of soluble sugars was shown in over 50% of patients (>10% of EI) (Figure S2).

Eventually, the adherence to the Mediterranean diet assessed by the MDS index did not show any significant differences between the GLP-1RAs and other medications group (MDS mean ± SD: 4.31 ± 1.39 vs. 4.39 ± 1.60; p = 0.78; median [range]: 4 [2,3,4,5,6,7] vs. 4 [2,3,4,5,6,7,8,9]; p = 0.82). Score distributions for both groups are presented in Figure 2.

Sperman's linear correlations showed that the MDS was directly correlated to age (r = 0.220, p = 0.025), soluble sugar (r = 0.201, p = 0.039), total fiber intake (r = 0.416, p < 0.001) and fiber adjusted for 1000 kcal (r = 0.480, p < 0.001) but inversely associated with protein (r = −0.197, p = 0.046) and cholesterol intake (r = 0.252, p = 0.010) in the whole sample. Repeating the analysis for each group, the same results were found in the other medications, apart from age, while MDS was positively correlated to age (r = 0.292, p = 0.036), total fiber (r = 0.317, p = 0.022) and fiber adjusted for 1000 kcal (r = 0.368, p = 0.007) in GLP-1RAs users.

To evaluate how dietary habits might be affected according to treatment duration, since patients started taking GLP-1RAs at different time points, a subgroups analysis was performed by splitting the whole sample into 3 subgroups as follows: patients taking medications for less than 1 year (<1 y), those taking mediations between 1 and 2 years (1–2 y) and the last group using GLP-1RAs for over 2 years (>2 y). Data on both energy and macronutrients are shown in Table 3. Overall, we did not find any significant differences in energy and macronutrient intake among the 3 subgroups.

Percentage values for macronutrients, SFA and soluble sugars were presented in Figure 3, where it seemed that those taking medications for less than 1 year showed a slightly higher intake of protein and lower intake of carbohydrates than the other two groups, without being statistically significant.

Finally, we looked into the weekly consumption of the main food or food groups retrieved from the FFQ according to the 3 subgroups of GLP-1RAs users, considering as reference portion for each food group the one proposed by the fifth version of LARN [17]. As reported in Figure 4, no significant differences were found in the number of consumed portions among patients, due to their huge variability. However, the median weekly consumption of different foods such as fish or legumes as well as of vegetables and fruit did not reach the dietary recommendations. Also, some patients did not eat some food at all.

Taken together, our results highlight an excessive intake of total and saturated fats and an inadequate fiber consumption, irrespective of medication type, suggesting that concurrent nutritional interventions remain essential. From a clinical perspective, the present findings reinforce the need for structured and individualized nutritional counseling to be systematically included into diabetes care management, even in patients treated with GLP-1RAs—both at therapy initiation and during follow-up, to monitor and address potential nutritional imbalances and optimize clinical outcomes.

Several drawbacks must be acknowledged. An intrinsic limitation of this study is its descriptive and cross-sectional design, which does not allow for inferences or conclusions regarding causal relationships between GLP-1RA treatment and dietary intake. Nevertheless, it provides valuable real-world evidence on dietary patterns in patients with T2D, helping to identify potential gaps that require further investigation.

Although the semi-quantitative FFQ used in this study is validated for the Italian population and provides an acceptable estimation of energy and macronutrient intake, it relies on participants' recall and their ability to estimate portion sizes used to quantify intake, which may sometimes not precisely correspond to the food actually consumed. This limitation could introduce reporting bias—particularly among individuals with obesity [35]. However, results from the FFQ were carefully checked together with participants to ensure the accuracy and internal consistency of self-reported data. Such bias may nonetheless have attenuated true differences between groups, potentially leading to an underestimation of dietary variation. Furthermore, while this FFQ allows for reliable estimation of macronutrient intake, it is not sufficiently accurate to assess micronutrient adequacy, which represents a recognized limitation for self-reported dietary assessment tools. Given these potential drawbacks, it might be useful to incorporate more objective measures of dietary intake, such as nutritional biomarkers or digital food tracking, to enhance data reliability and minimize bias in collecting dietary intake data.

Another limitation is the small sample size that could have prevented us to detect potential differences between the two groups as well as among subgroups, even after adjusting data for confounding factors such as sex, race/ethnicity, age, or BMI. Furthermore, the lack of information about prior structured nutritional counseling and variations in both duration and dosage of GLP-1RA treatment could have, respectively, influenced eating habits and food intake response, potentially masking differences between groups; thus, residual confounding cannot be ruled out.

Moreover, as the sample was drawn from a single diabetes clinic, the results cannot be generalized to other populations or healthcare settings. Last, but not least, due to the nature of the study, we did not include systematic measurements of nutritional biomarkers (e.g., vitamins or micronutrient levels), which would have provided objective support for our findings. This limitation highlights the need for future prospective studies combining dietary questionnaires with biochemical markers of dietary intake to achieve a more accurate and multidimensional evaluation of nutritional status in this population.

Overall, these factors highlight that the findings should be interpreted as descriptive real-world evidence rather than causal relationships, and they emphasize the need for larger, prospective studies to further clarify these associations.