Digital Food Purchase Signatures and Their Response to GLP‐1‐Based Treatment in People With Obesity
Freya Orban, Laura Fernandez Cerro, Melanie Stoll, Maurice Rupp, David Herzig, Lia BallyABSTRACT
Aims
To characterise digital food purchase signatures in adults with obesity (OB) versus normal‐weight (NW) controls and explore changes from pre‐treatment to early (5‐month) and later (10‐month) phases of glucagon‐like peptide‐1 receptor agonist (GLP‐1RA) therapy, including high and low weight‐loss responders.
Materials and Methods
We obtained household‐level loyalty‐card food purchase data at two major Swiss grocery retailers for 56 OB adults and 92 NW controls, covering 12–24 months. Purchases were matched to a nutritional database and analysed by purchase patterns using nutrient profiles, food group distributions, and composite diet quality scores, comparing OB and NW cross‐sectionally and, within the OB group, longitudinally following GLP‐1RA initiation.
Results
Compared to NW (BMI 22.0 ± 1.7 kg/m 2 ), OB participants (BMI 37.8 ± 5.2 kg/m 2 ) purchased a higher proportion of energy from protein (mainly animal‐based), higher energy‐adjusted salt content, but smaller energy shares of fruit, vegetables, nuts, whole foods and scored lower on the Dietary Phytochemical Index, Simplified Healthy Food Diversity Index and alternate Mediterranean Diet Score (all p < 0.05). GLP‐1RA initiation was followed by early increases in protein‐derived energy, fibre content, energy from fruit, vegetables and unprocessed foods (all p < 0.05). However, by 10 months, purchasing patterns no longer differed meaningfully from pre‐treatment levels in the full cohort. High/low weight‐loss responders did not differ in food purchase patterns, except for a greater increase in salt content among low responders at 5 months.
Conclusions
Distinct digital food purchase signatures were observed in OB versus NW adults. Within OB, no clear longitudinal shift was evident after 10 months of GLP‐1RA treatment, although the sample size and household‐level data limited individual‐level precision and interpretability.