GLP-1 Medications Reshape Food Buying: What the Data Show

Abstract

Study at a Glance

Study Snapshot

Study Facts Table

What Researchers Actually Did

Hristakeva and colleagues linked proprietary survey data on GLP-1 medication adoption to verified household transaction records maintained by Numerator, a market research firm tracking approximately 150,000 U.S. households. Surveys were fielded in four waves between October 2023 and July 2024, capturing self-reported GLP-1 use, adoption timing (reported in three-month intervals), and stated reason for use. The final estimation sample consisted of 22,691 households completing all four survey waves with internally consistent responses, including 2,602 that reported at least one household member initiating GLP-1 therapy between January 2023 and July 2024. Matching variables were type 2 diabetes status, household income, age, presence of children, preadoption per-member grocery spending, and a validated Healthiness Index score derived from semantic text embeddings of grocery item descriptions.

The primary analytic tool was a two-way fixed-effects (TWFE) difference-in-differences regression with household and calendar-month fixed effects, comparing adopter households against their nearest-neighbor-matched nonadopter controls across three postadoption windows: months 1–6 (short term), months 7–12 (medium term), and months 1–6 following confirmed discontinuation. The spend outcome was log-transformed as ln(spending + 1) at the household-by-month level. The identical specification was applied to 40 grocery product categories, to food-away-from-home spending, and to grocery item count as a quantity proxy. Six alternative DiD estimators, including methods from Callaway and Sant’Anna (2021), Borusyak, Jaravel, and Spiess (2024), and Sun and Abraham (2021), were used to validate the TWFE results against concerns about staggered adoption and treatment effect heterogeneity.

Key Findings: Primary Outcomes

• Overall grocery spending fell 5.3% in months 1–6 after adoption (95% CI: −6.8% to −3.7%; p < .001), translating to approximately $390 in annualized savings per household against a mean annual spend of $7,400.
• Income moderated the effect substantially: households earning $125,000 or more annually reduced grocery spending by 8.2% (approximately $700/year), more than double the 3.9% reduction ($260/year) seen in households earning below $125,000.
• Spending reductions reflect quantity reductions, not trading down: the nearly identical pattern seen when item count (rather than dollars) was used as the outcome variable indicates households are buying fewer items, not simply switching to cheaper alternatives.
• Medium-term persistence with attenuation: grocery spending remained −4.1% below baseline in months 7–12 for the overall sample (p < .001), though the weight-loss subgroup’s medium-term estimate was smaller in magnitude and not statistically significant.
• Postdiscontinuation reversion: among the 888 households (34%) that discontinued GLP-1s within the sample period, the postdiscontinuation grocery spending estimate was −1.9%, statistically indistinguishable from zero (wide confidence interval), consistent with return toward preadoption levels.
• No meaningful pre-adoption trend was detected across the six pre-adoption months; the joint Wald test of pretreatment coefficients was not significant (p = .87), and a placebo test reassigning adoption six months earlier produced a flat spending profile.

Key Findings: Secondary Outcomes and Subgroup Analyses

Category-level grocery shifts (months 1–6):

• Chips and savory snacks: −10.1%
• Sweet bakery: −8.8%
• Cookies: −6.5%
• Cheese: meaningful decline with high firm-level revenue concentration
• Spending declined across the majority of all 40 analyzed categories, including staples such as meat, eggs, bread, and vegetables, indicating a broad reduction in food volume rather than a targeted elimination of indulgent items alone.
• Only four categories showed directionally positive estimates: nutrition bars, fresh fruit, meat snacks, and yogurt (yogurt carrying the highest positive point estimate). None of these positive estimates were tested for significance individually and should be interpreted with caution.
• In the medium run (months 7–12), candy and chocolate increased 6.7% relative to preadoption; after discontinuation, the same category rose 11.4% above preadoption baseline, the most notable category-specific exception to general reversion.

Food away from home: Limited-service restaurant spending fell 8.0% (p < .001) in months 1–6, with broadly consistent magnitudes across income and motivation subgroups. Medium-term and postdiscontinuation estimates for food-away-from-home were imprecise and did not reveal a clear pattern.

Motivation subgroups: Weight-loss adopters showed a slightly larger short-term grocery decline (5.6%) than diabetes-management adopters (4.6%), though this difference was not statistically significant. Category-level patterns were similar across both groups. Diabetes-management users sustained medium-term grocery declines at approximately the same magnitude as their short-term decline, whereas weight-loss users showed attenuation.

Basket healthiness: The Healthiness Index showed directionally positive but statistically insignificant improvements during active GLP-1 use, and a measurable decline after discontinuation. The authors note the HI captures compositional change rather than total caloric load, so it likely understates dietary improvement during active use.

Aggregate economic projection: Applying the 5.3% household-level reduction to the 16.3% household adoption rate implies approximately a 0.9% reduction in total U.S. off-premise food and beverage spending, equating to roughly $9 billion annually against a $1 trillion market. An analogous calculation for limited-service restaurants implies approximately $7 billion in reduced annual spending.

Results: Adverse Events and Safety Profile

This study was not a clinical trial and did not assess medication tolerability or adverse events directly. The 34% in-sample discontinuation rate is consistent with published real-world discontinuation data of 35%–65% within the first year, cited from Do et al. (2024) and Rodriguez et al. (2025). The authors note possible discontinuation reasons include cost, side effects, changes in treatment plans, or achievement of weight-loss goals, but these were not directly observed. No safety or tolerability data were collected or reported in this study.

Statistical Approach and Rigor

The primary analytic approach, a TWFE DiD with nearest-neighbor matching on six predefined covariates, is methodologically sound for this observational setting. The use of six alternative DiD estimators (Callaway and Sant’Anna; Borusyak, Jaravel, and Spiess; Gardner; Roth and Sant’Anna; Sun and Abraham) to test for biases from staggered adoption and treatment effect heterogeneity represents a commendable robustness protocol. Parallel trends were assessed via event-study plots with a joint Wald test (p = .87) and a placebo reassignment test, both supporting the identifying assumption. Robustness to alternative functional forms was assessed via negative binomial, linear probability, and levels-based models. An empirical Bayes shrinkage analysis confirmed category-level estimates were not driven by noise. Standard errors were clustered at the household level throughout. The primary limitation of the statistical framework is the acknowledged inability to separate pharmacological effects from concurrent unobserved behavioral changes, a limitation the authors address transparently and repeatedly.

Clinical Takeaway

GLP-1 receptor agonist adoption is associated with a measurable and category-specific reduction in household food purchasing that extends across both grocery and restaurant channels. The reduction is driven by decreases in purchase quantity rather than price substitution, is largest in calorie-dense processed food categories, and persists for at least twelve months of active medication use. Spending reverts toward preadoption levels after discontinuation, which aligns with what randomized trials have shown about weight regain following GLP-1 cessation. For prescribers, this population-level transaction evidence reinforces the pharmacological basis for the appetite suppression these medications produce, while also highlighting that the dietary benefits are medication-dependent rather than durable habit changes.

Why This Matters Clinically

Prior research on dietary change following major health shocks, including new diagnoses of type 2 diabetes, cardiovascular disease, job loss, or childbirth, has consistently found minimal and short-lived modifications to food purchasing behavior. Behavioral interventions, nutrition labeling, and soda taxes produce modest effects at best. GLP-1 receptor agonists introduce an entirely different mechanism: biologically driven appetite suppression that overrides the social and psychological inertia sustaining poor dietary patterns. The magnitude of the food demand shifts documented here, concentrated in the processed and ultra-processed categories most strongly associated with cardiometabolic harm, is without precedent in the consumer dietary behavior literature. This has direct implications for prescribers managing obesity and diabetes: the metabolic signal is not limited to the scale and the hemoglobin A1c. It extends to the grocery cart.

CED Clinical Relevance

At CED Clinic, many patients using GLP-1 receptor agonists report changes in what foods appeal to them, what quantities feel tolerable, and how often they seek convenience foods. This study provides systematic, population-level transaction evidence that these anecdotal reports reflect real purchasing behavior: calorie-dense snack and bakery items decline sharply, staple food purchases decrease broadly, and yogurt and fresh fruit show modest gains. For CED patients managing weight alongside cannabis-related factors, the interplay of appetite modulation from GLP-1 therapy and appetite stimulation from cannabis is clinically relevant. The demonstrated reversibility of dietary gains upon GLP-1 discontinuation argues for careful, individualized discontinuation planning rather than abrupt cessation, particularly in patients whose dietary patterns were already suboptimal at baseline.

Fits What We Already Know

The findings align with and extend several lines of prior evidence cited within the paper itself. Randomized controlled trials from the SURMOUNT program (Garvey et al. 2023; Jastreboff et al. 2022; Wadden et al. 2023) established that GLP-1 receptor agonists produce 15%–20% body weight reductions causally attributable to appetite suppression, and that these gains partially reverse upon discontinuation (Aronne et al. 2024). Semaglutide specifically has been shown to reduce food cravings, appetite, and preference for high-fat foods in controlled settings (Blundell et al. 2017). The literature on dietary responses to health shocks provides the counterfactual context: diabetes diagnoses produce minimal dietary change (Oster 2018), with improvements in sugar intake offset by increases in fat and sodium and poor adherence to dietary guidelines (Ma, Ailawadi, and Grewal 2013; Ponzo et al. 2017). The magnitude of the GLP-1 dietary signal documented here is therefore exceptional relative to every previous intervention or life event studied in the consumer food demand literature.

What This Study Teaches Us

At the population level, GLP-1 adoption translates into measurable, category-specific reductions in food purchasing that are detectable in household transaction data. The reductions are not limited to “junk food” but extend across most grocery categories, including protein sources and staples, indicating a general reduction in caloric intake rather than targeted elimination of indulgent items. The small number of categories showing spending increases, yogurt and fresh fruit prominently, suggests a modest compositional shift toward nutrient-denser options, though the Healthiness Index changes were not statistically significant during active use. These patterns persist during medication use and reverse upon stopping, providing compelling observational support for the pharmacological basis of the behavioral changes. The income gradient is particularly striking: higher-income households show more than double the grocery spending reduction of lower-income households, a finding with implications for both equity and the aggregate economic impact on the food industry.

What This Study Does Not Show

• The study cannot isolate pharmacological effects of GLP-1 medications from concurrent behavioral changes (e.g., dietary counseling, increased physical activity, or health-conscious lifestyle adoption that may co-occur with GLP-1 initiation).
• Food purchasing is not equivalent to food consumption; waste and household food sharing are not captured.
• Individual-level dietary changes are not observable; only household-level aggregate spending is measured.
• Nutritional content of purchased items beyond the Healthiness Index proxy is not assessed; caloric density, macronutrient composition, and sodium load are not reported.
• Reasons for discontinuation are unobserved; the postdiscontinuation analyses cannot distinguish between patients who stopped due to side effects, cost, goal attainment, or physician direction.
• Long-term (beyond twelve months) purchasing patterns are not captured in the sample window.
• The directionally positive category estimates (yogurt, fresh fruit, nutrition bars, meat snacks) are not individually tested for statistical significance and should not be interpreted as confirmed increases.
• Specific GLP-1 agents are not differentiated; the analysis treats all GLP-1 receptor agonists as a class.

Fits the Broader Conversation

This paper arrives at a moment when GLP-1 adoption is accelerating rapidly, household-level incidence rose from 11.5% in October 2023 to 16.3% by July 2024 per this dataset, and the food industry, public health community, and payers are

Physician-Led, Whole-Person Care

A doctor who takes the time to truly understand you.

Personal care that starts with listening and is guided by experience and ingenuity.

Health, Longevity, Wellness

One-on-One Cannabis Guidance

Metabolic Balance

Leave a Message
Metabolic Care
Medical Consulting
Cannabis Care