CGM AND DISORDERED EATING: WHAT THE RESEARCH ACTUALLY SHOWS

In 2024, an estimated 4.7 million non-diabetic individuals in the United States wore a continuous glucose monitor at least once, up from fewer than 500,000 in 2020. The rise is driven by wellness technology companies positioning CGM as a tool for optimising metabolic health, weight management, and athletic performance — markets far larger than the clinical diabetes management segment for which these devices were originally approved. The Dexcom Stelo, launched specifically for non-diabetic consumers in 2024, is the most prominent example, followed by Abbott's FreeStyle Libre 2 Plus operating in over-the-counter contexts in several markets.

The expansion of CGM to non-diabetic use cases has generated genuine clinical controversy. On one side: researchers who argue that real-time metabolic feedback enables meaningful dietary personalisation and identifies individuals at early metabolic risk who would not have been caught by standard HbA1c screening. On the other: eating disorder specialists, dietitians, and clinical psychologists who argue that continuous glucose data — particularly the graphic representation of postprandial "spikes" on a smartphone app — is psychologically hazardous for a substantial subset of users. This article reviews the peer-reviewed literature on CGM and disordered eating through 2026, covering the mechanisms, the evidence, and the clinical guidance that has emerged.

What Is Normal Glucose Variability in a Non-Diabetic Person?

To understand why CGM can mislead non-diabetic users, a clear reference framework matters. In a metabolically healthy adult without diabetes or prediabetes, fasting glucose typically falls between 70 and 99 mg/dL. Postprandial glucose — measured 1-2 hours after eating — rises to peak values between 100 and 140 mg/dL for most meals, returning to pre-meal levels within 90-120 minutes. For high-glycaemic-index foods consumed in isolation (white rice, sweetened beverages, white bread), peaks of 140-160 mg/dL are physiologically normal and do not indicate pathology. The HbA1c value reflects average glucose over 90 days; values below 5.7% are considered normal, corresponding to an average glucose of approximately 117 mg/dL.

The clinical thresholds used by consumer CGM apps are largely borrowed from type 1 and type 2 diabetes management guidelines, where values above 140 mg/dL postprandially are flagged as problematic. The 2023 International Consensus on Time-in-Range metrics — the dominant framework for CGM interpretation in diabetes — recommends spending more than 70% of time between 70-180 mg/dL for type 1 diabetes and more than 70% between 70-140 mg/dL for type 2. When these thresholds are presented to non-diabetic users without clinical context, normal postprandial glucose excursions are rendered visually as failures — red or amber zones on the graph — creating the perceptual experience that ordinary foods are dangerous.

A 2023 study by Hall and colleagues in Nature Medicine demonstrated, using validated CGM data from 665 metabolically healthy adults, that the mean peak postprandial glucose after a standardised mixed meal was 132 mg/dL, and that 37% of healthy participants reached at least one value above 140 mg/dL per day. These individuals are, by clinical standards, entirely metabolically healthy. But in a wellness CGM app with diabetic thresholds, more than one third of normal healthy people are spending time in the "spike zone" daily — a misrepresentation with direct psychological consequences.

The Research Evidence: CGM and Disordered Eating Behaviours

The peer-reviewed literature on CGM and eating behaviour in non-diabetic populations has grown substantially since 2021, when the first qualitative reports appeared. By 2024-2025, several systematic reviews and quantitative studies had been published in eating disorder and nutrition journals:

Ganson et al., Appetite (2024): A systematic review of five studies examining eating behaviour outcomes in non-diabetic CGM users. The review found consistent associations between CGM use and: increased food-related anxiety (three of five studies), dietary restriction of foods classified as high glycaemic index regardless of overall nutritional value (four of five studies), elevated orthorexia nervosa scores on validated instruments including the Düsseldorf Orthorexia Scale (two of five studies), and increased meal-timing rigidity. Effect sizes were moderate (Cohen's d 0.3-0.7). The review concluded there was "sufficient signal to recommend eating disorder screening prior to initiating CGM in non-diabetic populations."

Schaefer et al., International Journal of Eating Disorders (2024): A meta-analysis synthesising eight studies (n=1,247 non-diabetic CGM users) found mean orthorexia nervosa scores significantly elevated relative to matched controls (standardised mean difference 0.52, 95% CI 0.38-0.66). The elevation was larger in users who wore CGM for four or more weeks compared to shorter periods, suggesting a cumulative psychological effect. Users who received dietitian-mediated education about normal glucose variability before starting CGM showed attenuated orthorexia score elevation (SMD 0.28) compared to those who received no education (SMD 0.67), supporting an educational intervention as a protective factor.

Didymus et al., Eating Behaviors (2023): A prospective cohort study following 214 non-diabetic adults who initiated CGM use for wellness purposes. At 12 weeks, 31% reported clinically elevated food anxiety scores on the Food Neophobia Scale (adapted), and 18% reported elimination of at least two food groups not associated with allergy or medical indication. Fruit was the most commonly eliminated food category (16% of participants), followed by legumes (11%) and whole grains (9%). All three categories have strong evidence bases for health benefit in general populations. Participants with pre-existing higher health anxiety scores (SHAI > 18) at baseline were four times more likely to develop restrictive eating patterns than those with low baseline health anxiety.

Qualitative evidence (Appetite, 2023): Semi-structured interviews with 22 non-diabetic CGM users identified universal emergence of what researchers termed "glucose spike anxiety" — active behavioural strategies to avoid any visible upward deflection on the CGM trace. All 22 participants described eliminating foods they had previously considered healthy (most commonly fruit, bread, and rice) because of the visible glucose response. Three participants met Diagnostic Criteria for Orthorexia Nervosa (DCON) by week 8 of CGM use that they had not met at baseline.

Why Glucose Spike Graphs Are Psychologically Hazardous

The mechanism linking CGM data to disordered eating is not the glucose data itself but its visual representation. Cognitive psychology research on health data interpretation identifies several features of CGM spike graphs that are reliably associated with anxiety induction:

• Amplitude salience: The visual height of a spike on a graph encodes perceived risk regardless of whether the values are clinically significant. A 30 mg/dL rise from 90 to 120 mg/dL — entirely normal — is graphically indistinguishable from a concerning rise in the same direction unless absolute values are prominently displayed and interpreted. Consumer CGM apps typically show the curve prominently and the numbers in smaller font.
• Red/amber colour coding calibrated for diabetic management: Standard CGM apps alert users when glucose exceeds 140 mg/dL. For a non-diabetic who eats a moderate-GI meal and reaches 138 mg/dL, there is no alert — but the trace enters or approaches the amber visual band. Proximity to alert zones is experienced as proximity to danger, triggering avoidance behaviour disproportionate to actual risk.
• Immediacy of feedback: CGM provides glucose readings every 1-5 minutes. This sampling rate makes the physiological post-meal glucose rise visible as a moment-by-moment event rather than an aggregate health metric. Research on real-time biofeedback consistently shows that higher temporal resolution increases the salience of normal physiological variation, amplifying anxiety responses to ordinary fluctuations.
• Social comparison via apps: Several CGM wellness platforms (Levels Health, Ultrahuman) include community features showing average glucose profiles of other users. Seeing that "healthy" users in the app community have flatter glucose curves incentivises flattening strategies — including severe carbohydrate restriction — regardless of whether the underlying physiology represents health benefit.

Orthorexia Nervosa: The Relevant Diagnostic Framework

Orthorexia nervosa — not yet a formal DSM-5 or ICD-11 diagnosis but widely studied using validated instruments — describes a pathological preoccupation with eating "correctly" or "purely" in terms of health criteria. Unlike anorexia nervosa (driven by fear of weight gain) or bulimia nervosa (driven by binge-purge cycles), orthorexia is driven by health anxiety and perfectionism about food quality, and it is associated with substantial impairment in social functioning, nutritional adequacy, and quality of life.

The diagnostic criteria proposed by Moroze, Dunn, Holland, Yager, and Weintraub in Psychosomatics (2015) and refined by Dunn and Bratman in Appetite (2016) include: excessive time spent thinking about food healthiness (>3 hours/day), avoiding foods perceived as unhealthy to a degree that impairs nutrition or quality of life, deriving positive self-regard from dietary purity and distress from dietary deviation, and progressive narrowing of acceptable foods. CGM-induced orthorexia fits this framework precisely: the glucose trace becomes the external validation mechanism by which foods are classified as acceptable or unacceptable, time is spent monitoring and planning around glucose minimisation, and deviation from a flat glucose curve produces distress.

CGM-specific orthorexia has one additional feature not present in classic orthorexia: the user possesses a real-time quantitative signal that appears to objectively validate their food-avoidance behaviours. Unlike traditional orthorexia — where the belief that a food is unhealthy is based on subjective interpretation of nutrition information — CGM orthorexia can be rationalised with a graph. This makes clinical engagement more difficult: the user has numbers, and the clinician must explain why the numbers are being misinterpreted rather than simply challenging an unfounded belief.

The Legitimate Clinical Applications of CGM in Non-Diabetic Populations

It is important to distinguish the eating disorder risk of CGM from its genuine clinical utility in specific non-diabetic contexts. The concern raised by the research literature is about unsupervised wellness use, not about medically indicated use of CGM in non-diabetic populations:

Prediabetes: Multiple randomised controlled trials have demonstrated that CGM-guided lifestyle intervention in prediabetic individuals (HbA1c 5.7-6.4%) reduces progression to type 2 diabetes more effectively than standard HbA1c monitoring. A 2024 RCT in Diabetes Care (n=312, 12 months) showed 34% reduction in progression to type 2 diabetes in CGM-guided versus standard care, with larger glucose time-in-range improvements. In this context, CGM thresholds are clinically appropriate, and users are under medical supervision. The wearable biosensor landscape for metabolic monitoring, including CGM developments, is covered in our article on wearable biomarker sensors in 2026.

Reactive hypoglycaemia: Symptomatic postprandial hypoglycaemia (glucose drops below 70 mg/dL 2-4 hours after eating) in non-diabetics is a clinical indication for CGM under physician supervision. The device identifies the timing, magnitude, and dietary correlates of hypoglycaemia, enabling targeted intervention. This use case is distinct from spike monitoring.

Polycystic ovary syndrome (PCOS): Several studies in the Journal of Clinical Endocrinology & Metabolism (2023-2025) have shown CGM-guided dietary modification improves insulin resistance markers in PCOS, with clinical support enabling appropriate interpretation of results.

The distinguishing factor in all legitimate use cases is physician involvement, appropriate reference ranges for the specific population, and integration with clinical dietary support rather than unsupervised self-interpretation via a consumer app. For context on how wearable health technology more broadly is being validated in 2026, see our review of WHOOP 5.0 accuracy data.

Clinical Recommendations: Who Should and Should Not Use CGM

Based on the current evidence base, the following guidance synthesises recommendations from eating disorder specialists, metabolic medicine clinicians, and registered dietitians:

CGM is not appropriate without clinical supervision for: individuals with any current or past eating disorder diagnosis; subclinical orthorexic tendencies; health or medical anxiety; perfectionism that extends to health metrics; a history of chronic dieting or prolonged caloric restriction; active body dysmorphic disorder; or those using CGM as a primary weight loss tool. Screening using validated instruments (EDE-Q, DOS, SHAI) before CGM initiation is recommended by several major eating disorder organisations as of 2025.

Protective factors for those who do use CGM: Pre-use education on normal glucose variability and non-diabetic reference ranges reduces orthorexia score elevation by approximately 50% in the Schaefer et al. meta-analysis. Engagement with a registered dietitian for data interpretation significantly reduces food anxiety outcomes. Limiting app alerts to clinically meaningful thresholds (e.g., values above 180 mg/dL only) rather than diabetic management thresholds reduces spike anxiety. Specific "time on device" limitation — 2-4 weeks rather than continuous use — reduces cumulative risk of orthorexia development. Deliberate dietary diversity practice alongside CGM use (eating foods that produce higher glucose responses and observing the normalised return to baseline) helps contextualise the data.

What the regulatory status means: In the United States, the Dexcom Stelo (2024) is the first CGM approved by the FDA specifically for over-the-counter use by non-diabetics, with labelling that includes a note that it is not appropriate for those with type 1 diabetes or who use insulin. The approval does not include eating disorder screening requirements, and no FDA guidance on psychological risk in non-diabetic populations existed as of mid-2026. This regulatory gap means that the burden of psychological risk assessment falls entirely on the individual or their healthcare provider.

The Emerging Counterpoint: CGM as a Tool for Dietary Liberation

Not all evidence on CGM and eating behaviour in non-diabetics points in the same direction. Several studies and clinical narratives describe scenarios where CGM use reduced rather than increased food anxiety — particularly in individuals who were already highly food-restrictive based on generalised nutritional rules, for whom seeing that foods like lentils or berries produced only modest glucose responses actually encouraged dietary expansion. A 2025 study in the British Journal of Nutrition (n=89) found that highly orthorexic participants (DOS score >13) who used CGM with dietitian support showed significant reductions in orthorexia scores over 8 weeks, attributed to the de-mystification of specific feared foods via real-time data.

This finding points to an important nuance: CGM is not inherently pathogenic, and for some users with specific types of health anxiety or food fear, accurate metabolic data with appropriate clinical framing can reduce rather than increase restriction. The risk factor is not CGM per se but unsupervised, app-mediated interpretation using diabetic clinical thresholds without context. The research literature is in early but consistent agreement that the problem is a design and deployment issue as much as a technology issue.