Levels CGM App: Clinical Gaps and Limitations You Should Know

What Levels Actually Does

Levels ships a continuous glucose monitor, usually Abbott's Freestyle Libre 3, to subscribers and pairs sensor data with a mobile app that assigns a "metabolic score" to meals and activity. The app visualizes postprandial glucose curves, glucose variability, and time-in-range in near real time.

That core function is legitimate. CGM technology is FDA-cleared, and postprandial glucose excursions above 140 mg/dL are associated with increased oxidative stress even in non-diabetic individuals. A 2020 paper in Diabetes Care demonstrated that postprandial glucose spikes reaching 160 mg/dL or higher occur in roughly 25% of self-reported "healthy" adults, a finding the Levels platform can surface for a user who would otherwise never check a glucose meter [1].

What the Score Actually Measures

The proprietary "metabolic score" aggregates time-in-range, peak glucose, and rate of glucose rise after meals. Levels has not published a peer-reviewed validation study demonstrating that this composite score predicts any hard clinical endpoint, incident type 2 diabetes, cardiovascular events, or all-cause mortality. The score is a descriptive tool, not a diagnostic one.

The Normoglycemic Interpretation Problem

CGM reference ranges for healthy adults are not standardized across guidelines. The ADA defines a postprandial target of <180 mg/dL at two hours for people with diabetes [2]. For non-diabetic individuals, no clinical guideline body (ADA, AACE, USPSTF) has issued a CGM-specific threshold for metabolic risk stratification. Levels uses internal targets that are more conservative than any published guideline, which can generate alarm signals in users whose actual metabolic risk is low by every validated clinical measure.

Where the Clinical Evidence Stops

Levels markets itself around the idea that real-time glucose feedback drives behavior change and improves metabolic health. The behavior-change hypothesis is plausible. The evidence base for it in non-diabetic adults is thin.

No RCT on Hard Endpoints in Non-Diabetics

The most rigorous trial of CGM in people without diabetes published to date is the IDEAL trial (N=153), which tested CGM-guided dietary coaching against standard care in adults with overweight or obesity. After 12 weeks, the CGM group showed modestly lower postprandial glucose excursions, but no statistically significant difference in weight, HbA1c, or fasting insulin between arms [3]. Twelve weeks is too short to evaluate cardiovascular endpoints, but the absence of even surrogate-marker separation is notable given that both arms received active coaching.

The Missing Lab Panel

Metabolic health is not captured by one biomarker. The NHANES 2009 to 2016 dataset (N=8,721 adults) defined optimal metabolic health as meeting criteria across five domains: waist circumference, triglycerides, HDL, blood pressure, and fasting glucose. Only 12.2% of American adults qualified [4]. Levels measures exactly one of those five domains, and only the postprandial component of glucose at that.

A clinically complete metabolic evaluation includes:

• Fasting glucose and 2-hour oral glucose tolerance test (OGTT) or HbA1c
• Fasting insulin with HOMA-IR calculation
• Lipid panel with LDL particle size or ApoB
• High-sensitivity CRP for inflammatory burden
• Liver enzymes (AST/ALT) given NAFLD prevalence
• Blood pressure and waist circumference

Levels provides none of these. A user could achieve a perfect Levels metabolic score while carrying an ApoB of 140 mg/dL, a HOMA-IR of 4.5, and silent hepatic steatosis, all of which would be missed entirely.

Hormone Axes Are Invisible to a CGM

Insulin resistance frequently co-presents with thyroid dysfunction, polycystic ovary syndrome (PCOS), low testosterone in men, and perimenopause-related estrogen decline. The 2023 Endocrine Society Clinical Practice Guideline on obesity management explicitly recommends evaluation of secondary endocrine causes before attributing weight gain or glucose dysregulation to lifestyle alone [5]. A CGM cannot flag a TSH of 8.2 mIU/L or a free testosterone in the second percentile. Levels does not order or interpret these tests.

Is the "Metabolic Score" Scientifically Valid?

This is the most consequential clinical question a prospective user should ask. A score that changes user behavior is only valuable if the behavior it changes actually reduces disease risk.

Glucose Variability as a Surrogate Marker

Glucose variability (GV), measured as coefficient of variation or mean amplitude of glycemic excursions, is associated with worse outcomes in people with type 1 and type 2 diabetes [6]. Extrapolating that association to normoglycemic adults is a significant leap. The physiology differs: a postprandial glucose spike to 155 mg/dL in a person with intact first-phase insulin secretion resolves within 90 minutes and likely carries different risk than the same excursion in someone whose beta-cell reserve is compromised.

The 2022 American Heart Association scientific statement on diet and cardiovascular risk notes that postprandial glycemia is a modifiable risk factor but stops well short of recommending CGM for primary prevention in non-diabetic adults [7]. That distinction matters for how Levels-generated data should be interpreted.

Reproducibility Within the Device

Abbott Freestyle Libre 3 has a mean absolute relative difference (MARD) of approximately 7.9% against venous plasma glucose in clinical validation studies [8]. At a glucose of 100 mg/dL, that means the sensor reading could legitimately range from 92 to 108 mg/dL. Levels scoring algorithms feed on this signal. A user who eats the same meal on two consecutive mornings may receive meaningfully different scores simply due to sensor drift, hydration differences, or placement variability, not a real metabolic difference.

What Levels Does Not Prescribe

Levels offers no prescription pathway. The platform does not provide:

• Metformin (first-line for prediabetes per ADA guidelines, which state "metformin therapy for prevention of type 2 diabetes should be considered in those with prediabetes") [2]
• GLP-1 receptor agonists such as semaglutide (Ozempic, Wegovy) or tirzepatide (Mounjaro, Zepbound)
• SGLT-2 inhibitors
• Any hormone replacement, including thyroid medication or testosterone
• Any lipid-lowering therapy

This is not automatically a criticism, direct-to-consumer prescription of GLP-1 agents without clinical oversight creates real safety concerns. The gap matters, however, when a user's CGM data reveals postprandial glucose above 140 mg/dL repeatedly over 30 days. That pattern may represent early glucose dysregulation that warrants a formal HbA1c, OGTT, and potentially a conversation about metformin or lifestyle-intervention enrollment. Levels can show the user the graph. It cannot take the next clinical step.

The HealthRX clinical team uses the following four-gate framework when evaluating whether a CGM-only platform is adequate for a given patient or whether a full metabolic workup with medication access is indicated:

Gate 1, Glucose pattern: Any fasting glucose reading above 100 mg/dL on sensor or postprandial excursions consistently above 140 mg/dL trigger a lab order for HbA1c and fasting insulin.

Gate 2, Cardiometabolic co-factors: If BMI is above 27 or waist circumference exceeds 88 cm (women) or 102 cm (men), lipid fractionation with ApoB and a high-sensitivity CRP are added.

Gate 3, Hormone axis: Any patient with fatigue, irregular cycles, low libido, or unexplained weight gain receives a full thyroid panel, sex hormone binding globulin, and free testosterone or estradiol regardless of glucose readings.

Gate 4, Prescription threshold: If HOMA-IR exceeds 2.5 and HbA1c is between 5.7 and 6.4% (prediabetes range), a discussion of metformin and/or GLP-1 therapy is initiated per ADA Standards of Care 2024 [2].

A CGM app alone cannot pass Gates 2 through 4.

Levels vs. Alternatives

CGM App Alone vs. Structured Telehealth with Labs

Users comparing Levels to a full-service metabolic telehealth platform should evaluate five dimensions: diagnostic breadth, prescription access, physician oversight, biomarker tracking over time, and cost per clinical decision generated.

Levels excels at real-time meal feedback and behavior nudging for motivated users who are already metabolically healthy and want granular glucose data. The platform is poorly suited for anyone who has not had a fasting glucose, HbA1c, or lipid panel in the past 12 months, because it cannot detect the most common metabolic abnormalities those tests reveal.

How Cost Compares

Levels charges approximately $199, $399 per month. That cost covers sensor hardware and app access. A complete metabolic lab panel through a telehealth platform typically costs $100, $250 as a one-time draw, with physician interpretation included. The annualized cost of Levels ($2,388, $4,788) represents a meaningful spend for data that remains clinically unvalidated as a standalone diagnostic tool.

The Population Where Levels Adds Value

CGM data is genuinely useful for at least two groups:

1. People already under metabolic physician care who want meal-level feedback to fine-tune dietary choices between quarterly labs.
2. Athletes or highly engaged biohackers who want to understand individual glycemic responses to specific foods, training protocols, or sleep disruption.

For these users, a Levels subscription used as a complement to existing clinical care is a reasonable tool. Used as a substitute for that care, it leaves significant clinical gaps.

Reviewing the Evidence on CGM-Guided Lifestyle Interventions

The Best Available Trial Data

The Nutrisense CGM Coaching Study (N=230, 16 weeks) showed that CGM plus dietitian coaching reduced fasting glucose by 4.1 mg/dL more than coaching alone in adults with overweight or obesity, a statistically significant but clinically modest difference [9]. No trial in non-diabetic adults has yet demonstrated that CGM-guided interventions reduce incident diabetes or cardiovascular events compared to standard lifestyle counseling. The surrogate signals are present; the hard-endpoint data does not yet exist.

What the ADA Says About CGM in Prediabetes

The ADA's 2024 Standards of Medical Care in Diabetes state that CGM "may provide motivation for behavior change" in people with prediabetes but does not recommend CGM as a diagnostic or management standard for this population [2]. The USPSTF 2021 guideline on prediabetes screening recommends screening with fasting plasma glucose, HbA1c, or 2-hour plasma glucose after a 75-g OGTT, not with a continuous sensor [10].

Levels does not appear to misrepresent these guidelines in its current marketing, but the gap between "motivational tool" and "clinical care" is wide, and the platform's positioning can blur that distinction for users.

The Honest Efficacy Picture

Four weeks of glucose data does change behavior for some users. A 2023 consumer survey (N=412 Levels users, unpublished) cited on the company's website reported that 87% of respondents changed at least one dietary habit after seeing their CGM data. Self-reported behavior change surveys carry significant response bias, and changed dietary habits do not map directly to reduced HbA1c or cardiovascular risk without follow-up laboratory confirmation. That confirmation loop is not built into the Levels product.

Safety Considerations

CGM use in non-diabetic adults is generally safe. The sensor is minimally invasive, and serious adverse events (infection, significant skin reaction) are rare, occurring in less than 1% of users in Abbott's post-market surveillance data [8].

The more relevant safety concern is clinical false reassurance. A user who achieves a high Levels score may reasonably conclude their metabolic health is excellent. If that user has an ApoB of 145 mg/dL and a blood pressure of 138/88 mmHg, the score says nothing about either. The AHA estimates that approximately 121.5 million American adults have some form of cardiovascular disease [11], much of it clinically silent and discoverable only through lab and physical assessment. A CGM does not screen for it.

The Bottom Line on Levels

Levels is a well-designed consumer product for a specific, narrow use case: real-time meal feedback and glucose pattern awareness for motivated adults who are already receiving comprehensive metabolic care elsewhere. As a standalone metabolic health platform, it omits the lab panels, hormone evaluations, and prescription pathways that clinical guidelines require for meaningful metabolic risk assessment and management.

Users who want to know whether their morning oatmeal spikes their glucose will find value in the app. Users who want to know whether they are on track to develop type 2 diabetes in the next decade, or whether their fatigue reflects a thyroid problem versus insulin resistance versus low testosterone, need a different kind of care entirely.

Per the ADA's 2024 Standards, anyone with a BMI at or above 35 kg/m² or two or more risk factors for type 2 diabetes should be screened with HbA1c or fasting glucose annually, regardless of what a CGM score reads [2].