GLOW UP Study: Protocol for an Observational Digital Biomarker Study for Prediabetes Screening and Digital Phenotyping

The GLOW UP study is a prospective observational case-control trial in Switzerland designed to evaluate the feasibility of using wearable and smartphone-derived digital biomarkers to screen for prediabetes and characterize lifestyle-metabolic relationships in adults aged 45 and older with elevated BMI.

The Gist

🌟 The Big Picture: Catching the "Silent Thief" Early

Imagine your body is a car. Type 2 Diabetes is like a major engine breakdown that happens suddenly. But before the engine blows, there is a warning phase called Prediabetes. It's like the "Check Engine" light flickering on and off. Most people ignore it, or don't even know it's there, until the car actually breaks down.

The GLOW UP Study (Glucose Observation and Wearable Use for Prevention) is a team of Swiss researchers trying to build a better, smarter dashboard for your car. They want to see if your daily habits—how you sleep, move, eat, and even your heart rate—can predict that "Check Engine" light before you ever need to go to the doctor for a blood test.

🕵️‍♀️ The Mission: The "Digital Detective"

The study is looking for 200 volunteers (people over 45 with a bit of extra weight). They are splitting them into two teams:

1. The "Warning" Team: People who already have prediabetes.
2. The "Green Light" Team: People who are perfectly healthy.

The goal? To see if the researchers can tell these two groups apart just by looking at their digital footprints (data from their phones and watches) without needing a needle in their arm every day.

🛠️ The Toolkit: Your Body's "Black Box" Recorder

For four weeks, participants will wear a "digital detective kit" that records everything. Think of it like a flight recorder for your body. Here's what they are wearing:

• The Smart Ring & Watch: These are like heart detectives. They don't just count steps; they listen to your heart's rhythm (Heart Rate Variability) and check your skin temperature. They are looking for tiny clues that your body is stressed or struggling to manage sugar.
• The "Blind" Glucose Monitor: This is a small sensor stuck on your arm (like a Band-Aid). It's called "blinded" because the participant cannot see the numbers. It's like a secret spy taking photos of your blood sugar every 15 minutes, day and night, so the researchers can see the full movie of your glucose levels, not just a single snapshot.
• The Food Camera: Instead of writing down what you ate (which is boring and often forgotten), participants just take a picture of their food with their phone. It's like a visual diary of your meals.
• The Smart Scale: This measures your body composition, acting like a 3D scanner for your weight and fat.

🗓️ The Journey: A 4-Week Road Trip

The study happens in four main stops:

1. The Check-In (T0): You get your blood tested to see if you qualify. It's like getting your driver's license to join the road trip.
2. The Gear Up (T1): You get your ring, watch, sensor, and app. You learn how to snap photos of your lunch and dinner.
3. The Road Trip (4 Weeks): You live your normal life! You sleep, work, eat, and exercise. Your devices quietly record everything in the background. Halfway through, you swap out the sensor (like changing a tape in a cassette player) to keep the recording going.
4. The Debrief (T3): You come back, get your blood tested again, and hand in the gear. The researchers compare your "digital diary" with your actual blood results to see if the digital clues matched the reality.

🧠 Why This Matters: The "Crystal Ball" Effect

Usually, doctors only check your blood sugar once a year or every few years. That's like checking your car's oil once a year and hoping nothing happened in between.

This study asks: Can our phones and watches act as a crystal ball?

If the study proves that a specific pattern of "poor sleep + high stress + sugary snacks" shows up on a smartwatch before a blood test turns red, we could have a superpower. We could get a notification on our phone saying, "Hey, your data looks like prediabetes is coming. Let's fix your sleep and diet today!"

This would allow us to stop the disease before it starts, saving millions of people from a lifetime of diabetes and saving the healthcare system billions of dollars.

⚠️ The Catch (Limitations)

The researchers are honest about the limits:

• It's a short trip: Four weeks is a long time for a study, but it's a blink of an eye for a disease that develops over years.
• It's a specific group: The results might only apply to people in Switzerland who are already interested in technology.
• No "Deep Dive" yet: They aren't testing gut bacteria or deep chemical markers, just the surface-level digital clues.

🏁 The Bottom Line

The GLOW UP Study is an experiment to see if we can turn our everyday gadgets into life-saving medical tools. It's about moving from "Wait until you get sick" to "Let's fix it while you're still healthy."

If they succeed, your smartwatch might one day be the most important doctor you have.

Technical Summary

Based on the provided protocol article, here is a detailed technical summary of the GLOW UP Study.

1. Problem Statement

• The Clinical Challenge: Prediabetes is a critical precursor to Type 2 Diabetes (T2D) and cardiovascular disease, yet it remains highly underdiagnosed. In Switzerland, approximately 64.5% of prediabetes cases in adults over 60 are undiagnosed.
• Limitations of Current Screening: Standard clinical screening relies on periodic blood tests (HbA1c and Fasting Plasma Glucose [FPG]) every 3–5 years. These tests capture only static snapshots or long-term averages, potentially missing transient preclinical metabolic dysregulation and failing to identify at-risk individuals between screening intervals.
• The Gap: While digital biomarkers (wearables, smartphones) show promise in detecting metabolic issues via physiological signals (e.g., Heart Rate Variability, sleep patterns), no study in Switzerland has yet validated whether commercial-grade lifestyle data collected in a free-living (real-world) setting can reliably distinguish prediabetes from normoglycemia.

2. Methodology

The GLOW UP study is a prospective, single-center, observational case-control study conducted in Switzerland.

• Study Design & Population:
  • Sample Size: $N=200$ adults (aged $\ge$45 years, BMI $\ge$25 kg/m²).
  • Groups: 100 individuals with prediabetes (Cases) and 100 age- and sex-matched normoglycemic controls.
  • Duration: 4 weeks of continuous remote monitoring.
• Data Collection Modalities (Multimodal Approach):
  1. Clinical Ground Truth:
     • Blood Tests: HbA1c and FPG measured at baseline (T1) and follow-up (T3).
     • Anthropometrics: BMI, waist/hip circumference, and body composition (visceral fat) via smart scale and full-body photography.
  2. Continuous Glucose Monitoring (CGM):
     • Device: Abbott FreeStyle Libre Pro iQ (blinded).
     • Protocol: Worn continuously for 28 days (two 14-day cycles) on the non-dominant upper arm, capturing interstitial glucose every 15 minutes.
  3. Wearable Sensors (Commercial & Medical Grade):
     • Finger-worn: Ultrahuman Ring AIR (measures HR, HRV, sleep, body temperature, stress).
     • Wrist-worn: Ametris Leap (formerly Actigraph) smartwatch (measures physical activity, sleep, HR).
  4. Smartphone-Based Digital Phenotyping:
     • Dietary Logging: Image-based meal logging (3x daily) via a custom app (MyDataHelps) to capture nutrition composition and timing.
     • Surveys: Daily mood/stress assessments and nightly questionnaires.
• Study Timeline:
  • T0 (Enrollment): Screening, consent, and eligibility verification.
  • T1 (Baseline): Device onboarding, baseline blood draw, and body composition scan.
  • T2 (Intermediate, Day 14-21): CGM device exchange and data upload.
  • T3 (Follow-up, Day 28): Final blood draw, device offboarding, and adherence assessment.
• Statistical Analysis Plan:
  • Primary Analysis: Machine learning classification models (Logistic Regression, Random Forests, Gradient Boosting) to predict prediabetes status using wearable-derived lifestyle metrics and demographics. Performance will be evaluated via AUC, accuracy, sensitivity, and specificity.
  • Secondary Analysis: Multivariate regression and network analysis to explore associations between lifestyle factors (HRV, sleep, nutrition) and glycemic variability (CGM metrics).

3. Key Contributions

• First-of-its-Kind Dataset in Switzerland: This is the first study in the region to combine continuous glucose monitoring, commercial-grade wearables (ring/watch), smartphone-based image logging, and clinical biomarkers in a free-living setting.
• Multimodal Digital Phenotyping: It creates a rich dataset linking high-resolution physiological data (CGM, HRV) with behavioral data (diet, sleep, activity) to characterize individual and metabolic subgroup-level variability.
• Validation of Commercial Devices: By using "blinded" medical-grade CGM as the ground truth alongside commercial wearables, the study aims to validate the efficacy of consumer devices for early metabolic screening.
• Longitudinal Ground Truth: Unlike many digital health studies that rely on self-report or single-timepoint data, GLOW UP includes baseline and follow-up clinical blood tests (HbA1c/FPG) to anchor digital markers to established diagnostic standards.

4. Results

• Status: As this is a study protocol, no empirical results or findings are reported yet. The paper outlines the design, recruitment strategy, and analysis plan for a study scheduled to run in the future (preprint posted March 2026).
• Expected Outcomes: The study aims to determine if a digital biomarker model can accurately classify prediabetes and identify specific lifestyle patterns (e.g., postprandial glucose spikes linked to specific sleep or nutrition patterns) that precede clinical diagnosis.

5. Significance and Implications

• Early Detection & Prevention: If successful, the study could demonstrate a scalable, non-invasive method for screening prediabetes using ubiquitous consumer technology, potentially identifying at-risk individuals years before traditional blood tests would flag them.
• Personalized Interventions: By characterizing "digital phenotypes," the research could enable Just-In-Time Adaptive Interventions (JITAI), where AI-driven health coaching provides personalized feedback based on real-time glucose and lifestyle data.
• Scalability: Moving away from controlled clinical settings to free-living environments makes the screening process more accessible and less burdensome for the general population.
• Future Research: The dataset will serve as a foundation for developing time-series foundation models for metabolic health and could evolve into a prospective cohort study to track progression to T2D.

Limitations Noted in Protocol:

• The 4-week duration is short relative to the 3-month biological window of HbA1c.
• Lack of direct measures for insulin resistance, lipid metabolism, or gut microbiome limits mechanistic insights.
• Potential selection bias toward individuals already interested in digital health tools.