RCTWikiRun RecoveryMagnesiumHigh evidence score

Wearables in Medicine

Read full paper →
Authors
Ali K. Yetisen, J. L. Martínez-Hurtado, Barış Ünal, Ali Khademhosseini, Haider Butt
Journal
Advanced Materials
Year
2018
Citations
542

TL;DR

This 2018 review article surveys the landscape of wearable medical devices—from smartwatches to electronic tattoos—and finds that while the technology for continuous physiological monitoring (heart rate, glucose, temperature, electrophysiological signals) exists and is rapidly improving, very few devices have been validated in randomized controlled trials, meaning their real-world clinical impact remains largely unproven for self-experimenters.

What they tested

This is a comprehensive review article, not a single experiment. The authors tested nothing directly. Instead, they systematically reviewed and synthesised the existing literature on:

**Interventions:** Wearable medical devices including smartwatches, wristbands, electronic/optical tattoos, head-mounted displays, subcutaneous sensors, electronic footwear, and electronic textiles. These devices measure electrophysiological signals (ECG, EEG, EMG), biochemical signals (glucose, lactate, pH, electrolytes), physical parameters (temperature, strain, pressure, motion), and deliver drugs via closed-loop systems.

**Comparators:** Traditional clinical measurement tools (e.g., fingerstick blood glucose, hospital ECG machines, thermometers) and non-wearable diagnostic methods.

**Outcome measures:** Accuracy of sensor readings compared to gold-standard clinical measurements; device durability (bending stiffness, effective modulus, cycles of stretching before failure); ability to detect specific physiological events (e.g., QRS waves in ECG, glucose fluctuations); and potential for closed-loop theranostic systems (sensor + drug delivery).

Who was studied

No human subjects were studied directly in this review. The authors analysed published studies on wearable devices, which collectively included:

Healthy volunteers (sample sizes ranging from 5 to 50 per study, typically aged 18–45)

Patients with specific conditions (diabetes, cardiovascular disease, Parkinson's disease, skin lesions) in clinical validation studies

No single pooled sample size—this is a narrative review of ~200+ referenced papers

How they measured it

The review does not report a single measurement protocol. Instead, it describes how individual studies measured device performance:

**Electrophysiological signals:** ECG (electrocardiogram) via electronic tattoos and patches, measuring depolarization of cardiac waves and QRS complexes; EMG (electromyogram) on the throat for vocalization recognition; EEG (electroencephalogram) via head-mounted sensors

**Biochemical signals:** Electrochemical biosensors measuring glucose (amperometric, range 0–30 mM, accuracy compared to fingerstick), lactate (0–20 mM), pH (range 4–9), and electrolytes (sodium, potassium) in sweat, tears, interstitial fluid

**Physical parameters:** Temperature sensors (thermistors, range 32–42°C, accuracy ±0.1°C); strain gauges and pressure sensors (measuring skin modulus, range 2–80 kPa for dermis); inertial measurement units (accelerometers, gyroscopes) for motion tracking

**Device mechanics:** Young's modulus (kPa), bending stiffness (nN m), effective modulus calculations, cycles of stretching before delamination (tested up to 1000 cycles)

**Validation:** Sensor readings compared to gold-standard clinical instruments (e.g., hospital ECG, blood glucose meters, thermocouples)

Methodology

**Study design:** This is a narrative review article published in *Advanced Materials* in 2018. It is not a systematic review, meta-analysis, or randomized controlled trial. The authors conducted a literature survey of wearable medical technologies, covering materials science, sensor design, fabrication methods, and clinical applications.

**No randomisation, blinding, or washout periods:** As a review, there is no experimental design to critique. The authors selected papers based on their relevance to wearable technology in medicine, not through a systematic search strategy with pre-registered criteria.

**Duration:** Not applicable—the review covers technologies developed over approximately 10–15 years (roughly 2003–2018), with no single study duration reported.

**Statistical approach:** No meta-analysis was performed. The authors report individual study results descriptively (e.g., "measurements of localized skin properties over lesion sites showed higher modulus ≈6 mPa than normal skin regions ≈5 mPa"). No pooled effect sizes, confidence intervals, or p-values are provided across studies.

**What this design can and cannot prove:**

**Can prove:** That wearable technologies exist, that they can measure specific physiological parameters in controlled laboratory settings, and that materials science has advanced to make flexible, conformable devices possible.

**Cannot prove:** That any wearable device improves clinical outcomes, reduces healthcare costs, or changes patient behaviour in real-world settings. The authors explicitly state: "The development of patient-oriented wearable technologies and their incorporation in randomized clinical trials will facilitate the design of safe and effective approaches"—acknowledging that such trials have not yet been done.

**Major methodological weaknesses:**

No systematic search strategy or inclusion/exclusion criteria

No assessment of study quality or risk of bias

Narrative synthesis prone to selection bias (authors may favour studies that support their narrative)

Industry funding not systematically reported (many cited devices were developed by companies or academic labs with commercial interests)

No quantitative synthesis of accuracy across studies

Publication bias likely (positive results more likely to be published and cited)

Key findings

**Materials and mechanics:**

Electronic tattoos can be fabricated on gas-permeable silicone sheets with Young's modulus ≈60 kPa (similar to human skin, which ranges 140–600 kPa for epidermis)

Devices achieve bending stiffness of 1–10 nN m, allowing conformal contact with skin

Effective moduli of less than 150 kPa allow devices to stretch with skin without delamination

Piezoelectric tattoos measured skin modulus: lesion sites ≈6 mPa vs. normal skin ≈5 mPa (a 20% increase in stiffness)

Devices survived >1000 cycles of stretching without delamination

**Sensor types and capabilities:**

ECG via electronic tattoos can distinguish phases of heartbeats (depolarization, QRS waves)

EMG on the throat can recognize vocalization of different words

Electrochemical biosensors in sweat can measure glucose (range 0–30 mM), lactate (0–20 mM), pH (4–9), and electrolytes

Contact lenses with biosensors can measure glucose in tears (range 0–50 mM, with accuracy comparable to fingerstick)

Temperature sensors achieve accuracy of ±0.1°C over range 32–42°C

**Drug delivery and closed-loop systems:**

Stimuli-responsive materials (hydrogels, microneedles) can release drugs in response to pH, temperature, glucose concentration, or enzymatic activity

Closed-loop theranostic systems combine sensors with drug delivery—e.g., a glucose sensor triggering insulin release

Microneedle patches (100–1000 µm depth) can deliver drugs through the stratum corneum without reaching pain receptors

**Consumer trends (as of 2018):**

Wearable device market projected to reach $34 billion by 2020

Smartwatches and wristbands dominated the market (Fitbit, Apple Watch, Samsung Gear)

Hearing aids were the most established medical wearable (global market $6.7 billion in 2016)

Electronic textiles and smart clothing were emerging but not yet commercially mature

**Data transmission:**

Body area networks can transmit data via Bluetooth, Wi-Fi, LTE, 3G, 4G, or 5G

Internet of Things (IoT) architecture allows data to be sent to healthcare providers or emergency services

No data on transmission reliability, latency, or data loss rates in real-world settings

Effect magnitude

Since this is a review, effect sizes are reported from individual studies rather than pooled:

**Skin modulus difference:** Lesion sites showed ~20% higher stiffness (6 mPa vs. 5 mPa) compared to normal skin—a small but measurable difference that could help dermatologists distinguish benign from malignant lesions

**ECG detection:** Electronic tattoos could detect QRS complexes with sufficient fidelity to distinguish individual heartbeats—comparable to clinical ECG electrodes but without the need for gel or adhesive

**Glucose sensing:** Contact lens sensors achieved accuracy "comparable to fingerstick" in laboratory settings—but no specific correlation coefficient or MARD (mean absolute relative difference) was reported

**Temperature accuracy:** ±0.1°C—sufficient for detecting fever (typically >38°C) but may miss subtle circadian temperature variations of 0.3–0.5°C

**Device durability:** >1000 stretching cycles without failure—equivalent to roughly 3 years of daily wear if the device is stretched 1–2 times per day (e.g., putting on/taking off a patch)

Limitations

**What the authors acknowledge:**

"The development of patient-oriented wearable technologies and their incorporation in randomized clinical trials will facilitate the design of safe and effective approaches"—implying current evidence is insufficient

Data safety and privacy concerns are mentioned but not quantified

Energy management (battery life) remains a challenge for continuous monitoring

Device conformity to skin varies with body location and individual anatomy

**What a critical reader would note:**

**No clinical outcome data:** The review focuses on technical performance (can the sensor measure X?) rather than clinical impact (does using the sensor improve health outcomes?). No study cited shows that wearing a device reduces mortality, hospitalizations, or disease progression.

**Publication date:** Published in 2018—the wearable technology landscape has changed dramatically since then. Devices like continuous glucose monitors (CGMs) and smartwatch ECG (Apple Watch Series 4, 2018) were just emerging. The review does not cover modern AI-powered analytics or large-scale validation studies.

**Selection bias:** The authors are materials scientists and engineers, not clinicians. The review emphasises device fabrication and materials properties over clinical validation. Studies showing poor sensor accuracy or device failure may be underrepresented.

**No systematic search:** Without a pre-registered search strategy, the review may miss important negative studies or studies from non-English journals.

**Industry ties:** Multiple authors have affiliations with companies (Triton Systems Inc., TUM Incubator) or have filed patents on wearable technologies. Conflicts of interest are not declared.

**Sample sizes in cited studies:** Individual studies cited typically had 5–50 participants—far too small to generalise to diverse populations (different skin types, ages, health conditions, activity levels).

**Laboratory vs. real-world:** Most sensor validation was done in controlled laboratory settings. Sweat composition, skin temperature, and motion artefacts in real-world conditions may degrade accuracy significantly.

**No head-to-head comparisons:** The review does not compare different wearable brands or form factors (e.g., wristband vs. patch vs. tattoo) for the same measurement.

**Cost and accessibility:** No discussion of device cost, insurance coverage, or whether patients can afford or will use these devices long-term.

Practical takeaways

For someone running their own n=1 experiment:

### What to test

**Heart rate variability (HRV) and resting heart rate** using a wrist-worn optical sensor (photoplethysmography, PPG) or chest strap ECG. Compare accuracy of wrist vs. chest strap vs. manual pulse check.

**Sleep stages and duration** using a wrist-worn accelerometer and heart rate sensor. Compare to a sleep diary or (if available) a research-grade actigraphy device.

**Continuous glucose monitoring (CGM)** using a subcutaneous sensor (e.g., Freestyle Libre, Dexcom G6). Test how different meals, exercise, stress, or sleep affect glucose levels.

**Skin temperature** using a wearable patch or smart ring. Test how temperature varies with menstrual cycle, illness, exercise, or time of day.

**Step count and activity intensity** using a wristband or smartphone. Compare different devices (e.g., Fitbit vs. Apple Watch vs. phone in pocket) for the same activity.

### Minimum meaningful duration

**Heart rate/HRV:** At least 7–14 days to establish baseline, then 7–14 days per intervention. HRV has high day-to-day variability (coefficient of variation 20–40%), so longer baselines improve reliability.

**Sleep:** At least 14 days baseline, then 14 days per intervention. Sleep patterns vary by day of week (work vs. weekend), so include both.

**CGM:** 7–14 days per sensor (sensor lifespan). Run at least 2–3 sensors to account for sensor-to-sensor variability.

**Skin temperature:** At least one full menstrual cycle (28 days) for women; 14–21 days for men. Temperature varies 0.3–0.5°C across the day, so measure at the same time each day.

**Step count:** At least 7 days per condition. Include both weekdays and weekends.

### What to measure (specific metrics)

**Heart rate:** Resting heart rate (upon waking, before getting out of bed), average daily heart rate, maximum heart rate during exercise, heart rate recovery (drop in HR in 1–2 minutes after exercise)

**HRV:** RMSSD (root mean square of successive differences, measured in ms), SDNN (standard deviation of NN intervals, ms). Measure upon waking, before eating, or at the same time each day. Use a dedicated app (e.g., HRV4Training, Elite HRV) for consistency.

**Sleep:** Total sleep time (minutes), sleep onset latency (minutes to fall asleep), wake after sleep onset (WASO, minutes), sleep efficiency (time asleep / time in bed × 100%). Compare device-reported sleep stages (light, deep, REM) to your subjective feeling upon waking.

**Glucose:** Fasting glucose (mmol/L or mg/dL), postprandial glucose (1–2 hours after meals), time in range (TIR, % of time glucose is 3.9–10.0 mmol/L or 70–180 mg/dL), glucose variability (coefficient of variation, %)

**Temperature:** Baseline temperature (upon waking, before movement), temperature change after exercise or meals, temperature during illness

**Activity:** Steps per day, active minutes per day (minutes of moderate-to-vigorous activity), sedentary time (hours per day)

### Key confounds to control for

**Time of day:** Measure at the same time each day. HRV is highest upon waking and decreases through the day. Temperature follows a circadian rhythm (lowest ~4 AM, highest ~6 PM).

**Caffeine, alcohol, and drugs:** These affect HRV (caffeine decreases HRV, alcohol decreases HRV), sleep (alcohol reduces REM), and glucose (caffeine increases glucose). Log all intake.

**Exercise:** Acute exercise increases HRV for 24–48 hours (if moderate) or decreases it (if intense). Log exercise type, duration, and intensity (RPE or heart rate).

**Meal timing and composition:** Glucose and HRV are affected by what and when you eat. Log meal times and macronutrient composition.

**Stress and mood:** Psychological stress decreases HRV and can raise glucose. Log daily stress on a 1–10 scale or use a validated questionnaire (e.g., Perceived Stress Scale).

**Menstrual cycle (for women):** HRV, temperature, and glucose vary across the cycle. Track cycle phase (follicular, ovulatory, luteal) using calendar or ovulation tests.

**Device placement:** Wrist-worn optical sensors are less accurate during movement (exercise) and in people with darker skin or tattoos. Chest straps are more accurate for HRV. Keep device placement consistent.

**Device battery and data syncing:** Low battery can cause data gaps. Sync devices daily and check for missing data.

### What a positive result would look like

**Heart rate:** A consistent decrease in resting heart rate of 3–5 bpm after an intervention (e.g., meditation, increased cardio, better sleep) suggests improved cardiovascular fitness or reduced stress. A change of <2 bpm is likely noise.

**HRV:** An increase in RMSSD of 10–20% from baseline (e.g., from 40 ms to 48 ms) suggests improved parasympathetic tone (relaxation, recovery). A decrease of similar magnitude suggests stress, overtraining, or illness. Changes <5% are within day-to-day variability.

**Sleep:** An increase in total sleep time of 30–60 minutes, a decrease in sleep onset latency of 10–15 minutes, or an increase

Test it on yourself

Run a structured run recovery experiment

The research gives you a prior. Your own data tells you what actually works for you.

Wearables in Medicine | Steady Practice | SteadyPractice