Wearable Medical Device Development

The short answer. We design wearable medical devices: wrist monitors, skin-worn patches, knee and gait monitors, stimulators and controllers for implants. On a wearable the sensor is the cheap part. The engineering goes into a clean signal from a moving body, a measurement schedule the battery can survive, and a device patients keep wearing. In mid-2026 about 30% of our ongoing projects were wearables, and one of our wearable programs, a single-use glucose monitoring patch, reached clinical trials.

What we build

Projects from our portfolio

Engineering decisions that make or break a wearable

So first one like signal not sensor. So the sensor is a cheap part. The hard part is getting a clean signal from the body.

Build the interference matrix before the block diagram

On our wrist monitor the cuff pump injected vibration into the optical path and noise into an ECG front end that resolves microvolts, so ECG and SpO2 could not be sampled during inflation. The processor with Wi-Fi active heated the enclosure enough to bias the skin temperature reading, and the 5 GHz radio had to stay silent during ECG capture. Five concurrent measurements became a schedule. Listing every sensor pair and what one does to the other is cheap at the architecture stage and expensive if it is discovered at design for manufacturing.

The measurement schedule sets the battery and the size

Measurement cadence sets the duty cycle, the duty cycle sets the power budget, the power budget sets the cell, and the cell sets the enclosure. A requirement that reads “smartwatch size, clinical blood pressure, one week of battery” cannot be met with a cuff today, and it is cheaper to say so in discovery than at the second enclosure revision.

Microcontroller or embedded Linux

Most wearables run a low-power microcontroller. Our wrist monitor ran Yocto Linux on an i.MX8M with containers, because the client needed server-side data exchange, remote configuration in the field and measurement applications updated separately from the firmware. That choice costs runtime, thermal headroom and boot time, so we decide it against the integration requirement and write the runtime cost down in hours.

Claim what the sensors can support

Inertial sensors measure knee flexion with an error of 4.4 to 9.4 degrees against optical motion capture, and the gait deficits after knee replacement are 4.8 to 6.6 degrees. That gap is the lesson of our knee monitor program, where we designed the PCB and embedded software: a knee monitor should claim change within one patient over weeks rather than an absolute angle, keep the flexion estimate independent of the magnetometer, which in arthroplasty patients sits above a metal implant, and check how the device is put on at the start of each session.

Close fast loops locally, not over the radio

Bluetooth core specifications up to version 6.1 set the minimum connection interval at 7.5 milliseconds; version 6.2 lowers it to 375 microseconds. For a neural implant controller with a millisecond loop we kept the motion sensor wired to the hub and sent stimulation patterns over the radio, so the loop closed at the implant against a synchronized clock.

Skin, wear time and the consumable

Skin contact is a design input, not a test at the end. On the glucose patch the second prototype phase covered the applicator, tissue and model testing and small trials to find a design that does not irritate the skin over long wear. On the stimulator the disposable electrode set both the load on the output stage and the economics: our sourcing put a pre-gelled pad at $0.18 to $0.45 at 100,000 units, with the hydrogel the largest single line at $0.05 to $0.14.

We think that the next shape is a patch. So think of something like a band-aid like thin sticker you put right on the spot you care about

Published benchmarks we design against

Parameter	Published value	Where it comes from
Wrist blood pressure monitor with a cuff	115 g, 48 mm case, 14 mm thick	Omron HeartGuide instruction manual
Smartwatch without a cuff	29 to 42 g, 9.7 mm thick	Apple Watch Series 10 specifications
Cuffless blood pressure bracelet runtime	6 to 15 days	Aktiia battery documentation
Blood pressure accuracy to pass ISO 81060-2	Mean error within 5 mmHg, standard deviation 8 mmHg or less	ISO 81060-2:2018
Knee flexion error of body-worn inertial sensors	4.4 to 6.0 degrees slow, 5.4 to 9.4 degrees fast movements	Validation against optical motion capture, 2026
Gait deficit more than a year after knee replacement	4.8 to 6.6 degrees	Systematic review, 2024
Minimum BLE connection interval	7.5 ms up to Core 6.1, 375 microseconds in Core 6.2	Bluetooth core specifications
Disposable stimulation electrode at 100,000 units	$0.18 to $0.45 per pad, hydrogel $0.05 to $0.14	OVA Solutions sourcing data

FDA and Medicare rules that shape a wearable in 2026

Where a wellness wearable ends and a medical device begins

FDA reissued its general wellness guidance on January 6, 2026. It now says some non-invasive wearables that estimate blood pressure, oxygen saturation, blood glucose or heart rate variability can be general wellness products, if they are not implanted, not intended for diagnosis or treatment, do not prompt clinical action and do not show values that mimic clinical ones unless those values are validated. Products for screening, diagnosis, monitoring, alerting or management of a disease stay regulated, and FDA names blood pressure monitors, continuous glucose monitors and ECG devices among them. A microneedle glucose wearable is not low risk, because it penetrates the skin.

The practical line runs through the labeling, the user interface and the alerts as much as through the hardware. Words such as “clinical grade”, disease thresholds or alerts that require action move a product into device territory.

Standards and guidance for connected home-use wearables

• Wireless: FDA’s radio frequency wireless guidance is still the August 2013 final; EMC follows IEC 60601-1-2 Edition 4.1 and FDA’s June 2022 EMC guidance.
• Home use: IEC 60601-1-11 Edition 2.1 and FDA’s 2014 guidance on design considerations for devices intended for home use.
• Cybersecurity: a wearable with Bluetooth, Wi-Fi, cellular or USB and software is likely a cyber device under Section 524B and needs a vulnerability plan, update processes and an SBOM. See FDA cybersecurity requirements.
• Skin contact: ISO 10993-1:2025 is the new edition; FDA recognized it in part in May 2026 and accepts the 2018 edition until July 1, 2029.
• Batteries: IEC 62133-2 Edition 1.1 for lithium-ion cells.
• Pulse oximetry: FDA’s January 2025 draft guidance ‘Pulse Oximeters for Medical Purposes’ (performance testing and labeling) is pending; finalizing it is on FDA’s FY2027 guidance list.

How many wearable-type devices FDA cleared in 2025

Product code	Device type	510(k)s cleared in 2025 (2024)
DXN	Non-invasive blood pressure system	28 (17)
DQA	Oximeter	15 (14)
MHX	Physiological monitor with arrhythmia detection or alarms	8 (5)
DPS	Electrocardiograph	6 (8)
FLL	Continuous measurement thermometer	6 (13)
DSH	Ambulatory ECG recorder (Holter)	4 (2)
MSX	Network and communication system for physiological monitors	4 (4)
DRG	Radiofrequency physiological signal transmitter and receiver	3 (3)
QBJ	Integrated continuous glucose monitor	2 (3)

What Medicare pays for remote monitoring in 2026

Since January 1, 2026 Medicare pays for remote physiologic monitoring when a device sends data on as few as 2 days in a 30-day period, and the set-up code needs 2 days instead of 16. For a device team that changes the requirement: the device has to log and transmit reliably on the days it is worn, and adherence features matter more than continuous streaming.

Code	What it pays for	2026 national amount
99453	Remote physiologic monitoring set-up and patient education (now needs 2 days of data, not 16)	$21.71
99454	Device supply with data on 16 to 30 days in 30	$52.11
99445	Device supply with data on 2 to 15 days in 30 (new in 2026)	$52.11
99457	Treatment management, first 20 minutes in a month	$51.77
99458	Each additional 20 minutes	$41.42
99470	Treatment management, first 10 minutes (new in 2026)	$26.05
98977 / 98985	Remote therapeutic monitoring, musculoskeletal device supply, 16 to 30 days / 2 to 15 days	$51.44

How a wearable program runs: a real example

Phases of the glucose monitoring patch we developed for a pharma company, as Lisa Voronkova described them on the Global Medical Device Podcast:

1. Discovery, proof of concept and design

   Requirements, proof of the core function and the design that led to the first prototype.

2. Prototype 1

   Electronics, firmware and enclosure; a batch of 100 units; bench testing and accuracy work.

3. Prototype 2

   Optimization, the applicator, tissue and model testing, and small trials to find a design that does not irritate the skin during long wear.

4. Prototype 3

   The clinical-ready device that goes into the study.

5. Clinical phase about 2,000 hours

   About 30 devices; we covered the software, the protocol, support during the study and analysis of the results.

6. Prototype 4

   Updated on the clinical findings. Total for eight phases: 11,600 engineering hours and one and a half years.

7. Next: design for manufacturing 5,000 to 7,000 hours

   Estimated separately and priced after the clinical results, together with certification and manufacturing.

Questions

Related pages

• Wrist monitor caseFive vitals, one wrist
• Glucose patch caseEight phases, 11,600 hours
• Stimulator caseElectrode as the product
• FDA cybersecurity, Section 524BConnected devices need an SBOM and a patch plan
• Orthopedic devicesRehab aids and knee monitors
• What development costs$125 per hour, FDA fees

Sources

Updated on October 9, 2026.