Medtronic’s Pulse Oximeter Fix Began in the Cohort

A nurse reads 95 percent off the monitor while the patient in the bed is quietly losing oxygen. The pulse oximeter clipped to the finger is doing exactly what it was built to do. The catch is what it was built against. For years the reference data behind these sensors leaned on lighter-skinned volunteers, so the device learned to read one kind of body well and other bodies less well. This month Medtronic cleared a version meant to narrow that gap. It turns on a decision most hardware founders never write down: the population you validate on is a product call, and the product owner owns it.

What Medtronic actually cleared

Medtronic disclosed on June 9 that the FDA cleared its Nellcor pulse oximetry system with a new component it calls the Nell-EQ processor, through the 510(k) pathway. The processor uses patient- and sensor-specific signal characteristics to turn the raw optical reading into a blood-oxygen number more consistently across different patients, skin tones, and care settings. This was not a rush job. The agency had already granted the device its Safer Technologies Program designation in 2025, and in January Medtronic published a verification study it says covered a full range of skin tones, reporting accuracy that beat the threshold in the FDA’s draft guidance for pulse oximeters.

The gap lived upstream

None of this was a physics surprise. A pulse oximeter infers oxygen from how light passes through tissue, and skin pigment changes that absorption. The FDA wrote its draft guidance after years of concern that the readings run high more often in patients with darker skin, which can hide a drop steep enough that a clinician would otherwise act on it. That failure mode was knowable from the requirement, not discovered in the field. When your validation cohort skews toward one population, your real intended-use claim quietly narrows, even while the label still says patients.

Deciding who the device is for, and who you prove it on, is product strategy, the same kind of call as how you frame the indication itself. I tell founders not to hand that one to the regulatory team. The regulatory team executes against the framing. Choosing the body the product has to hold up for belongs to whoever owns the product, and getting it wrong does not show up as a regulatory problem until much later, when it is also a clinical one.

What skipping it costs

For a founder, the validation cohort is cheap to scope at the protocol stage and brutal to fix after launch. Medtronic can absorb this. It is a large incumbent retrofitting a processor and re-running studies on a product it has sold for decades. A seed-stage company measuring something on a human body does not get that runway. If your device reads a signal off a patient and makes a call, whether it is a wearable, a diagnostic, an imaging system, a monitor, or software that counts as a medical device, the body you test on is part of the spec, not a recruiting detail you settle later.

The commercial side is moving the same way. Danaher agreed in February to buy Masimo, the other major name in this market, for $9.9 billion. Accuracy across every patient is turning into a place companies compete, not only a box they check for the FDA.

Dave Saunders is the founder of Base Reality Group and a Fractional CPO for hard-tech founders. He was a founder and operator at Galen Robotics, where the surgical-robotics platform earned FDA De Novo authorization in 2023, and he managed a 35-patent portfolio licensed from Johns Hopkins. He wrote Founders Who Finish and publishes The Build. More about Dave →