Intraosseous access is what a paramedic reaches for when the veins have collapsed and there are seconds left. You drill a needle into the bone, pull out the stylet that kept the tip from clogging, and run fluid straight into the marrow. For a run of BD needle sets, the stylet wouldn't come out.
On July 31 BD recalled five catalog numbers of its Intraosseous Vascular Access System Needle Sets. The affected lots shipped between September 30, 2024 and June 12, 2026. Through June, BD had logged 75 complaints and 45 serious injury reports, along with four deaths, all four during out-of-hospital cardiac arrest resuscitations where the obturator would not come out. FDA posted an early alert. MedTech Dive wrote it up on August 5.
BD gives the cause in one line. The affected lots were manufactured with out of tolerance dimensions. That reads like a factory problem, and treating it as one is how a company earns the right to repeat it.
The tolerance held everywhere except a real resuscitation
FDA is specific about when the parts lock together. The obturator can seize if the drill isn't pulled straight back during insertion, or if the obturator gets rotated on the way out. Neither of those is abuse. Both are what hands do when the patient is on the floor and the clock is the only thing in the room anybody is thinking about.
So the dimension was in spec for a clean axial withdrawal by a steady operator. That describes a bench, not the device's job.
I ran a shipping test on a surgical robot once. The company doing the test dropped the container off a raised forklift, and I'm not naming names. The robot didn't bang around inside at all. But there was enough force that one of the actuators tore itself off its mounts. I never would have thought about internal kinetic energy issues. We rebuilt the container, the roll-in and roll-out process, and the entire internal securing approach off that one drop. There's video. It was terrifying to watch.
Nothing in our analysis predicted that failure, because the analysis assumed the payload was the thing at risk. The payload was held perfectly still. That was the problem. A tolerance stack behaves the same way. Every part passes on its own, and the failure lives in what happens between two of them under conditions nobody wrote down.
Seventy-five complaints arrived before the recall did
BD's own count is 75 complaints and 45 serious injury reports through June. Those did not land in a single week. They accumulated across a shipping window of roughly twenty months, and each one arrived at a company with a functioning complaint system, because BD has one and it is audited.
The mechanism is never the interesting part. What differs between device companies is how many reports of a stylet being hard to remove it takes before somebody stops evaluating them one at a time and goes back to check whether the drawing is wrong. Every one of those complaints is individually explainable. The pattern is what carries the information, and patterns only surface if the person who owns the dimension ever sees the pile.
Part of BD's fix is a label
BD says it implemented corrective and preventive action, including product and labeling improvements, to enhance device robustness and reduce the likelihood of recurrence. The product improvement is the real work. The labeling half is worth staring at.
ISO 14971 ranks risk controls in a fixed order. Inherently safe design first. Protective measures second. Information for safety, which means labels, warnings and the instructions for use, comes last. It ranks last because a label only works if somebody reads it, understands it, and recalls it at the moment of use.
Intraosseous access exists for the moments when nobody is reading anything. A control that asks a paramedic to remember a withdrawal angle during CPR has a known failure rate, and the rate is not low.
To be fair to BD, they found the root cause and they changed the product, and a labeling update running alongside a design fix is legitimate practice. The useful question for anyone building a device in this category is about what their own file looked like before the complaints started, not about what BD did once it knew.
Dave's take
I've never seen a risk file that handles this shape well. The hazard that gets you is rarely one part failing. It's a part sitting at the edge of its tolerance meeting a user doing something reasonable under pressure, and describing it takes two entries instead of one, which is exactly why it doesn't get written. Go read yours. Every control that ends in "per the instructions for use" is a bet that somebody reads the label during the worst ten minutes of their week.
From Dave’s video library
Dave walks through what a fractional product lead does week to week, what it costs against a full-time hire, and why six months is the floor before the work produces anything real.
I’m here to help you scale.
Work With DavePrefer a smaller first step? Book a $500 one-hour working session →
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 →