Aletta’s 94.5% blood-draw rate starts after the robot accepts a vein

Clinical denominator audit · Health

The robot’s safest answer can be no

Aletta’s 94.5% first-stick result is real, but it begins only after the system has found a suitable vein. The rejected arms belong to a different operational question.

1. Two denominators answer two questionsAmong 1,633 people the robot attempted, 1,543 had a first-stick draw. Across all 1,743 people screened, the same numerator produces 88.5%.
2. Refusal is a safety functionThe device uses near-infrared imaging, ultrasound and Doppler to locate a vein and distinguish it from an artery. It does not puncture when it cannot identify a suitable target.
3. Autonomy moves the human boundaryOne trained phlebotomy supervisor can oversee up to three devices, but still initiates the session, verifies tube order and fill, and handles exceptions.

The United States Food and Drug Administration has authorised Aletta as the first standalone robotic device able to draw blood without hands-on operator intervention. That is a consequential milestone. The machine can place a tourniquet, prepare the skin, identify a vein, insert a needle, change and invert tubes, remove the needle, dispose of it and apply a bandage. Calling this merely “a robot arm” understates how much of routine phlebotomy has been assembled into one regulated system.

The more revealing story, however, sits inside the performance headline. The multicentre trial reported a 94.5% first-stick success rate. That figure describes what happened after Aletta found a suitable vein and proceeded to puncture. It does not describe every person who arrived at the device. Once the people the robot declined are restored to the denominator, overall collection success was 88.5%.

The headline rate excludes the arms the robot declined

The A.D.O.P.T. multicentre trial in Clinical Chemistry screened 1,743 participants in its large performance cohort. Aletta did not find a suitable vein in 110 people, or about 6%. It attempted a puncture in the remaining 1,633 and completed 1,543 first-stick draws. The published 94.5% rate is therefore 1,543 divided by 1,633. The 88.5% overall rate is 1,543 divided by all 1,743 people screened.

Same clinical outcome1,543 first-stick draws

The numerator does not change. The question changes when the denominator changes.

÷ 1,633 attempted arms94.5%How often did the first stick work after the robot accepted a vein?

÷ 1,743 screened people88.5%How often did the service produce a first-stick draw for everyone who arrived?

Neither number is fraudulent and neither should replace the other. The conditional rate is the better measure of puncture performance. The population rate is the better measure for capacity planning, patient routing and the promise a collection centre can make. A laboratory buying automation needs both because an arm rejected by the robot still creates work: the person needs a manual pathway, a different site or another clinical decision.

A safe refusal belongs inside the automation claim

Aletta combines near-infrared imaging with ultrasound to locate candidate veins, then uses Doppler ultrasound to confirm that the selected vessel is a vein rather than an artery. The FDA’s authorisation notice says the system refuses to proceed when it cannot find a suitable vein. That refusal is not an embarrassing gap around the autonomous procedure. It is one of the procedure’s safety outputs.

This matters because conventional automation rhetoric treats every non-completion as failure. Clinical automation has a more demanding objective. It must complete routine cases consistently while recognizing when its own evidence is insufficient. A machine that always attempts a puncture can produce a superficially impressive throughput figure while converting uncertainty into patient risk. Aletta’s selection stage makes uncertainty visible before the needle moves.

The distinction also prevents a common policy error. “Autonomous” describes how the device performs an accepted routine draw. It does not mean every body is eligible for the routine pathway, every sample is analytically interchangeable or every visit finishes without human judgement.

One supervisor can oversee three devices because exceptions remain human work

According to the FDA authorisation announcement, a trained phlebotomy supervisor can oversee up to three Aletta devices. A patient or laboratory manager initiates a session. The device then performs the physical sequence without hands-on intervention, while sensors can stop the process if the patient moves and can alert the supervisor to other problems.

Inside the accepted routine

Aletta performs the physical draw

Tourniquet · skin preparation · multimodal vein selection · needle insertion · tube changes and inversion · needle disposal · bandage

Across the service

The supervisor owns context and exceptions

Initiate the session · confirm tube order · verify adequate fill · respond to alerts or detachment · route refusals · preserve transport and laboratory handling

One-to-three supervision is a staffing model, not zero-human phlebotomy.

The labour moves from continuous manual execution toward oversight, preparation and exception recovery.

The human boundary remains concrete. The supervisor verifies the requested order of draw and whether tubes are adequately filled. People remain responsible for transporting samples and for the pre-analytical steps outside the device. A refusal, patient movement, sensor alert or detached needle returns the case to trained judgement. The device therefore changes the shape of labour more convincingly than it eliminates labour.

A successful stick is not the same as a completed laboratory service

The trial’s large cohort found a median procedure time of 2 minutes 37 seconds and mild device-related adverse events in 0.6% of participants, with no moderate or serious events reported. First-stick success remained above 92% in subgroups described as having difficult venous access, obesity or age 65 and older. These are encouraging results for the physical collection stage.

They do not collapse the rest of phlebotomy into one percentage. The study’s analytical-equivalence cohort evaluated selected laboratory tests, not every analyte and every pre-analytical condition. Potassium, a test especially sensitive to haemolysis, was not included in the equivalence panel. Haemolysis rates were low overall but varied by site, a reminder that local handling can affect quality after a technically successful puncture.

Question	Best evidence in the current study	What remains outside the claim
Can the robot place a first successful stick after accepting a vein?	94.5% in 1,633 attempted participants	People with no suitable vein and later manual routing
Can the service complete a first-stick draw for everyone screened?	88.5% in 1,743 screened participants	Alternative collection after refusal or failed first stick
Are samples analytically suitable?	Equivalence for selected tests in a separate 153-person cohort	Every analyte, every site workflow and all transport conditions
Do patients accept the experience?	Survey respondents generally reported similar or less pain and no aversion to future use	Non-responders and preference after routine US deployment

The trial supports deployment questions more than replacement claims

The performance cohort was single-arm. Its first-stick rate was compared with published manual-phlebotomy benchmarks rather than with a simultaneous manual draw in the same participants. The patient-experience survey covered a subset of the cohort, so its 90% “less or similar pain” result and 82% preference-or-no-preference result may not represent every screened person. The study sites were in the Netherlands, while the FDA authorisation applies to adult outpatient use in the United States under its specified supervision model.

Those boundaries do not make the device unimportant. They make the next questions better. A prospective deployment should publish the suitable-vein rate, first-stick rate among attempts, overall completion rate, manual-rescue rate, sample-rejection rate, procedure time and supervisor interventions together. Reporting only the most flattering percentage would hide where the service actually transfers work.

This is a different question from the broad promise of smart medical hardware. Aletta is already a regulated workflow with a specific refusal state, measured patient funnel and named supervisor role. It also differs from a continuous sensor such as Libre Duo, which changes when information becomes visible but does not perform the intervention. Here, the evidence problem is not whether the machine produces a signal. It is which patients enter the autonomous action at all.

Questions about Aletta robotic blood draws

Primary regulatory and clinical sources

Primary materials checked 1 September 2026

1. US FDA authorisation announcement — authorised outpatient use, autonomous steps, imaging, refusal behaviour and one-to-three supervision model.
2. FDA De Novo record DEN250046 — decision date, device classification and regulatory pathway.
3. Giesen et al., Clinical Chemistry — multicentre performance, safety, analytical-equivalence and patient-experience results.
4. ClinicalTrials.gov record NCT05878483 — A.D.O.P.T. study design and registration.
5. Vitestro’s official Aletta product page — device configuration, multimodal imaging and the manufacturer’s supervised-workflow description.