Autoinjector Use Errors vs. Device Defects: Evidence from MAUDE

Autoinjector use errors vs device defects: what MAUDE and FDA sources actually show

Autoinjectors and prefilled pens are combination products designed to let untrained patients and caregivers self-administer injections. That design goal also makes them unusually exposed to use error, meaning harm that arises from how a person handles the device rather than from a manufacturing or mechanical defect. This review pulls hard data from FDA MAUDE (Manufacturer and User Facility Device Experience) and cross-references FDA human factors guidance, combination-product review documents, warning letters, and device recalls to characterize the dominant use-error modes, illustrate them with concrete MDR case examples, and show how often use error is invoked relative to confirmed device failure.

Why use errors are hard to surface in MAUDE

There is no clean "use error" checkbox in the MAUDE data model. Device problem codes describe device failures or problems, and event types default to Malfunction, Injury, Death, or Other, so a handling mistake is typically logged as a malfunction or injury and only revealed inside the free-text narrative and the manufacturer's investigation conclusion ¹⁷⁹. FDA's own guidance acknowledges the attribution problem directly: its human factors prioritization draft notes that device types of concern include those where adverse events, problem reports, or complaints are "attributed to use error or use error is the only explanation," conceding that the cause is often uncertain ¹⁸⁰. The hemodialysis quality guidance similarly observes that a large share of incidents are user error while another large segment is device malfunction, and instructs manufacturers to determine whether device design or incomplete labeling contributed, which is exactly the ambiguity that makes MAUDE coding unreliable for this question ¹⁷⁴. The practical consequence for anyone mining MAUDE: you must read narratives and manufacturer conclusions, and multiple synonym-driven keyword passes are required because the same failure is described inconsistently across reporters and firms.

How FDA defines use error versus device failure

FDA draws a bright-line definition that governs how these events should be classified. In the URRA guidance (Purpose and Content of Use-Related Risk Analyses for Drugs, Biological Products, and Combination Products), use error is defined as a user action or lack of action that differed from what the manufacturer expected, produced a different-than-expected result, was not caused solely by device failure, and did or could result in harm ²³³. The corollary matters for MAUDE triage: if a problem is caused solely by a device failure, it is not a use error ²³³. FDA expects sponsors to build a use-related risk analysis that lists every task required for use, the potential use errors and harms for each task, whether each task is critical, the user-interface risk controls, and the evaluation methods, updated across the product lifecycle including postmarket ²³⁶. The companion combination-product HFE guidance stresses that the "user interface" includes packaging, labels, carton labeling, instructions for use, and any training, and that user interactions can result in harm including medication errors ²³². Human factors validation testing exists precisely to identify use errors that may cause serious harm and to evaluate risk-control effectiveness, with any residual use-related risk analyzed to decide whether further user-interface changes are warranted ²³⁴²²⁸.

Notably, FDA's URRA appendix uses a notional autoinjector as its worked example, and the two task-level use errors it calls out map exactly onto the dominant real-world MAUDE modes below: failure to remove the cap because of excessive pull force, leading to delayed administration, and holding the activation button for less than the required time, leading to an incomplete dose and underdose ²²⁷.

Dominant use-error modes in MAUDE

The autoinjector MDR record clusters into a recognizable set of handling-failure modes. The modes below are ordered roughly by how prominently they appeared across the sampled narratives, with concrete report numbers.

1. Premature removal and insufficient hold time (incomplete dose / underdose)

This is the archetypal autoinjector use error and the one FDA's own guidance example anticipates ²²⁷. The device is pulled off the skin before the full dose is delivered, so medication is deposited on the skin or lost.

Humira Pen Starter Kit, report MW5108166: the patient held the pen to the skin, waited at least 15 seconds, then pulled it away to check it, at which point the dose released and liquid went everywhere. The report was a clear early-removal handling error, not a confirmed device defect ⁴¹. This same report also appeared in the premature-activation search as a dose-on-removal event, underscoring how one handling mistake can be coded several ways ⁴¹.
BD AutoShield Duo safety pen needle, report 9616656-2019-00631: held on the skin for 10 seconds, then removed, with medication leaking onto the skin; the manufacturer could not confirm the complaint and root cause was undetermined ⁶⁹.

FDA combination-product reviewers see the same pattern in human factors validation before approval. For BIMZELX (bimekizumab-bkzx), 6 of 75 participants removed the device early, including one who pulled away too early on a second attempt; FDA called this a known autoinjector usability issue, judged residual risk acceptable, and noted it could not be further mitigated ¹⁵³¹⁵⁵. ALHEMO (concizumab-mtci) participants removed the prefilled pen before the injection completed, creating single-underdose risk ¹⁵⁰, and ENTYVIO (vedolizumab) reviewers recommended clarifying end-of-injection cues to reduce underdose ¹⁶².

2. Failure to remove the cap and wrong-end / needle-end confusion

Cap and end-handling errors were among the most explicitly adjudicated use errors in the record.

Prevent Safety Pen Needle 31G x 6mm, report 9613304-2023-00017: nurses reported leakage during insulin administration, and the manufacturer concluded the event was solely the result of user error because the outer cap had not been removed before priming ¹⁰³.
BD AutoShield Duo, report 9616656-2020-00639: the non-patient-end shield fell off and insulin was found inside the protective cap; the row supports cap-related handling or a failed safety mechanism but does not confirm which ¹⁰¹.

Reviewer caveat: the MAUDE sample did not contain a report explicitly describing an autoinjector held fully backwards with the wrong end against the skin, so that specific "needle-end confusion" scenario is better supported by the pre-market human factors record than by these narratives ²²⁷. In human factors validation, cap and shield removal recur as critical-task failures: ADLYXIN (lixisenatide) ran a supplemental study specifically because of prior failures removing the pen cap, attaching the needle, and removing caps ¹⁴⁴; ofatumumab products (Arzerra/Kesimpta) listed "remove rigid needle shield/needle cap" and "remove cap" among critical tasks with observed errors ¹⁴³; and Plegridy Pen participants tried to pull the needle guards from the pen after removing the cap ¹⁴⁵.

3. Premature or accidental activation

Firing the device before it is correctly placed wastes the single dose, which is especially dangerous for a rescue product such as epinephrine.

Twinject, report MW1040476: accidental activation during a father's demonstration to a babysitter, characterized as user handling error with no accidental injection and no adverse event ³⁴.
The Humira Pen dose-on-removal case (MW5108166) also reads as a premature release before the device was removed correctly, with device defect not confirmed ⁴¹.

4. Needle bending or breaking from incorrect attachment or rough handling

A large, coherent cluster of pen-injector reports involved bent or broken needles that manufacturers attributed to how the user attached the needle to the pen or handled the device, rather than to a manufacturing defect.

Report 3006948883-2020-00106: the needle bent and broke off during use and lodged below the skin, requiring ultrasound and a 45-minute procedure with stitches to remove; the returned sample showed a broken patient-end cannula with signs of bending and re-straightening, and the manufacturer concluded user error with no manufacturing defect ⁵⁷.
BD AutoShield Duo, report 9616656-2019-00724: the safety shield did not fully engage and bent over; the sample showed a bent non-patient-end cannula and BD attributed the probable cause to user error when attaching the pen needle to the pen injector ⁶⁴.
Reports 9616656-2019-00287, 2243072-2020-01133, and 2243072-2020-01179: bent non-patient ends preventing insulin flow, each attributed by the manufacturer to user error in mounting the needle ⁶³⁶¹⁵⁵.
Report 9681835-2018-00020: the outer cap was forced off during rough handling, and the needle was found damaged/missing; the manufacturer found no evidence of device defect ⁵⁸.

Across this sampled set, user-error root causes were stated far more often than confirmed defects, and even a report whose product-problem field said "defective device" carried an investigation narrative concluding no manufacturing issue and citing user error ⁵⁶.

5. Accidental needlestick and self-injection into a finger or thumb

Accidental sticks while handling, recapping, or removing pen needles are common, and the classic epinephrine mechanism of injecting one's own thumb appears in the record.

In a set of finger/thumb/hand stick reports, roughly 12 of the sampled rows involved accidental sticks while handling pen needles, and most were handling, recapping, or removal accidents with no proven defect ¹²³⁹⁵¹²⁷⁸¹³²¹³⁶. Examples include a caregiver stuck while detaching a used needle with no evidence of defect ⁹⁵ and a care worker punctured while removing a used needle, where the IFU had warned about needlestick injury ⁸.
Epinephrine autoinjector, report 3002919960-2025-00014: a nurse accidentally injected the autoinjector into a thumb because the tip was not placed on the patient's outer thigh; the investigator concluded the later-observed bent needle was most likely post-use user error from improper reinsertion into the carrying case ¹⁶⁹.

Distinguishing these from defects is genuinely hard. The clearest device-side needlestick cases were safety-mechanism failures on the BD AutoShield Duo, where the protective cover failed to deploy or the shield separated and exposed the needle ¹²⁹¹³³⁹⁶, but even some of those were adjudicated toward loose-skin or minimal-force user technique rather than a confirmed defect ⁸⁴. Human factors reviews flagged needlestick and sharps-disposal errors pre-market too: BONSITY (teriparatide) participants failed to dispose of the needle, often because they did not remove it ¹⁴⁶, and an ALHEMO participant suffered a needlestick because they did not expect a double-sided needle and missed the IFU warning ¹⁵¹.

6. Wrong site, wrong angle, and technique errors

Site and angle errors were present but less frequently and less cleanly documented in MAUDE than the modes above.

Report 2243072-2020-01076 (teriparatide prefilled pen): the patient did not pinch the skin before injecting and medication came out of the injection site, explicitly labeled wrong technique; the manufacturer could not confirm a device defect ⁶.

Again the pre-market human factors record is richer here: AJOVY (fremanezumab-vfrm) had 4 participants inject into the front of the upper arm instead of appropriate sites, with FDA noting IM/IV injection and pharmacokinetic risk ¹⁴²; RIVFLOZA and SEMGLEE reviews flagged users defaulting to a 90-degree angle or not injecting to the specified depth ¹⁵⁸¹⁵⁶. Wrong-angle handling also drives the bent-needle cluster in section 4.

7. Reuse or a second attempt with a spent device

Users repeatedly tried to re-inject with an already-fired or empty device, or believed a dose had been delivered when it had not, producing underdose, missed dose, and downstream hyperglycemia or DKA.

Greenstone epinephrine auto-injector pen, report 2161650: the nurse made two injection attempts with the same auto-injector, which failed to deploy, before a second brand was used successfully ³².
Aranesp SureClick, report MW1041599: the reporter tried to activate the injector, then tried again into an empty vial, and the report notes this was the second wasted product in a month ²⁰.
HumaPen Ergo reports 1819470-2001-00004, -2002-00005, and -2001-00037: patients believed insulin had been delivered when it had not, dialed incorrectly, or did not connect the pen correctly, missing doses; several were judged user error rather than malfunction ¹⁷¹⁸¹⁹.

Use error versus pure device failure: relative frequency

Within the device-defect baseline sample of 17 autoinjector/pen reports, the split is instructive. Roughly 5 were strong, often return-confirmed device malfunctions: a jammed spring/actuation mechanism (Omnitrope Pen, 2243072-2019-01026) ⁸⁵, failure of the safety shield to activate from a bent cannula (9616656-2020-01311) ⁸⁹, shield separation exposing a clean needle (9616656-2018-00375) ⁹⁶, and broken cannulas causing failure to deliver (9616656-2020-00669 and 2243072-2021-00135) ⁸⁷⁹⁷. About 8 were explicitly framed by the manufacturer as user error or improper handling despite being filed as malfunctions ⁸⁴⁹¹⁹²⁹³⁵⁸⁹⁸. In that sample, use-error-attributed reports outnumbered confirmed device defects by roughly 2 to 1 ⁸⁵⁸⁷⁸⁹⁹⁶⁹⁷⁸⁴⁹¹⁹²⁹³⁵⁸⁹⁸.

That ratio should be read as a directional signal, not a population rate, but it is consistent across every keyword pass run for this review: when a device was actually returned and evaluated, manufacturers most often reported no manufacturing defect and cited user technique, particularly incorrect needle attachment, bending the needle, changing angle during injection, insufficient hold time, or failing to remove the cap ⁵⁵⁵⁶⁵⁷⁵⁸⁶¹⁶³⁶⁴⁶⁷. It is also consistent with the pre-market human factors record, where early removal, cap-removal failure, incomplete injection, wrong-site injection, dose-setting errors, and needlestick/disposal errors are the recurring critical-task failures across a wide range of approved products ¹⁵³¹⁵⁵¹⁴³¹⁴⁴¹⁴⁵¹⁵⁰¹⁴²¹⁶¹¹⁴⁶¹⁵¹.

A structural caveat cuts the other way, however. Because manufacturers control the complaint investigation and a returned sample is often unavailable, "root cause undetermined" and "no device anomaly found, user error" are the path-of-least-resistance conclusions, and several MAUDE narratives explicitly note the complaint could not be confirmed because nothing was returned ⁶²⁶⁹¹²³. FDA guidance tells firms to determine whether device design or incomplete labeling contributed to a user error rather than stopping at the user ¹⁷⁴, so a use-error label in MAUDE does not exonerate the design.

What pre-market human factors reviews predict about the post-market signal

The combination-product review record functions as a predictive index of the MAUDE modes. FDA DMEPA and human factors reviewers repeatedly logged, before approval, the exact errors that later populate MAUDE: premature removal and insufficient hold time ¹⁵³¹⁵⁵¹⁵⁰¹⁶²; cap and needle-shield removal failures ¹⁴⁴¹⁴³¹⁴⁵; incomplete injection and missed end-of-injection cues ¹⁴³¹⁶²; wrong site, angle, and incomplete needle insertion ¹⁴²¹⁵⁸¹⁵⁶; dose-setting and dose-omission errors, including one participant setting 24 mg instead of 10 mg from decimal misreading and the risk of omitting the second device in a two-injection 320 mg dose ¹⁶¹¹⁵²¹⁵⁴; and needlestick/sharps-disposal errors ¹⁴⁶¹⁵¹¹⁶⁰. Reviewers frequently accepted the residual risk as low while still recommending IFU and labeling clarifications, and in several cases noted the risk was not unique to the device class ¹⁴³¹⁵¹¹⁶⁰¹⁵⁵.

Enforcement and recall context

Two adjacent FDA data streams round out the picture. On enforcement, warning letters cite injection-device firms for inadequate complaint handling and MDR reportability rather than for the use errors themselves: Robbins Instruments was cited for failing to review and investigate complaints that Dermo-Jet needleless injectors failed to fire, penetrate, or inject as intended ²⁴⁸; Sol-Millennium for not evaluating confirmed complaints, including a needle embedded in deep subcutaneous tissue, for MDR reporting ²⁴⁷; and Stryer Biotech for at least 33 MDR-reportable events it failed to file or filed late ²⁴⁴. These letters matter because weak complaint handling is precisely what degrades the MAUDE narrative quality that use-error analysis depends on.

On recalls, the autoinjector actions split cleanly into device-build failures versus labeling/IFU issues that can drive use error. Owen Mumford recalled Autoject EI (Model AJ1310) and a related veterinary product for a syringe carrier missing a damper and spring and a possible assembly error (Class II) ¹⁹⁸¹⁹⁹, a genuine device malfunction. In contrast, BD recalled Ultra-Fine Pen Needles for carton labeling that omitted U.S. compatibility information, and Owen Mumford recalled Unifine Pentips Plus for mixed-up 4 mm and 8 mm needle configurations, both labeling/packaging issues that could induce use error rather than mechanical failures ²¹²²¹¹.

Practical takeaways for regulatory and post-market teams

The dominant autoinjector use-error modes, in order of prominence in this review, are premature removal/insufficient hold time causing underdose, cap-removal and end-handling errors, needle bending/breaking from incorrect attachment, accidental needlestick/self-injection, premature activation, wrong site/angle technique, and reuse of a spent device ⁴¹¹⁰³⁵⁷⁹⁵³⁴⁶³².
Use error, not confirmed device defect, is the more common adjudicated conclusion when a device is investigated, on the order of roughly 2 to 1 in the sampled defect-baseline set, though attribution is biased by unreturned samples and manufacturer-controlled investigations ⁸⁵⁸⁷⁸⁹⁹⁶⁹⁷⁸⁴⁵⁸.
MAUDE will systematically undercount use error because there is no discrete code for it; robust surveillance requires reading narratives and manufacturer conclusions and running multiple synonym-based queries ¹⁷⁹¹⁸⁰.
Pre-market human factors validation reliably predicts the post-market modes, so the URRA critical-task list is the best available map of what to monitor in MAUDE and complaints after launch ²³⁶²²⁷.
A use-error label does not close the loop: FDA expects firms to test whether device design or labeling contributed, and warning letters target the complaint-handling and MDR failures that would otherwise hide these signals ¹⁷⁴²⁴⁸²⁴⁷.

Deeper follow-ups a reader may want to pose to Rhizome directly: a device-specific MAUDE pull and time trend for a single product (for example EpiPen or a named biologic pen), a full extract of DMEPA critical-task findings for one molecule's human factors file, or a quantified event-type breakdown for a specific product code once the exact autoinjector product code is fixed.