A soft robotic device for rapid, self-guided intubation

Abstract

Endotracheal intubation is a critical medical procedure for protecting a patient’s airway. Current intubation technology requires extensive anatomical knowledge, training, technical skill, and a clear view of the glottic opening. However, all of these may be limited during emergency care for trauma and cardiac arrest outside the hospital, where first-pass failure is nearly 35%. To address this challenge, we designed a soft robotic device to autonomously guide a breathing tube into the trachea with the goal of allowing rapid, repeatable, and safe intubation without the need for extensive training, skill, anatomical knowledge, or a glottic view. During initial device testing with highly trained users in a mannequin and a cadaver, we found a 100% success rate and an average intubation duration of under 8 s. We then conducted a preliminary study comparing the device with videolaryngoscopy, in which prehospital medical providers with 5 minutes of device training intubated cadavers. When using the device, users achieved 87% first-pass and 96% overall success rate, requiring an average of 1.1 attempts and 21 s for successful intubation, significantly (P = 0.008) faster than with videolaryngoscopy. When using videolaryngoscopy, the users achieved 63% first-pass and 92% overall success rate, requiring an average of 1.6 attempts and 44 s for successful intubation. This preliminary study offers directions for future clinical studies, the next step in testing a device that could address the critical needs of emergency airway management and help democratize intubation.

One Sentence Summary:

A soft robotic device autonomously and rapidly guides intubation in models and cadavers and requires minimal skill and no visualization.

Editor’s Summary:

Airway Autopilot. Endotracheal intubation requires extensive training and clear anatomical visualization, making it challenging during emergencies where first-pass failure rates are high. Here, Haggerty et al. developed a soft robotic device that guides a breathing tube into the trachea, reducing the need for extensive specialized training. Testing of the device by prehospital medical professionals showed the soft robotic device improved success and average intubation duration compared with the gold standard videolaryngoscope. The findings suggest that this device could enable improved outcomes, broader use of intubation in emergency settings, and use by a broader segment of medical personnel. ---Molly Ogle

INTRODUCTION

Establishing an artificial airway is an integral component of care in operating rooms (ORs), emergency departments (EDs), and prehospital settings. An endotracheal tube (ETT) is usually placed through manual guidance, while lifting the tongue and epiglottis with a laryngoscope or videolaryngoscope (VL) (1) to visualize the entrance to the trachea. Manual ETT placement requires skill, training, and anatomical knowledge. This procedure is well-suited for the OR where pre-procedure airway exams and a controlled environment usually result in clear visualization (2), and caregivers are highly trained and well-practiced. Outside the OR, manual placement is less suitable because visualization can be challenging due to fluids in the airway, inadequate lighting, and non-ideal patient body position (3). Further, emergency medical personnel, especially those in the prehospital setting, may perform intubation only a few times a year, corresponding to limited psychomotor skills and anatomical knowledge (4, 5). Consequently, emergency intubation has higher first-pass failure rates of approximately 35% in the prehospital setting and 15–20% in the ED (1, 6) compared with 3–15% in the OR (2, 7, 8). The delayed oxygenation and increased trauma associated with repeated intubation attempts and failure in the critically ill are associated with increased severe adverse events ranging from hypoxia to cardiac arrest, which are in turn associated with increased mortality (9).

An alternative to the ETT is the supraglottic airway (SGA) (10), which passes air towards the lungs from outside the glottis. The SGA can be placed “blindly,” without need for visualization. Therefore, less skill and anatomical knowledge is necessary, resulting in higher success rates compared with intubation in prehospital settings. For example, one large study showed that the SGA had a 90% pre-hospital success rate versus 52% for the ETT (6). However, since the airway is not sealed as it is with the ETT, the use of the SGA is often contraindicated for patients at risk of aspiration, with intubation as the preferred alternative (11). To address this limitation, a modification of the SGA, the ‘intubating SGA’, affords the insertion of an ETT blindly through the core of the device into the glottis (12). The procedure often requires advanced manual manipulations, resulting in varied success in the OR of 50–70% for first-pass or 73–89% for overall success and requiring an average of more than 50 s to complete (13).

Considering these current technologies, there remains a need for an intubation device that, regardless of the anatomical view or provider skill, achieves a high success rate and minimizes intubation duration. Such a device could potentially benefit the millions of emergency intubations that occur each year in the US (14, 15), many of which currently result in substantial morbidity (16). Outside of the civilian sphere, such a device could benefit military far-forward medical care, as poor airway control is the second most common preventable cause of death (17, 18). Finally, half of the world’s population currently is without access to essential health services, airway management included (19). A simple-to-use, electronics-free intubation device could help democratize airway management in regions without access to highly-trained medical personnel.

To address this need, we present a soft robotic intubation device that is designed to autonomously guide the ETT into the trachea using the mechanism of tip extension. This mechanism distinguishes the device from current airway devices and is characterized by a tube that lengthens from its tip, rather than being pushed from its base. Tip extension is found in diverse organisms that navigate constrained environments, such as fungal hyphae, animal nerve cells, and plant roots (20), and has been previously explored for medical applications including endovascular surgery (21, 22), mammary gland access (23), colonoscopy (24), and laryngeal tubes (25). Tip extension can reduce contact forces between the tube and the environment, enable the tube to form into 3D shapes matching a pathway like the airway, and allow the tube to passively and autonomously “self-guide” into openings and pathways (20, 26–29).

RESULTS

Soft robotic intubation system (SRIS) working principle and design

The SRIS is made of two separable components: (i) the “introducer,” which positions the self-guided tube in the correct location for entering the trachea, and (ii) the “self-guided tube” comprised of a soft ETT and a soft everting overtube which navigates into the trachea (Fig. 1).

Fig. 1. The working principles of the Soft Robotic Intubation System (SRIS).

(A) The “self-guided” tube comprises an inflated overtube and a soft ETT, attached at two points (black dots). As the soft ETT is advanced, the overtube everts and extends from the tip, while its outer wall remains stationary (red dots connected by grey dashed line). (B) Tip extension enables following of tortuous trajectories in space without environmental interactions. (C) The self-guided tube can passively deform with minimal force to “find” openings. (D) Images of the SRIS, comprising the self-guided tube and the introducer. Two steps are required to intubate: (i) insert the introducer, and (ii) advance the self-guided tube. (E) Illustration of the process of intubating using the SRIS device. Illustrations created with Adobe Illustrator and KeyShot.

The introducer is designed to follow the contours of the pharynx during insertion, utilizing a similar shape and stiffness to an i-gel SGA (Intersurgical, Ltd.) (Fig. 1, A and B). The shape and dimensions were defined through an iterative design and testing process with dozens of cycles and over 100 cadaveric specimens. The “anatomical stop” seats in the funneling anatomy of the hypopharynx, allowing for the appropriate positioning of the port through which the self-guided tube grows. A soft, thin leading edge is designed to slip posterior to the epiglottis without causing downward folding of the epiglottis, and an epiglottic elevating bar (Fig. 1A) is designed to help lift the epiglottis as the self-guided tube emerges, for cases where the epiglottis may become partially or fully downfolded.

The self-guided tip-extending tube is made of soft, thin-walled tubing, which is formed into a closed torus and inflated by the user before insertion. This overtube is attached at two internal connection points to a custom ETT, which is designed to be much softer than a traditional ETT (Fig. 1C). As the user advances the soft ETT through the introducer, the torus rolls, everting from its tip and pulling the ETT forward, eventually delivering the ETT tip into the trachea. To successfully navigate the anatomy, the key geometric parameters of the overtube (Fig. 1D) were selected through a similar iterative design and testing process in cadaver models as was used with the introducer. This geometry, combined with the tip extending mechanism, results in four differences from a standard push-from-the-base tube. First, there is minimal relative motion of the exterior of the overtube with respect to the tissue, which potentially reduces tangential forces between the device and tissue. Second, because the inflated overtube pulls the ETT, the ETT can be made very soft without buckling like a soft push-from-the-base tube would. The softness of the inflated overtube and ETT potentially reduces normal forces between the device and tissue. Third, the overtube can be formed into a sigmoid shape to approximately match the airway anatomy, naturally following both the primary and secondary curves of the pharynx and larynx (30) as the overtube grows (Fig. 1D), which may further reduce normal forces and help the device autonomously navigate the anatomy. Fourth, the overtube can passively deform to search for the path of lowest energy (27, 28) (Fig. 1E), potentially aiding in the ability to autonomously navigate to the tracheal opening. Examples of these behaviors can be seen in movie S1.

To intubate with the SRIS, the user inserts the introducer into the mouth like a traditional SGA until it stops (Fig. 1, A and B). The user then advances the self-guided tube to place the soft ETT in the trachea. After intubation is complete, the user can inflate the distal cuff of the soft ETT to seal the trachea and begin ventilating. The self-guided can be detached from the back face of the introducer, allowing the introducer to be optionally removed without interrupting ventilation.

Soft self-guided tube reduces tool-tissue axial force during tube insertion

To evaluate the potential of the soft self-guided tube to reduce forces on tissues during intubation, we measured the tangential force applied by either a standard ETT or the self-guided tube to an anatomically-correct model glottis (31), 3D-printed in a resin that has similar durometer as normal cartilage (32) (Fig. 2A). The average axial force applied by the self-guided tube to the model tissue was significantly less than that applied by the standard ETT (1.5 ± 1.3 N vs. 10.3 ± 5.7 N, r = 0.71, 95% CI 0.38–0.88, P = 0.003, n = 5 each, Table 1, Fig. 2B). The peak force, averaged across all five trials at each device’s respective peak, was significantly less for the self-guided tube than the traditional ETT (3.7 ± 0.8 N vs. 14.7 ± 2.4 N, r = 0.83, 95% CI 0.50–0.95, P = 0.008, n = 5 each, Table 1).

Fig. 2. Benchtop force and misalignment testing of the self-guided tube.

(A) Photo of the model larynx force testing platform with a model larynx mounted to a force-sensing platform. (B) Force applied to advance the traditional ETT (grey) and the self-guided tube (blue) plotted by distance into the larynx. Data are presented as scatter plot with average (left) and box-whiskers plot with median and IQR (right). Error bars indicate the range. Analysis done with Wilcoxon Rank Sum test. **P < 0.01. n = 5 trials for each device at each test point. (C) Depiction of benchtop misalignment testing done in coronal (left) and sagittal (right) planes. (D) The success region in both planes for traditional ETT (top) and self-guided tube (bottom). Data presented as interpolated success (green) and failure (red) envelopes analyzed with the Kolmogorov–Smirnov test. (E) Depiction of cadaveric superior-inferior misalignment testing with traditional ETT (top) and self-guided ETT (bottom); red “X” symbols indicate failure while green check marks indicate success. (F) Percent success for traditional ETT (grey) and the self-guided tube (blue) vs offset from nominal SGA placement. Raw data are presented as a histogram (left) and mean ± SD (right). Analysis by Kolmogorov–Smirnov test. **P < 0.01. n = 4 trials for each device at each test point. Illustrations created with Adobe Illustrator.

Table 1:

Provider background information.

Provider Characteristic	Number of Providers
Level of experience
Paramedic	4
EMT	4
Age
Less than 25	3
25–34	3
35–44	2
Years of Experience
Less than 1	2
1–5	3
6–10	2
11–15	1
Frequency of airway management
Rarely	1
Monthly	6
Weekly	1
Intubation Experience
Zero	2
1–20	2
21–50	1
Greater than 50	3

Soft self-guided tube increases tolerance to misalignment

Next, we evaluated whether the standard ETT and the self-guided tube could enter the trachea in the 3D-printed model glottis when misaligned from the tracheal axis in either the coronal or sagittal plane (Fig 2C). The two devices performed similarly (D = 0.27, 95% CI 0–0.64, P = 0.736, Table 1). In the coronal plane, the two devices had equivalent performance when the angular offset was in the direction of the ETT bevel, whereas, when the offset was opposite the bevel, the traditional ETT failed at every increment. In the sagittal plane, the traditional ETT succeeded up to a 15-degree offset, whereas the self-guided device succeeded up to 30 degrees (Fig 2D).

We evaluated whether each tube would enter the trachea in a cadaver when offset along the superior-inferior axis from the nominal glottic opening. A Size 3 i-gel SGA (Intersurgical, Ltd.) was placed at a range of depths distal to and proximal to the cadaver’s glottic opening, and either the traditional ETT or the self-guided tube was passed through it (Fig. 2E). The self-guided tube had a significantly larger successful range of approximately 9 cm vs. 2.5 cm for the traditional ETT (D = 0.70, 95% CI 0.40–0.90, P = 0.007, n = 40 each, Fig. 2F, Table 1). This finding indicates that the device could have more tolerance to misplacement compared with traditional devices.

A device-expert user has high success rate using SRIS in mannequin and cadaver models

We next characterized the performance of the SRIS (self-guided tube plus introducer) in a mannequin and cadaver with device-expert users (greater than 100 SRIS uses). To test the repeatability of the SRIS, a single device-expert user attempted 50 intubations sequentially in a single Laerdal Airway Management Trainer mannequin with approximately one minute between tests (Fig. 3A). The mannequin was equipped with an endoscope in the trachea viewing superiorly to visualize the emergence of the overtube and soft ETT (Fig. 3B). The device-expert user had a 100% success rate (50/50 attempts) using the SRIS, with an average intubation duration of 7.4 ± 1.7 s (Fig. 3C). This contrasts with the 23–28 seconds required for physicians using VL in similar simulation environments (33).

Fig. 3. Initial testing of the SRIS in a mannequin by an expert user.

(A) External view of intubation. From left to right: introducer placed in mouth; introducer is advanced until stopped; soft ETT is advanced; intubation complete. (B) Internal tracheal view, from left to right: depiction of the location of the internal camera; introducer becomes visible as advanced into the pharynx; self-guided tube begins advancing into the trachea; self-guided tube is fully deployed with soft ETT tip exposed in trachea. (C) Per-attempt intubation duration and success across 50 trials. Data are presented as scatter plot with average (blue shaded line). n = 50 trials.

To test in a more translational model, we further characterized the intubation duration and introducer removal duration of the SRIS in a single cadaver with a different device-expert user (Fig. 4, A and B). Using the SRIS, the device-expert user had a 100% intubation success rate (10/10 attempts) with an average intubation duration of 5.9 ± 0.7 s, average removal duration of 2.3 ± 0.6 s, and average overall procedure duration of 8.3 ± 1.3 s (Fig. 4C).

Fig. 4. Preliminary validation of the SRIS in a cadaver by an expert user.

(A) View from inside the trachea looking superiorly through the vocal cords into the hypopharynx. From left to right: procedure begins; introducer rounds oropharynx (blue tips visible); self-guided tube is advanced and passes cords; tube enters trachea and contacts camera before ETT is revealed. (B) External view of intubation, from left to right: introducer rounds oropharynx; self-guided tube is advanced; introducer is separated from the self-guided tube and removed without interrupting the ventilation pathway; self-guided tube remains in trachea. (C) Per-attempt intubation duration and introducer removal time. Data are presented as a bar graph of the average duration, with each attempt (dots) overlaid to describe the range; green signifies intubation success and red signifies intubation failure (no failures recorded). n = 10 trials.

Preliminary study of the SRIS versus videolaryngoscopy in cadavers with non-expert users

The primary goal of this study was to test the viability of the presented device for real-world, non-expert users in multiple cadaveric models. A secondary goal was to do a preliminary, non-powered comparison to VL. The primary endpoints were intubation success rate (first-pass and overall), intubation duration, and number of attempts (Fig. 5, A and B). Eight emergency medical services (EMS) healthcare providers tested the presented device and a control device (VL) in eight different unembalmed adult cadavers. Four of these users were paramedics who intubate regularly with VL, and four were emergency medical technicians (EMTs) with a scope of practice that does not include endotracheal intubation (see Table 1). Two of the EMTs had no previous experience with VL. We provided each user with approximately 5-minute standardized training for both devices.

Fig. 5. Preliminary study of the SRIS in 8 cadavers by novice users after approximately 5 minutes of training.

(A) Intubation success percentage (left), duration (middle), and number of attempts up to the maximum allowed of 3 (right) for all users across all airways for both VL (grey) and the SRIS (blue). Intubation success rate presented as mean data are presented as a bar graph with error bars to signify standard deviation; intubation duration and attempts data are presented as box-and-whisker plots with median, IQR, and outliers. Analysis by Chi-Square for success percentage and Wilcoxon Rank Sum otherwise. *P < 0.05, **P < 0.01. N = 24 trials for VL and 23 trials for SRIS. (B) Intubation performance for all users across most challenging airways for both VL (grey) and the SRIS device (blue). Intubation success rate data are presented as histograms with error bars to signify standard deviation; intubation duration and attempts data are presented as box-and-whisker plots with median, IQR, and outliers. Analysis by Chi-Square for success percentage and Wilcoxon Rank Sum otherwise. *P < 0.05, **P < 0.01. N = 14 trials for VL and 14 trials for SRIS. (C) Photo of an EMT completing representative intubation procedure from t = 0s to t = 20s.

The first-pass intubation success rate for SRIS was 87% across all user attempts (n = 20 successes across 23 attempts by 8 users) versus 63% across all user attempts for VL (n = 15 successes across 24 attempts by 8 users) (P = 0.054, odds ratio (OR) 4, 95% CI 0.92–17.37, Table 2). The duration of first-pass attempts was 16 s for SRIS compared with 21 s for VL (P = 0.08, r = 0.30, 95% CI −0.03–0.63). When including all trials, the overall success rate with SRIS was not different from VL (96% vs 92%, n = 22 per group, OR 2, 95% CI 0.17–23.70, P = 0.58, Table 2); however, the time to complete intubation was significantly lower with the SRIS than with VL overall (21 s vs. 44 s, P = 0.008, r = 0.40, 95% CI 0.11–0.70, Table 2). The SRIS required, in aggregate, an average of 1.1 attempts compared with 1.6 attempts for VL (P = 0.06, n = 22 per group, r = 0.27, 95% CI −0.01–0.56) (Fig. 5A, Table 2). The two EMTs with no VL experience achieved a first-pass success rate with the SRIS of 84% and 100% overall success, as compared to with 33% and 67%, respectively, with VL. Data showing the comparative performance of the EMT and paramedic subgroups are shown in fig. S1 and table S1.

Table 2: Comparative outcomes of SRIS vs. existing devices.

Findings from each statistical test completed. Effect sizes for success rates are Odds Ratios (OR), and correspond to a 95% confidence interval (CI). Effect sizes for force and duration is the correlation coefficient r, and correspond to a 95% confidence interval. Effect sizes for ranges are the Kolmogorov Smirnov test statistic, D, and correspond to a 95% confidence interval.

Variable	Device Results		Absolute difference (trad. – SG)	Effect size (OR, r, D)	CI	P -value
	Traditional	Self-guided
Intubation force
Average force	10.29 N	1.52 N	8.77N	0.71	0.38–0.88	0.003
Peak force	14.75 N	3.74N	11.00 N	0.83	0.50–0.95	0.008
Intubation range
Bench range	45.5%	72.7%	−27.2%	0.27	0–0.63	0.736
Cadaver range	2.5 cm	9.0 cm	−6.5 cm	0.70	0.40–0.90	0.007
User study, All users / all airways
First-pass success	62.5%	86.9%	−24.5%	4.0	0.92–17	0.055
Overall success	91.7%	95.7%	−4.0%	2.0	0.17–24	0.576
First-pass duration	21.1s	16.4s	4.7s	0.30	−0.03–0.63	0.079
Overall duration	44.4s	20.8s	23.6s	0.40	0.11–0.70	0.008
Average # attempts	1.6	1.1	0.5	0.27	−0.01–0.56	0.062
User study, All users, difficult airways
First-pass success	35.7%	92.9%	−57.1%	23.4	2.23–236	0.002
Overall success	85.7%	100%	−14.3%	5.80	0.25–133	0.224
First-pass duration	22.6s	17.7s	4.9s	0.19	−0.27–0.57	0.456
Overall duration	64.4s	23.7s	40.7s	0.48	0.15–0.71	0.012
Average # attempts	1.8	1.1	0.7	0.50	0.18–0.73	0.008

Since some anatomies can be more challenging for intubation, we conducted a post hoc exploratory analysis of 14 of the 24 total intubations performed on the three cadavers that were most challenging for traditional intubation techniques using VL to develop hypotheses for future, prospective, clinical testing. The subgroup intubation analyses were performed on the three most challenging anatomies with the highest overall failure rates, intubation durations, and the number of attempts needed for success (Fig. 5B, fig. S2, table S2). Within this subgroup, the first-pass success rate was higher with the SRIS when compared with VL (93% vs. 36%, n = 13 vs. 5 out of 14 attempts, OR 23.4, 95% CI 2.3–236, P = 0.002, Table 2), and the overall success rate was similar (100% vs. 86%, n = 14 vs. 12 out of 14 attempts, OR = 5.8, 95% CI 0.25–132, P = 0.224, Table 2). First-pass intubation duration was similar (18 s vs. 23 s, r = 0.19, 95% CI −0.27–0.57, P = 0.46, Table 2), and overall intubation duration was lower than VL (24 s vs. 64 s, r = 0.48, 95% CI 0.15–0.71, P = 0.012, Table 2). The number of intubation attempts was lower with the SRIS than with VL in this subgroup as well (1.1 vs. 1.8, n = 14 vs. 12, r = 0.50, 95% CI 0.18–0.73, P = 0.008, Table 2).

Further details of the study implementation can be found in Methods, and a representative intubation by an EMT with the SRIS device is depicted in Fig 5C.

DISCUSSION

In this study, we introduced a soft robotic tip-extending intubation device designed to not require visualization, extensive training, or anatomical knowledge. This contrasts with the current modality of manual placement, in which a push-from-the-base tube is guided through the vocal cords. The presented device may provide new benefits when supporting emergency airway management in austere and resource-limited environments.

In characterization tests of the self-guided tube component, our results showed a reduction in tool-tissue tangential forces when passing through a constriction and an expansion of the operational range in pre-clinical models compared with a traditional ETT. The reduced tool-tissue tangential force could be explained by the mechanism of tip extension, which minimizes the relative motion of the tube body with respect to the environment. Reduced tool-tissue forces during intubation could reduce tissue irritation. We also found an expanded operational range during misalignment testing, which may be a result of how tip extension enables passive path following. Such misalignment could occur in practice due to non-ideal patient body positions, anatomical differences, injuries, growths, or poor placement of the device. Unlike the standard ETT, which would catch on various anatomical features when misaligned, the soft, inflated overtube was able to deform to follow the path into the trachea, and the everting tip was able to “roll” over catchpoints in the anatomy, such as the petiole. However, the testing was limited to two pre-clinical models (one mannequin and one cadaver), and for each, no anatomical variation was tested. Further testing with diverse anatomies, injury, or alternate body positions should be conducted before conclusions can be made about use in live humans.

The initial tests of the full device (self-guided tube plus introducer) demonstrated 100% first-pass success in a mannequin (Fig. 3) and a cadaver (Fig. 4) when used by a device expert, with average intubation durations of less than 8 seconds. This contrasts with one study reporting approximately 50 seconds with existing devices outside the hospital (34). These results suggest that the SRIS has the potential to offer reliable and rapid intubation in controlled environments with expert users. However, the sample of airway anatomy in this testing was limited to two variations of pre-clinical models, and clinical evidence is required before conclusions can be made about performance in live humans in uncontrolled circumstances.

Finally, the preliminary cadaveric study comparing SRIS performance to VL in the hands of emergency medical providers showed a significant reduction (P = 0.008) in overall intubation duration, and similar performance on first-pass success rate, first-pass intubation duration, and the number of attempts required for successful intubation. Post hoc exploratory analysis showed that in the most challenging airways tested by this cohort of users, the SRIS required fewer attempts, shorter intubation duration, and increased first pass success compared with VL, suggesting future clinical testing specifically targeting difficult airways. Such performance in a mechanical system with minimal training requirements, if validated in a powered clinical trial, could represent a promising step forward in emergency airway management capabilities.

However, this preliminary study has several limitations. First, it was completed in fresh, recently deceased cadavers, which are superior to embalmed samples (27), but do not precisely represent the live human airway. The cadaveric model lacks the hemodynamic properties and tissue response of a live patient. Second, the testing was conducted in a controlled environment, rather than the stressful environment of an emergency situation. Additionally, the preliminary nature of this testing lacks predefined and powered endpoints, includes post hoc analysis of much of the data, and reports from a relatively small sample size. Future testing should evaluate the device in diverse anatomies to better quantify the SRIS performance across a wider sample of human airways and clinical providers.

Despite these limitations, we demonstrated that the SRIS can successfully intubate models and cadavers with a high success rate and short intubation duration, even with minimal training and lower-skilled operators. These results are a promising first step, with future work planned to develop clinical data, obtain 510(k) regulatory clearance in the USA, and market the device. Such efforts seek to develop a device that has the potential to benefit emergency intubation, enabling safe, rapid, and reliable intubations without requiring manual guidance or anatomical knowledge and complex psychomotor skills.

MATERIALS AND METHODS

Study Design

The objective of this study was to conduct a preliminary characterization and evaluation of the presented intubation device in pre-clinical models. Testing of the self-guided tube component and full-device characterization tests were completed by study authors in benchtop human larynx models and fresh-frozen cadavers. All data were included in the statistical analysis, including outliers, with methods for each test described below. The preliminary user study was completed with eight device testing participants and eight cadavers under Institutional Review Board (IRB) oversight, Protocol Number 20240922. All experimental protocols were reviewed and approved by Western-Copernicus Group (WCG) IRB, and were carried out in accordance with all relevant guidelines and regulations. User inclusion criteria were adults over 18 years old, credentialed to practice medicine as a paramedic or EMT, with no advanced training (MD or equivalent). Sample sizes for each study arm were selected as three unique cadaver trials for each of 8 users, for 24 total trials per arm; due to the preliminary nature of the study, no power analysis was used in sample size determination. Prospective endpoints were intubation success rate (both first-pass and overall), intubation duration, and intubation attempts. Users were grouped in sets of two, comprising one paramedic and one EMT, and each group was randomly assigned three cadavers. For each user, the device sequence in each cadaver was also randomized; sequence and allocation randomization were blinded. The conduct of the experiment was not blinded due to the nature of device testing. There was one exclusion from the SRIS for gross protocol deviation. Otherwise, all data were included in the analysis.

Self-guided Tube Force Testing

To evaluate the force to advance the traditional ETT and the self-guided tube into the airway, five tests were conducted for each through a 3D printed, anatomically-correct, durometer-matched model larynx (31) with force measurements captured at 10mm intervals. The model was printed on a FormLabs Form 3 in translucent Flexible 80A resin, with this durometer selected to match the mechanical properties of human cartilage, as tested and published in (32). The model was graduated at 10mm increments to aid in data collection. The test setup included a custom-fabricated, free-rolling cart that carried the model larynx. This cart was placed against a grounded force gauge (Mark-10 25N Force Gauge), ensuring that all forces applied to the model larynx were directly transferred to the force gauge.

For the traditional ETT (7mm ID, 10mm OD), prior to each test, a manual counterbend was imparted to straighten the ETT to match the shape of the straight, self-guided tube. Each device was slowly passed through the center of the model larynx while force data was gathered at 60 Hz. An aperture was secured to a linear slide to facilitate alignment with the laryngeal inlet, through which the devices were passed, allowing each device to pass straight through the axis of the model larynx.

Each device was placed posterior of the epiglottis with the epiglottis in its natural, open position, with first datapoint collected when the tube reached the base of the epiglottis posterior of the hyoid bone, and successive data points at 10mm increments thereafter. One-minute breaks were held between each test to ensure proper relaxation of the 3D-printed material

Self-guided Tube Misalignment Testing

Misalignment testing was conducted in two different environments: (i) in the durometer-matched, anatomically correct 3D printed model used for force testing, and (ii) in a cadaver. For (i), the model used was the same 3D-printed model larynx used for force testing, which was designed to capture a wide variety of misalignments. The misalignments this fixture were intended to test include two different configurations: coronal plane (lateral-medial plane and left-right misalignment) and sagittal plane (anterior-posterior plane and posterior misalignment).

(ii) Inferior-superior offset misalignment testing in cadaver: This test was completed with one unembalmed, fresh-frozen cadaver bust (including the cephala and torso, dissected at the shoulders and nipple line) sourced from the UC Irvine Willed Body Program. “Fresh-frozen,” rather than embalmed, cadavers result in substantially more life-like tissue qualities (35). The epiglottis was not lifted out of the way and was part of the challenge of intubating in this test.

A commercially available SGA (i-gel, Intersurgical, size 3) was used to place the intubating tubes. The SGA was placed and its nominal location in the hypopharynx was marked on the handle. From here, graduations at 1cm increments both superior and inferior were added, and approximately 3 cm of material was removed from its tip (without modifying the tube egress port geometry) to allow for deeper-than-standard placement to better explore the range. The SGA was placed at increments of 1 cm, as measured by markings on the handle and referenced from the front teeth. Successful intubation was determined with endoscopic confirmation through a port cut in the trachea. Four trials at each depth with each device were completed.

Full device mannequin testing

Using the Laerdal Airway Management Trainer mannequin, 50 tests were conducted to measure success rate and intubation duration under controlled conditions. To do so, one expert operator conducted 50 consecutive tests in this industry-standard mannequin. “Expert” designation was based on the number of attempts with the SRIS device exceeding 100 uses. An internal camera (Ambu aScope 4 bronchoscope, with an Ambu aView 2 Advance monitor) was placed in the trachea internally, from which the user was blinded. An external camera was set up on a tripod to observe the initiation and completion of a trial to register the intubation duration. Intubation success was defined as the ETT fully deploying through the cords and into the trachea. Intubation duration was defined as the time from the user or device first interacting with the mannequin to completion of placement (fully advanced), subject to confirmation of success. Cuff inflation time was not included for the SRIS or VL in any test.

A single user completed all 50 tests, with the mannequin lubricated with manufacturer-supplied lubricant every 3 runs. In each test, the user inserted the introducer until it stopped, and then advanced the self-guided tube through its full throw. Once completed, the user confirmed proper placement on the internal camera feed. If successful, the intubation duration was recorded from the external camera as the time between the tip of the device entering the mouth and the completion of the ETT advancement. The device was reset in between tests, and no failures were experienced.

Full device cadaver testing with an expert user

Expert-user testing was completed with one unembalmed, fresh-frozen cadaver bust (including the cephalus and torso, dissected at the shoulders and nipple line) sourced from the UC Irvine Willed Body Program. The time to deploy the device and the time to remove the introducer were recorded, and the test was repeated 10 times.

Similar to the mannequin testing, an internal camera (Ambu aScope bronchoscope powered by the Ambu aView 2 Advance) was placed through a tracheostomy cut at the cadaver’s clavicle. An external camera was set up to capture the entire attempt, from insertion to removal. The user was blinded from the internal camera and was tasked to seat the device, advance the ETT, and remove the introducer without dislodging the ETT from the trachea, as confirmed with the internal camera.

Intubation duration was measured with the external camera and defined as the time from the user or device first interacting with the cadaver to completion of placement, subject to confirmation of success. The external camera measured removal duration as the time between completion of successful placement and the complete removal of the introducer. Success was contingent upon the ETT entering the trachea and not getting dislodged during introducer removal.

Full device testing in cadavers with novice users

The emergency provider cadaver tests were conducted at the Centre for Emergency Health Sciences in San Antonio, Texas in one day. Participants were recruited from the local emergency medical services agency (Bulverde Spring Branch Emergency Services).

Nine cadavers were made available by the Centre for Emergency Health Sciences for testing through their ethical procurement process. The testing team intubated each specimen with the SRIS device, during which each cadaver was successfully intubated and the appropriately-sized device was selected (from one of three sizes that follow the sizing convention of existing SGAs, which does not affect the sizing of the self-guided tube). This step was completed to minimize the number of testing variables for this preliminary study.

Of the nine cadavers made available, eight were used for testing, and one was used exclusively for training purposes. Due to the lack of vascular pressure and flow, repeated laryngoscopy attempts have been shown to create tongue grooves in cadavers (36). Efforts were made to minimize the number of attempts per cadaver to preserve the quality of the specimens made available by the Centre for other uses, and to maximize realism for each participant. Due to the minimization of attempts, coupled with the randomization and sequencing, this tongue groove issue was not observed in this study.

Each participant received a standardized training of a 5-minute video for video laryngoscopy training (movie S2), a 3-minute video for SRIS training (movie S3), and one attempt with each device in the training cadaver. Participants were allowed to ask the testing team clarifying questions about the use of each device during this training, but no unsolicited instruction was provided. The video laryngoscope used was the UEScope VL460 as it is the standard issue VL for the Bulverde Spring Branch Emergency Services system from which the participants were recruited.

For testing, the order of participants and devices was randomized. First, a digital coin flip decided the paramedic/EMT order. Next, a series of coin flips determined which device would be used for each cadaver first for that participant, and this process was repeated for the second participant. Each participant completed their tests in two sequences: the first sequence with the devices selected in the coin flip, the second sequence with the alternate devices. Only one participant was allowed in the testing room at a time, and a break was given in between these sequences.

There were 4–5 proctors in the room with the participant. Two of these investigators have financial conflicts of interest (conflicted investigators, CIs), and were only allowed to set up the test devices and support the unconflicted investigators (UIs) in protocol implementation, but not to make decisions on data collection. These UIs were medical professionals (two EMS-trained medical doctors and one PhD researcher with a paramedic background). The UIs provided the participants with a standardized case (a drowning victim in respiratory failure with a weak pulse that did not require CPR) and recorded all data collected.

An external camera was mounted on a tripod to record each test. The participant was given their test device and the medical scenario and allowed to complete the intubation when ready. Recorded quantities were the device employed, intubation success/failure, intubation duration, number of intubation attempts, provider completing the intubation, and cadaver used. Representative intubations with both VL and SRIS are shown in movies S4 and S5. Intubation success was defined as proper placement in the trachea, and verified by passing the Ambu aScope bronchoscope through the core of the ETT to provide a confirming view of tracheal rings. Intubation duration was defined as the time from the user or device first interacting with the cadaver to completion of placement, subject to confirmation of success. Time was paused to confirm proper placement with the endoscope; if improperly placed, time was restarted and the participant was instructed to retry, up to the limit of attempts for SRIS (3 attempts) or time for VL (180s). An attempt was defined as the device entering and exiting the body, or undergoing substantial modifications of the device by the user (such as using the teeth of the cadaver to bend the traditional, styled ETT into a new shape without removing it from the mouth). Success rates were calculated by dividing the number of successes by the number of trials, on both a first-pass and overall basis. All these quantities were evaluated and recorded by the UIs only.

Device Design and Fabrication

The device design was arrived at through dozens of iterative testing cycles through over 100 unique cadaveric airways. Through that process, we finalized the key geometric parameters for the overtube s-curve geometry: the angle of the rising section with respect to the horizontal, θ₁ (approximately 80 degrees), the height of the rise, h (15mm), and the angle of the distal section with respect to the horizontal, θ₂ (approximately 15 degrees). The overtube diameter was similarly selected to be 12 mm to allow free passage of the ETT within it. The overtube was fabricated out of 0.005” thick TPU film with a laser cut template mask and a thermal welding process. The mask produces the s-curve geometry, which is characterized by the above values. This overtube is then adhered at two locations along the body of the ETT with UV-curing adhesive to allow eversion.

The introducer was similarly developed through iterative design cycles across the same cadaveric specimens, starting with inspiration from commercially available supraglottic airways. Key values were found to be the length of the anatomical stop (25mm), the radius of curvature of the handle (9 cm), and the stiffness of the handle portion (9 N at a 4 cm tip deflection). It was then fabricated by insert overmolding, wherein a two-part polyurethane mixture was injected into a 3D-printed mold with a polycarbonate insert to define the geometry of the anatomical stop. This composite was then overmolded with a lubricious silicone material to enable low-friction interaction with the oropharynx. The overtube and ETT assembly were then affixed to the introducer with UV curing glue.

Statistical Analysis

Data is presented as mean ± SD unless otherwise stated. Statistical analysis was completed in MATLAB. Intubation duration, device force application, and intubation attempts statistical testing used the two-sided, unpaired, Wilcoxon Rank Sum test to evaluate the null hypothesis of no difference in median performance compared with existing methods, at a 5% significance level. Intubation success rate, both first pass and overall, was evaluated with a Chi-squared test for statistical significance. Misalignment performance testing was completed using the two-sided Kolmogorov-Smirnov (K-S) two-sample test to compare distributions. All data was assumed to be non-normal but were not statistically evaluated for normality; no statistical test employed relies on the normality assumption. Effect sizes for each test were reported: the correlation coefficient r (Wilcoxon Rank Sum test); Odds Ratio (OR) (Chi-squared test, with Haldane-Anscombe correction added in the cases of zero-cell circumstances); or the K-S coefficient D (K-S test). These effect sizes were supported with 95% confidence intervals, calculated using the Wald Method. A logarithmic transform and Fisher’s Z-transform was applied to the CI calculation for OR and r, respectively, for small sample sizes to improve normality of the CI. All individual-level data is available in data file S1.

Supplementary Material

List of Supplementary Materials

Materials and methods

Figures S1 and S2

Data file S1

Movie S1 to S5

MDAR reproducibility checklist

Data and materials availability

All data associated with this study are available in the main text or the supplementary materials. Prototype devices for additional testing can be provided upon request to the corresponding authors.