Background: Orotracheal intubation is an essential procedure in critically ill patients requiring ventilatory support, and the use of videolaryngoscopy is recommended as the gold standard in clinical practice. In recent years, the development of videolaryngoscopes through additive manufacturing has demonstrated lower production costs, and their integration with artificial intelligence (AI) has shown promising results in improving procedural success and safety.

Aim: To evaluate the clinical performance and usability of a videolaryngoscope developed through additive manufacturing and integrated with artificial intelligence (AI) and computer vision, compared with a commercially available gold-standard videolaryngoscope.

Methods: This is a multicenter, randomized, controlled, superiority Phase I/II clinical trial with a single-blind design. A total of 120 critically ill patients admitted to intensive care units at four hospitals in southern Brazil will be enrolled. Participants will be individually randomized in a 1:1 allocation ratio to parallel groups. The interventions will be performed by at least 10 physicians. All physicians will receive standardized training on device use, and procedures will be monitored by trained research staff. Data will be recorded in electronic Excel spreadsheets, and statistical analyses will be conducted using IBM SPSS Statistics. All ethical principles will be strictly observed, including obtaining informed consent from participants or their legal representatives. A Data and Safety Monitoring Committee will be established to oversee the study.

Conclusions: If the expected results are confirmed, patients, healthcare professionals, and healthcare institutions may benefit from the improved performance, safety, and usability of the experimental device, particularly in critical care settings and public health emergencies.

Backgorund Orotracheal intubation (OTI) is an essential procedure in critically ill patients requiring mechanical ventilation to maintain gas exchange. It is estimated that OTI is the third most commonly performed procedure in hospitals worldwide, with more than 15 million people intubated annually in surgical centers and 650,000 in other clinical settings, generating over 2.1 billion dollars in global expenditures on intubation and airway protection devices.

Traditionally, a laryngoscope is used for direct visualization of the upper airways and insertion of the endotracheal tube. However, scientific evidence has demonstrated the superiority of videolaryngoscopy (VL), recommending its widespread use in all intubation scenarios as the primary technique in clinical practice.

During the COVID-19 pandemic, given the large number of patients with severe respiratory dysfunction, videolaryngoscopy became essential for managing difficult intubations. In this context, our research team developed a VL prototype using additive manufacturing with polylactic acid (PLA) and carbon fiber-reinforced PLA (PLA-Carbon), validated in realistic simulations for mechanical resistance and usability, with a patent registered at the Brazilian National Institute of Industrial Property (INPI) under number BR 10 2020 026194.

Currently, this VL is being enhanced through the integration of artificial intelligence (AI) using computer vision techniques based on machine learning and deep learning, such as convolutional neural networks, capable of automatically recognizing upper airway anatomical structures from pre-trained images. Given the high demand for orotracheal intubation and the inherent risk of severe complications, the limited availability of videolaryngoscopes due to high import costs in low- and middle-income countries represents a significant gap in care. To address this challenge, the development of a low-cost device using 3D printing combined with artificial intelligence is proposed. This technological integration is justified by its potential to expand access to clinical decision support and enhance procedural safety by providing real-time automated anatomical recognition to the operator.

Primary Objective To evaluate the effectiveness of an experimental videolaryngoscope integrated with artificial intelligence (AI-VL), developed using additive manufacturing, compared with a standard commercial videolaryngoscope (C-VL) during orotracheal intubation in adult patients admitted to the intensive care unit.

Study Design This is a multicenter, randomized, controlled, superiority, phase I/II, single-blinded clinical trial with individual randomization and 1:1 allocation into parallel groups (60 in the AI-VL group and 60 in the C-VL group). Interventions will be performed by at least 10 physicians.

Interventions and Comparator The study compares two devices for orotracheal intubation: a videolaryngoscope produced using additive manufacturing with embedded artificial intelligence and computer vision (AI-VL), and a standard commercial videolaryngoscope (C-VL), used as the comparator (gold standard). Participating physicians will receive prior training in the use of the AI-VL. In the intensive care unit, after confirmation of the indication for intubation and eligibility criteria, patients will be allocated to one of the study groups.

Administration of Interventions Each intensive care unit will include previously trained field researchers responsible for data collection and for the use of both AI-VL and C-VL devices. These researchers will oversee recruitment, physician training, intervention implementation, troubleshooting, and quality assurance. Following allocation, the assigned device will be used, and intubation will be performed as promptly as possible. The same physician may perform more than one intervention.

In both groups, patients will receive standard intubation care, including preoxygenation, monitoring, equipment preparation, selection of appropriate videolaryngoscope blade and endotracheal tube size, confirmation of tube placement, adequate ventilation (SpO₂ > 92%), tube fixation, and connection to mechanical ventilation.

Procedures for Individual Adaptation Both videolaryngoscopes will be available in different blade sizes to accommodate anatomical variability. For the AI-VL, the artificial intelligence software will be trained using an image database to enable automated glottic identification across different individuals.

Physical and Informational Materials Both devices will be accompanied by technical manuals and clinical use guidelines in physical and digital formats (QR code), as well as video tutorials covering assembly, operation, and cleaning. The AI-VL will additionally include instructions regarding the use of the embedded AI system.

Sample Size The study follows ISO 14155:2020, ISO 62366-2, IEC 62366, and ANVISA RDC No. 837/2023 guidelines for research involving medical devices. The sample size calculation, based on FDA CDRH (2016) and ISO 62366-2 recommendations, considered 80% power and a significance level of 5%. A total of 120 patients will be included, equally distributed across groups and institutions, along with at least 10 physicians, aiming at preliminary analyses of effectiveness and usability.

Recruitment Informational materials will be made available in advance within participating institutions and through digital media. Recruitment of patients and professionals will take place in the ICU, according to eligibility criteria. Professionals and patients capable of providing consent will be invited through direct contact. For patients unable to consent, consent may be obtained from family members or legal representatives, either prior to or following the procedure (deferred consent). Further details on consent are provided in the ethics section.

Sequence Generation The randomization sequence will be generated in Microsoft Excel by an external statistician using computer-generated random numbers, in equal-sized blocks and stratified by participating center.

Allocation Concealment Mechanism The allocation sequence will be placed in opaque, sealed envelopes by the principal investigator and distributed to field researchers at each of the four participating hospitals.

Implementation Field researchers, blinded to the allocation sequence, will enroll and assign participants individually in the ICU immediately before intubation to either the intervention group (AI-VL) or the comparator group (C-VL), in a 1:1 ratio.

Data Collection Methods

Data collection will be conducted by trained field researchers under the supervision of the research team. Data will be collected as follows:

• During the procedure: Field researchers will perform direct observation during intubation using a critical task checklist;
• Immediately after the procedure: Physicians will complete a sociodemographic characterization form, the System Usability Scale (SUS), the NASA Task Load Index (NASA-TLX), and a user perception questionnaire;
• After the procedure: Field researchers will collect patient sociodemographic and clinical data from medical records.

Data will be obtained through electronic questionnaires (tablets or QR codes) or paper forms, which will be compiled and stored in Microsoft Excel and SPSS.

A pilot study with the first 10 patients will be conducted to assess the reliability and validity of the instruments; these participants will be included in the final study sample. Reasons for non-adherence or loss to follow-up will be recorded.

Statistical Methods for Data Analysis Data will be analyzed using descriptive statistics according to variable type: continuous variables will be presented as mean ± standard deviation or median (interquartile range), after normality assessment using the Shapiro-Wilk test; categorical variables will be presented as absolute numbers and proportions (%).

Comparisons between non-parametric groups will be performed using the Kruskal-Wallis test. Predictors of intubation success will be analyzed using logistic regression, including subgroup analyses through regression models with interaction terms.

The internal consistency of the SUS scale will be assessed using Cronbach's alpha, and associations between usability (SUS), satisfaction, and overall perception will be examined using Spearman's correlation. For NASA-TLX score calculation, the unweighted version (Raw TLX) will be used; internal consistency will be assessed using Cronbach's alpha, and normality will be evaluated using the Shapiro-Wilk test. For inferential analysis, Wilcoxon and/or Mann-Whitney tests will be applied.

Missing data in Likert scales will be handled using multiple imputation (MICE), or alternative methods when appropriate. In cases where multiple imputation is not suitable, methods such as Last Observation Carried Forward (LOCF) will be used, if appropriate. Interim analyses will assess data consistency and outcome trends.

A significance level of 5% (p < 0.05) will be adopted for all tests, with 80% statistical power. Data will be analyzed using SPSS software. Responses from the user perception questionnaire will be subjected to qualitative analysis.

Data and Safety Monitoring Committee

The Data and Safety Monitoring Committee, established in accordance with recommendations from the Brazilian Ministry of Health (Brazil, 2008), will be responsible for overseeing study conduct, evaluating the effectiveness and safety of interventions, and protecting participants. The committee will be appointed by the project coordinator and composed of four members with the following qualifications:

• Physician: experience in conducting studies related to the study topic;
• Nurse: experience in conducting studies related to the study topic;
• Statistician: undergraduate or postgraduate degree in a relevant field and experience in health data analysis;
• Other healthcare professionals: experience in the design, conduct, or analysis of clinical trials.

Members will be independent of the funding source and will not receive financial compensation. The committee will meet monthly, remotely, to review study progress and issue recommendations regarding continuation, modification, suspension, or termination, which will be communicated to the Research Ethics Committees.

An interim analysis will be conducted by the statistician after inclusion of 50% of the sample (n = 60) to assess the need for early termination of the study if statistical (p < 0.05) and clinical superiority of the AI-VL is demonstrated, according to the criteria in Tables 1 and 2. The committee, not blinded, will evaluate safety issues and will be notified in cases of serious adverse events.

ANTICIPATED RESULTS It is expected that the results of this clinical trial will demonstrate superior effectiveness and usability of the AI-integrated videolaryngoscope (AI-VL) compared to the standard commercial videolaryngoscope (C-VL) for orotracheal intubation in critically ill patients. If this hypothesis is confirmed, healthcare systems may benefit from improved performance and safety, reduced complications, cost optimization, enhanced response in high-stress or crisis situations, reduced risk of pathogen exposure, and positive impacts on professional training.

Inclusion Criteria:

Patient:

• Age ≥18 years.
• Both sexes.
• Clinical indication for orotracheal intubation.

Physician:

• Both sexes.
• Working in emergency or ICU settings.
• Minimum of one year of experience in laryngoscopy and/or orotracheal intubation.

Exclusion Criteria:

Patient:

• Clinically unstable or with contraindications to participation in experimental research.
• Significant airway malformations.
• Previously intubated or with a functioning tracheostomy.
• Pregnant.

Physicians:

• On vacation or leave during the study period.
• Orthopedic, visual, or cognitive limitations that impair participation.

All IPD that underlie results in a publication.

Individual patient data (IPD) will be available at the time of publication of the study results (expected in 2028), for an indefinite period.

Individual patient data (IPD) will be accessible to researchers via a link that will be available along with the study results.

The statistician will be masked.