Skip to main content

Foundational Trials

DEVICE

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Video versus Direct Laryngoscopy for Tracheal Intubation of Critically Ill Adults
  • Acronym: DEVICE
  • Year: 2023
  • Journal published in: The New England Journal of Medicine
  • Citation: Prekker ME, Driver BE, Trent SA, et al. Video versus direct laryngoscopy for tracheal intubation of critically ill adults. N Engl J Med. 2023;389(5):418-429.

Context & Rationale

  • Background
    Emergency tracheal intubation in the ED/ICU frequently requires repeated attempts and is associated with peri-intubation hypoxaemia and cardiovascular instability in observational cohorts6.
    Video laryngoscopy improves glottic visualisation, but earlier ICU and ED trials showed mixed effects on first-pass success and complications, including an ICU multicentre RCT reporting similar first-pass success with higher severe complications with video laryngoscopy1 and smaller single-centre trials suggesting benefit in some operator groups23.
    Contemporary systematic reviews/meta-analyses in critically ill adults continued to show heterogeneity by setting, operator experience, and device type, with uncertainty about whether improved visualisation translates into fewer complications or better patient-centred outcomes45.
  • Research Question/Hypothesis
    In critically ill adults undergoing emergency tracheal intubation in the ED/ICU, does initial use of a video laryngoscope (vs a direct laryngoscope) increase successful intubation on the first attempt without increasing peri-intubation severe complications?
  • Why This Matters
    First-attempt success is a pragmatic, operator- and system-sensitive endpoint linked to hypoxaemia and haemodynamic collapse during airway management; determining the optimal default laryngoscopy strategy has implications for equipment purchasing, training curricula, and safety bundles in high-risk ED/ICU intubations.

Design & Methods

  • Research Question: Among critically ill adults requiring emergency tracheal intubation in EDs and ICUs, does assigning video laryngoscopy for the first attempt increase first-attempt success compared with direct laryngoscopy?
  • Study Type: Pragmatic, multicentre, parallel-group, randomised superiority trial conducted in 17 emergency departments and ICUs; open-label (no blinding) with concealed allocation using sealed opaque envelopes; outcome data collected by trained independent observers when feasible7.
  • Population:
    • Inclusion: Adults (≥18 years) located in a participating unit with a planned orotracheal intubation using a laryngoscope, performed by a clinician who routinely performs tracheal intubation7.
    • Key practical eligibility constraint: The operator believed that either laryngoscopy approach could reasonably be used (i.e., no perceived requirement for a specific approach for that patient)7.
    • Exclusions (protocol-level): Age <18 years; known pregnancy; known prisoner status; an immediate need for intubation that precluded safe performance of the randomisation process; operator believed one approach was necessary or that one approach was contraindicated7.
  • Intervention:
    • Assignment to a video laryngoscope for the first intubation attempt (first attempt defined by a single laryngoscope blade insertion and either a single tracheal tube insertion or a bougie insertion followed by tracheal tube insertion); video laryngoscope type/blade geometry left to clinician discretion (pragmatic).
  • Comparison:
    • Assignment to a direct laryngoscope for the first intubation attempt; subsequent attempts and adjuncts (e.g., bougie, stylet) at clinician discretion (pragmatic).
  • Blinding: Unblinded (operators and bedside team not blinded); the primary outcome is operationally defined and intended to be objectively observed; lack of blinding increases risk of performance bias (choice of adjuncts/backup strategy) and detection bias for more subjective complications.
  • Statistics: Power calculation targeted a 5% absolute increase in first-attempt success (from 80% to 85%) with 90% power at a two-sided 5% significance level; planned enrolment 1920 patients (inflated to 2000 allowing for up to 4% missing data), with one interim analysis at 1000 patients and a Haybittle–Peto-style efficacy boundary of P≤0.001; primary analysis by intention-to-treat with absolute risk differences and 95% CIs (with adjusted sensitivity analyses accounting for site as a random effect)7.
  • Follow-Up Period: Peri-procedural outcomes measured during the intubation sequence (from induction through shortly after intubation); ICU-free days, ventilator-free days, and in-hospital death assessed up to 28 days after enrolment.

Key Results

This trial was stopped early. Enrolment was stopped after a prespecified interim analysis for efficacy, at 1420 enrolled (1417 included in analysis).

Outcome Video laryngoscope Direct laryngoscope Effect p value / 95% CI Notes
Successful intubation on first attempt (primary) 600/705 (85.1%) 504/712 (70.8%) Absolute difference +14.3 pp 95% CI 9.9 to 18.7; P<0.001 First attempt = 1 laryngoscope blade insertion and 1 tube insertion (or bougie + tube insertion).
Severe complication during intubation (prespecified) 151/705 (21.4%) 149/712 (20.9%) Absolute difference +0.5 pp 95% CI −3.9 to 4.9 Composite from induction to 2 min after intubation: SpO2 <80%, SBP <65 mmHg or new/increased vasopressor, or cardiac arrest.
Median duration of intubation attempt 38 s (IQR 27–60) 46 s (IQR 32–72) Median difference −8 s 95% CI −12 to −4 Duration from laryngoscope insertion to laryngoscope removal (or end of first attempt).
Successful intubation on first laryngoscope blade insertion 636/704 (90.3%) 546/706 (77.3%) Absolute difference +13.0 pp 95% CI 9.1 to 16.9 Exploratory.
Successful first attempt without severe complication 484/705 (68.7%) 420/712 (59.0%) Absolute difference +9.7 pp 95% CI 4.5 to 14.8 Exploratory composite of procedural success and safety.
Oesophageal intubation 6/705 (0.9%) 9/712 (1.3%) Absolute difference −0.4 pp 95% CI −1.5 to 0.7 Operator-reported complication.
Aspiration 7/705 (1.0%) 12/712 (1.7%) Absolute difference −0.7 pp 95% CI −2.1 to 0.8 Operator-reported complication.
ICU-free days at 28 days 20 (IQR 0–25) 19 (IQR 0–24) Median difference +1 day 95% CI −1 to 3 Death = 0; discharge before day 28 = 28.
Ventilator-free days at 28 days 24 (IQR 0–26) 23 (IQR 0–26) Median difference +1 day 95% CI 0 to 2 Death = 0; discharge before day 28 = 28.
In-hospital death within 28 days 184/705 (26.1%) 191/712 (26.8%) Absolute difference −0.7 pp 95% CI −5.5 to 4.0 Post hoc clinical outcome.
In-hospital death within 1 hour 15/705 (2.1%) 27/712 (3.8%) Absolute difference −1.7 pp 95% CI −3.6 to 0.2 Post hoc; low event counts.
  • Video laryngoscopy increased first-attempt success by 14.3 percentage points (85.1% vs 70.8%), meeting an early-stopping efficacy threshold.
  • The prespecified composite of severe complications was similar between groups (21.4% vs 20.9%), with wide CIs compatible with modest harm or benefit.
  • Video laryngoscopy shortened the median intubation attempt duration (38 s vs 46 s) and improved single blade-insertion success (90.3% vs 77.3%).
  • Subgroup (operator experience; exploratory): absolute difference in first-attempt success (video vs direct) was 26.1 pp (95% CI 15.4 to 36.8) for operators with <25 prior intubations, 12.3 pp (6.8 to 17.7) for 25–100 prior intubations, and 5.9 pp (−4.1 to 16.0) for >100 prior intubations; CIs not adjusted for multiple comparisons8.

Internal Validity

  • Randomisation and allocation concealment: Block randomisation stratified by site, implemented via sealed opaque envelopes; allocation revealed only immediately before laryngoscopy, supporting concealment and reducing selection bias7.
  • Post-randomisation exclusions: Three patients identified after enrolment as prisoners were excluded from subsequent data collection and analysis (pragmatic exclusion that introduces a small risk of bias if not random).
  • Performance/detection bias: Unblinded operators and teams could influence adjunct choices and escalation strategies; primary endpoint is largely procedural/objective, and observer-based capture was intended to reduce misclassification7.
  • Protocol adherence (separation): Laryngoscope on first attempt was video in 705/705 (100%) of the video group and direct in 704/712 (98.9%) of the direct group (video crossover 8/712, 1.1%).
  • Delivered “dose” of intervention: Randomisation applied only to the first attempt; subsequent attempts were at clinician discretion, which typically biases estimates towards the null for downstream clinical outcomes but preserves pragmatic relevance for “default first device”.
  • Baseline comparability: Key baseline variables were well balanced (e.g., age median 55 vs 55 years; APACHE II median 16 vs 16; ED location 69.7% vs 70.2%).
  • Heterogeneity and effect modification: Broad operator mix (residents/fellows predominated) increases generalisability but also introduces heterogeneity; prespecified effect modification by operator experience showed larger absolute benefit among less experienced operators (exploratory; not multiplicity-adjusted)8.
  • Timing: Randomisation occurred in proximity to airway management (workflow-concordant); however, very time-critical intubations were excluded by design (potential selection effect).
  • Missingness for physiological components: Severe-complication components used available data (e.g., SpO2 and SBP denominators smaller than total enrolled); missing induction vitals were reported in the trial (oxygen saturation at induction missing in 8.6%; systolic blood pressure at induction missing in 14.7%), which may attenuate precision for complication estimates.
  • Outcome assessment: Primary outcome definition was explicit; severe complications were anchored to prespecified thresholds and a narrow time window (induction to 2 minutes post-intubation), increasing objectivity but potentially missing later downstream deterioration.
  • Statistical rigour: Prespecified power, interim stopping rule, and ITT analysis; early stopping for efficacy can inflate effect size estimates for the primary endpoint and leaves rarer harms underpowered relative to procedural outcomes.
  • Separation of the variable of interest (numeric): Better glottic view (Cormack–Lehane grade 1: 538/705 [76.3%] vs 318/712 [44.7%]); shorter median attempt duration (38 s vs 46 s); higher single blade-insertion success (90.3% vs 77.3%).

Conclusion on Internal Validity: Overall, internal validity appears strong for the primary procedural endpoint (concealed randomisation, minimal crossover, explicit endpoint definitions), with moderate limitations for safety/clinical outcomes due to open-label conduct, some missing physiological data, and early stopping.

External Validity

  • Population representativeness: ED/ICU adults undergoing emergency intubation reflect a common real-world critically ill airway population; however, very time-critical cases (where randomisation could not be safely performed), pregnancy, and prisoners were excluded, and clinicians could exclude patients when they believed one approach was required or contraindicated.
  • Operator mix: Most intubations were performed by emergency medicine residents or critical care fellows (91.5%), which is typical of many academic ED/ICU systems but may differ from settings where anaesthetists are the primary operators.
  • Device/ecosystem dependency: Benefit may depend on access to modern video laryngoscopes, training, and routine use of adjuncts (bougie/stylet) and physiological optimisation; translation to resource-limited settings or systems without reliable video equipment/training may be constrained.
  • Applicability to other settings: Findings directly apply to ED/ICU (not operating theatre, elective intubation, prehospital, or paediatric practice) and should be interpreted in the context of local airway algorithms and rescue capabilities.

Conclusion on External Validity: Generalisability is moderate to high for adult ED/ICU airway management in similarly resourced systems, but is limited for very time-critical airways, settings lacking video equipment/training, and non-ED/ICU contexts.

Strengths & Limitations

  • Strengths: Large, pragmatic multicentre design; concealed randomisation; high adherence to assigned first-attempt device; clinically meaningful, operationally defined primary endpoint; broad operator and setting mix (ED and ICU); detailed procedural and safety outcome capture.
  • Limitations: Unblinded trial with potential performance/detection bias for some outcomes; early stopping for efficacy (risk of overestimating effect size for the primary endpoint); severe-complication composite may be underpowered for modest differences and relies on available physiological measurements; device type/blade geometry and adjunct strategies were not standardised (pragmatic heterogeneity).

Interpretation & Why It Matters

  • Default device choice
    As a pragmatic ED/ICU trial, DEVICE supports video laryngoscopy as a default first-attempt strategy to improve first-pass success, with no clear increase in the prespecified severe-complication composite; the procedural benefit is large enough to plausibly improve airway safety in systems where repeated attempts are common.
  • Safety signal interpretation
    Absence of a detected difference in severe complications does not equal equivalence: the confidence interval includes modest harm/benefit, and complication ascertainment depends on physiological measurement completeness and a narrow peri-intubation window.
  • Training and implementation
    Exploratory heterogeneity suggests the largest absolute benefit may accrue to less experienced operators; implementation should prioritise structured training in video laryngoscopy, clear rescue pathways (including direct laryngoscopy competence), and physiological optimisation bundles.

Controversies & Subsequent Evidence

  • Discordant prior trial evidence: The earlier ICU multicentre MACMAN RCT reported similar first-pass success (67.7% video vs 70.3% direct) but higher severe life-threatening complications with video laryngoscopy (45.2% vs 27.1%)1; DEVICE’s contrasting findings highlight how operator experience, device/blade selection, adjunct use (bougie/stylet), and local airway workflows can modify real-world effectiveness.
  • Early stopping and effect inflation: Stopping after interim efficacy increases the probability that the observed primary effect size is overestimated; this is particularly relevant for technology adoption decisions where training and workflow change can alter baseline performance.
  • Choice of endpoints: First-attempt success is a plausible safety surrogate but is not a patient-centred endpoint; DEVICE did not demonstrate clear differences in ICU-free days, ventilator-free days, or 28-day in-hospital mortality (all post hoc or exploratory), leaving uncertainty about downstream clinical benefit.
  • Safety composite and measurement: The severe-complication composite mixes hypoxaemia, haemodynamic collapse, and cardiac arrest; it is clinically coherent but sensitive to missing vital sign data and to the chosen observation window (induction to 2 minutes post-intubation).
  • Editorial framing: Contemporary commentary argued that the magnitude of the first-pass benefit in DEVICE is practice-changing and supports a “default video” paradigm, while emphasising that implementation requires training and system-level preparation rather than device substitution alone9.
  • Subsequent syntheses: Updated systematic reviews/meta-analyses in critically ill adults incorporating newer trials (including DEVICE) generally report improved first-pass success with video laryngoscopy but persistent uncertainty regarding major clinical outcomes, with heterogeneity by setting and operator experience45.
  • Guideline alignment: Guidance for critically ill airway management has increasingly endorsed video laryngoscopy as a first-line option alongside physiological optimisation, structured preparation, and rescue planning, consistent with DEVICE’s pragmatic implications1011.

Summary

  • DEVICE was a pragmatic multicentre RCT in 17 EDs/ICUs comparing video vs direct laryngoscopy for the first intubation attempt in critically ill adults.
  • The trial stopped early after an interim analysis; video laryngoscopy improved first-attempt success (85.1% vs 70.8%; absolute difference +14.3 pp; 95% CI 9.9 to 18.7; P<0.001).
  • The prespecified severe-complication composite was similar (21.4% vs 20.9%; absolute difference +0.5 pp; 95% CI −3.9 to 4.9).
  • Video laryngoscopy shortened attempt duration (median 38 s vs 46 s) and improved single blade-insertion success (90.3% vs 77.3%).
  • Exploratory analyses suggested greater benefit among less experienced operators; interpretations should account for early stopping, open-label conduct, and heterogeneity in devices/adjunct strategies.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Effect estimates are absolute differences (percentage points) or median differences, as reported; confidence intervals for secondary/exploratory outcomes were not adjusted for multiplicity.
  • DEVICE’s pragmatic design (operator choice of device geometry and adjuncts) increases real-world relevance but means effects may vary with local training and workflow.

Overall Takeaway

DEVICE is a landmark pragmatic ED/ICU airway trial showing that a “video-first” strategy substantially improves first-attempt intubation success in critically ill adults. The prespecified severe-complication composite was similar between groups, and clinical outcomes were not clearly different, underscoring that device choice is one component of a broader, physiology-focused intubation safety system.

Overall Summary

  • Video laryngoscopy improved first-attempt success by 14.3 percentage points in ED/ICU intubations.
  • Severe peri-intubation complications (composite) were similar, with CIs allowing modest harm or benefit.
  • Implementation should pair “video-first” with training, rescue planning, and physiological optimisation.

Bibliography