Skip to main content

Trial Summaries & Critiques

NOTACS

Image: Shutterstock / Nisitmicrostock

Pulmonary angiogram showing a pulmonary embolism

Publication

  • Title: High-Flow Nasal Oxygen Therapy After Cardiac Surgery: A Randomized Clinical Trial
  • Acronym: NOTACS
  • Year: 2026
  • Journal published in: JAMA Network Open
  • Citation: Litton E, Parke RL, McGuinness SP, Dawson SN, Villar SS, Shetty SS, et al; Nasal High-Flow Oxygen Therapy After Cardiac Surgery Investigators. High-flow nasal oxygen therapy after cardiac surgery: a randomized clinical trial. JAMA Netw Open. 2026;9(4):e265447.

Context & Rationale

  • Background
    • Postoperative pulmonary complications after cardiac surgery are common, especially in patients with COPD, asthma, recent lower respiratory tract infection, severe obesity, or heavy smoking.
    • High-flow nasal oxygen was physiologically attractive because it provides heated humidified oxygen, low-level positive airway pressure, and dead-space washout, and it was increasingly being used after extubation despite limited definitive outcome data.
    • A single-centre pilot RCT in a similar high-risk cardiac surgical population reported shorter hospital stay and fewer ICU readmissions with prophylactic high-flow nasal oxygen. 1
    • A later meta-analysis of 9 mostly small studies suggested less escalation of respiratory support, but not consistent benefit in mortality, reintubation, ICU stay, or hospital stay. 2
  • Research Question/Hypothesis
    • In adults undergoing elective or urgent cardiopulmonary-bypass cardiac surgery who were at increased risk of postoperative pulmonary complications, does prophylactic high-flow nasal oxygen started immediately after extubation improve days alive and at home by 90 days compared with standard oxygen?
    • The protocol anchored this around a patient-centred DAH90 framework, extending a validated perioperative outcome concept into this pragmatic postoperative cardiac surgery setting. 34
  • Why This Matters
    • Routine prophylactic HFNO has cost, equipment, and staffing implications; a definitive neutral trial would be as practice-shaping as a positive one.
    • Before NOTACS, the evidence base was largely physiologic, single-centre, or meta-analytic synthesis of small studies, so confirmation at scale was needed. 12
    • The trial was also designed to inform enhanced-recovery style postoperative pathways rather than merely rescue respiratory support decisions.

Design & Methods

  • Research Question: Whether prophylactic high-flow nasal oxygen therapy (HFNOT), delivered for at least 16 hours immediately after extubation, increases days alive and at home at 90 days without increased support compared with baseline, versus standard oxygen therapy, in adults at high risk of pulmonary complications after nonemergent cardiopulmonary-bypass cardiac surgery.
  • Study Type: Investigator-initiated, adaptive, international, multicentre, open-label, parallel-group randomised clinical trial with concealed allocation and blinded outcome assessment; conducted across 17 cardiac surgical centres in the UK, Australia, and New Zealand, beginning in ICU or postoperative recovery and continuing onto the ward.
  • Population:
    • Adults aged 18 years or older undergoing elective or urgent first-time or redo cardiac surgery performed with cardiopulmonary bypass.
    • Required at least 1 pulmonary-risk criterion: COPD, asthma, lower respiratory tract infection within the previous 4 weeks, body mass index ≥35 kg/m², or current/recent heavy smoking >10 pack-years.
    • Key exclusions: home oxygen; home respiratory support including CPAP or BiPAP; planned deep hypothermic circulatory arrest; contraindication to HFNOT; emergency surgery within 24 hours of the decision to operate; patients not fluent in English. 3
  • Intervention:
    • HFNOT started immediately after extubation.
    • Initial settings: inspired oxygen fraction 40% and flow 30 L/min, increased to 50 L/min over 5-10 minutes.
    • Protocol-mandated minimum duration of 16 hours, with up to 1 total hour off the allocated therapy for transfer around hospital or physiotherapy/mobilisation.
    • Subsequent oxygen titration and escalation followed a prespecified respiratory escalation protocol. 3
  • Comparison:
    • Standard oxygen therapy started immediately after extubation.
    • Typical starting settings: 30%-40% inspired oxygen at 2-6 L/min via nasal prongs or nonrebreathing mask; not heated and not humidified.
    • Same minimum 16-hour treatment window and the same escalation framework if deterioration occurred. 3
  • Blinding: Treating clinicians and participants were unblinded. Outcome assessors and central follow-up staff were blinded; discharge decisions affecting the primary outcome were intended to be made by clinicians independent of the research team. 5
  • Statistics: Adaptive design. Initial power calculation: 850 patients were required to detect a 2-day increase in median DAH90 with 90% power at the 5% significance level after inflating for expected crossover (12% SOT to HFNOT, 25% HFNOT to SOT) and 5% loss to follow-up; a blinded nuisance-parameter sample-size re-estimation after 300 patients increased the final maximum and achieved sample size to 1280. Primary analysis was intention-to-treat using a Mann-Whitney-Wilcoxon test; prespecified per-protocol, time-on-treatment, complete-case, and alternative-death-handling sensitivity analyses were also planned. 356
  • Follow-Up Period: 90 days after surgery, with in-hospital data collection plus discharge, 30-day, and 90-day follow-up. 3

Key Results

This trial was not stopped early.

Outcome HFNOT SOT Effect p value / 95% CI Notes
DAH90 without increased support (primary) Median 0 (IQR 0-79) Median 0 (IQR 0-87) Median difference 0 P=0.75; 95% CI 0 to 0 Primary outcome available for 1224/1280 patients (95.6%). Final endpoint incorporated support level as well as location.
Days alive and out of hospital to day 90 Median 82 (IQR 77-84) Median 82 (IQR 78-84) Not reported Not reported No stated between-group difference.
Days alive and out of hospital to day 30 Median 22 (IQR 18-24) Median 23 (IQR 18-24) Not reported Not reported No stated between-group difference.
Hospital stay Median 7.1 d (IQR 6.0-10.1) Median 7.1 d (IQR 5.9-10.7) Not reported Not reported No LOS benefit reproduced from the earlier pilot study.
ICU stay Median 2.0 d (IQR 0.9-3.9) Median 1.8 d (IQR 0.8-3.6) Not reported Not reported Includes postoperative recovery, ICU, and high-dependency unit time.
ICU readmission 31/640 (4.8%) 27/640 (4.2%) Not reported Not reported No evidence of reduced ICU readmission.
Readmission to hospital 124/598 (20.7%) 128/604 (21.2%) Not reported Not reported Mean readmissions per patient: 0.3 (SD 0.6) vs 0.3 (SD 0.7).
Reintubation 21/625 (3.4%) 15/630 (2.4%) Not reported Not reported No stated between-group difference.
CPAP use 50/625 (8.0%) 59/630 (9.4%) Not reported Not reported No stated between-group difference.
Postoperative pulmonary complications 84/638 (13.2%) 108/637 (17.0%) Not reported Not reported Numerically lower with HFNOT, but authors reported no overall clinical benefit.
Mortality by day 90 14/640 (2.2%) 11/640 (1.7%) Not reported Not reported Median timing of death: 14.5 d (IQR 5.0-43.8) vs 6.0 d (IQR 1.5-18.5).
Major complications Stroke 2/640 (0.3%); sepsis 16/640 (2.5%); AKI 18/640 (2.8%); MI 3/640 (0.5%) Stroke 8/640 (1.2%); sepsis 11/640 (1.7%); AKI 24/640 (3.8%); MI 7/640 (1.1%) Not reported Not reported All were uncommon; no consistent harm or benefit pattern emerged.
    • The primary patient-centred endpoint was completely neutral: median difference 0 days; 95% CI 0 to 0; P=0.75.
    • No country-level or UK-centre subgroup signal was reported; interaction estimates were not reported numerically.
    • Safety was broadly similar. One non-serious adverse event in the HFNOT arm was judged possibly device-related; no definitely or probably device-related adverse or serious adverse events were reported.

Internal Validity

  • Randomization and Allocation
    • Allocation was concealed and web-based, with 1:1 randomisation stratified by centre using variable block sizes of 4 or 6.
    • This provides strong protection against selection bias.
    Drop out or exclusions
    • Four HFNOT patients and 7 SOT patients died before extubation and therefore never received allocated therapy.
    • Primary outcome data were complete for 610/640 HFNOT patients and 614/640 SOT patients, leaving 56/1280 (4.4%) without DAH90 data.
    • This degree of missingness is modest and unlikely to conceal a clinically important benefit given the wholly neutral primary estimate.
    Performance/Detection Bias
    • The trial was open-label for bedside clinicians and patients, so escalation and some aspects of in-hospital care were vulnerable to performance bias.
    • Blinded outcome collection mitigated detection bias, but the support-dependent DAH90 construct is less purely objective than mortality or simple hospital length of stay.
    Protocol Adherence
    • Compliance with allocated therapy occurred in 522/640 (81.6%) HFNOT patients and 498/640 (77.8%) SOT patients.
    • Adherence was acceptable for a pragmatic device trial, but nonadherence and switching would tend to dilute any true treatment effect.
    Baseline Characteristics
    • Groups were well matched: age 62.7 vs 63.2 years, men 70.8% vs 68.6%, ARISCAT 51.8 vs 51.3, EuroSCORE II median 1.6 vs 1.7, first-time surgery 96.3% vs 95.3%.
    • The cohort was enriched for pulmonary risk while operative mortality risk remained comparatively modest, which is appropriate for a DAH-based efficacy question.
    Heterogeneity
    • Seventeen centres across 3 countries inevitably introduce practice variation.
    • Centre stratification at randomisation and the absence of a subgroup signal by country or UK centre reduce concern that the null result is explained by one dominant practice environment.
    Timing
    • Randomisation occurred during surgery or immediately postoperatively before extubation, and therapy began immediately after extubation.
    • This is the mechanistically relevant window if prophylactic HFNOT is to work.
    Dose
    • HFNOT was protocolised at 30 to 50 L/min for at least 16 hours.
    • This tested a realistic routine-prophylaxis strategy, although it does not exclude benefit from longer duration, different flow settings, or symptom-triggered therapeutic use.
    Separation of the Variable of Interest
    • Early physiological separation was present: ROX index medians were 16.84 vs 17.99 at 2 hours (difference -1.19; 95% CI -1.86 to -0.51), 16.96 vs 19.05 at 6 hours (difference -1.98; 95% CI -2.72 to -1.26), 16.67 vs 18.54 at 12 hours (difference -1.73; 95% CI -2.44 to -1.03), 16.96 vs 18.05 at 24 hours (difference -1.40; 95% CI -2.09 to -0.70), and 19.25 vs 19.54 at 48 hours (difference -0.60; 95% CI -1.32 to 0.08).
    • Therefore, lack of effect is unlikely to be explained simply by failure to deliver the assigned therapies at all.
    Key Delivery Aspects
    • The control arm was credible contemporary standard care rather than an artificially weak comparator.
    • Both groups followed the same prespecified escalation pathway, which strengthens fairness of comparison.
    Outcome Assessment
    • Mortality, ICU readmission, reintubation, and lengths of stay were objective.
    • The primary endpoint depended on patient diaries and adjudication of residence plus support status, which introduces more scope for approximation or misclassification than simple hard endpoints.
    • The co-primary outcome of health economic analysis was not reported in this manuscript.
    Statistical Rigor
    • The protocol, original SAP, and updated pre-lock SAP were all published in advance, and the trial used an independent blinded nuisance-parameter re-estimation to preserve type I error. 356
    • The main caveat is that the final primary endpoint differed from the original DAH90 construct used for trial planning and interim re-estimation. 56

Conclusion on Internal Validity: Internal validity is moderate to strong. Concealed randomisation, balanced groups, high follow-up completeness, and prepublished analysis plans support the result, while open-label care, moderate nonadherence, and the mid-trial refinement of the primary endpoint are the main tempering factors.

External Validity

  • Population Representativeness
    • The cohort was representative of a selected high-risk adult cardiopulmonary-bypass cardiac surgery population in large tertiary centres: mean age 62.9 years, BMI 32.1-32.6, ARISCAT approximately 51, and substantial prevalence of obesity, asthma, COPD, and smoking.
    • Most patients underwent first-time and nonemergent surgery, so the findings are less informative for emergency surgery, salvage cases, or patients already dependent on chronic oxygen or home noninvasive support.
    Applicability
    • The results apply directly to routine prophylactic postextubation HFNOT in comparable high-income cardiac surgical systems with early extubation pathways and ready access to HFNO equipment.
    • They are less generalisable to off-pump surgery, noncardiac surgery, patients with established postoperative hypoxaemia being treated rather than prevented, or settings without robust follow-up infrastructure.
    • The protocol exclusion of non-English-speaking participants also narrows inclusiveness. 3

Conclusion on External Validity: External validity is good for the specific question of routine prophylactic HFNOT after nonemergent adult cardiopulmonary-bypass cardiac surgery in patients at increased pulmonary risk. It should not be over-extended to rescue HFNO use, emergency surgery, or markedly different health-system contexts.

Strengths & Limitations

  • Strengths:
    • Largest randomised trial in this niche, with 1280 participants across 17 centres in 3 countries.
    • Pragmatic multinational design with concealed allocation, centre stratification, and blinded outcome assessment.
    • High follow-up completeness for the primary endpoint at 95.6%.
    • Patient-centred 90-day follow-up rather than short-term physiologic end points alone.
    • Protocol and SAPs were published prospectively, and adaptive re-estimation protected against an underpowered final trial. 356
  • Limitations:
    • Open-label bedside care and moderate nonadherence/crossover.
    • The final primary endpoint was refined during the trial, making comparison with prior DAH literature and with the interim planning metric less straightforward. 46
    • The modified DAH90 was highly zero-inflated, giving a median of 0 in both groups and reducing intuitive clinical interpretability.
    • The trial was not powered for small differences in rare events such as stroke, myocardial infarction, or death.
    • The prespecified economic evaluation, a co-primary outcome, was not reported in the index publication. 7

Interpretation & Why It Matters

  • Clinical practice
    NOTACS does not support routine prophylactic HFNOT after extubation for all adult cardiac surgical patients at increased pulmonary risk when the aim is to improve DAH, reduce length of stay, avoid ICU readmission, or prevent reintubation.
    How it changes prior assumptions
    The trial materially weakens the earlier optimistic narrative created by physiology, a positive pilot study, and small-study meta-analyses. In this specific postoperative cardiac surgery context, plausible respiratory mechanics did not translate into better patient-centred outcomes. 12
    Research implications
    Future studies, if pursued, should probably focus on rescue treatment of established postoperative hypoxaemia, biologically plausible subgroups, or different exposure intensity and duration, while preserving simple and reproducible patient-centred endpoints.

Controversies & Other Evidence

  • Primary endpoint refinement
    • The main methodological controversy is the change from conventional DAH90 to DAH90 without increased support compared with baseline.
    • This was prespecified before database lock and justified by the investigators on the basis of baseline abode, patient/public input, and the desire to distinguish health-related from non-health-related time away from home.
    • Even so, it makes comparison with the original validated DAH construct less straightforward, and the interim sample-size re-estimation was based on the earlier definition. 456
    Why NOTACS differed from earlier optimism
    • The pilot study found a 2-day shorter median hospital stay and ICU readmission 1/49 vs 7/45 in favour of HFNO, while the 2024 meta-analysis suggested reduced escalation of respiratory support (RR 0.67; 95% CI 0.48 to 0.93).
    • NOTACS, however, was much larger, multicentre, and explicitly patient-centred, and found no meaningful benefit across primary or major secondary outcomes. The most likely interpretation is that earlier evidence overestimated benefit because it was smaller, more heterogeneous, and more vulnerable to centre effects and random error. 12
    Relationship to existing HFNO guidance
    • Broader ACP and ERS guidance supported HFNO in selected acute respiratory failure and postextubation scenarios.
    • Those recommendations were not cardiac-surgery-specific and should not be extrapolated to justify routine prophylactic use after cardiac surgery now that NOTACS is available. 89
    Economic interpretation remains incomplete
    • NOTACS prespecified co-primary economic analyses using both DAH and QALYs, but the JAMA report presented clinical results only.
    • Given the lack of clinical benefit, routine adoption is already difficult to justify, but a full cost-effectiveness report remains relevant for policy and resource allocation. 7
    Subsequent or parallel evidence
    • A 2026 retrospective cohort in patients with established post-cardiac-surgery hypoxaemia reported better oxygenation, shorter ICU and hospital stay, and lower reintubation with HFNC.
    • That study addressed treatment of hypoxaemia rather than universal prophylaxis, and its observational design leaves major residual confounding; it does not overturn NOTACS. 10

Summary

    • NOTACS randomised 1280 adults at increased pulmonary risk after nonemergent cardiopulmonary-bypass cardiac surgery to HFNOT or standard oxygen for at least 16 hours after extubation.
    • The primary endpoint was neutral: median DAH90 without increased support was 0 in both groups; median difference 0; 95% CI 0 to 0; P=0.75.
    • There was no convincing benefit in days out of hospital, hospital stay, ICU stay, ICU readmission, reintubation, CPAP use, readmission, postoperative pulmonary complications, quality of life, or Barthel function.
    • Internal validity is good but not pristine, limited chiefly by open-label delivery, moderate nonadherence, and refinement of the primary endpoint during the trial.
    • The best practical reading is that routine prophylactic HFNOT should not become standard care after adult cardiac surgery solely to improve postoperative outcomes.

Overall Takeaway

NOTACS is an important effectiveness trial. Its value lies in testing a highly plausible and increasingly used postoperative strategy at adequate scale and showing that, in high-risk adult cardiac surgery, routine prophylactic HFNOT does not improve patient-centred recovery or major clinical outcomes compared with standard oxygen therapy.

Overall Summary

  • Routine prophylactic HFNOT after high-risk adult cardiac surgery did not improve DAH90, length of stay, ICU readmission, reintubation, or other major outcomes.
  • The trial’s scale and methodology make it more credible than the earlier small positive studies in this field.
  • After NOTACS, HFNO after cardiac surgery should be selective rather than default.

Bibliography