Medical-AI trial: predicting the success of a spontaneous breathing trial with machine learning — ICU ventilator weaning from biosignals and biomarkers (NCT05886803)
An observational study of whether machine learning can predict, from biosignals and biomarkers, the success of the spontaneous breathing trial (SBT) used to test weaning from a ventilator in the ICU. Weaning takes up much of an ICU stay, and the first attempt reportedly fails in about 20%. Target 500; recruiting.
Trial overview (primary data)
- StatusRecruiting
- ConditionsWeaning From Mechanical Ventilation in Care Unit
- InterventionsOTHER: Spontaneous ventilation test
- SponsorCentre Hospitalier Universitaire de Nice
- Target enrollment500 participants
- Period2023-01-01 〜 2027-06-25
Key points
- An observational study of whether ML can predict, from biosignals and biomarkers, the success of the spontaneous breathing trial (SBT) for ICU ventilator weaning.
- Weaning can take up to half of an ICU stay; the first attempt is unsuccessful in about 20% of patients.
- Mortality can reach 38% in the most difficult cases; weaning too early risks reintubation, too late invites complications and longer stays.
- Framed as one example of AI decision support for clinicians (alongside septic-shock management and renal-failure prediction).
- Target 500; recruiting. A predictive-performance evaluation, not a decision on extubation or an establishment of clinical use.
A critically ill patient who cannot breathe well on their own is supported by a ventilator. As they recover, the ventilator is withdrawn (weaning), but judging the right moment is hard. Weaning too early can force reintubation, adding burden and risk, while weaning too late invites ventilator-associated complications and a longer stay.
Weaning indeed occupies much of an ICU stay; per the summary, the first attempt fails in about 20% of patients, and mortality can reach 38% in the most difficult cases.
So this study examines, as an observational study, whether the success of the spontaneous breathing trial (SBT, in which ventilator support is reduced to see how well the patient breathes on their own for a set time) can be predicted in advance with machine learning. Per the registry summary, it uses biosignals such as heart rate and respiration and biomarkers from blood as cues to estimate whether the SBT will succeed.
As research applying AI to clinical decision support grows, from managing septic shock to predicting renal failure, this study focuses that lens on the weaning decision. The target is 500, and it is recruiting.
Weaning decisions rely heavily on experience and can vary between institutions and clinicians. Estimating SBT success in advance from objective measures could help reduce reintubation and unnecessary delays and allocate scarce critical-care resources more appropriately. That said, this is a study evaluating predictive performance; it does not decide actual extubation or establish clinical use.
Why it matters
Estimating in advance, from objective measures, a weaning decision that otherwise leans on experience could help reduce reintubation and unnecessary delays and allocate scarce critical-care resources. It offers a window into how ICU AI decision support is being validated (this trial evaluates predictive performance, not clinical use).
FAQ
What is a spontaneous breathing trial (SBT)?
Does the AI decide extubation?
Sources (primary)
Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.
- ClinicalTrials.gov (study record, original)
- NCT ID: NCT05886803