A New Approach to Predicting Extubation Failure?

Abstract & Commentary

Synopsis: This single-center prospective observational study reveals that the presence of 3 factors (low cough peak flow, high secretion volume, and poor neurologic score per a simple 4-task test) may be useful in predicting extubation failure.

Source: Salam A, et al. Intensive Care Med. 2004; 30(7):1334-1339.

Though a spontaneous breathing trial (sbt) may demonstrate readiness for liberation from ventilator support, it does not provide information about continued need for an artificial airway. There are limited and conflicting data on the utility of assessment of cough, secretions and neurologic status (by Glasgow Coma Scale [GCS]) in predicting extubation outcome.1,2 Salam and colleagues address whether combination of a different neurologic score with quantitative measures of cough strength and secretion volume may be useful in predicting extubation outcome.

An 11-month prospective observational study was performed in a Connecticut medical-cardiac intensive care unit (ICU). All patients receiving invasive mechanical ventilation were eligible after passing a 30-60 minute protocol-guided SBT. Patients with tracheostomies were excluded from the study. Extubations that were part of life-support withdrawal were also excluded. Demographics, APACHE II scores and a variety of respiratory variables were gathered for all study patients. Cough peak flow (CPF) was recorded using a pneumotachograph-calibrated Aztech peak flow meter placed in series with the endotracheal tube. Endotracheal secretions were suctioned and quantitated hourly for 2-3 hours before anticipated extubations. Finally, a simple 4-task neurologic score (1 point per task) was assigned by asking the patient to: open eyes, follow observer with eyes, grasp hand and stick out tongue. Patients’ caregivers were blinded to the results of these tests. Decision to extubate was per the attending physician. Patients who remained extubated at 72 hours were classified as "successful extubations." Standard statistical analyses were employed.

Eighty eight patients (mean age, 62 years; mean APACHE II score, 24) with 100 extubations were enrolled. Median time of intubation was 4 days. Pneumonia was the most common reason for intubation. Fourteen (15.9%) patients failed the first extubation, with hypoxia and increased work of breathing being the most common reasons. There was no statistically significant difference in age, gender, duration of intubation or secretion volume between patients successfully extubated and those who failed. CPF and neurologic score were both significantly worse in patients with extubation failure. Neurologic score 0/4 was independently associated with extubation failure (RR, 3.2; CI, 1.6-6.1). Severity of illness was also significantly worse in patients with extubation failure (median APACHE II 28 vs 23). Patients with all 3 risk factors (CPF = 60L/min, secretions > 2.5 mL/hr, neurologic score 0/4) had a 100% extubation failure rate. There was no difference in results between analysis of first extubations vs both first and repeat extubations.

Salam and colleagues conclude that quantitative measures of cough strength and secretions may be more useful than qualitative ones in predicting extubation outcome. They also propose that a neurologic score such as theirs may be more predictive than GCS of extubation outcome. They suggest replication of their study in other centers and patient populations.

Comment by Saadia R. Akhtar, MD, MSc

Intensivists clearly understand that in the daily care and assessment of intubated, mechanically ventilated ICU patients, the continued need for ventilator support should be considered separately from the continued need for an artificial airway. The best approach to the latter though remains unclear.

Salam et al’s report adds some data to the limited literature on this topic and offers interesting new hypotheses that deserve to be explored further. Their overall results suggest that quantitative rather than qualitative measures of ability to protect the airway may have reasonable predictive value for extubation outcome. It is still quite surprising that such a small volume of secretions (> 2.5 mL/hr) would be significant, even in combination with other factors. Alone, the volume of secretions did not differ between patients successfully extubated and those who failed, suggesting this should not be a very important or reliable predictor. Further study must follow to clarify this. One of the most intriguing observations from this report is that a neurologic score that is more specific for abilities required for airway protection may be more useful than a usual measure such as the GCS. Only reproduction of these findings in other studies will reveal whether Salam et al’s 4-task score is adequate. Greater consideration may need to be given to what neurologic features are required for airway control and maintenance and how to measure them before concluding that this is not a valid predictor of extubation outcome.

Though some of the findings in this report conflict with those of prior studies, they do not detract from those data. The patient populations evaluated and the aims of the studies are quite different and simply cannot be directly compared. Coplin and colleagues described timing of extubation in patients with acute brain injury as well as the impact of delayed extubation on ICU and hospital lengths of stay and pneumonia incidence.1 They examined GCS and airway issues only as secondary outcomes: they found neither had significant association with extubation result in their cohort. Namen and colleagues’ aim was to evaluate a respiratory-therapy driven weaning protocol in neurosurgical patients.2 Their findings were similar to those of Coplin et al: they observed significant extubation delays. Post-hoc analysis suggested that higher GCS was associated with increased likelihood of extubation success but this was not further assessed or independently validated. Both of these studies emphasize the fact that our current approaches to assessing readiness for extubation have considerable weaknesses and that relying on them may lead to unnecessary extubation delays. Neither, though, provides the ultimate solution.

I believe that all 3 of these publications present compelling hypotheses that, as a first step, must be confirmed in larger prospective studies of similar patient populations. The next step then will be to apply and validate the approach across other patient populations. Perhaps then we will have a tool for predicting extubation success that is at least equivalent to the SBT for liberation from mechanical ventilation: a tool with reasonably good (but far from 100%) negative and positive predictive value that must be accompanied by a disclaimer reminding us that a failed extubation rate of 0% means we may be delaying extubation unnecessarily and bringing harm to some of our patients.


1. Coplin WM, et al. Am J Respir Crit Care Med. 2000:161:1530-1536.

2. Namen AM, et al. Am J Respir Crit Care Med. 2001:163:658-664.

Saadia R. Akhtar, MD, MSc, Pulmonary and Critical Care Medicine Yale University School of Medicine, is Associate Editor for Critical Care Alert.