Introduction
Post-extubation stridor associated with post-extubation laryngeal oedema is one of the most frequent causes of reintubation in the intensive care unit (ICU) [
1‐
7]. Reintubation may result in increased morbidity (for example, nosocomial infection, prolonged length of ICU stay, additional costs) and mortality [
1‐
4,
6,
7]. The prevalence of post-extubation stridor ranges between 6 and 37% of intubated ICU patients [
5,
8‐
13], depending on the studied population (those at high risk or not). Controversy still exists about the effectiveness of prophylactic steroid therapy to prevent occurrence of both post-extubation stridor and related reintubation in both patients selected because they are at high risk of stridor and reintubation [
8,
9,
13] and non-selected patients [
10‐
12,
14].
Two recent meta-analyses [
15,
16], based on original papers published up to 2007, have been performed. They report contradictory conclusions regarding the efficacy and safety of prophylactic steroid therapy in preventing post-extubation laryngeal oedema and the need for reintubation in adult ICU patients. Fan and colleagues [
15] have suggested, regarding the most recent clinical trials, that prophylactic steroid therapy can reduce the incidence of post-extubation laryngeal oedema and the subsequent need for reintubation in mechanically ventilated patients. In contrast, Markovitz and colleagues [
16] concluded that using steroids to prevent (or treat) stridor after extubation has not proven effective for neonates, children or adults. By reporting conflicting results, recent trials [
8,
9,
13] and the two meta-analyses [
15,
16] have intensified the debate surrounding the use of prophylactic steroid therapy to prevent both post-extubation stridor occurrence and reintubation. Moreover, the meta-analyses results were pooled from trials which included selection of patients at risk of post-extubation stridor development [
8,
9,
13] and unselected patients with an no risk of post-extubation stridor development [
10‐
12,
14] and allowed for very different steroid administration regimens (well in advance of extubation or immediately before). Indeed, the anti-inflammatory effect of steroids, the main mechanism responsible for reduction of post-extubation laryngeal oedema, is time-course dependant [
17,
18]. Although the two meta-analyses [
15,
16] allowed for these differences, they did not perform subgroup analyses of the early versus late steroid administration nor for selected high-risk patients versus unselected patients. Finally in 2007, two additional randomised clinical trials (RCTs) were presented in abstract form but were not included in these two meta-analyses [
8,
14]. Thus, we performed a quantitative meta-analysis to evaluate the effectiveness of prophylactic steroid therapy to prevent reintubation and post-extubation stridor, taking into account the studied populations (at risk to develop post-extubation stridor or not) and the steroid administration regimen (pre-extubation early versus late).
Discussion
The present meta-analysis documents that steroid administration before a planned extubation decreases the risk of post-extubation stridor and reintubation both in high-risk and unselected patients. The beneficial effect of steroids to prevent post-extubation stridor and reintubation was clear in the subgroup of patients at high-risk for development of post-extubation stridor as identified by a cuff-leak test (a low level of leak less than 110 ml or less than 25%).
The discrepancies observed in studies that evaluated the interest to administer steroids before extubation could be due to several factors including patient inclusion criteria, duration of intubation, dosage, timing of treatment and risk levels of developing stridor. Only the last two criteria (risk levels of developing stridor and timing of administration initiation) could be extensively evaluated in the present meta-analysis, allowing their importance to be reported for the first time. Post-extubation stridor is commonly the result of oedema of the subglottic area or the vocal cords. The difficulty in defining the relationship between laryngo-tracheal injury and post-extubation stridor is that the presence of the endotracheal tube precludes direct visualisation of the upper airway before extubation.
The ability to predict which patients will develop stridor following extubation, possibly culminating in reintubation, is obviously a desirable goal. Beyond assessment of risk factors, clinicians have long used the cuff-leak test to predict post-extubation airway patency, wherein the endotracheal tube cuff is deflated and a leak of air around the tube is sought during either spontaneous ventilation (with the endotracheal tube lumen occluded) or positive-pressure ventilation. The cuff-leak test may be performed using the 'qualitative method' (presence or absence of air leak around the tube when the cuff is deflated) or the 'quantitative method' by reporting the leak volume (inspired minus exhaled tidal volume during positive-pressure ventilation when the cuff is deflated) or the fraction of leak volume (inspired minus exhaled volume divided by inspired tidal volume when the cuff is deflated). Several cuff-leak test studies [
5,
9,
27‐
30] suggest that the presence of an air leak is associated with a low likelihood of clinically important post-extubation stridor, whereas the absence or a low level of leak (less than 110 to 140 ml in absolute value or less than 12% to 25% in relative value) is associated with a high incidence of stridor and reintubation. The use of the cuff-leak test should be standardised and take into account a possible discrepancy between inspired and exhaled tidal volume measurement devices together with significant breath by breath variability.
A more reliable identification of patients at high-risk of developing post-extubation stridor and reintubation would appear desirable not only to decrease the risk of reintubation, but also to avoid excessive steroid treatment as it may induce adverse effects in patients for whom there is no need. Indeed as shown in the present meta-analysis, the NNT to prevent one stridor episode decreased from 11 in the overall population (selected and unselected) to five in a population determined to be at high-risk of developing post-extubation stridor as determined by the cuff-leak test (Figure
3 and Table
2). However, steroids did not significantly reduce the incidence of post-extubation stridor when patients were not selected (that is, unselected patients) for their risk of post-extubation stridor. The NNT to avoid one reintubation decreased from 28 in the overall population (selected and unselected) to nine in patients at high risk (Figure
2 and Table
2). On the other hand, the benefit of steroids is unclear when trials did not use the cuff-leak test to selected patients. In this case, the NNT increased to 44 and the upper limit of the CI is infinity (Figure
2 and Table
2).
Although steroids are potentially associated with several adverse effects (such as hyperglycaemia, arterial hypertension, agitation and infection) when they are administered for a few days (more than 48 hours) [
31], side effects associated with steroid treatment for less than 24 hours are minimal [
17,
18]. The studies included in the present meta-analysis reported no side effects related to steroids, but detection of steroid-related adverse events was not specifically studied in these trials.
Laryngotracheal injury related to intubation may cause narrowing of the airway mainly due to inflammatory oedema. The potential capacity of steroids to relieve laryngeal oedema is mainly due to its anti-inflammatory effects, which inhibit the release of inflammatory mediators and decrease capillary permeability [
9,
11,
13,
18]. The initial anti-inflammatory effects start at least one to two hours after intravenous administration and maximal effects appear between 2 and 24 hours, depending on steroid type and administered dose [
9,
11,
17,
18]. Indeed, a single injection of dexamethasone (1 mg/kg) one hour before extubation had no effect on subglottic histological injury in a rabbit model [
32,
33]. Moreover, in the two trials [
10,
12] included in the present meta-analysis in which steroids were administered one hour before extubation, no significant difference was observed between control and steroid groups for post-extubation stridor and reintubation rates. The same is true for the study by Gaussorgues and colleagues [
25] for which steroids were also administered one hour before extubation and no significant difference was observed between control and steroid groups for post-extubation stridor and reintubation rates. Although the study by Gaussorgues and colleagues [
25] was excluded because the quality assessment score was less than three, the inclusion of this study [
25] would not change the conclusions of the present meta-analysis. Except for one trial presented in abstract form at a congress [
14], all the published RCTs in which steroids were administered at least 4 to 24 hours before extubation (Table
1 and Figures
5 and
6) reported a significant decrease in post-extubation stridor [
8,
9,
11,
13] and reintubation [
8,
9,
11].
It might be argued that the use of corticosteroids in adult critical care for planned extubation is unnecessary, because objectively the incidence of reintubation is low and symptomatic laryngeal oedema has self-limited symptoms. However, stridor and laryngeal dyspnoea increase care needs because of the administration of adrenaline or corticosteroid aerosol and associated nursing time. Similarly, reintubation increases cost, morbidity, care needs, and both ICU and hospital lengths of stay. Unfortunately, trials included in the current meta-analysis evaluated the benefit of corticosteroids only during the first 48 hours and no information on the outcome of reintubated patients was provided. Further studies on this topic are needed; using standard criteria for the assessment of readiness to extubate and a well-defined evaluation on the relation between post-extubation laryngeal oedema and re-intubation.
The quality of the trials included in a systematic review may alter the results [
34], because meta-analyses are often handicapped by the heterogeneity of the included trials. Moher and colleagues [
34] demonstrated that meta-analyses with low-quality trials (Jadad assessment scale of two or less) compared with high-quality trials (Jadad assessment scale above two) were associated with a 33% increase in the estimated benefit. Similarly, trials using inadequate allocation concealment may also overestimate the benefit of treatment by as much as 37% [
34]. Therefore, multiple scales have been proposed to assess the quality of trials included in a meta-analysis in order to decrease bias due to the inclusion of low-quality trials. We used the Jadad composite scale [
20] to assess quality, using the following items: randomisation, double-blinding and patient withdrawals.
Meta-analyses of trials with low quality, as evaluated with this scale, significantly exaggerate benefits [
19,
34]. All seven trials selected for our systematic review have a scale reflecting high quality [
34] and, consequently, were double-blinded and randomised. Patients included in trials have variable risks for post-extubation stridor or reintubation. Interestingly, the reduction of risk for stridor appears to be similar (approximately 50%), regardless of the risk of post-extubation laryngeal dyspnoea, suggesting that the effect is the same in the presence of oedema. Dosage, duration and type of corticosteroids differed from one trial to another. Pooling RCTs with varying designs may be interesting because the current meta-analysis appears to demonstrate that the timing of the first administration influences the risk of reintubation.
The current meta-analysis suggests an effect of administration timing on the efficacy of corticosteroids, because steroids appear to prevent reintubation more effectively if they are administrated at least four hours before planned extubation. As stridor and reintubation, secondary to upper obstruction airway obstruction, occur soon after extubation [
5,
11], it may be reasonable to suggest starting steroid treatment at least four hours before planned extubation to prevent prolongation of weaning from mechanical ventilation.
Further studies should be conducted to better define the optimal use of steroids to prevent extubation failure. In patients selected at high risk for postextubation stridor (for example, traumatic intubation, low cuff-leak value or previous extubation failure) steroids should be used but the optimal steroid to use before extubation without delay remains to be established, as does steroid type, dosing regimen, administration timing and duration. Dose response should also be established to achieve the lowest effective dose. Moreover, the risk of steroid use remains a source of concern in critical care patients. The side effects of steroid administration to prevent reintubation are unknown and were not investigated clearly in all trials included in this meta-analysis. The current meta-analysis showed no benefit when trials that did not select patients at risk for reintubation were pooled. In this group, only one trial [
5,
11] found a significant benefit of steroid use but the others found no benefit. The study by Francois and colleagues [
11] appears to be the main cause of heterogeneity between the trials that did not select patients at risk. The timing of administration does not seem to be the major reason for heterogeneity because the study by Shih and colleagues [
14] administrated steroid sooner than Francois and colleagues [
11] (24 hours compared with 12 hours, respectively). Another hypothesis may be the dose of steroid used by Francois and colleagues [
11] because they administrated the highest dose among all trials studied. Finally, all trials have the possibility of giving a significant result even if one is not available (Type I error). Thus, the evidence for steroid administrated in unselected patients remains unclear and additional studies are warranted to clearly determine the benefits, but also the potential adverse effects, of this group of drugs.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SJ designed and supervised the research, collected, analysed and interpreted the data, drafted and revised the manuscript. BJ contributed to the conception of the study and approved the final version of the manuscript. GC made substantial contributions to the conception and design of the study and approved the final version of the manuscript. FB participated in the design of the study and helped to draft the manuscript. EM co-designed and supervised the research, collected and analysed the data and performed the statistical analysis. All authors read and approved the final manuscript.