The extubation of patients who have required invasive mechanical ventilation (MV) is common practice in Intensive Care Units (ICUs), but is not without problems. Between 25 and 40% of all patients develop signs of breathing difficulty after extubation.1–3 Extubation failure, defined as the need for re-intubation in the first 48h after extubation, occurs in 5–20% of all patients, depending on the population studied.4 In this context, the risk is greater in clinical and in neurological patients.

Re-intubated patients suffer greater in-hospital mortality and longer stays in the ICU and in hospital, with an increased need for tracheotomy, and often require greater long-term medical care.5–8 Avoidable delays in extubation prolong the stay in the ICU, increase the risk of pneumonia and in-hospital mortality,9 and generate total and daily hospital costs that double the costs generated by patients who are successfully extubated.10

A review fundamented upon medicine, based on evidence, and identified over 50 objective physiological variables (weaning predictors) as tools for evaluating the degree of preparation for spontaneous breathing among patients on MV. Of these weaning predictors, only 5 (negative inspiratory force, minute ventilation (Vm), respiratory frequency (Fr), tidal volume (Vt) and Fr/Vt ratio) were significantly associated to the probability of weaning success or failure — though their predictive capacity was not very high.11–13

The dead space (Vd) is the component of ventilation that is wasted, since it does not participate in gas exchange.14 The dead space fraction (Vd/Vt) has already been used in the identification of pulmonary thromboembolism,15,16 in the management of chronic obstructive pulmonary disease (COPD),17 in the determination of pulmonary dysfunction in adults with sepsis,18 and as a discriminant factor in weaning from MV among children.19,20

The main objective of the present study was to analyze the usefulness of Vd/Vt as a predictor of extubation success in patients subjected to MV in the ICU.

A prospective, observational cohort study was carried out.

Study population

The study population comprised patients admitted to the general ICU (GICU) of a third-level hospital (Marques of Valdecilla University hospital, Santander, Spain) and who met the following inclusion criteria: (a) MV for over 12h; (b) a weaning process following a low-level pressure support protocol in continuous positive airway pressure-assisted spontaneous breathing (CPAP+ASB) mode with an Evita 4 ventilator. The exclusion criteria were: age under 18 years, tracheostomy ventilation, and patients considered to be unable to collaborate for different reasons.

During the study period, a total of 392 patients were admitted to the GICU. Of these, 214 required MV, and weaning was started in 154 cases. We excluded 54 patients: 14 subjects who did not require MV for more than 12h, and 40 in which weaning did not involve a low-level pressure protocol. Of the 100 patients who started weaning, 24 failed to pass the test and were not extubated from MV. A total of 76 patients were therefore finally studied (Fig. 1).

Figure 1.

Flow chart showing the distribution of the patients eligible for inclusion in the study.

(0.18MB).

Registered variables. Analyzed variables

For each patient we recorded personal and demographic data such as age and sex. The disease leading to the need for MV was registered, along with personal antecedents of chronic obstructive pulmonary disease (COPD) and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score of the patient. Patient height was measured, and the ideal body weight was calculated using the Devine method:

• Ideal male weight=50kg+0.91 (height [cm] – 152.4)

• Ideal female weight=45.5kg+0.91 (height [cm] – 152.4)

Before the CPAP+ASB test, we recorded the patient heart rate (HR), SBP and DBP, and SaO2 (measured by pulsioximetry) and the Fr.

The parameters used during the test were registered: inspiratory oxygen fraction (FiO2), PEEP, ASB, and its duration in minutes. When the test was considered successful, we recorded the values corresponding to peak pressure, Vm, Vt and the expired CO2 volume (VCO2), obtained from the screen of the Evita 4 ventilator.

After completing the test we again measured the patient HR, SBP and DBP, SaO2 (measured by pulsioximetry) and Fr, as well as the arterial oxygen pressure (PaO2) and the arterial carbon dioxide pressure (PaCO2).

All the variables were entered in a datasheet, with calculation of the dead space fraction (Vd/Vt) based on the following equation:

where PaCO2 is obtained from the arterial blood gas measurements, and PeCO2 (carbon dioxide pressure in exhaled air) is calculated from the data obtained by the CO2 sensor of the Dräger respirator.

Accordingly: PeCO2=FeCO2×Patm

FeCO2 (carbon dioxide fraction in exhaled air) is calculated from the following equation:

The values of VCO2 and Vm are obtained directly from the respirator screen. Atmospheric pressure (Patm) was taken to be a fixed value of the subtraction 760–47mmHg (the water vapor pressure value): 713mmHg. With the data obtained we also calculated the ratio between Fr and Vt in each patient, as well as the PaO2/FiO2 ratio.

All the variables were recorded by medical personnel blinded to the Vd/Vt value obtained at the time of deciding both extubation and re-intubation, where required.

The most important conclusion of the present study is the definition of the dead space fraction (Vd/Vt) as a potent predictor of mechanical ventilation as an extubation success.

However, the analysis made has a series of limitations. The main limitation may be the heterogeneity of the study sample, based on the disease leading to the indication of MV. Nevertheless, we consider that a non-selected population of patients in the GICU is the best way to reflect the daily clinical activity of many ICUs in our setting.

Two important aspects should be considered in this study. One aspect is the use of a low-level pressure support protocol. This choice was based on the fact that the latter afforded simple breathing, with unlimited flow upon demand, and allowed control of the patient inspiration index. Likewise, this type of protocol facilitated data measurement and registry, fundamentally of the variables obtained from the capnography sensor. We consider that the pressure support mode is comfortable for the patient and facilitates patient–ventilator synchronization. In addition, it is recognized as one of the weaning methods offering the best results.2–4

The second aspect requiring mention is the percentage of extubation failure observed in our series. In effect, the recorded 22% may be considered too high. However, as was commented in “Introduction” section, the poorest results correspond to the clinical and neurological cases, and these were the largest groups in our cohort. Moreover, it must be pointed out that extubation failure was considered not only when re-intubation proved necessary but also when there were clear and strict indications to the effect. In a meta-analysis of 2486 patients, the proportion of cases of failed weaning was found to be 31.2%, with a 13% re-intubation rate.21

The fact that Vd/Vt was seen to be a good predictor of extubation failure can be explained by its behavior as an indicator of the possible imbalance between the demand and the actual ventilation capacity of the patient. Ventilation demand depends not only on the ventilation requirements, but also on the respiratory system of each patient. Thus, when the MV needed to maintain normal alveolar ventilation exceeds the capacity of the patient, weaning will not be satisfactory.

These ventilation requirements are determined by VCO2, the dead space, and the respiratory stimulus at central nervous system level. Since Vd/Vt in a single parameter comprises two of the components that determine the ventilation requirements, its efficacy in predicting extubation failure can be explained.

However, the work of breathing (WOB) as another of the main determinants of ventilation demand or load has not been weighted in the measurement of Vd/Vt. The mechanical characteristics of the respiratory system, airway resistance and thoraco-pulmonary compliance determine WOB through its two components: resistive work and elastic work. In other words, situations of bronchospasm, upper airway obstruction, bronchial hypersecretion and certain characteristics of the external circuits of the ventilator that imply an important increase in resistive work by raising airway resistance are not contemplated by the Vd/Vt based prediction. The same applies to situations involving an increase in the elastic component of WOB (loss of compliance, effusion, pneumothorax, etc.).22

In the medical literature, the first investigations in relation to Vd/Vt date back to 196523 with studies of its role as a marker of acute lung injury (ALI). More recently, in 1983, Pierson24 identified values of Vd/Vt>60 as indicators of intubation. However, despite these old studies, Vd/Vt traditionally has not been included in the clinical guides as a parameter to be considered in the extubation of adults subjected to MV. This may be attributed to the complex task of collecting the measurements needed to calculate the fraction, before technological advances simplified this work by introducing new ventilator and sensor measurement and calculation techniques.

Recently,25 emphasis has been placed on the role of Vd/Vt in predicting extubation. Based on a cohort of 35 patients, it has been postulated that the highest Vd/Vt values measured on the first day of hospitalization could be an early predictor of extubation failure. In this case, the sensitivity, specificity, positive predictive value, negative predictive value and a Vd/Vt cutoff point of ≥0.60 in predicting extubation failure according to the ROC curve were 70%, 72%, 58%, 81% and 71%, respectively.

In this same sense but with a different working philosophy we have the study published by Farah et al.,26 which attempts to clarify the effect of Vd/Vt in the weaning from MV of patients with COPD. On analyzing their registry of 49 patients, the authors concluded that Vd/Vt does not afford a reliable value in estimating the duration of MV in patients with acute respiratory failure due to COPD.

The studies with a methodology and physiological basis more akin to those of this article have been carried out in the pediatric setting. In this sense, Hubble et al.,20 in an analysis of 45 pediatric patients, found Vd/Vt values of ≤0.50 calculated 30min before extubation to be associated with success, while Vd/Vt>0.65 was identified with the need for additional respiratory support after extubation. For a cutoff point of Vd/Vt=0.50, the study showed a sensitivity of 0.75% and a specificity of 0.92%.

In contrast to the above are the results published by Bousso et al.19 in another pediatric study. In this case the Vd/Vt value obtained in the analysis of the ROC curve was found to be poor as a predictor of extubation failure (with an area under the curve (AUC) of 0.62). Using a cutoff point of 0.65 yielded no significant differences in the correlation of this point with the incidence of extubation success or failure, as established from the logistic regression analysis.

However, there are important differences between the two studies commented above, on considering the populations involved. The population in the study by Bousso et al. was younger (mostly neonates) and with greater lung damage prior to extubation, compared with the study published by Hubble et al. Accordingly, the younger the patients, the lesser their resistance to muscle fatigue, cough proves less effective, the patients are less able to eliminate secretions, and their airway is moreover narrower. All these characteristics are not quantified in the measurement of Vd/Vt.

Lastly, Mohr et al.27 studied Vd/Vt 24h before and after tracheotomy in 45 adults admitted to a surgical ICU. The authors concluded that MV weaning success or failure could not be predicted on the basis of the Vd/Vt value, since Vd/Vt did not vary following tracheotomy. In relation to this study, it must be mentioned that the patients were ventilated in intermittent mandatory ventilation (IMV) pressure support mode from the planning of tracheotomy, and that the values for calculating Vd/Vt were registered while the patients were being ventilated with a minimum prefixed Vt adjusted to body weight, and with pressure support according to the needs of the patient. We consider that the fact that the patients received ventilatory support apparently sufficient to maintain adequate alveolar ventilation may have masked the Vd/Vt values obtained in this study. Nevertheless, we consider that this study should be taken into account in that its conclusions are opposed to the data obtained in our own work.

On comparing the results obtained for Vd/Vt with the multiple predictors analyzed in the literature, we could rank the dead space fraction at a level comparable to the CROP index (compliance, respiratory rate, oxygenation and pressure) and the IWI (integrative weaning index). In this sense, the CROP index, developed by Yang and Tobin in a prospective study with a cutoff point of 13ml/rpm, yielded a positive predictive value and a negative predictive value of 0.71% and 0.70%, respectively. In turn, the IWI, studied by Nemer et al. and calculated as the product of static compliance and SaO2 divided by the Fr/Vt ratio, with a cutoff point of 25ml/cmH2O/FiO2 yielded a sensitivity of 0.97% and a specificity of 0.94% in predicting extubation failure.28,29 Undoubtedly, these values imply a very high level of success, though when extubation failure predictors are compared, not only the numerical values of the statistical analyses must be taken into account. In this field a very important consideration is the complexity of application, integration and interpretation of the values obtained. On the other hand, in calculating Vd/Vt, use of the CO2 sensor can be considered, as well as the conduction of arterial blood gas measurements—this being an already standardized and routine practice in ICUs of third-level hospitals. However, indices such as the CROP or IWI involve increased complexity of calculation, based on the integration of several formulas, with the added difficulty of measuring static compliance of the respiratory system in spontaneously breathing patients.30

Lastly, our analysis shows that improved patient oxygenation, as measured with the PaO2/FiO2 ratio, behaves as a protective factor against extubation failure. Our data are consistent with the literature. Thus, in the study of Khamiees et al.,31 involving a majority of clinical patients (89%), a PaO2/FiO2 ratio of between 120 and 150 predicted extubation success in four out of every 5 patients. Likewise, Krieger et al.32 found a PaO2/FiO2 ratio of 238 to have a positive predictive value of 90%.

It can be concluded that the present article is the first study in adults methodologically designed to determine whether Vd/Vt is a good predictor of extubation failure in patients subjected to at least 12h of MV in an ICU. Based on the results obtained, it can be affirmed that the dead space fraction (Vd/Vt) is a valid tool in support of the difficult task of deciding the best moment for extubation.

The authors have no conflicts of interest to declare.