INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Mechanical ventilation is a common intervention used to support critically ill children in the pediatric intensive care unit (ICU). Determining the optimal time to discontinue mechanical ventilation is usually based on the clinical and laboratory evidence available at the time of extubation that indicate a patient's ability to sustain adequate gas exchange with spontaneous ventilation. Extubation failure rates in patients when extubation is based on clinical criteria are reported to be 17 to 19% in adults, 22 to 28% in premature infants, and 16 to 19% in children (1). Premature extubation places the patient at risk for emergent reintubation (1). However, unnecessary prolongation of mechanical ventilation increases the risk of airway trauma, nosocomial infection, discomfort and increases the cost of intensive care (5).

A number of indices that measure oxygenation, inspiratory muscle strength, lung function, minute ventilation (E), and ventilatory reserve have been proposed as useful predictors of weaning outcome in adults (6, 7). Some commonly used weaning indices in adult ICUs include airway occlusion pressure, the ratio of tidal volume (VT) to respiratory frequency (rapid shallow breathing index [RSBI]), and the compliance, rate, oxygenation, and pressure index (CROP index). However, objective criteria to predict successful extubation in children have not been established. Furthermore, a previous study showed that adult weaning indices were not predictive of extubation success in children (1). The availability of objective criteria to predict successful extubation in children may prevent inadvertent premature extubation and unnecessary prolongation of mechanical ventilation (1).

Our goal was to study factors associated with extubation failure in infants and children and to compare measures of oxygenation, ventilation, lung mechanics, and inspiratory muscle strength in children successfully extubated with those failing extubation. We also wanted to evaluate the commonly used adult integrated weaning indices (RSBI and CROP index) as predictors of successful extubation in infants and children.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The Hospital Institutional Review Board approved this study and the need for informed consent was waived. A total of 472 patients admitted to the pediatric ICU at Children's Hospital and Regional Medical Center, from September 1996 to September 1997, receiving mechanical ventilation were eligible for the study. In order to decrease variability in data collection, patients were enrolled only when one of the investigators was present on site at the time of their extubation. All patients in the study received mechanical ventilation on Siemens Servo 900C or Servo 300 ventilators (Siemens, Solna, Sweden) at the time of extubation. The patient's primary physician made the decision to wean and extubate based on the patient's clinical status, blood gas determination, and the amount of ventilator support. There was no set protocol for weaning a patient from mechanical ventilation. However, the usual practice in our institution is to wean patients down to a low ventilator rate of 6 to 8 breath/min before extubation. It is also our institutional practice to wean the rate of continuous infusions of narcotic and sedative agents to low doses and to stop the administration of any intermittent doses of narcotic and sedative agents for at least 2 h before extubation. The patient's primary physician also made all decisions regarding reintubation and reinstituting mechanical ventilation.

After the decision to extubate was made, patients were allowed to breathe spontaneously on continuous positive airway pressure (CPAP) of 4 cm H₂O for measuring weaning parameters. After quiet breathing was assured, the patient's respiratory rate (RR, breaths/ min) was counted. The patient's spontaneous tidal volume (sVT) and E were obtained from the digital output of the ventilator. For sVT, the largest volume of four consecutive spontaneous breaths was taken. Next, the patient's maximal negative inspiratory force (NIF; cm H₂O) was measured using the method described by Baumeister and coworkers (2). A manometer (Rusch, Chicago, IL) and a unidirectional valve system (Rescal inspiratory force adaptor; DHD, Canastota, NY), which allowed exhalation but not inhalation was attached to the patient's endotracheal tube. The patient's airway was occluded and the maximal negative deflection during inspiration of a single breath was recorded. The largest negative deflection of three trials was recorded as the maximal NIF. The patient was allowed to rest for 1 min, on CPAP, between each trial. Respiratory therapists caring for patients in the pediatric ICU collected all data under the supervision of one of the authors. The patient was extubated after the data collection was complete. The patient's primary physician was blinded to the data collected for the purposes of this study.

Demographic data collected at the time of extubation included the patient's age, weight, sex, admitting diagnosis, date of intubation, and date and time of extubation. Data collection also included the patient's endotracheal tube size, presence of air leak around the endotracheal tube, mode of ventilation prior to extubation, pre-extubation arterial blood gases, peak inspiratory airway pressure (PIP; cm H₂O), positive end-expiratory pressure (PEEP; cm H₂O), mean airway pressure (; cm H₂O), corrected exhaled VT from mechanical ventilator breaths (vVT; ml/kg), total E on the ventilator (vE; ml/kg · min), and fraction of inspired oxygen concentration (FI_O2). Extubation failure for purposes of this study was defined as reintubation within 24 h of extubation. The reason for reintubation as noted by the patient's primary physician was recorded on the data sheet.

Because RR vary with age, a standardized RR (RRstd) was calculated using the mean and SD of RR normal for a given age (8). Both the sVT and spontaneous minute ventilation (sE) were standardized to body weight. Dynamic compliance (Cdyn, ml/kg/cm H₂O) was calculated using the formula: vVT/(PIPPEEP). Alveolar oxygen concentration (PA_O2, mm Hg) was calculated using the alveolar gas equation, PA_O2 = [FI_O2 × (barometric pressure [P_B]  partial pressure of water vapor [PH₂O])]  Pa_CO2/respiratory quotient (RQ), where P_B = 760 mm Hg, PH₂O = 47 mm Hg and RQ = 0.8. The RSBI (breaths/ml/ kg) was calculated using the formula: RR/sVT. The CROP index (ml/ kg/breaths/min) was calculated using the formula: Cdyn × NIF × (Pa_O2/PA_O2)/RR (6).

Statistical analyses of the data collected from the two outcome groups (extubation success and extubation failure) were compared using SPSS version 7.5 for Windows (SPSS Inc., Chicago, IL). Data are presented as means and SD unless specified otherwise. Data regarding PaO₂, oxygenation, and CROP indices were only compared for patients without cyanotic congenital heart disease at the time of extubation. Continuous variables were compared using Student's t test for normally distributed data and the Mann-Whitney U test for data that were not normally distributed. Categorical variables were compared using chi-square test and Fisher exact test. A p value  0.05 was considered statistically significant. For each spontaneous breathing parameter and integrated weaning index showing statistically significant difference between the outcome groups, an attempt was made to identify a threshold value which best predicted success and failure. Standard formulae were used to calculate sensitivity, specificity, and positive and negative predictive values for each variable (9).

Finally, a logistic regression model was constructed to predict extubation failure using variables that were significantly related to failure in a univariate analysis. Variables were entered into the regression model in a stepwise fashion. The inclusion and exclusion criteria for the regression model were defined as a p value of  0.05 and  0.1, respectively. The logistic regression equation was used to construct a probability table for extubation failure based on different values for variables in the equation (10).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 227 patients underwent 254 episodes of extubation. Demographic data of patients in the two outcome groups are presented in Table 1. Seventy-one percent of patients were extubated in the postoperative period after cardiac surgery. Two hundred twenty-six episodes (89%) were successful; 28 (11%) were failures. Twenty-three (82%) reintubations were the result of increased work of breathing from respiratory or cardiovascular causes, four (14%) were due to increased work of breathing associated with upper airway obstruction, and one (4%) was caused by acute hemorrhage in an otherwise stable patient, who was reintubated during resuscitation. Those who required reintubation for upper airway obstruction and the patient who was reintubated during resuscitation were excluded from the failure group. Patients extubated successfully were significantly older than those failing extubation. Patients in the 0 to 30 d age group had the highest failure rates compared with the 31 d-1 yr, > 1-5 yr, and > 5 yr age groups. Patients who failed extubation had a variety of diagnoses. Failure was significantly higher in patients with cyanotic congenital heart disease compared with patients with other diseases.

The pre-extubation ventilator settings and derived measurements of patients in both outcome groups are shown in Table 2. Patients successfully extubated received mechanical ventilator support for significantly shorter duration compared with those who failed extubation. Patients extubated successfully received larger VT from mechanical ventilator breaths for significantly lower ventilator peak airway inflation pressures, suggesting better lung compliance. This is also reflected by the calculated Cdyn which was significantly higher for patients successfully extubated compared with those who failed. Patients successfully extubated had significantly higher Pa_O2, Pa_O2/FI_O2 ratio with similar levels of FI_O2, PEEP, and significantly lower . Patients successfully extubated also demonstrated significantly lower Pa_CO2 and significantly lower E on the mechanical ventilator suggesting better distribution of ventilation. Thus, patients successfully extubated demonstrated significantly better lung compliance, oxygenation, and ventilation compared with those failing extubation.

The spontaneous breathing parameters and calculated integrated weaning indices obtained on CPAP of 4 cm H₂O from the two groups are shown in Table 3. Patients successfully extubated had significantly lower RR, lower RRstd, and significantly larger sVT than those failing extubation. Although patients in both outcome groups attained similar sE, patients in the failure group achieved their E requirements by increased RR implying higher energy expenditure to sustain adequate gas exchange. The maximal NIF was significantly lower in patients failing extubation. This suggests a difference in muscle strength between patients successfully extubated and those who failed. The calculated indices RSBI and CROP index were also significantly different in patients succeeding and failing extubation. The higher RSBI in children failing extubation demonstrates that patients failing extubation exhibited rapid shallow breathing. The significantly larger CROP index in patients successfully extubated suggests better compliance, muscle strength, oxygenation, and lower RR in these patients compared with those failing extubation.

Threshold values that showed statistically significant differences and best differentiated patients succeeding and failing extubation were identified for the indices shown in Table 4. A RSBI value of  8 was the most specific index (74%) and a CROP index value of  0.15 the most sensitive index (83%) for predicting extubation success. A higher threshold RSBI value was identified for the 0 to 30 d age group. A RSBI value of  11 had higher sensitivity (78%) but lower specificity (56%) for newborns compared with using a RSBI value of  8 (sensitivity 49%, specificity 83%). Age-specific threshold values could not be identified in the other age groups because of the small number of extubation failure rates in the older groups. The threshold values for each index had better positive predictive value than negative predictive value suggesting that the indices predicted extubation success better than failure. Finally, the logistic regression model showed that the duration of mechanical ventilation and the RSBI were significantly associated with extubation outcome. The relative risk for extubation failure based on the duration of mechanical ventilation and the RSBI is presented in Table 5. We constructed a regression equation using these variables. The probability of extubation failure was calculated for different RSBI values, and increasing duration of mechanical ventilation is shown in Table 6.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Successful extubation in patients receiving mechanical ventilation is dependent on the resolution of the primary process, presence of intact airway reflexes and ability to clear secretions, an intact central inspiratory drive, ability to exchange gases efficiently, and respiratory muscle strength to meet the work associated with respiratory demand (11, 12). Several other factors such as the nutritional status, acid-base balance, hemodynamic stability, and psychological factors can also influence extubation outcome (13). Accurate prediction of extubation is difficult because multiple factors determine a patient's ability to breathe spontaneously. Although several clinical and physiological characteristics such as age, weight, duration of mechanical ventilation, ventilation mode, RR, VT, Cdyn, maximal inspiratory force, and airway resistance have been associated with extubation success in infants and children, only few studies have examined their usefulness for predicting extubation outcome (14). The indices that have been evaluated for predicting extubation outcome include crying vital capacity, maximal inspiratory force, maximal transdiaphragmatic pressure, and the integrated adult indices (RSBI and the CROP index) (1, 2, 17).

Our study demonstrates that children failing extubation were younger, received mechanical ventilation for a longer duration, had increased ventilatory demand, poorer respiratory system compliance, and had defects both in oxygenation and ventilation compared with children who were successfully extubated. When allowed to breathe spontaneously, patients failing extubation demonstrated rapid RR with smaller VT to achieve the required E. Thus, their energy expenditure was likely to be higher compared with patients who were successfully extubated. The pattern of breathing exhibited by children failing extubation in our study, namely rapid shallow breathing, resembles the description of the pattern of breathing seen with adults failing extubation (3). We found that the adult weaning indices (RSBI and CROP index) were significantly different in children failing extubation compared with those successfully extubated and were predictive of extubation outcome. We found that the duration of mechanical ventilation and RSBI were independently associated with extubation outcome. Because previous adult studies have shown that the duration of mechanical ventilation can influence the predictive value of weaning indices, duration of mechanical ventilation should be considered when defining a threshold value for RSBI when used to predict extubation outcome (6, 20).

In a recent study, Khan and coworkers evaluated the use of several bedside measures of respiratory function as predictors of extubation success and failure in infants and children (1). In this study of 208 patients, 34 failed extubation resulting in a failure rate of 16.3%. They showed that the rate of extubation failure increased with worsening sVT, FI_O2, , oxygenation index, fraction of support provided by the mechanical ventilator, peak inflation pressure, Cdyn, and mean inspiratory flow. Although they did not identify a single threshold value that best predicted extubation success or failure, they defined low (< 10%) and high ( 25%) risk threshold values for extubation failure for these variables. However, they were unable to define low- or high-risk threshold values for both the RSBI and CROP index. They concluded that the prediction of extubation outcome was possible using bedside weaning parameters, but that the RSBI and CROP index were poor predictors of extubation outcome in children. In another study, Baumeister and coworkers evaluated the use of RSBI and CROP index for the prediction of extubation outcome in infants and children (2). In their study, nine of 47 extubation trials resulted in failure, for a failure rate of 19%. In contrast to the study by Khan and coworkers, they found that both the integrated indices reliably predicted extubation success. The CROP index had better predictive value than the RSBI.

The extubation failure rate in our study of 11% is lower than the failure rates quoted in both the previous studies. Like the study by Khan and coworkers, we found children who failed extubation had poorer indices of respiratory function. In contrast to the study by Khan and coworkers but like the study by Baumeister and coworkers, we found that the RSBI and CROP index were predictive of extubation success. However, we found different threshold values for the RSBI (8 and 11 breaths/min/ml/kg) and the CROP index (0.15 and 0.1 ml/ kg/breaths/min). The positive predictive value for RSBI was higher in our study, whereas the positive predictive value for the CROP index and the negative predictive value for the RSBI and CROP index were lower. The differences in the threshold values may be due to the measurement technique used and the age distribution of the study populations in the two studies. Our study population was evenly distributed among the age groups compared with the study by Baumeister and coworkers, which contained younger children who have more rapid RR.

Several other important differences exist between our study and the studies by Khan and coworkers and Baumeister and coworkers. In the study by Khan and coworkers, extubation was considered unsuccessful if the patient was reintubated within 48 h after extubation. In the study by Baumeister and coworkers, failure was defined as reintubation within 24 h after extubation or need for increased ventilator support after the decision to extubate was made and collection of weaning data. Our definition of extubation failure, as reintubation within 24 h after extubation, was conservative and may explain the lower failure rate in our study. The sVT and sE in our study were obtained from the digital output from the patient's ventilator, whereas the other studies measured volumes using a pneumotachograph attached to the patient's endotracheal tube. We made spontaneous breathing measurements on CPAP of 4 cm H₂O pressure, whereas Khan and coworkers measured spontaneous breathing variables with patients disconnected from the ventilator and attached to a Mapleson bag. Baumeister and coworkers measured spontaneous breathing with the patient remaining on the mechanical ventilator on a low set ventilator rate. In our study measurement errors may have occurred because of errors in ventilator calibration, operator errors, or from loss of volume associated with air leak around the endotracheal tube (4). However, it is common practice to use volumes and pressure derived from the ventilator dials to help guide mechanical ventilator therapy in patients receiving mechanical ventilation. In addition, mechanical ventilators in our institution undergo calibration prior to use as specified by the manufacturer and the accuracy of exhaled VT measured by the ventilator is said to be within ± 0.5 ml of the actual exhaled volume for both the Siemens Servo 900C and Servo 300 ventilators (21, 22). Furthermore, standardized measurement techniques were used and experienced respiratory therapists who staffed the ICU on a regular basis, made all measurements. It is also possible that the increased imposed work of breathing, for patients breathing spontaneously on Servo 900C ventilators compared with those on Servo 300 ventilators, owing to differences in the CPAP characteristics of the two ventilators, may have influenced our measurements. Because patients in our study were only subjected to spontaneous breathing on CPAP for a few minutes, for making measurements, it is unlikely that the difference in work of breathing in patients breathing spontaneously on CPAP on the Servo 900C ventilators caused significant errors in our measurements. Finally, our method is easy to use, requires little additional equipment and respiratory therapist time, and does not require disconnecting the patient from the ventilator.

A wide range of sensitivities and specificities for predicting extubation success using the RSBI has been published in the adult literature (6, 20, 23, 24). The differences in the predictive value of the RSBI may be the result of factors unrelated to respiratory load, capacity, and endurance. Several factors including gender, endotracheal tube size, anxiety, agitation, mental stress, endogenous opioids, duration of mechanical ventilation before extubation, timing of measurements, and measurement technique used have been known to alter the predictive value of the index (6, 20, 23, 25, 26). Although similar studies in children are unavailable, the influence of these factors should be considered when using the RSBI for predicting extubation success.

In conclusion, our study demonstrates that extubation failure in children receiving mechanical ventilation was associated with worsening of several indices of respiratory function. Extubation failure is more common in neonates and infants, in patients with cyanotic congenital heart disease, and in those receiving prolonged mechanical ventilation. Close attention to the duration of mechanical ventilation, respiratory mechanics, support received from the ventilator, oxygenation indices, and Pa_CO2 may help prevent inadvertent premature extubation. The adult integrated weaning indices RSBI and CROP index can be reliably used to predict extubation outcome. However, the RSBI and CROP indices predict extubation success more accurately than failure. The threshold values for the RSBI for predicting extubation outcome appears to vary depending on duration of mechanical ventilation. The accuracy of these indices should be prospectively evaluated prior to their introduction into clinical practice.