INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Several prospective studies of weaning protocols have been performed in mechanically ventilated medical intensive care unit (MICU) and surgical intensive care unit (SICU) patients (1). One approach, incorporating a daily screen (DS) of weaning parameters, physician prompts, and spontaneous breathing trials (SBT) performed by respiratory care practitioners (RCP), has been associated with improved outcomes in coronary intensive care unit (CICU) and MICU patients with respiratory failure (4, 5), and has been extended safely to general surgical patients (6). In addition, passing the DS of weaning parameters has been shown to be independently predictive of hospital survival, even after adjustment for severity of illness (5). It is unclear whether this protocol would be equally effective and have similar prognostic significance in other patient populations.

As with MICU and general SICU patients, neurosurgical (NSY) patients are predisposed to a number of complications related to mechanical ventilation. Retrospective and prospective investigations have documented increased rates of reintubation, pneumonia, tracheostomy, and prolonged mechanical ventilation among patients with acute brain injury (7). Clinical indications for intubation and mechanical ventilation in this setting generally include impaired neurologic function and the need for airway protection and/or manipulation of intracranial pressure, as well as pulmonary physiologic abnormalities. Considerable variability among clinicians' strategies for liberation from mechanical ventilation and extubation in NSY patients has been documented, and the optimum approaches to these measures have not been clearly defined. A recent prospective observational study of selected mechanically ventilated, brain-injured patients did not support delaying extubation when impaired neurologic status was the sole reason for prolonging intubation (18). The authors suggested that a randomized extubation trial, applying standard respiratory weaning criteria, was justified in these patients.

To determine whether a ventilator management protocol previously validated in medical patients was useful in NSY patients, we introduced an RCP-driven weaning protocol involving DS, prompts, and SBT in a prospective, randomized, controlled investigation involving mechanically ventilated NSY patients. We monitored the incidence of complications related to mechanical ventilation, mortality, length of stay, and cost. In addition, we appraised overall adherence to the protocol, barriers to its implementation, and factors associated with successful extubation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients were assigned randomly to Control or Intervention groups after informed consent was obtained by study physicians not involved in their routine care. Study physicians recorded demographic and laboratory data, initial ventilator settings, minute ventilation (E), respiratory rate ( f ), fraction of inspired oxygen (FI_O2), static compliance, set tidal volume (VT), spontaneous VT, peak pressure, plateau pressure, endotracheal tube size, acute lung injury (ALI) score (19), and Acute Physiology and Chronic Health Evaluation (APACHE) II score (20). Vital signs, including Glasgow Coma Scale (GCS) score (21, 22), were recorded prospectively by ICU nurses.

DS and Protocol

The protocol approved for the study by the Wake Forest University Baptist Medical Center institutional review board included a DS of pulmonary physiologic parameters and SBT as described previously (4, 23) (see APPENDIX), with the following modifications for NSY patients: (1) Intervention-group patients with an intracranial pressure-monitoring device did not undergo SBT until its removal; and (2) attending physicians, all of whom were board-certified neurosurgeons blinded to DS results and group assignments, decided the mode of mechanical ventilation and discharge from the ICU and hospital. Data for f, VE, P, FI_O2, Pa_O2/FI_O2 (P/F ratio), and f/VT ratio, GCS score, and use of sedation or pressors were recorded on a flow sheet for each patient by RCP, and were monitored daily by study physicians. After a patient completed the study, data were entered into a computerized data base.

A priori power analysis showed that an accrual of 188 patients over a 12-mo period would have allowed detecting a 20% difference in duration of mechanical ventilation with 80% power at the 5% two-sided level of significance. When a planned 12-mo interim analysis revealed an obvious lack of efficacy, the study was discontinued at 100 patients.

Outcomes

Primary outcomes, defined a priori, included overall duration of mechanical ventilation, length of ICU stay, and time to successful extubation. Secondary outcomes were the frequency of complications (reintubation, self-extubation, tracheostomy, and mechanical ventilation exceeding 21 d); costs of mechanical ventilation, respiratory and ICU care, and overall hospitalization; hospital stay; and mortality. The existence of ventilator-associated pneumonia (VAP) was determined independently by the hospital epidemiologist and nurses, using criteria from the Centers for Disease Control and Prevention (CDC) (25). Financial data were calculated as described previously (4).

Statistical Analysis

We performed an intention-to-treat analysis, with all randomized patients included in all primary analyses. Treatment differences in categorical variables were assessed with chi-square tests and Fisher's exact tests as appropriate (26). Logistic regression (for binary outcomes), and analysis of covariance, with raw or transformed data (for continuous outcomes), were used to assess the effect of interactions after adjusting for baseline characteristics. Log or rank transformations were used when assumptions (e.g., normality, equal variances) were not met. Kaplan-Meier tests (27) were used to estimate the distributions of time to discharge, time on mechanical ventilation, and time to successful extubation. Log-rank tests were used to assess differences between treatment groups in these outcomes. Cox's proportional hazards regression model (28) was used to assess the effect of intervention after adjusting for patients' baseline characteristics. The accuracy of predicting extubation results was calculated for all possible cutpoints of GCS score, f/VT ratio, and P/F ratio, in order to determine the best values to use in subsequent protocols. A generalized estimating equations (GEE) analysis, done with each day's information for each patient, was used to determine factors associated with successful extubation. Receiver operator characteristic (ROC) curves were derived for each GCS score. After an ROC cutoff was established for GCS score, f/VT ratio, and P/F ratio, an odds ratio (OR) of successful extubation was determined with both univariate and multivariate testing, including and censoring those patients withdrawn from care.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Characteristics

Between July 1997 and August 1998, 141 mechanically ventilated NSY patients were admitted to our 806-bed university medical center. One hundred (71%) patients entered the study after the family and/or patient had given signed consent, 31 patients and/or families declined to participate, and in the case of 10 patients a separate decisionmaker was unavailable from whom to obtain informed consent. Of the 100 patients who consented to the protocol, 51 were randomized to the Control group and 49 to the Intervention group. Demographic characteristics of these patients are summarized in Table 1. The patients' ages ranged from 16 to 91 yr, with a median of 58.5 yr. Fifty-four percent of the patients were male and 79% were white. Reasons for admission included head trauma (23%), subarachnoid hemorrhage (19%), intracerebral hemorrhage/arteriovenous malformations (34%), tumor (8%), spinal trauma (4%), and other (10%). Neurologic injuries were the primary reason for intubation and mechanical ventilation in 95% of these patients; chronic obstructive pulmonary disease (1%), pneumonia (3%), and aspiration (1%) accounted for the remainder. A majority of patients (52%) received volume-controlled intermittent mandatory ventilation (IMV) with pressure-support (PS) ventilation; another 22% received IMV alone, and 25% received PS ventilation alone; one patient (1%) received assist-control ventilation. No significant differences were noted between the Control and Intervention groups in the baseline characteristics listed in Table 1 (all p > 0.2).

Outcomes

Outcomes are summarized and compared in Table 2. There were no significant differences between the Intervention and Control groups with regard to complications, mortality, or cost, with or without adjusting for pretreatment patient characteristics. Patients took a median of 2 d to pass the DS and another 2 d to be extubated. The median duration of ventilation (6 d) and time to the first successful extubation (10 d) were similar for both study groups. Control-group patients spent a median of 32 d in the hospital, as compared with 40 d for the Intervention-group patients. Complications in the 100 patients studied included reintubation (16%), self-extubation (6%), and tracheostomy (29%). Overall mortality was 36% (36 of 100 patients), and 28 of the patients who died had care withdrawn because of a poor prognosis. The cost of care per patient was accrued primarily during the ICU stay. The costs of ICU stay were $43,995 and $55,689 in the Control and Intervention groups, respectively. This accounts for slightly more than 80% of the total costs of hospitalization, which averaged $57,700 for Control-group and $64,494 for Intervention-group patients. The costs of mechanical ventilation in the two groups were $1,782 and $1,985, respectively.

Protocol Implementation

DS were performed on 794 patient days, of which passing was noted on 403 (51%) and failure on 391 (49%) patient days. SBT were performed in 42 of 49 (86%) patients in the Intervention group who successfully passed the DS. RCP performed an SBT on 99% (199 of 201) of occasions on which DS were passed, but an attempt at extubation was made for only 25% of patients who passed an SBT (Figure 1). Neurosurgeons' primary reasons for not extubating such patients included concerns about the patient's level of consciousness (84%), a decision to perform tracheostomy (10%), and other reasons (6%).

View larger version (10K): [in this window] [in a new window]   Figure 1.   Frequency of ordering extubation after a successful SBT. Physicians ordered an extubation in 25% of those patients who successfully passed an SBT. During the initial 2 mo of the study there was a statistically insignificant increase in the frequency of attempted extubation.

Days on Mechanical Ventilation and Time to Successful Extubation

A total of 813 d of mechanical ventilation were observed, with a mean of 9.1 d in the Control and 8.7 d of mechanical ventilation in the Intervention group. The median time to an attempted extubation was 4 d in the Control and 5 d in the Intervention group. Both groups had a median elapse of 2 d after passing SBTs before an attempt at extubation was made. Although early attempts were made, the median time to successful extubation was 10 d in both of the study groups. There was no difference observed between the two groups in the duration of mechanical ventilation, time to first attempt at extubation, and time to successful extubation (Table 2).

Factors Associated with Successful Extubation

Among the 100 patients enrolled in the study, 117 attempts at extubation were made, of which 73 (62%) were successful and 44 (38%) failed. The 44 failed extubations were followed by 22 reintubations, which occurred in 16 patients. No deaths or other complications of reintubation occurred. In the remaining 22 cases the patients and/or their families elected not to have further heroic life-support measures. These patients were not reintubated, and all died. We examined the first extubation for each patient and assessed which factors on that day were associated with success. Ninety-eight of the 100 patients enrolled in the study had at least one extubation while in the hospital; the other two patients were transferred to a nursing care facility while receiving mechanical ventilation. Sixty (61%) of the 98 initial extubations were successful. Factors considered in the analysis were the GCS, P/F ratio, f/VT ratio, and E measured on the day of attempted extubation. Descriptive statistics for these variables, by extubation outcome and OR from univariate logistic models, are shown in Table 3.

Successful extubations were associated with higher GCS score and P/F ratio values, and with lower E values. The f/VT ratio was similar for those patients having successful and unsuccessful extubations. The presence of an intact cough reflex or cough during suctioning, appraised during DS in all patients, did not show an association with successful or unsuccessful extubation in either univariate or multivariate analysis. The odds of having a successful extubation increased by 39% for every increment in the GCS score. A plot of the frequency of successful extubations for each GCS score is shown in Figure 2. The odds of response decreased by 13% for every unit increase in E. A multivariate logistic regression model showed GCS score and E as significant (p = 0.008 and p = 0.013, respectively) factors associated with successful extubation. In this analysis, the odds of success increased by 30% for a unit increase in GCS score, and decreased by 14% for a unit increase in E. A GCS score  8 provided the greatest accuracy in predicting successful extubation relative to other GCS score values. ROC analysis showed that a GCS score  8 had an area under the curve of 0.681 (Figure 3). Patients with a GCS score  8 at the time of extubation had a successful extubation in 75% (60 of 80) cases, whereas those with a GCS score  7 had a successful extubation in 36% (14 of 38) cases.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Relationship between successful extubation and GCS score. With incremental increases in the GCS score, the probability of successful extubation increased by 20 to 39%.

View larger version (13K): [in this window] [in a new window]   Figure 3.   ROC curve of GCS score for predicting successful extubation in NSY patients. A GCS score  8 was associated with best prediction for successful extubation in NSY patients with an area under the curve of 0.681.

All 813 d of mechanical ventilation were then included in the GEE analysis, which examined the association of factors mentioned earlier and the results of extubation. Successful extubations occurred on 73 (9%) of the 813 observed days. Results of the multivariate models are shown in Table 4. GCS score and P/F ratio were associated with increased odds of success, whereas f/VT ratio and E were negatively associated with success. The multivariate GEE logistic model showed that the GCS score, P/F ratio, and f/VT ratio were jointly associated with extubation results, with OR that were similar to those obtained in the univariate models.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recognizing a patient's readiness to wean and eventually become extubated successfully has been a priority since the introduction of mechanical ventilation (6, 23, 24, 29, 30). Because few objective studies have addressed the course of mechanical ventilation or the decision to extubate NSY patients (7, 11- 18), we implemented (without modification) a previously validated protocol (4, 23) in NSY patients. The patients in our study group represent one of the largest prospective cohorts of mechanically ventilated NSY patients, about whom we collected comprehensive outcome data including length of stay, complication rates, and cost of care, and our study provides important insights about these patients' ventilatory management. In contrast to our previous experience with MICU and CICU patients, primary and secondary outcomes, such as number of days of mechanical ventilation and hospitalization, cost, morbidity, and mortality, were not statistically different between the Intervention and Control groups in the present study. Our RCP-focused protocol, incorporating respiratory parameters, was not associated with improved overall outcomes of the Intervention group.

Readiness Criteria and Protocol Adherence

Not surprisingly, most (95%) NSY patients in our study were intubated because of their primary neurologic problem, and not because of a respiratory disorder. GCS score, a measure of neurologic status, emerged as the best predictor of successful extubation upon logistic regression analysis. This finding is consistent with the primary indication for intubation and mechanical ventilation, and may also be attributed to neurosurgeons' use of the GCS score in their daily patient assessments. Clinicians have generally synthesized such information in their bedside determinations, but the experience in the present study provides objective criteria. The observations in the study might also suggest overconfidence and/or overreliance on the rapid shallow breathing index alone as a predictor of extubation, and underscore the need to interpret this measurement in the context of the patient's overall status and the initial indications for intubation and mechanical ventilation. Therefore, in order to apply a weaning protocol effectively in this setting, an objective measure not only of traditional respiratory parameters, but also of the level of neurologic impairment, should be included as a criterion of readiness for extubation.

Integrating the screening information in the study with their knowledge of the patient's neurologic status, underlying pathology, and other clinical data, neurosurgeons extubated only 25% of patients who passed SBT. When the decision was made not to extubate after a successful SBT, neurosurgeons expressed concerns about patients' mental status in 84% of instances. Our observations revealed that our study protocol, as designed, did not influence physician behaviors. We believe that the most likely explanation for the lack of success of the protocol was related to the absence of neurologic assessment parameters from the screening criteria. As a result, the protocol did not encompass the unique needs of the NSY population, which led to a lack of adherence to the protocol.

Complications

Complications have not been prospectively assessed specifically in mechanically ventilated NSY patients. Retrospective analyses have suggested that the duration of mechanical ventilation may be prolonged (7, 17). We observed that the median duration of ventilation was 6 d and the time to successful weaning was 2 d, with only 6% of our patients receiving prolonged mechanical ventilation (> 21 d). These findings are remarkably similar to our experience with other critically ill groups, and suggest a similar course of recovery despite differing indications for mechanical ventilation. Other ventilator- associated complications, including reintubations, pneumonias, tracheostomies, and self-extubations, were relatively infrequent. Although protocol adherence was suboptimal, it is important that we observed no untoward outcomes of NSY patient participation. Neither deaths nor other complications of reintubation were noted, perhaps because of the close monitoring of these patients. This low morbidity is consistent with recent observations by Coplin and colleagues, who reported the safety of timely extubation in their patients with impaired neurologic function (18).

In a prospective study of MICU/SICU patients, Vallverdu and colleagues reported a reintubation rate of 36% among those with central nervous system (CNS) impairment (16), whereas in a retrospective analysis, Koh and coworkers observed a 22% reintubation rate (7). These observations contrast with a reintubation rate of only 16% in our patients. This low percentage in our study may reflect both a higher threshold for extubation and our high frequency of tracheostomy. Epstein and colleagues identified several factors associated with an increased risk of reintubation, including non-airway-related causes, increased APACHE II scores, and comorbid conditions (11), and underscored the importance of altered neurologic function and characterization of reintubation as due to extubation failure or weaning failure. Our findings support the importance of this distinction. It is interesting that despite the conservative approach taken toward extubation, in our study, only 61% of initial extubations were successful. This low frequency contrasts with that seen in other populations at our institutions, in whom 85% of initial extubations are successful (4), and reflects the potential vulnerability of intubated NSY patients and/or other undefined aspects of their care.

We used CDC criteria for the diagnosis of VAP, monitoring for which was done by hospital epidemiologists and nurses independently of our study, and found a lower incidence (6%) than expected. Cook and colleagues identified VAP in 17.5% of 1,014 MICU patients, with an OR of 3.4 among patients with a CNS insult (13). Baraibar and colleagues reported an OR of 10 for VAP among NSY patients in their prospective cohort of MICU/SICU patients (14). Reasons for the comparably low frequencies of ventilator-associated complications in our NSY patients are unclear.

Integrating Neurologic Measurements into the Extubation Decision

Multiple investigations have found a pivotal role of respiratory parameters in identifying a patient's readiness to be extubated (6, 24, 29, 31). Some studies suggest that markers of neurologic impairment may help predict a patient's readiness (18, 35), but interactions between lung mechanics and neurologic function in weaning patients are incompletely defined. We performed a secondary analysis in our NSY cohort to identify respiratory parameters and/or neurologic measures associated with successful extubation. In doing this, we sought to answer the following questions: (1) Do physiologic weaning parameters validated in the medical intensive care unit (f/VT ratio, P/F ratio, E) predict the success of extubation in NSY patients? (2) Do other measures (e.g., GCS) predict extubation success? (3) What are the joint effects of respiratory parameters and neurologic functional measures in predicting outcomes?

Although the f/VT ratio, P/F ratio, and E were significantly related to successful extubation in the present study, OR suggest that they may not be as clinically useful as other predictors. Chevron and associates, in a prospective case-control study, found that GCS score correlated with unplanned extubation and the need for reintubation (17). Vallverdu and colleagues advocated assessment of neurologic measures in patients with CNS impairment, but did not relate extubation to GCS score (16). Coplin and colleagues noted that a GCS score  8 was associated with delayed extubation, but did not preclude successful extubation (18). Using multivariate analysis, we sought to determine the strongest predictor of successful extubation in NSY patients. Although several factors were predictive, GCS score was consistently the best independent predictor in all models considered. The odds of successful extubation rose with increments in GCS score by 20 to 39%. ROC analysis suggested that a GCS score  8 established a cutoff associated with a likelihood of successful extubation, whereas a GCS score  8 was more likely to be associated with extubation failure and an increased likelihood of complications. Because our neurosurgeons perform a daily GCS in their patient assessment, the decision to extubate may not have been completely independent of the recorded GCS. We did not assess the ways in which our neurosurgeons incorporated into their decisions about extubation their personal measures of GCS, its components, or those GCS data charted at the bedside by nursing staff. Accordingly, the utility of incorporating GCS and other validated neurologic assessment tools (35) into assessment for extubation should be further investigated, to better determine the optimum timing for an extubation attempt in patients with impaired neurologic function. Other limitations of our investigation may relate to sample size. Although it remains possible that a type II error existed, we believe that the small observed differences between our study groups make it unlikely that additional patients (88) would have made a difference in our findings. A further consideration is that our investigation did not control for the timing of and criteria for tracheostomy, a procedure performed in nearly one third of our patients.

In this study, the focal nature of our patients' diseases provided a unique circumstance in which to administer a weaning protocol. Our experience in the study underscores the practical challenge of implementing and the importance of refining protocols for weaning from extubation, using the unique clinical characteristics of a specific group of critically ill patients. An approach validated in a group of primarily coronary and respiratory patients will probably be met with resistance and an appropriate lack of confidence if it is applied without modification to an entirely different patient population. Those who manage NSY patients are in need of additional prospective data regarding objective measures of neurologic status and their relationships to outcomes of mechanical ventilation. We believe that future multidisciplinary studies should modify existing protocols so as to incorporate the combination of respiratory parameters and measures of neurologic function in order to determine the optimum management approach and timing of extubation in NSY patients receiving intensive care.