The main findings from our study are the following: compared with DIS, the use of an SA was associated with (a) decreased duration of MV, (b) more rapid resolution of critical illness, and (c) shorter ICU and hospital lengths of stay. (d) Additionally, DIS was associated with less deep sedation levels but was not associated with the administration of less sedative medication compared with the use of SA.
Early in the study, a patient developed serious study-related adverse events. When an interim analysis demonstrated increased mortality rates in the DIS group, the DSMB, guided by safety concerns, examined the primary endpoint as this could potentially explain the mortality findings. When it was noted that the a priori-defined difference in time to successful extubation across the two groups exceeded 2 days, the DSMB recommended study termination. It should be noted that the study was not terminated because of the mortality findings.
More patients treated by DIS were withdrawn from the study. After study withdrawal, these patients were treated using the ICU SA. This could have resulted in eliminating differences in total MV duration between the two groups. However, the DIS group had a longer total MV duration by 2.8 days (P = 0.0003), suggesting a true effect of DIS on MV outcome. Ventilator-free survival was also improved by 7 days in the SA group, although this did not reach significance (P = 0.004, which is greater than the preset alpha of 0.001 for interim analysis).
DIS has been advocated by the Surviving Sepsis Campaign [19
] and the Institute for Healthcare Improvement [20
] as a tool to improve patient outcome. Kress and colleagues [21
] previously demonstrated that use of DIS shortens the duration of MV, decreases sedative administration, increases frequency of wakefulness from 9% to 86%, and decreases the frequency of neurodiagnostic testing. The recent ABC trial also found that sedation interruption, when combined with an SBT, resulted in improved ventilator-free survival [2
]. Our study differs significantly from that of Kress and colleagues [21
] and the ABC trial in a number of ways. First, we compared DIS with a carefully implemented nurse-driven SA rather than with a non-algorithmic approach. Although some patients in the ABC trial may have been treated by an algorithm, this was not a requirement. Second, the majority of our patients were women and half of patients were African-American. Racial and gender differences in medication metabolism are well described, and studies have found that patient characteristics influence medication benefit [22
]. Third, in the ABC trial, clinicians could continue the use of analgesics if they felt this was necessary, and this occurred in 15% of patients whose sedatives were discontinued [2
]. Additionally, if patients required escalating doses of sedatives, they were not considered candidates for sedation interruption. Our protocol was modeled after that of Kress and colleagues [21
]. In our protocol, escalating doses of sedatives did not preclude the discontinuation of sedative and opioid medications unless a patient was agitated at the time the medications were due to be interrupted.
Although DIS patients had higher RASS scores, they did not receive less sedation than the comparison group, an observation that may result from several factors. It is conceivable that the SA used in our study resulted in more medication administration than the control group in the study of Kress and colleagues. Alternatively, our DIS group may have received more sedation because of the protocol modification requiring sedation resumption because of changes in vital signs. Indeed, in a third of occasions, sedation was resumed because of vital sign criteria. Finally, it is also possible that the higher rates of agitation in the DIS group resulted in an increased need for larger doses of sedation, thereby eliminating dosage differences between the two groups.
Another possibility is that our patient population contained a high proportion of patients with alcohol and other drug use disorders. In Richmond, the prevalence of these disorders is 18%, approximately twice the national rate, and our institution cares for many patients with these disorders [26
]. In a retrospective study, we found that 39% of our mechanically ventilated critically ill patients have alcohol and other drug use disorders, and this rate likely under-represents the true rate due to the retrospective nature of the study [28
]. Additionally, alcohol and other drug use disorders typically are coexisting diagnoses in our patients and are not usually the primary reason for requiring MV. The prevalence of alcohol and other drug use disorders in our medical ICU patients is likely to be substantially higher than in the study of Kress and colleagues and the ABC trial. Neither of these two studies reported rates of coexisting alcohol and other drug use disorders. Although one patient in the study of Kress and colleagues required MV for a drug overdose, the ABC trial reported that only 1% of patients with alcohol withdrawal were enrolled in the study.
Alcohol withdrawal has been shown to be associated with longer duration of MV, and patients with alcohol use disorders can develop withdrawal syndromes if they are undersedated or have early withdrawal of sedation [29
]. Additionally, sedative agents have been found to reduce the duration of alcohol withdrawal delirium, and opioids have been shown to decrease the stress response in critically ill patients with alcohol use disorders [31
]. Patients with alcohol and other drug use disorders require a 2.5-fold increase in dosage administration of sedatives and a 5-fold increase in opioid dosage administration to achieve sedation levels similar to patients without these disorders [28
]. It is possible that patients with alcohol and other drug use disorders may well be patients who require escalating doses of sedatives and opioids and who would not have had their sedation interrupted in the ABC trial. Additionally, it is possible that patients with these disorders did not have their analgesic medications discontinued in the ABC trial, thereby resulting in a blunted stress response during sedation interruption and minimizing the symptoms of withdrawal during sedation interruption [32
]. The assessment of withdrawal syndromes rests principally on patient self-reporting of subjective sensations of irritability, nausea, headache, and tactile, visual, and auditory hallucinations. No objective criteria exist for assessing withdrawal in the non-verbal mechanically ventilated ICU patient [33
]. We believe the hypertension, tachycardia, and tachypnea (that is, 'autonomic agitation') experienced by the third patient randomly assigned to DIS may be explained by withdrawal symptoms.
Based on the results of our study, DIS may not be the sedation strategy of choice in all mechanically ventilated patients. We cannot be sure that the sedative strategy alone was responsible for the difference in outcome as there are many uncontrolled factors (that is, comorbidities, severity of illness, organ failure, and so on) and temporally some patients died after leaving the ICU. However, our study raises some concerns about the applicability of DIS in all patients and highlights the need for additional randomized control trials. Previous trials examining DIS were done at institutions with expertise in sedation research and with a research coordinator at the bedside, which may limit generalizabiltity.