Daily sedative interruption versus intermittent sedation in mechanically ventilated critically ill patients: a randomized trial

This trial was conducted in a closed multidisciplinary six-bed ICU that admits patients from emergency department, surgical room and ward in an academic tertiary hospital from January 2009 to December 2011. The ICU has one intensive care physician, one intensive care resident and three second-year internal medicine residents on daytime shifts. During the night, there is one intensive care physician and one second-year internal medicine resident on charge. This intensive care physician is also responsible for another 14-bed ICU, where an intensive care resident and a first-year internal medicine resident are on charge. The nurse-to-patient ratio is 1:6 and the nursing assistant-to-patient ratio is 1:2 on all shifts. Nursing assistants are only responsible for the patients’ hygiene and administration of drugs in peripheral venous and nasoenteral catheters. There is one respiratory and physical therapist only on daytime shifts.

Eligible patients were those who required mechanical ventilation within the last 24 hours and were expected to need mechanical ventilation for more than 24 hours. Patients were excluded if they were younger than 18 years-old, were pregnant, needed deep levels of sedation (intracranial hypertension, status epilepticus, hypothermia after cardiac arrest, severe asthma exacerbations, and severe hypoxemic respiratory failure (PaO₂/FiO₂ ratio < 50), were not expected to survive for more than six months (for example, metastatic cancer, NY functional class IV heart failure, Child C hepatic cirrhosis, oxygen-dependent chronic obstructive pulmonary disease), were previously cognitive impaired (for example, advanced dementia), or were readmitted to the ICU after participating in the trial.

Research staff randomized included patients in a 1:1 ratio to intermittent sedation or continuous infusion of sedatives with daily interruptions. Treatment allocation was concealed by random selection of opaque sealed envelopes for consecutive patients from a box with 120 envelops. Every envelope contained a paper in which the sedation strategy of that patient would be allocated. The study was unblinded. Therefore, attending clinicians and research staff were aware of which group the included patients were allocated to.

The study (protocol number 0284/08) was approved by the Ethical Committee for Research Projects Analysis of Hospital das Clínicas, University of São Paulo, in September 2008. Written informed consent was obtained from patients or their legal representatives. If consent was given by a representative, patients were contacted for their consent as soon as possible.

All patients had their level of sedation monitored with a Portuguese version of Sedation Agitation Score (SAS) [12] every six hours by bedside nurses. Sedation was defined as the infusion of sedative drugs, that is, midazolam or propofol, at the discretion of the attending physician. Patients in the intermittent sedation group would be kept without continuous infusion of sedatives if the intubation had been performed in the ICU, or would have their infusion interrupted after randomization if they had been admitted already intubated from emergency department, surgical room, wards or another ICU. Patients would be kept without sedatives infusion until they awoke. After patients were awake, if they were calm and collaborative (SAS of 4), they would be kept without infusion of sedatives. If the patient was uncomfortable or agitated (SAS ≥ 5), the physician (attending or resident) would be consulted, and possible causes of discomfort would be investigated in a standardized method (pain, patient-ventilator asynchrony, thirst, hunger, and position on the bed, all of them using a poster which included figures expressing these uncomfortable sensations) and treated. Pain was treated with boluses of fentanyl (50 to 150 μg). If the pain recrudesced in less than two hours or there was a persistent pain stimulus (for example, surgical scar, drains) a continuous infusion of fentanyl would be initiated and titrated by the attending nurse using numeric pain scale (which measures pain from 0 = no pain, to 10 = the worst pain ever experienced) aiming a value ≤ 4. If agitation had no visible cause and pain was already empirically treated with a bolus of fentanyl, then delirium would be suspected and haloperidol administrated (bolus of 2.5 or 5 mg). After 15 minutes, if the patient was still uncomfortable or agitated, a continuous infusion of midazolam or propofol would be initiated to achieve a SAS of 3 to 4. The choice between midazolam and propofol was at the discretion of the attending physician. Sedative dosing would be titrated every two hours thereafter or sooner if the patient was agitated (that is, SAS ≥ 5). Interruption of sedatives infusions would then be performed during the next shift (morning, afternoon or night) in order to try to keep the patient without sedation again.

The daily interruption group would receive sedatives (midazolam or propofol, at the discretion of the attending physician) aimed to a SAS target of 3 to 4, which means that the patient is awake or easily arousable with verbal or gentile physical stimulus. SAS would be recorded at least three times a day. Every morning, after changing shifts (7 am), sedative and opioid infusions would be interrupted by bedside nurses until patients were awake and could follow simple commands (open their eyes, look at the clinician, squeeze the hand, open their mouth). The sedative infusion would be restarted at half the previous dose only if the patient was agitated (SAS ≥ 5). If the patient could not follow commands after sedative interruption because of agitation, the infusion would also be restarted at half the previous dose and titrated to a SAS of 3 to 4, if possible. Sedative infusions would only be titrated during that day if the patient was agitated (SAS ≥ 5). Pain was evaluated with the numeric scale every two hours. It was treated the same way as in the intermittent sedation group. In case of suspected delirium, that is, agitation with no visible cause after the administration of fentanyl, haloperidol would be administrated as described above.

All patients had their sedative infusions interrupted when positive end-expiratory pressure (PEEP) was set at 5 cmH₂O and inspiratory fraction of oxygen ≤ 0.4, when a spontaneous breathing trial (T-tube trial) would be performed if the patient was hemodynamically stable (mean arterial pressure ≥ 65 mmHg, no substantial use of norepinephrine or dobutamine, respiratory rate < 35 per minute, PaO₂/FiO₂ ratio > 150). If the patient was re-intubated, he or she would be treated again according to the previously randomized group protocol.

Baseline demographic data, Acute Physiology and Chronic Health Evaluation (APACHE II) score [13], the reason for admission and for intubation were recorded for all patients. Minimal and maximal heart rate, mean arterial pressure and respiratory rate were recorded from day 1 to day 5 or extubation, whichever came first. Dosage of inotropic, vasopressors and daily fluid balance were also recorded. Ventilation variables such as ventilator mode, peak pressure, driving pressure, PEEP, tidal volume, P/F ratio and arterial pressure of carbon dioxide (PaCO₂) were also recorded from day 1 to 5. Delirium was assessed by the internal medicine resident on charge twice daily using Confusion Assessment Method for ICU (CAM-ICU) [14], also on the first seven days or until ICU discharge. Nurse workload was evaluated daily until day 5 with the Portuguese version of Nursing Activity Score (NAS) in morning shifts [15]. Total sequentialorgan failure assessment (SOFA) scores were also calculated daily until day 5 [16]. All patients who survived until hospital discharge were contacted after six months and evaluated for the level of psychological stress using the Impact of Event Scale (IES) [17].

The primary study outcome was ventilator-free days in 28 days. Secondary outcomes were ICU and hospital mortality, ICU and hospital length-of-stay, incidence of delirium, delirium and coma-free days in seven days, percentage of time on target SAS, self-extubation and reintubation, accidental removal of catheters, tracheostomy rates, total sedatives doses per patient, differences in hemodynamic and ventilator variables, SOFA score, nurse workload and level of psychological stress six months after ICU discharge.

We calculated that to provide 90% power with P-values of less than 0.05, a sample size of 106 patients would be required to detect a mean difference of 2 days of mechanical ventilation with a standard-deviation of 3.15 days in favor of the intermittent sedation approach. We intended to accomplish two interim analyzes after 35 and 70 patients, in order to evaluate the safety of the interventions (that is, self-extubations and accidental removal of catheters). We postulated that a difference in adverse events with a significance level P < 0.01 in these analyses would preclude the trial from recruiting new patients. Nevertheless, since the previewed period of enrollment ended, and only 60 patients were included, a new sample size analysis was conducted using the current data. Moreover, a sample size of 170 patients would be necessary to reach the programmed difference in mechanical ventilation-free days within 28 days. At this moment, based on the slow recruitment rate, the decision to stop the trial was made.

Data were analyzed on an intention-to-treat basis. Continuous data is presented as median (percentile 25th, percentile 75th) and compared with the Mann-Whitney U-test. Categorical data are presented as absolute values and percentages, and compared with chi-square or Fischer’s exact tests as appropriate. Interaction analyses were performed using a mixed generalized model with the patient as random factor in order to account for the within-subject correlation among repeated observations. The Markov Chain Monte Carlo procedure with 1,000 simulations to reach the equilibrium of distributions was used to retrieve a fixed probability of each resulting independent variable. The post hoc analyses for interactions were performed using the Mann-Whitney or Wilcoxon’s tests as appropriated. Kaplan-Meier survival analysis was used to assess the effects of the sedation strategy on the duration of mechanical ventilation, while log-rank tests were used to examine differences in survival curves. The R free source statistical package was used to build the graphics and to perform all the statistical analyses.