The primary outcome was duration of mechanical ventilation, beginning on Day 0 at the time of endotracheal intubation, initiation of assisted breathing for patients with tracheostomies, or PICU admission for patients intubated at an outside hospital, and continuing until the first time the endotracheal tube was continuously absent for at least 24 hours or, in patients with tracheostomies, the first time pressure support was <5 cm H2O (continuous or bi-level) for at least 24 hours. Patients were assigned 28 days if they remained intubated or were transferred or died prior to Day 28 without remaining extubated for >24 hours, therefore making the primary outcome equivalent to ventilator- 21 (link) free days. 21 (link)All secondary outcomes were selected a priori and included time to recovery from acute respiratory failure, duration of weaning from mechanical ventilation, neurological testing, PICU and hospital lengths of stay, in-hospital mortality, sedation-related adverse events, sedative exposure, and occurrence of iatrogenic withdrawal (eTable 4 ). Sedative exposure and sedation outcomes included measures of wakefulness, pain, and agitation; specifically, percentage of study days awake and calm (daily modal SBS=-1 [responsive to gentle touch or voice] or 0 [awake and able to calm]), days to first awake/calm, percentage of study days with modal pain score <4, and percentage of study days with any episode of pain (highest daily pain score ≥4) or agitation (highest daily SBS=+1/+2 [restless and difficult to calm/agitated]).
Sedation-related adverse events were also defined a priori and prospectively monitored 22 (link)(eTable 4 ). These included inadequate pain management (pain score>4 [or pain assumed present if receiving neuromuscular blockade] for 2 consecutive hours), inadequate sedation management (SBS>0 [or agitation assumed present if receiving neuromuscular blockade] for 2 consecutive hours), clinically significant iatrogenic withdrawal in patients weaning from ≥5 days of opioids (rescue therapy to manage an increase in WAT-1 symptoms), extubation failure (reintubation within 24 hours), post-extubation stridor, unplanned extubation, unplanned removal of any invasive tube, ventilator-associated pneumonia, catheter-associated bloodstream infection, immobility-related stage 2+ pressure ulcers, and new tracheostomy.
The primary analysis compared the duration of mechanical ventilation in intervention vs. control patients using Kaplan-Meier curves and proportional hazards regression adjusting for age group (2 weeks to 1.99 years, 2.00 to 5.99 years, 6.00 to 17.99 years), Pediatric Risk of Mortality (PRISM) III-12 score, 23 (link) and Pediatric Overall Performance Category (POPC)>1 24 (link) at enrollment and accounting for PICU as a cluster variable with generalized estimating equations. 25 Exploratory analyses of secondary outcomes used logistic, multinomial logistic, cumulative logit, linear, and Poisson regression accounting for PICU as a cluster variable using generalized estimating equations for binary, nominal, ordinal, continuous, and rate variables, respectively. Statistical analyses were performed with SAS software (Version 9.4, SAS Institute), using two-sided 0.05 level tests.
A priori, the study team determined that a 20% reduction in the duration of mechanical ventilation, or a hazard ratio of 1.25, was clinically important for patients managed with the sedation protocol and plausible based on our pilot study. 26 Assuming independent observations, proportional hazards between groups, and that up to 15% of patients would not be successfully extubated by Day 28, 26 ,27 (link) 1050 patients were required for a two-sided 0.05 level log-rank test to achieve 90% power to detect a 20% reduction assuming three interim analyses to assess efficacy or futility using an O’Brien-Fleming 28 (link) stopping rule (East, Version 5.3, Cytel Statistical Software). To account for the intraclass correlation coefficient (ICC) in our cluster-randomized design and using ICC=0.01 from previous experience 26 ,27 (link) and conservatively assuming 22 sites, 1990 patients were required. 29 (link),30 We chose 2448 patients as our target sample size to guarantee 90% power to detect a 20% reduction in length of ventilation controlling for patients not successfully extubated by Day 28, three formal interim analyses, and modest within-site correlations. This allowed for moderate site-to-site variability in cluster sizes and adjustment for age group, PRISM III-12 score, and POPC>1.
Sedation-related adverse events were also defined a priori and prospectively monitored 22 (link)(
The primary analysis compared the duration of mechanical ventilation in intervention vs. control patients using Kaplan-Meier curves and proportional hazards regression adjusting for age group (2 weeks to 1.99 years, 2.00 to 5.99 years, 6.00 to 17.99 years), Pediatric Risk of Mortality (PRISM) III-12 score, 23 (link) and Pediatric Overall Performance Category (POPC)>1 24 (link) at enrollment and accounting for PICU as a cluster variable with generalized estimating equations. 25 Exploratory analyses of secondary outcomes used logistic, multinomial logistic, cumulative logit, linear, and Poisson regression accounting for PICU as a cluster variable using generalized estimating equations for binary, nominal, ordinal, continuous, and rate variables, respectively. Statistical analyses were performed with SAS software (Version 9.4, SAS Institute), using two-sided 0.05 level tests.
A priori, the study team determined that a 20% reduction in the duration of mechanical ventilation, or a hazard ratio of 1.25, was clinically important for patients managed with the sedation protocol and plausible based on our pilot study. 26 Assuming independent observations, proportional hazards between groups, and that up to 15% of patients would not be successfully extubated by Day 28, 26 ,27 (link) 1050 patients were required for a two-sided 0.05 level log-rank test to achieve 90% power to detect a 20% reduction assuming three interim analyses to assess efficacy or futility using an O’Brien-Fleming 28 (link) stopping rule (East, Version 5.3, Cytel Statistical Software). To account for the intraclass correlation coefficient (ICC) in our cluster-randomized design and using ICC=0.01 from previous experience 26 ,27 (link) and conservatively assuming 22 sites, 1990 patients were required. 29 (link),30 We chose 2448 patients as our target sample size to guarantee 90% power to detect a 20% reduction in length of ventilation controlling for patients not successfully extubated by Day 28, three formal interim analyses, and modest within-site correlations. This allowed for moderate site-to-site variability in cluster sizes and adjustment for age group, PRISM III-12 score, and POPC>1.
