Early Percutaneous Tracheotomy Versus Prolonged Intubation of Mechanically Ventilated Patients After Cardiac Surgery A Randomized Trial*

Background: Whether early percutaneous tracheotomy in patients who require prolonged mechanical ventilation can shorten mechanical ventilation duration and lower mortality remains controversial. Objective: To compare the outcomes of severely ill patients who require prolonged mechanical ventilation randomly assigned to early percutaneous tracheotomy or prolonged intubation. randomization. primary end the number of ventilator-free days during the first 60 days after randomization. Secondary outcomes included 28-, 60-, or 90-day mortality rates; durations of mechanical ventilation, intensive care unit stay, and hospitalization; sedative, analgesic, and neuroleptic use; ventilator-associated pneumonia rate; unscheduled extubations; comfort and ease of care; and long-term health-related quality of life (HRQoL) and psychosocial evaluations. Results: There was no difference in ventilator-free days during the first 60 days after randomization between early percutaneous tracheotomy and prolonged intubation groups (mean, 30.4 days [SD, 22.4] vs. 28.3 days [SD, 23.7], respectively; absolute difference, 2.1 days [95% CI, (cid:1) 4.1 to 8.3 days]) nor in 28-, tracheotomy; PI prolonged intubation.

T racheotomy is a common procedure for patients who require prolonged mechanical ventilation (1,2). The development of the percutaneous dilatation technique, which allows easy and rapid bedside placement of the tracheal cannula by trained intensivists, has markedly increased its implementation in intensive care units (ICUs) over the past decade (3,4). Tracheotomy may be beneficial by lowering airway resistance, improving oral hygiene, and improving pulmonary toilet and airway security (4 -6). It might also be associated with less sedative administration (6,7), less time of heavy sedation (6,7), enhanced patient autonomy and comfort, and fewer pulmonary infections (6,8). However, translaryngeal intubation avoids local tracheotomy complications (for example, bleeding, perforation, and pneumothorax) and preserves cough strength. The effect of early tracheotomy on hospital mortality, mechanical ventilation duration, and ventilator-associated pneumonia rates of long-term ventilated ICU patients remains hotly debated (2, 6, 9 -14).
To assess whether early percutaneous tracheotomy really improves the outcomes of patients who require longterm mechanical ventilation, it must be tested on patients with prolonged mechanical ventilation (exceeding more than 14 to 21 days) whose confounding factors that might affect mechanical ventilation duration, for example, sedation and mechanical ventilation weaning policies, are tightly controlled. Prolonged mechanical ventilation after cardiac surgery is becoming more and more common (15)(16)(17) because patients referred for heart surgery have ever poorer risk profiles with more comorbid conditions, and patients still on mechanical ventilation for more than 3 to 4 days after cardiac operations are infrequently weaned within the following 7 to 10 days (16,18). As a result, this group of critically ill patients might represent the population of choice for testing potential early tracheotomy benefit.
This prospective, randomized, controlled study compared the outcomes of cardiac surgery patients who still require mechanical ventilation 4 or more days after surgery who are randomly assigned to immediate early tracheotomy or prolonged intubation with possibly late tracheotomy.

Design
This prospective, randomized, controlled, singlecenter trial, conducted from June 2006 to March 2009, included patients who required prolonged mechanical ventilation after heart surgery and were randomly assigned to immediate early tracheotomy or prolonged intubation with tracheotomy only when mechanical ventilation exceeded day 15 after randomization. Sedation and mechanical ventilation weaning policies were rigorously controlled. The local ethics committee approved the protocol. Pa-tients' relatives provided written informed consent before randomization.

Patients
Patients were eligible if they had undergone cardiac surgery; were still on mechanical ventilation 4 days thereafter; had not successfully passed a mechanical ventilation weaning screening test or spontaneous breathing trial on the day of randomization, according to the Ely protocol (19); and were expected to require mechanical ventilation for 7 or more days (16). Exclusion criteria included persons who were younger than 18 years, were pregnant, were previously enrolled in this or other trials evaluating morbidity or mortality, had received more than 48 hours of mechanical ventilation preoperatively, had previous tracheotomy within 6 months, had received an artificial heart device, had a prothrombin time greater than 1.5 times the upper limit of normal or platelet count less than 50 ϫ10 9 cells/L despite replacement, had an irreversible neurologic disorder, had a Simplified Acute Physiology Score II greater than 80 (20), or had decided to limit care. Patients with soft-tissue neck infections or anatomical deformities, making percutaneous tracheotomy unsafe, or those who had concomitant neck or carotid surgery were also excluded.
Demographic, preoperative and perioperative physiologic and radiographic (21,22) features, coexisting conditions (23,24), and mechanical ventilation characteristics at randomization were recorded. Preoperative predicted operative mortality was calculated by using the European System for Cardiac Operative Risk Evaluation (25).

Randomization and Intervention
After screening for inclusion and exclusion criteria on the day of randomization, we used an independent, computer-generated randomization sequence (Unité de Recherche Clinique, Pitié-Salpêtrière Hospital, Paris, France) to assign patients, in a 1:1 ratio, to either early tracheotomy (before the end of calendar day 5 after surgery) or prolonged intubation. Randomization was stratified (minimization) by the Simplified Acute Physiology Score II (either Յ45 or Ͼ45) calculated on the day of randomization; the randomization procedure was password-protected and accessed by the principal investigators or study coordinator after the patient had met selection criteria and the surrogate gave consent. The patient's initials were entered, and treatment allocation was assigned.
Tracheotomies were done bedside by experienced intensivists in the ICU using the Ciaglia percutaneous technique (Ciaglia Blue Rhino, Cook Critical Care, Bloomington, Illinois). Any side effect or technical problem that occurred during the procedure was recorded. All patients were managed with goal-directed sedation, guided by the Richmond Agitation Sedation Scale (RASS) (26). Only 2 sedatives (propofol or midazolam) and 1 narcotic (sufentanil) were used. Sedation was monitored and evaluated by nurses 8 times daily to maintain that the patient was calm

Context
It is not clear whether early tracheotomy improves the outcome of patients who are expected to require prolonged mechanical ventilation.

Contribution
In this randomized trial of patients who still required mechanical ventilation 4 days after cardiac surgery, immediate tracheotomy did not decrease mortality or length of intensive care unit or hospital stay or increase the number of days off the ventilator compared with waiting 2 weeks to consider tracheotomy. Early tracheotomy did, however, reduce the use of sedatives, ease nursing care, and improve patient comfort.

Caution
Patients in the control group were free of mechanical ventilation sooner than anticipated, possibly limiting the ability to detect a benefit from early tracheotomy.

Implication
More study is required to determine whether routine early tracheotomy is beneficial.

-The Editors
Original Research Early Versus Late Tracheotomy After Cardiac Surgery and cooperative or lightly sedated (respective RASS scores, 0, Ϫ1, and Ϫ2). Our ICU's nurse-patient and nurse's aide-patient ratios were always 1:2.25 and 1:5 respectively. When the treating physician considered the sedation level too profound (RASS score, Ϫ3 to Ϫ5) and not justified by the severity of the patient's respiratory distress, the study protocol recommended 50% reduction of the initially prescribed sedative dose. Sedatives, narcotics, or both were interrupted daily at 9 a.m. (27). Sedative infusions were stopped when the patient remained calm and cooperative or, if agitation prevented successful weaning, were restarted at half the previous dose and adjusted according to need. After stopping intravenous sedation, oral benzodiazepines, neuroleptics, or both were prescribed as needed. In particular, haloperidol (1-to 5-mg increments) was used to treat agitation, delirium, or both.
For both groups, mechanical ventilation weaning was conducted according to a strict protocol, beginning as early as 1 day after randomization (Appendix, available at www.annals.org).

Outcomes and Follow-up
The primary end point was the number of days alive and breathing without assistance (ventilator-free days) during the first 60 days after randomization. Ventilator-free days were counted from the last day that a patient received mechanical ventilation during the 60-day period. In particular, patients not receiving mechanical ventilation who died before day 60 were assigned 0 ventilator-free days, and each day a patient required noninvasive mechanical ventilation for more than 4 hours after extubation or decannulation was counted as 1 mechanical ventilation day, therefore resetting ventilator-free days at 0. However, days with the tracheotomy cannula in place but without mechanical ventilation before decannulation were counted as ventilator-free days. Sensitivity analyses were also done assigning ventilator-free days when death occurred before day 60 and using the exact number of days of mechanical ventilation for counting ventilator-free days (Appendix).
Secondary end points included the number of ventilator-free days at 28 and 90 days (based on data through 28 and 90 days); 28-, 60-, and 90-day mortality rates; durations of mechanical ventilation and length of ICU and hospital stays; the number of endotracheal prosthesis-free days at day 60; frequencies of unscheduled extubations, decannulations (tracheal cannula removed), and reintubations or recannulations; 7-, 14-, 21-, and 28day Sequential Organ Function Assessment (28) scores in the ICU; and durations of vasopressor (exclusively epinephrine or norepinephrine) and renal replacement therapy. The sedatives, analgesics, and neuroleptics used and their cumulative daily doses during the first 15 days were recorded. Sedation-free days at day 28 were calculated. In addition, RASS scores were recorded 8 times every 24 hours during days 1 to 15 after randomization. Three sedation status categories based on RASS scores were defined a posteriori: heavily sedated (RASS score, Ϫ5, Ϫ4, or Ϫ3); calm, awake, or lightly sedated (RASS score, Ϫ2, Ϫ1, 0, or ϩ1); and agitated (RASS score, ϩ2, ϩ3, or ϩ4). The number of hours per day spent in each of these consciousness states was calculated. If 2 consecutive scores were different (for example, heavily sedated at 9 a.m. and awake and calm at 12 p.m.), the patient was considered to have had the former score 50% of the time and the latter score for 50% (that is, 90 minutes heavily sedated and 90 minutes awake and calm). If the same sedation level was recorded twice, the patient was considered to have remained at this level during the time between these 2 assessments. Analyses of time spent in each consciousness state were based on these daily scores.
Ventilator-associated pneumonia was suspected when a new and persistent radiographic infiltrate was accompanied by purulent secretions, temperature of 38.3°C or more, or a leukocyte count greater than 10.0 ϫ 10 9 cells/L, or when 1 of these occurred in patients with baseline diffuse, dense infiltrates. Ventilator-associated pneumonia was diagnosed before administration of antibiotics by quantitative distal bronchoalveolar lavage cultures growing at 10 4 colony-forming units/mL or greater (29,30). Other complicating infections (bloodstream, sternal wound, and tracheotomy stomal site, which required antibiotics, surgical debridement, or both) and duration of antibiotic treatment were also recorded.
Nurses evaluated ease of care and patient comfort daily by using subjective scores; a trained physiotherapist evaluated muscle strength with the Medical Research Council score; and laryngeal, tracheal, and other complications were also recorded (Appendix).

Long-Term Follow-up, Health-Related Quality of Life, and Psychosocial Evaluations
We designed a posteriori a cross-sectional study of long-term outcomes, which was conducted from May 2010 to June 2010, on all patients who were alive on day 90. After explaining the study objective and asking for informed consent during a telephone call, we administered the following questionnaires to survivors: activities of daily living (31, 32); 36-Item Short Form Health Survey (SF-36) (33, 34) evaluating health-related quality of life (HRQoL); the Hospital Anxiety and Depression Scale (35,36); and the Impact of Event Scale (37), which assessed posttraumatic stress disorder (Appendix).
Lastly, patients were asked if the endotracheal tube or cannula had caused pain or discomfort; if they had swallowing problems, phonation problems, or both after ICU discharge; and if the tracheotomy scar engendered aesthetic embarrassment.

Statistical Analysis
On the basis of personal data, we predicted the prolonged intubation group would have a mean of 23 days (SD, 17) of ventilator-free days in the first 60 days. To demonstrate that early tracheotomy achieved an absolute increase in ventilator-free days over 7 days, with 80% power and 5% type I error, 198 patients were required (38). To account for persons who withdrew, 216 patients were enrolled. No interim analysis was planned, but an independent safety monitoring board reviewed the trial's progress and evaluated adverse events according to treatment assignment. All randomly assigned patients were included in the analyses according to their randomized treat-ment assignment, all patients received the allocated intervention, and none was lost to follow-up during the first 90 days. Patients randomly assigned to the prolonged intubation group who had late tracheotomy were always analyzed in the prolonged intubation group. Binary variables were compared with chi-square tests and continuous variables with the t test, as appropriate. The between-group differences were assessed by using risk difference or mean differences, presented with 95% CIs. Post hoc adjustment for between-group differences in baseline characteristics (repeated operation, heart transplantation, and renal replacement therapy) was achieved with multivariable regression analyses. Sedation administration and number of hours spent in each consciousness state (heavily sedated; calm, awake, or lightly sedated; or agitated) were evaluated daily during days 1 to 15 after randomization. Comparisons between groups were done using a mixed model in which intervention group and day were included as fixed effects while random effects for the subject was taken into account through the repeated measures over time. Paired t tests or Wilcoxon tests were used to compare survivors' mean SF-36 scores with age-and sex-matched French population normative values (33,34). For survival, follow-up started at randomization, and cumulative event curves were estimated with the Kaplan-Meier method; survival curves were compared with the logrank test. All tests were 2-sided. We used SAS software, version 8.02 (SAS Institute, Cary, North Carolina), for statistical analyses.

Role of the Funding Source
This study was sponsored by an academic grant from the French Ministry of Health (Programme Hospitalier de Recherche Clinique régional 2005). The study sponsor did not participate in the study design, data collection, data analysis, data interpretation, or writing and the decision to submit this manuscript for publication.

RESULTS
Among the 3484 consecutive patients who had undergone heart surgery from June 2006 to March 2009, a total of 287 were still intubated 4 or more days after surgery, and 216 were enrolled in the study: 109 assigned to early tracheotomy and 107 to prolonged intubation (Figure 1). None was lost to follow-up on day 90. Table 1 and Appendix Table 1 (available at www.annals.org) show baseline characteristics of enrolled patients. Characteristics at randomization were similar for the 2 groups, except for higher rates of heart transplantation, repeated cardiac surgery, and renal replacement therapy in the early tracheotomy group. Twenty-nine (27%) patients in the prolonged intubation group had late tracheotomy because of the expected need for more prolonged mechanical ventilation.
There was no difference in ventilator-free days during the first 60 days after randomization ( Table 2) between   (Figure  2, top); ventilator-free days on days 28 and 90; endotracheal prosthesis-free days; durations of mechanical ventilation and ICU and hospital stays; rates of ventilatorassociated pneumonia, stomal or sternal infections, or positive blood cultures; and number of patients who were administered catecholamine or antibiotics ( Table 2 and Appendix Table 2, available at www.annals.org) were also similar. Different definitions of ventilator-free days did not alter these findings (Appendix Table 3, available at www.annals.org). However, patients in the prolonged intubation group had more frequent unscheduled extubations and reintubations or recannulations. After adjustment for repeated operations, heart transplantations, and renal replacement therapy at baseline, between-group differences in primary and secondary outcomes remained similar (Appendix Table 4, available at www.annals.org). Intravenous sedative and analgesic use at randomization were similar but decreased more rapidly thereafter for early tracheotomy patients (Figure 3). The early tracheotomy group had lower cumulative sufentanil, propofol, and midazolam consumption during days 1 to 15 after randomization and significantly more sedation-free days during days 1 to 28 ( Table 2). More haloperidol was used to treat agitation, delirium, or both in the prolonged intubation group. During days 1 to 15 after randomization, 20 290 sedation scores were recorded. Patients in the early tracheotomy group spent less time heavily sedated; more time calm, awake, or lightly sedated (Appendix Figure 1, available at www.annals.org); and more days comfortable and with care deemed "easy" according to the nurses' subjective scores ( Table 2). Patients in early tracheotomy group received oral nutrition and were transferred from bed to chair sooner than patients in the prolonged intubation group. By day 15, more of the patients in the early tracheotomy group were receiving oral nutrition and undergoing bed-to-chair transfers. Muscle strength scores were similar between the 2 groups ( Table 2). One patient in each group had a major tracheotomy or intubation complication, but both had favorable outcomes. The numbers and severities of other adverse events were similar for the 2 groups (Appendix Table 5, available at www.annals.org).
At long-term follow-up of 873 days after randomization (interquartile range, 547 to 1201 days; range, 93 to 1491 days), 29 patients had died, 3 were lost to follow-up, and 119 were long-term survivors (Figure 1). According to the Kaplan-Meier method, between-group long-term survival probability was similar (Figure 2, bottom). No between-group differences were found for activities of daily living, anxiety, depression, or posttraumatic stress disorder ( Table 3). Both groups had similar HRQoL scores at longterm follow-up compared with age-and sex-matched population normative values; however, they had worse physical functioning scores (Appendix Figure 2, available at www .annals.org). Severe anxiety and depression symptoms were present in 16% to 30% of long-term survivors ( Table 3). Only 1 survivor in each group was deemed at risk for posttraumatic stress disorder. Only 3 patients in the early tracheotomy group and 5 in the prolonged intubation group remembered feeling pain or discomfort associated with endotracheal prosthesis. None of the long-term survivors had problems with swallowing, phonation, or both, and only 1 patient reported being embarrassed by the tracheotomy scar.

DISCUSSION
In this large, randomized, controlled trial of patients who require prolonged mechanical ventilation after cardiac surgery, early tracheotomy provided no benefit in terms of mechanical ventilation duration, length of hospital stay, mortality rate, or frequency of infectious complications over prolonged intubation possibly followed by late tracheotomy. However, it was associated with less intravenous sedation; less time of heavy sedation; less haloperidol use to treat agitation, delirium, or both; fewer unscheduled extubations and reintubations; better comfort; and earlier oral nutrition and bed-to-chair transfers. After a median follow-up exceeding 2 years, survival, psychosocial variables, and HRQoL did not differ between groups, with the latter being similar to that of age-and sex-matched French population normative values. Therapeutic practices decreasing sedation and improving whole-body rehabilitation of patients on mechanical ventilation were recently shown to limit delirium episodes (39,40) and shorten the duration of mechanical ventilation (27,40). As previously reported by our group (7) and others (6), early tracheotomy was associated with less sedative and analgesic administration; less time of heavy seda-tion; and earlier oral nutrition, out-of-bed mobilization, or both. Less haloperidol use might also indicate that less sedation and analgesia led to less agitation, delirium, or both, although this was not specifically assessed herein. However, the early tracheotomy group's 20% significantly higher number of sedation-free days did not translate into a shorter duration of mechanical ventilation and more ventilator-free days. Possible explanations are that polyneuropathy acquired in the ICU might have prolonged mechanical ventilation for both groups or the strict mechanical ventilation weaning protocol applied to the prolonged intubation group might have compensated for higher sedative administration.
Less sedative consumption has also been postulated to improve long-term outcomes of patients surviving critical care (41). Despite lower use and better nurse-assessed comfort levels during critical care, we found no difference in  Original Research Early Versus Late Tracheotomy After Cardiac Surgery long-term, between-group HRQoL. However, we cannot exclude the possibility that early tracheotomy might provide short-term HRQoL and emotional health benefits because evaluations were conducted long after ICU discharge. Psychosocial variables and HRQoL did not differ from those of French population normative values, confirming that the long-term outcomes of patients who require prolonged mechanical ventilation after car-diac surgery are better than those of other ICU patients, probably because of the surgical cure of severe coronary or valve diseases (17,18,42). However, although the frequency of posttraumatic stress disorder was negligible, 15% to 30% of long-term survivors had signs of anxiety, depression, or both and therefore might benefit from strategies aimed at attenuating emotional and psychological distress.  Another potential benefit attributed to early tracheotomy is a lower ventilator-associated pneumonia rate (6,43,44), which might result from a combination of factors, for example, easier oral hygiene and bronchopulmonary toilet or less time spent deeply sedated. Although extreme vigilance for ventilator-associated pneumonia was maintained throughout our study, early tracheotomy did not modify the frequency of this infection. Four other trials (45)(46)(47)(48) and the randomized Early Versus Late Tracheotomy Trial, whose primary end point was ventilator-associated pneumonia (14), obtained similar results. Early tracheotomy also did not affect the rates of other infectious complications after cardiac surgery (for example, bloodstream infection or mediastinitis).
The effect of early tracheotomy on the outcomes of patients requiring prolonged mechanical ventilation has been the matter of heated debates, with fewer ICU and hospital deaths after tracheotomy observed in some retrospective, cohort studies (9,10,13,49,50) but not others (44 -48). A prospective, randomized study on a mixed population of 120 ICU patients established that patients who had early tracheotomy had lower mortality rates in the ICU and hospital and spent less time on mechanical ventilation (6). However, a recent study of 419 mechanically ventilated, mixed medical and surgical ICU, adult patients demonstrated that early tracheotomy did not lower the frequency of ventilator-associated pneumonia, affect the length of hospital stay or 28-day and 1-year mortality, or influence the need for care at a long-term health facility but did increase the number of ICU-free days and ventilator-free days (14). Our analysis of 216 patients with projected prolonged mechanical ventilation also found no mortality benefit for early tracheotomy. The results of another randomized, controlled trial of early tracheotomy in ICU patients should help to further clarify the role of tracheotomy in critically ill patients (51).
Several limitations of our study should be noted. First, it was conducted within a single center, and hence, our results may not be applicable to patients who receive mechanical ventilation in other centers with different case mixes, sedation practices, mechanical ventilation weaning strategies, and ICU discharge policies. Conversely, this monocentric design might be an advantage, with minimization of biases, because of the population's relative homogeneity and our application of strict protocols for those practices. Second, according to personal data, we had based our study sample calculation on the hypothetical mean of 23 ventilator-free days (SD, 17) during the first 60 days after randomization for the prolonged intubation group and an expected benefit of 30% more ventilator-free days for the early tracheotomy group, which was considered a meaningful clinical difference. The prolonged intubation group had a higher mean number of ventilator-free days during the first 60 days after randomization than predicted, close to the expected benefit in the early tracheotomy group. Although chance cannot be excluded, another possible explanation is the implementation of modifications of our ICU practices during the 2 years between the pilot study and this trial's start. The higher number of ventilator-free days observed might also reflect the standardized weaning and sedation strategies for mechanical ventilation and more focused attention on sedation reduction and early identification of mechanical ventilation  Original Research Early Versus Late Tracheotomy After Cardiac Surgery weaning criteria for patients included in this trial. Third, 27% of patients in the prolonged intubation group had late tracheotomy, and we cannot exclude that it might have diluted the overall early tracheotomy effect on the entire cohort. Lastly, because tracheal examinations were not done systematically in the weeks after extubation or decannulation, we perhaps missed laryngotracheal complications related to tracheotomy (or possibly intubation). However, rates of tracheal stenosis, tracheal symptoms, or both (for example, dysphonia or swallowing disorders) were low and equally distributed in patients with or without tracheotomy in previous studies (6,47).
In conclusion, early tracheotomy in patients who require prolonged mechanical ventilation after cardiac surgery provided no benefit in terms of mechanical ventilation duration, length of hospital stay, and mortality and infectious complication rates but was associated with diminished sedative, analgesic, and neuroleptic consumption and better comfort, ease of care, and earlier oral nutrition and bed-to-chair transfers. Future studies might explore the potential contribution of later (after mechanical ventilation day 10 to 15) tracheotomy.
Reproducible Research Statement: Study protocol: Available at www .reamedpitie.com. Statistical code: Available from Dr. Guiguet (e-mail, mguiguet@ccde.chups.jussieu.fr). Data set: Certain portions of the analytic data set are available to approved individuals through written agreements with the authors.

Mechanical Ventilation Weaning Protocol
For both groups, mechanical ventilation weaning was conducted according to a strict protocol, beginning as early as the first day after randomization. Weaning criteria were defined as regression or resolution of the underlying cause of acute respiratory failure; oxygen saturation greater than 90%, with a fraction of inspired oxygen (FIO 2 ) less than 40% and a positive endexpiratory pressure of 5 cm H 2 O or more; correction of electrolyte anomalies; a good level of consciousness; no further need for high doses of vasoactive and sedative agents, and respiratory frequency to tidal volume ratio less than 105 after 1 minute of spontaneous breathing (19). Patients fulfilling these criteria had a 1-hour spontaneous breathing trial (T-piece trial), and those who successfully passed the trial were extubated within 6 hours or temporarily left with the tracheotomy cannula for 24 or 48 hours without mechanical ventilation before decannulation. If they fulfilled clinical poor tolerance criteria (respiratory rate Ͼ35 breaths/min or increased by 50%, use of accessory respiratory muscles, diaphoresis, heart rate, or blood pressure increased by Ͼ20% or altered consciousness), confirmed by blood gas anal-ysis, assist control ventilation was reintroduced. T-piece trials were repeated every day until extubation or decannulation. When a patient developed signs of respiratory failure after extubation or decannulation (respiratory rate Ͼ30 breaths/min, oxygen saturation as measured by pulse oximetry Ͻ90%, or hypercapnia), noninvasive mechanical ventilation was applied by using the pressure support method for 1-to 2-hour sessions.

Outcomes and Follow-up
Ventilator-free days were counted from the last day that a patient received mechanical ventilation during the 60-day period. In particular, patients who did not receive mechanical ventilation and died before day 60 were assigned 0 ventilator-free days, and each day a patient required noninvasive mechanical ventilation for more than 4 hours after extubation or decannulation was counted as 1 mechanical ventilation day, therefore resetting ventilator-free days at 0. Sensitivity analyses with 3 other definitions of ventilator-free days were also done. Definition 2: counting ventilator-free days from the last day that a patient received mechanical ventilation and assigning ventilator-free days when death occurred before day 28, 60, or 90. Definition 3: using the exact numbers of days on mechanical ventilation to calculate ventilator-free days and assigning 0 ventilator-free days when death occurred before day 28, 60, or 90. Definition 4: using the exact numbers of days on mechanical ventilation to calculate ventilator-free days and assigning ventilator-free days when death occurred before day 28, 60, or 90.

Ease of Care, Patient Comfort, and Muscle Strength Assessment
Nurses evaluated the ease of care daily by using a subjective score specifically designed for this trial. Results are expressed according to a numerical scale ranging from very easy (score ϭ 1) to very difficult (score ϭ 4). Two ease-of-care categories were defined a posteriori: easy (score ϭ 1 or 2) and difficult (score ϭ 3 or 4). Nurses also assessed the patient's comfort with a similar scale specifically designed for this trial ranging from very comfortable (score ϭ 1) to very uncomfortable (score ϭ 4). Similarly, 2 patient comfort categories were defined a posteriori: comfortable (score ϭ 1 or 2) and uncomfortable (score ϭ 3 or 4). In addition, durations of enteral feeding through a nasogastric tube, central venous catheter, and bladder catheterization and the times to the first bed-to-chair transfer, the first oral nutrition, definitive urinary catheter removal, and central venous catheter removal were recorded.
Muscle strength was evaluated by using the Medical Research Council score. A trained physiotherapist assessed muscle groups of the upper (shoulder, elbow, wrist, and hand) and lower (hip, knee, and ankle) limbs. Each muscle group score ranges from 0 (paralysis) to 5 (normal muscle strength). The global score, obtained by adding all the muscle group scores, ranges from 0 to 280. In an attempt to limit interexaminer variability, only 1 investigator (the same qualified physiotherapist) made 85% of the assessments. This score was determined the day of inclusion and then every 14 days until the ICU discharge.

Adverse Events
Late laryngeal and tracheal complications were systematically evaluated during fibroscopy by the intensivist 1 to 5 days after extubation or decannulation for the first 29 patients. For the subsequent patients, this exploration was done only when unusual clinical symptoms or an evident complication developed and completed with an examination by an ear, nose, and throat specialist, when necessary. All serious adverse events not related to tracheotomy were also recorded.

Long-Term Follow-up, HRQoL, and Psychosocial Evaluations
A cross-sectional study of long-term outcomes, designed a posteriori, was conducted from May 2010 to June 2010 on all patients who were alive on day 90. After explaining the purpose of the study and asking for informed consent during a telephone call, we administered the following questionnaires to survivors.
First, activities of daily living were assessed with the basic and instrumental scales (31,32). The basic scale assesses the level of difficulty and receipt of assistance with bathing, dressing, toileting, getting in or out of bed or chairs, controlling bowel and bladder continence, and eating, and its total score ranges from 0 (major) to 6 (no) disability. The instrumental scale contains 8 questions (ability to use the telephone, shop, prepare food, maintain a home, do laundry, drive a car or take a bus or a taxi, be responsible for own medications, and handle finances), and its total score ranges from 0 (major) to 8 (no) disability. Then, HRQoL was assessed with the French version of the SF-36 (33). Its 36 items are combined to evaluate 8 domains (physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health). The aggregate physical and mental component summary measures were then computed, as recommended by developers. Individual component and overall physical and emotional domain scores range from 0 (poor) to 100 (excellent). Our patients' mean SF-36 levels were compared with age and sex-matched French population normative values (34).
Anxiety and depression symptoms were assessed with the Hospital Anxiety and Depression scale (35), which contains 14 questions-7 to assess anxiety (subscale A) and 7 to assess depression (subscale D)-and uses a 4-point Likert scale (range, 0 to 3), giving a possible score of 0 (none) to 21 (severe) for each of the 2 subscales. Subscale scores of 8 or more indicated clinically significant anxiety or depression (35,36).
Symptoms related to posttraumatic stress disorder were assessed with the Impact of Event Scale (37), which includes 15 questions divided into 2 subscales: intrusion (7 items) and avoidance (8 items). The total score ranges from 0 (no symptoms) to 75 (severe symptoms). In agreement with previous reports, patients with a total score of 30 points or more were considered at high risk for symptoms of posttraumatic stress disorder (36).

Results: Adverse Events
Only 1 major complication related to tracheotomy (that is, perforation of the posterior tracheal wall) occurred in the early percutaneous tracheotomy group. The procedure was stopped; the patient was temporarily left intubated; and a tracheotomy was done the next day by an ear, nose, and throat surgeon. Alternatively, a large tracheal tear during the initial intubation for cardiac surgery was diagnosed in the prolonged intubation group 5 days after randomization. The endotracheal tube was left in place, and a percutaneous tracheotomy was done on day 16 after randomization without any subsequent complication. Both patients with initial tracheal complications had favorable local outcomes and were alive on day 90. Only 1 stomal infection and no significant hemorrhage were related to the tracheotomy procedure. Laryngeal symptoms (for example, swallowing disorders or dysphonia) and abnormalities at laryngeal examination were rare and similar for the 2 groups. In addition, the numbers and severities of other adverse events did not differ between the 2 groups (Appendix Table 4).