Association of Early vs Late Tracheostomy Placement With Pneumonia and Ventilator Days in Critically Ill Patients

Key Points

Question

Is the timing of tracheostomy placement in critically ill patients associated with the rate of ventilator-associated pneumonia and duration of mechanical ventilation?

Findings

This meta-analysis assessed findings from 17 randomized clinical trials with 3145 participants and found that early tracheotomy in adults undergoing ventilator support for critical illness was associated with improved clinical outcomes.

Meaning

These findings suggest that tracheostomy placement no more than 7 days after ventilator support may lower the rates of ventilator-associated pneumonia and ventilator duration.

This meta-analysis compares clinical outcomes of early vs late tracheotomy among patients undergoing ventilator support for critical illness.

Abstract

Importance

The timing of tracheostomy placement in adult patients undergoing critical care remains unestablished. Previous meta-analyses have reported mixed findings regarding early vs late tracheostomy placement for ventilator-associated pneumonia (VAP), ventilator days, mortality, and length of intensive care unit (ICU) hospitalization.

Objective

To compare the association of early (≤7 days) vs late tracheotomy with VAP and ventilator days in critically ill adults.

Data Sources

A search of MEDLINE, CINAHL, Cochrane Central Register of Controlled Trials, references of relevant articles, previous meta-analyses, and gray literature from inception to March 31, 2020, was performed.

Study Selection

Randomized clinical trials comparing early and late tracheotomy with any of our primary outcomes, VAP or ventilator days, were included.

Data Extraction and Synthesis

Two independent reviewers conducted all stages of the review. The Preferred Reporting Items for Systematic Reviews and Meta-analyses guideline was followed. Pooled odds ratios (ORs) or the mean difference (MD) with 95% CIs were calculated using a random-effects model.

Main Outcomes and Measures

Primary outcomes included VAP and duration of mechanical ventilation. Intensive care unit days and mortality (within the first 30 days of hospitalization) constituted secondary outcomes.

Results

Seventeen unique trials with a cumulative 3145 patients (mean [SD] age range, 32.9 [12.7] to 67.9 [17.6] years) were included in this review. Individuals undergoing early tracheotomy had a decrease in the occurrence of VAP (OR, 0.59 [95% CI, 0.35-0.99]; 1894 patients) and experienced more ventilator-free days (MD, 1.74 [95% CI, 0.48-3.00] days; 1243 patients). Early tracheotomy also resulted in fewer ICU days (MD, −6.25 [95% CI, −11.22 to −1.28] days; 2042 patients). Mortality was reported for 2445 patients and was comparable between groups (OR, 0.66 [95% CI, 0.38-1.15]).

Conclusions and Relevance

Compared with late tracheotomy, early intervention was associated with lower VAP rates and shorter durations of mechanical ventilation and ICU stay, but not with reduced short-term, all-cause mortality. These findings have substantial clinical implications and may result in practice changes regarding the timing of tracheotomy in severely ill adults requiring mechanical ventilation.

Introduction

Tracheotomy is a commonly performed procedure for patients requiring prolonged mechanical ventilation.¹ Benefits from a tracheotomy include patient comfort and less exposure to sedatives.^2,3 Evidence also suggests that tracheotomies improve pulmonary recruitment and decrease the length of hospitalization in severely ill patients.^4,5 Accordingly, earlier tracheostomy placement may provide benefit in adults requiring prolonged assisted ventilation.

Despite multiple trials, the optimal timing of tracheostomy placement is still an area of contention. Previous meta-analyses assessing early vs late tracheotomy^6,7,8,9,10 have demonstrated differing results on clinical outcomes. For instance, no difference was found for the duration of mechanical ventilation, incidence of ventilator-associated pneumonia (VAP), or short-term mortality between early vs late tracheotomy in studies by Szakmany et al¹¹ and Hosokowa et al.¹² However, a study by Siempos and colleagues⁶ showed that early tracheotomy lowered the incidence of VAP, but no difference was noted for short-term mortality. Similar to previous findings, a Cochrane review¹³ concluded that mortality at 28 days was similar between early and late tracheotomy, yet survival was improved in the early intervention group when observed at long-term follow-up.

Starting in 2014, several large-scale trials^2,14,15 have attempted to untangle these questions. These trials have not been incorporated in most meta-analyses, and incorporating these studies may address the inconsistencies reported on the topic. Thus, our objective was to collate, critically appraise, and analyze all randomized clinical trials examining the effects of timing of tracheotomy on the primary outcomes of incidence of VAP and duration of mechanical ventilation and secondary outcomes of short-term, all-cause mortality and intensive care unit (ICU) days.

Methods

We conducted a systematic review and meta-analysis according to recommendations from the Cochrane Handbook for Systematic Reviews of Interventions and adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria. The protocol was submitted to PROSPERO international prospective register of systematic reviews.

Search Strategy

Two investigators (K.C. and A.M.) systematically searched MEDLINE, CINAHL (Cumulative Index to Nursing and Allied Health Literature), and the Cochrane Central Register of Controlled Trials from inception of the database to March 31, 2020. Search terms included (tracheostomy OR tracheotomy) AND (early OR late OR timing) (eMethods in the Supplement). Searches within MEDLINE and Cochrane Registry were filtered for clinical trials. Furthermore, we manually searched references from the retrieved articles and reviewed conference proceedings.

There was no limit for language, location, or sample size for included studies. Last, we reviewed previously published meta-analyses on early vs late tracheotomy and supplemented additional articles that were not found in our original search. Two investigators (K.C. and A.M.) independently reviewed the titles and abstracts of all citations to determine suitability based on our primary outcomes. This was followed by independent review of the full-text articles to confirm eligibility. Disagreements were reviewed by a third author (A.H.).

Study Selection

Studies were included if they were randomized clinical trials that compared early vs late tracheotomy and reported the incidence of VAP or ventilator days. Studies that randomized patients to no tracheotomy (eg, prolonged intubation) were assigned to the late tracheotomy group. We included all approaches (eg, percutaneous vs surgical tracheotomy) and accepted studies that enrolled critically ill adult patients (≥18 years) admitted to any ICU setting (eg, medical, surgical, or burn). We excluded studies in children and neonates because the threshold for tracheostomy placement in this population differs from that for adults. Retrospective studies, case reports, clinical overviews, editorials, commentaries, and practice guidelines were also excluded.

Data Extraction

Two authors (K.C. and A.M.) independently extracted study data specific for patient characteristics, interventions, and clinical outcomes using a standardized collection form in Excel, version 2007 (Microsoft Corporation). Disagreements were resolved after thorough discussion with the panel of investigators. We collected the following data: first author, country of origin, study period, clinical setting, sample size, age of patients, sex, and day of tracheostomy placement.

Risk of Bias

To assess the risk of bias in randomized clinical trials (RCTs), the investigators used the Cochrane Collaboration tool. This guide is structured into 5 domains that evaluate different aspects of trial design, conduct, and reporting. The domains are characterized by a series of questions examining selection, performance, detection, attrition, and reporting bias. The risk of bias in each domain can be judged as high, low, or unclear. Studies were deemed as having moderate risk of bias if they had 2 high-risk components and high risk of bias with 4 or more high-risk components. Studies with low risk of bias and those with no more than 3 risk of bias domains judged as unclear were included in the meta-analysis.

Definitions and Outcomes

Early tracheotomy was defined as intervention no more than 7 days after initiation of mechanical ventilation. We defined late as tracheostomy placement after 7 days or no tracheotomy. If a study defined early tracheotomy after 7 days, we did not include the study in this review. We had 2 distinct primary outcomes: VAP and mechanical ventilator days. Studies that provided data on at least 1 of our primary outcomes were included in the review. If the primary outcome was assessed multiple times during the hospitalization (eg, 30-day VAP vs VAP during the entire hospitalization), we chose the earliest point. Secondary outcomes included (1) short-term, all-cause mortality, defined as death in the first 30 days of hospitalization, and (2) ICU days.

Statistical Analysis

Data from each study were tabulated and checked by investigators (K.C. and A.M.). Continuous data were recorded as mean with SD. If outcome data were presented as median with interquartile range, we calculated the mean (SD) per Wan et al.¹⁶

For categorical data, we aggregated the number of patients who had each outcome per total number of patients in that group. Continuous variables were presented as mean differences (MDs) with 95% CI, whereas dichotomous variables were presented as pooled odds ratios (ORs) with 95% CI. All analyses used random effects (DerSimonian and Laird approach) with significance defined as 2-sided P < .05.¹⁷

Given the range of participants, inclusion criteria, timing of procedure, and outcome measures, we expected our review to have variability (eg, heterogeneity). As such, we calculated the statistical heterogeneity using the I² statistic, which provides a percentage of the variability in effect estimates. If I² was ≥50%, we performed a sensitivity analysis (eg, recalculated the meta-analysis) by removing 1 article at a time (guided by highest I²) until the sensitivity was below our threshold of 50%.¹⁸ To evaluate publication bias, we created funnel plots. Funnel plots scatter the studies according to their size, effect estimate, and SE of the effect estimate. Because precision improves with larger studies, their representation in funnel plots is typically scattered toward the peak. Ideally, the plot should have a symmetrical distribution (eg, low publication bias). All statistical analyses were performed in R, version 3.6.2 (R Program for Statistical Computing).

Results

Identification of Eligible Studies

Our electronic search yielded 474 publications, of which 47 were reviewed in full. A total of 17 trials^{3,4,14,19,20,21,22,23,24,25,26,27,28,29,30,31,32} met inclusion criteria and were described in the systematic review. Meta-analyses were performed in the 14 studies^{3,4,14,22,23,24,25,26,27,28,29,30,31,32} that were deemed low risk of bias. The flow diagram of selected articles is shown in Figure 1.

Figure 1. PRISMA Flow Diagram.

Study Characteristics

The Table provides study details of the combined population of 3145 patients (1619 in the early tracheotomy group vs 1526 in the late tracheotomy group). Fourteen trials were conducted in the medical and/or surgical ICU, and 9 studies^{14,22,23,24,25,26,28,30,31} (52.9%) were performed in Europe. Patients included had a variety of indications for being intubated, and trials varied with respect to definitions of early vs late tracheotomy. The patient population was predominantly male (1975 of 3070 [64.3%], with 2 trials not reporting sex) with a mean (SD) age ranging from 32.9 (12.7) to 67.9 (17.6) years.

Table. Characteristics of Individual Trials, Patient Populations, and Interventions (Early vs Late Tracheotomy).

Source	Criteria		Patient characteristics, early vs late tracheotomy groupsa	Severity scoring system, early vs late tracheotomy groupsa	Center (No. in early/late tracheotomy group)
	Inclusion	Exclusion
Blot et al,25 2008	Age >18 y; expected intubation >7 d	Previous tracheotomy or enrollment in the trial; major risk of bleeding; infection or anatomical deformity of the neck; severe respiratory insufficiency or neurological failure; high severity of illness scores	Age: 54.3 (14.9) vs 56.0 (14.6) y Reason for admission: respiratory failure (34% vs 32%); neurology (21% vs 24%); trauma (18% vs 19%)	SAPS II: 50 (17-103) vs 50 (15-96)b	Multiple centers (61/62)
Bösel et al,26 2013	Age ≥18 y; admission to neuro-ICU; nontraumatic ICH/SAH/acute ischemic stroke; expected intubation ≥2 wk	Ventilation >3 d; severe chronic cardiopulmonary comorbidities; anatomical or clinical conditions jeopardizing PDT; expected to require a permanent ST; enrolled in other trials; life expectancy <3 wk; pregnant	Age: 61 (12) vs 61 (13) y Reason for admission: ICH (43% vs 43%); acute ischemic stroke (37% vs 30%); SAH (20% vs 27%)	SET: 13 (10-14) vs 13 (11-16)c GCS: 9 (7-11) vs 8 (5-10)d NIHSS: 21 (15-23) vs 20 (11-33)e APACHE II: 17 (13-19) vs 16 (11-19)f APS: 12 (10-14) vs 12 (8-15)g mRS: 4 (4-5) vs 4 (3-4)h	Single center (30/30)
Bouderka et al,3 2004	Isolated head injury with GCS ≤8 on admission; cerebral contusion on CT scan; GCS <8 on fifth day without any sedation	Not reported	Age: 41.1 (17.5) vs 40 (19) y	SAPS: 5.4 (1.5) vs 6 (3.8)b	Single center (31/31)
Bylappa et al,27 2011	Prolonged intubation due to noncorrosive poisoning, snakebites, head injuries, and respiratory paralysis due to neurological disease	Trauma to the neck; previous neck surgery; tracheostomy; scar/keloid/previous radiotherapy in the neck; chemotherapy; fungating growth in neck; granulomatous disease; infection to neck	Age: 33.1 (16.4) vs 32.9 (12.7) y Reason for admission: organophosphorus poisoning (59.1% vs 68.2%); MVC (13.6% vs 18.2%); cerebral malaria (18.2% vs 4.5%)	Not reported	Single center (22/22)
Diaz-Prieto et al,14 2014	Age >18 y; intubated >48 h	Prior tracheotomy; included in another trial; technical difficulty in performing PDT	Age: 58.8 (12.7) vs 59.5 (12.3) y Reason for admission: respiratory failure (68% vs 72%); coma (24% vs 19%)	SAPS II: 38 (3-78) vs 37.5 (10-85)b SAPS III: 62 (32-114) vs 61 (29-105)b APACHE II: 20 (5-40) vs 19 (4-38)f ISS: 29 (9-66) vs 30 (25-59)i SOFA: 6 (1-17) vs 6 (0-15)j	Single center (245/244)
Dunham et al,21 2014	Age 18-65 y; blunt trauma with admission GCS≤8; ICH on brain CT scan	Cardiac arrest, near-brain death, preexisting coagulopathy, or severe obesity	Age: 33 (13) vs 37 (16) y Reason for admission: increased ICP (40.0% vs 55.6%); craniotomy (46.7% vs 44.4%)	ISS: 28 (11) vs 35 (9)i GCS: 4 (2.5) vs 4 (0.9)d Head Abbreviated Injury Score: 4.7 (0.6) vs 4.9 (0.3)k	Single center (17/14)
Filaire et al,23 2015	Age 18-79 y; preoperative diagnosis of lung cancer or high suspicion of lung cancer; predicted postoperative DLco ≥30%; 30% ≤ ppoFEV1 < 50%; ppoVo2max ≥10 mL/kg/min; surgical approach by lateral or posterolateral thoracotomy	Pregnant; preoperative tracheotomy; vocal cord paralysis; phrenic nerve paralysis on the operated side; neuromuscular disorders; previous pharyngeal or laryngeal surgery; anatomical deformity of the neck; video-assisted thoracoscopic surgery; lung resection less important than planned at the inclusion (ppoFEV1 ≥50%).	Age: 63.5 (8.2) vs 59.9 (7.8) y Reason for admission: pneumonia (47.4% vs 64.3%); sepsis (47.4% vs 69.0%)	DLco: 60.0% (15.8%) vs 63.6% (16.7%) Vo2max: 19.2 (3.4) vs 19.7 (3.6) ppoFEV1: 41.3% (4.7%) vs 41.7% (6.7%)	Single center (39/39)
Karlović et al,24 2018	Age >18 y; patients in surgical and trauma units; intubated >48 h; expected duration of mechanical ventilation ≥14 d based on diagnosis; SOFA score >5, APACHE II scores >10, Pao2 ≤60 mm Hg with FIO2 0.5 and PEEP of at least 8 cm H2O	Previous tracheotomy; anatomical deformity of the neck; hematologic malignant neoplasms; respiratory infection within the first 48 h of mechanical ventilation	Age: 60.0 (13) vs 61.5 (28) y Reason for admission: pneumonia (47.4% vs 64.3%)	APACHE II: 23.6 (8.1) vs 22.4 (7.3)f SOFA: 14.0 (2.7) vs 14.8 (3.2)j	Single center (38/42)
Koch et al,28 2012	Age >18 y; expected time of ventilation >21 d (decided by 2 independent intensivists not involved in the study)	Anatomical variants or deformities of the larynx/trachea; preexisting tracheostomy; preexisting pneumonia; critical trauma of the cervical vertebral column; coagulopathy (thrombocyte level <60 × 103/μL; prothrombin time >40 s; INR>1.4); estimated to die within the next 24 h; planned permanent tracheostomy; >3 d of ventilation before entry into the study	Age: 61.7 (9.1) vs 57.3 (6.3) y Reason for admission: trauma (19% vs 6%); neurosurgical (10% vs 18%); GI tract (8% vs 6%); sepsis (4% vs 7%); data only provided for 50 patients receiving intervention	APACHE II: 21 (IQR 12-31) vs 22.3 (IQR 10-33)f	Single center (50/50)
Rodriguez et al,20 1990	Patients with multiple injuries requiring mechanical ventilation	Patients who did not require ventilator therapy >1 d; patients being actively disengaged from the ventilator; patients who died in the first 24 h	Age: 36 (2) vs 39 (2) y Reason for admission: cranial neurological injury (63% vs 53%); chest injury (39% vs 40%); abdominal injury (22% vs 22%); extremity injury (43% vs 40%); fracture injury (57% vs 69%)	ISS: 28 (2) vs 27 (1)i GCS: 10 (1) vs 7.1 (2.7)d APACHE II: 10 (1) vs 10 (1)f	Single center (51/55)
Rumbak et al,4 2004	Age >18 y; projected to need ventilation >14 d; APACHE II score >25	Anatomical deformity of the neck; previous tracheotomy; platelet count <50 × 103/μL, activated partial thromboplastin time/prothrombin time >1.5 times, or bleeding time >2 × normal; soft tissue infection of the neck; mechanical ventilation with a PEEP >12 cm H2O; intubated >48 h; neck on which it was technically difficult to perform a PDT	Age: 63 (10.4) vs 63 (9.3) y Reason for admission: severe sepsis (70% vs 66%); renal failure (45% vs 41%); multiorgan failure (58% vs 55%)	APACHE II: 27.4 (4.2) vs 26.3 (2.6)f	Multiple centers (60/60)
Saffle et al,29 2002	Age >18 y; hospitalized within 24 h of acute burn injury; ongoing mechanical ventilatory support on postburn day 2	Pregnant women; preexisting significant renal or hepatic disease; corticosteroid use before admission; patients who did not have cutaneous burn injuries	Age: 44.5 (4.3) vs 51.3 (4.0) y Reason for admission: full-thickness burn (34.0% vs 21.7%); inhalation injury (86.0% vs 87.0%)	Not reported	Single center (21/ 23)
Sugerman et al,19 1997	Intubated and required mechanical ventilation for 3 d; anticipated need for ventilatory support for ≥7 d	Age <18 y; patients with major burns or inhalation injury	Age: 40 (2.4) y in head trauma cohort; 61 (3.4) y in non–head trauma cohort; 62 (4.6) y in nontrauma cohort. Details not provided for early vs late cohorts. Reason for admission: head trauma (n = 67), non-head trauma (n = 41), non-trauma (n = 18)	APACHE III: 66 (3) vs 55 (3)l	Multiple centers (53/59)
Terragni et al,22 2010	Age >18 y; intubated for 24 h; SAPS II score 35-65; SOFA score ≥5; did not have a pulmonary infection (CPIS<6)	COPD; anatomical deformity of the neck; cervical tumors; history of esophageal, tracheal or pulmonary cancer; previous tracheotomy; soft tissue infection of the neck; hematologic malignancy; pregnant	Age: 61.8 (17.4) vs 61.3 (16.8) y Reason for admission: respiratory failure (45.9% vs 47.1%); CNS (22.9% vs 25.7%); cardiovascular (24.4% vs 20.0%)	SAPS II: 51.1 (8.7) vs 49.7 (8.6)b SOFA at enrollment: 7.9 (2.6) vs 7.6 (2.9)j SOFA at randomization: 10.1 (1.3) vs 9.8 (1.5)j	Multiple centers (209/210)
Trouillet et al,30 2011	Undergone cardiac surgery; still mechanical ventilation 4 d after surgery; unsuccessful mechanical ventilation screening test result or spontaneous breathing trial on the day of randomization; expected to require mechanical ventilation for ≥7 more d	Age <18 y; pregnant; previously enrolled in this or other trials of morbidity or mortality; received >48 h of mechanical ventilation preoperatively; previous tracheostomy within 6 mo; received an artificial heart device; prothrombin time >1.5 × upper limit of normal; platelet count <50 × 103/μL; irreversible neurological disorder; SAPS>80; decided to limit care; soft-tissue neck infections or anatomical deformities or concomitant neck or carotid surgery	Age: 64.1 (13.3) vs 66.0 (12.4) y Reason for admission: CABG (25% vs 25%); valve (28% vs 32%); CABG plus valve (13% vs 19%)	SAPS II: 47.2 (12.4) vs 45.8 (11.4)b SOFA: 11.6 (3.5) vs 10.9 (3.6)j GCS: 10.9 (3.3) vs 11.4 (3.0)d Lung injury score: 1.7 (0.7) vs 1.8 (0.7)m Charlson Comorbidity Index: 2.9 (1.9) vs 2.6 (1.8)n	Single center (109/107)
Young et al,31 2013	Mechanical ventilation in adult critical care units; identified by the treating clinician in the first 4 d after admission; likely to require ≥7 d of ventilatory support	Requiring an immediate, life-saving tracheotomy; tracheotomy contraindicated for anatomical or other reasons; respiratory failure due to chronic neurological disease	Age: 63.6 (13.7) vs 64.2 (13.3) y Reason for admission: respiratory (59.9% vs 59.0%); GI tract (19.1% vs 19.4%); cardiovascular (10.6% vs 13.2%); neurological (5.8% vs 4.4%). Primary reason for admission not recorded for 3.7% of patients	APACHE II: 19.6 (6.5) vs 20.1 (6.0)f	Multiple centers (451/448)
Zheng et al,32 2012	Age >18 y; treated with mechanical ventilation via endotracheal intubation	Anatomical neck deformity; thyromegaly; cervical tumors; hematologic malignant neoplasm; previous tracheotomy; pregnant; weaned or died 48 h after mechanical ventilation onset	Age: 67.5 (14.7) vs 67.9 (17.6) y Reason for admission: chronic lung disease (24.1% vs 23.0%); cardiac disease (24.1% vs 31.1%); neurological (29.3% vs 24.6%); chronic kidney disease (17.2% vs 18.0%)	APACHE II: 19.6 (2.3) vs 19.6 (2.5)f SOFA: 7.4 (1.5) vs 7.3 (1.7)j	Single center (58/61)

Abbreviations: APACHE II, Acute Physiology and Chronic Health Evaluation II; APS, Acute Physiology Score; CABG, coronary artery bypass graft; CNS, central nervous system; COPD, chronic obstructive pulmonary disease; CPIS, clinical pulmonary infection score; CT, computed tomography; DLco, diffusing capacity for carbon monoxide by single breath; FEV₁, forced expiratory volume in 1 second; FIO₂, fraction of inspired oxygen; GCS, Glasgow Coma Score; GI, gastrointestinal; ICH, intracranial hemorrhage; ICP, intracranial pressure; ICU, intensive care unit; INR, international normalized ratio; IQR, interquartile range; ISS, Injury Severity Score; mRS, modified Rankin Scale score; MVC, motor vehicle collision; NIHSS, National Institute of Health Stroke Scale; PDT, percutaneous dilational tracheotomy; PEEP, positive end expiratory pressure; ppo, predicted postoperative; SAH, subarachnoid hemorrhage; SAPS, Simplified Acute Physiology Score; SET, Stroke-related Early Tracheotomy score; SOFA, sequential organ function assessment; ST, surgical tracheotomy; Vo₂max, maximal oxygen consumption.

SI conversion factor: To convert platelet and thrombocyte counts to ×10⁹/L, multiply by 1.0.

^aUnless otherwise indicated, age and severity scoring system data are expressed as mean (SD) or median (interquartile range).

^bScores range from 0 to 163, with higher scores indicating higher risk for mortality in patients in the intensive care unit (ICU).

^cScores range from 8 to 48, with higher scores indicating tracheotomy need in patients with stroke.

^dScores range from 3 to 15, with higher scores indicating neurologically intact after trauma injury.

^eScores range from 0 to 42, with higher scores indicating severe stroke.

^fScores range from 0 to 71, with higher scores indicating in-hospital mortality.

^gEquivalent to SAPS.

^hScores range from 0 to 6, with higher scores indicating increased disability in people with neurological injury.

ⁱScores range from 0 to 75, with higher scores indicating severe traumatic injury.

^jScores range from 0 to 24, with higher scores indicating mortality in the ICU.

^kScores range from 0 to 6, with higher scores indicating greater injury severity.

^lScores range from 0 to 299, with higher scores indicating in-hospital mortality.

^mScores range from 0 to 4, with higher scores indicating greater severity.

ⁿScores range from 0 to 5 or greater, with higher scores indicating greater cumulative mortality attributable to comorbid disease.

Risk of Bias

Overall, 14 clinical trials^{3,4,14,22,23,24,25,26,27,28,29,30,31,32} were deemed to have low risk of bias, and 3 trials^19,20,21 had moderate to high risk of bias. Most of the RCTs (14 trials^{3,4,14,19,20,21,23,24,25,26,27,29,30,32} [83.4%]) were classified as unclear risk for the domain pertaining to detection bias. Three trials (17.6%)^19,20,21 had a high risk for random sequence generation, whereas selective reporting was the domain with the lowest risk of bias. A more detailed description of the risk of bias table can be found in the eTable in the Supplement.

Meta-analysis

Primary Outcomes

The incidence of VAP was lower in patients with mechanical ventilation who underwent early tracheotomy compared with late tracheotomy (OR, 0.59 [95% CI, 0.35-0.99]; I² = 64%; 1894 patients) (Figure 2). Our other primary outcome was duration of mechanical ventilation, and this was expressed as ventilator days (10 trials^{3,4,23,24,25,26,27,28,29,32} [58.8%]) and ventilator-free days (4 trials^14,22,30,32 [23.5%]). Early tracheotomy was not associated with any change in ventilator days (MD, −2.40 [95% CI, −5.09 to 0.29] days; I² = 92%; 1610 patients) (Figure 3A). In contrast, early tracheotomy was associated with more ventilator-free days with an MD of 1.74 (95% CI, 0.48-3.00 days; I² = 0%; 1243 patients) (Figure 3B).

Figure 2. Association of Early Tracheotomy With Ventilator-Associated Pneumonia.

Forest plots demonstrate pooled odds ratios (ORs) and 95% CI with a random-effects model. The vertical dashed line represents the point estimate of the overall effect (as it meets with the middle of the diamond).

Figure 3. Association of Early Tracheotomy With Duration of Mechanical Ventilation and Ventilator-Free Days .

Forest plots demonstrating pooled mean differences (MDs) in days and 95% CIs with a random-effects model. NA indicates not applicable; pts, patients. The vertical dashed line represents the point estimate of the overall effect (as it meets with the middle of the diamond).

Secondary Outcomes

Early tracheotomy was not associated with reduced short-term, all-cause mortality. The pooled OR for 2445 patients was 0.66 (95% CI, 0.38-1.15; I² = 59%) (Figure 4). Duration of ICU stay was decreased in the early tracheotomy cohort, with a pooled MD of −6.25 (95% CI, −11.22 to −1.28) days (I² = 96%; 2042 patients) (eFigure 1 in the Supplement). Two RCTs^22,32 (11.8%), with a total of 538 individuals, showed improvement in ICU-free days in the early group (MD, 2.09 [95% CI, 0.58-3.60]; I² = 15%) (eFigure 2 in the Supplement).

Figure 4. Mortality Outcome in Early vs Late Tracheotomy.

Forest plots demonstrating pooled odds ratios (ORs) and 95% CIs with a random-effects model. The vertical dashed line represents the point estimate of the overall effect (as it meets with the middle of the diamond).

Sensitivity Analysis and Publication Bias

We conducted sensitivity analyses on primary outcomes with high heterogeneity (I² ≥50%). After removal of 2 RCTs^14,27 with high heterogeneity, the incidence of VAP remained lower in the early tracheotomy group compared with the late tracheotomy group (OR, 0.60 [95% CI, 0.38-0.96]; I² = 49%) (eFigure 3 in the Supplement). For duration of mechanical ventilation, 4 trials^4,26,27,29 had high heterogeneity and were removed for sensitivity analysis. Early tracheotomy was associated with decreased mechanical ventilation time (MD, −2.80 [95% CI, −4.65 to −0.95]; I² = 30%) (eFigure 4 in the Supplement). Sensitivity analysis showed no benefit in short-term, all-cause mortality in the early tracheotomy group (OR, 0.75 [95% CI, 0.43-1.31]; I² = 37%) (eFigure 5 in the Supplement). For ICU days, 3 trials^26,28,30 remained after removing those with high heterogeneity. Early tracheotomy reduced ICU days by an MD of −4.48 (95% CI, −7.94 to −1.02) days (I² = 37%) (eFigure 6 in the Supplement).

Publication bias was low for all primary and secondary outcomes. Plots can be viewed in eFigures 7 to 11 in the Supplement.

Discussion

Our systematic review and meta-analysis of 17 RCTs showed that early tracheotomy was associated with improvement in 3 major clinical outcomes: VAP, ventilator-free days, and ICU stay. No difference was noted in mortality between early vs late tracheotomy. Herein, we supply an updated comprehensive systematic review and robust meta-analysis comparing clinical outcomes for early and late tracheotomy in critically ill adult patients.

Our findings indicate that early tracheotomy (≤7 days) may reduce the incidence of VAP. This is in congruence to the meta-analysis by Siempos and colleagues.⁶ Our results are also in accordance with the findings by Wang et al,¹⁰ in which their OR of 0.65 favored early tracheotomy. This outcome is clinically important because VAP is the most common nosocomial infection in the ICU setting.³³ Evidence supports that patients diagnosed with VAP have longer ICU and hospital stays, incur higher hospital costs, and have an increased risk for mortality.^22,34

Other notable findings in this review included the benefit of early tracheotomy on ventilator-free days and ICU stay. These findings align with the meta-analysis conducted by Hosokawa and colleagues.¹² Their MD in ventilator days was slightly higher than ours (2.12 vs 1.74 days), but most likely owing to differences on how medians and/or means were handled prior to analysis. Unlike a meta-analysis published in 2015,¹¹ we observed a decrease in ICU stay when patients underwent early tracheostomy placement. This is most likely attributable to the addition of several trials after 2014.^14,21,23,24 A more recent review also concluded that early tracheotomy reduced ICU days.³⁵ Our initial analysis of early tracheotomy on ventilator days did not show benefit; however, the heterogeneity was significantly high at an I² of 92%. After removing outliers, sensitivity analysis suggests that early tracheotomy may also reduce ventilator days.

Timing of tracheostomy placement on overall mortality is a controversial topic in the field. Our study demonstrated that early tracheostomy placement was not associated with improved short-term, all-cause mortality (eg, first 30 days in the ICU). We are fully cognizant that this outcome is speculative given that mortality was a secondary outcome and our study inclusion focused on VAP and/or ventilator days. Regardless of this shortcoming, many previous meta-analyses have failed to show that early tracheotomy improves the rate of short-term mortality. On the contrary, early tracheotomy may provide benefit in the long-term assessment of survival. For instance, a Cochrane review by Andriolo et al¹³ and the meta-analysis by Hosokowa et al¹² concluded that early tracheotomy reduced mortality at the longest follow-up time and when evaluated as a composite of 6-month, 1-year, and hospital mortality, respectively.

In view of the current coronavirus disease 2019 (COVID-19) pandemic, important messages from the present study can be translated. For example, a substantial number of patients positive for COVID-19 require long periods of mechanical ventilation and care in the ICU, posing tremendous challenges to health care systems.³⁶ It is possible that early tracheostomy placement may facilitate weaning from the ventilator and potentially increase the availability of ICU beds, mechanical ventilators, and clinicians.

Limitations

This study has several limitations. First, we included patients with diverse diseases and admitted in multiple ICU settings. Although this led to a more global approach to our primary question, the differences between the ages and disease processes may have influenced our outcomes. For instance, the inclusion of patients in the neuro-ICU may have led to some of the positive findings observed. Many of the studies that we include are single-center trials, which inherently affects the generalizability of our findings. Because some RCTs took more than 5 years to complete, the role new protocols, guidelines, or clinical decisions have with the overall effect is also brought into question. The implementation of multidisciplinary teams that consist of clinicians, respiratory therapists, otolaryngologists, speech language pathologists, nurses, and family members have improved overall tracheotomy outcomes.^37,38

Another limitation involved the range in severity of illness for participants. In some trials, the mean Acute Physiology and Chronic Health Evaluation II score was as low as 10 whereas in others it was as high as 27.4. In addition, we limited the study population to adults and therefore excluded 2 studies involving children.^39,40

Conclusions

Using a systematic review and meta-analysis approach, we examined whether clinical differences were evident in early vs late tracheotomy in critically ill adults. In a combined analysis of 17 trials (3145 individuals), early tracheotomy improved VAP, ventilator, and ICU days, but not short-term all-cause mortality. These findings may influence current clinician attitudes and ICU/surgical guidelines entailing the timing of tracheostomy placement.

Supplement.

eMethods. Database Search

eTable. Risk of Bias Assessment

eFigure 1. Length of ICU Hospitalization in Early vs Late Tracheotomy

eFigure 2. ICU-Free Days in Early vs Late Tracheotomy

eFigure 3. Sensitivity Analysis of VAP

eFigure 4. Sensitivity Analysis of Mechanical Ventilator Days

eFigure 5. Sensitivity Analysis of Mortality

eFigure 6. Sensitivity Analysis of ICU Days

eFigure 7. Publication Bias Assessed by Funnel Plot for VAP

eFigure 8. Publication Bias Assessed by Funnel Plot for Duration of Mechanical Ventilation

eFigure 9. Publication Bias Assessed by Funnel Plot for Duration Ventilator-Free Days

eFigure 10. Publication Bias Assessed by Funnel Plot for Mortality

eFigure 11. Publication Bias Assessed by Funnel Plot for Length of ICU Hospitalization

eReferences.