Timing of tracheostomy in acute traumatic spinal cord injury: A systematic review and meta-analysis

Supplemental digital content is available in the text.

P atients with acute traumatic spinal cord injury (SCI) at the cervical or high thoracic level typically experience severe respiratory complications, resulting in the need for mechanical ventilation (MV). [1][2][3] In cervical SCI, patients have significantly reduced vital capacity and ventilatory reserve because of interruption of neural pathways to the diaphragm and respiratory muscles of the chest and abdomen, leading to a restrictive ventilatory impairment, while the loss of sympathetic innervation results in increased bronchial tone and mucous secretions. 4,5 Many patients, therefore, require endotracheal intubation and initiation of MV. In thoracic SCI, respiratory insufficiency and MV are more commonly related to direct chest trauma and pulmonary injury. 1,6 Tracheostomy is typically preferred in situations where prolonged MV is required or weaning from MV is anticipated to be prolonged. 2,7 Tracheostomy may facilitate weaning by reducing airway resistance and may prevent complications from prolonged orotracheal intubation, such as ulceration, granulation tissue formation, subglottic edema, and tracheal and laryngeal stenosis. [8][9][10] Other posited benefits of tracheostomy include improved patient comfort, swallowing, early phonation, and ease of access for tracheal suctioning to manage respiratory secretions. 2,11,12 It is, however, an invasive procedure with the potential for multiple complications. [13][14][15] The decision to convert an endotracheal intubation to a tracheostomy, therefore, requires anticipation of the expected duration MV and a careful assessment of the benefits and risks of the procedure.
For patients with acute traumatic SCI, there is no consensus on the optimal time to perform a tracheostomy. 16 Clinicians may wait in the hope that the patient will be extubated or may delay the placement of a tracheostomy following anterior cervical spine fixation. 17 Guidelines for the respiratory management after SCI were published in 2005; however, these recommendations were not specific to the acute care setting, focused mainly on evidence from noncritically ill SCI patients, and did not provide recommendations on the optimal timing of tracheostomy. 18 Although individual studies have investigated whether early (within 7 d of intubation) or late (after 7 d of intubation) tracheostomy improves outcomes, including mortality, ventilator-associated pneumonia, and length of hospital and intensive care unit (ICU) stay, the results remain inconclusive. In addition, whether more patient-orientated outcomes, such as the ability to speak or maintain oral intake, are improved by early tracheostomy remain unknown. We performed a systematic review to evaluate and synthesize evidence regarding the timing of tracheostomy in patients with acute traumatic SCI.

METHODS
This systematic review was performed in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses and guided by an a priori protocol registered with PROSPERO (ID: CRD42020162488). 19

Search Strategy
Studies were identified by searching MEDLINE, EMBASE, CINAHL, Scopus, Web of Science, and the Cochrane Central Register of Controlled Trials (CENTRAL) from their inception to January 2020, with no limitations on time or language of publication. An experienced health sciences librarian assisted in development of the strategy (Supplemental Appendix 1, http:// links.lww.com/TA/C132). The reference lists of retrieved articles were investigated to identify additional studies. Abstracts were searched from the conference proceedings listed in Supplemental Appendix 1, http://links.lww.com/TA/C132, within the past 10 years, and the gray literature was searched using Google Scholar. A sensitivity analysis was performed using 10 preidentified studies (Supplemental Appendix 2, http://links.lww.com/TA/C132).

Study Selection
Studies were initially screened for eligibility by title, keywords, and abstract using the Covidence software (Melbourne, Australia) by the primary reviewer (S.J.F.). 20 Studies passing the initial screen were subsequently reviewed in full by two reviewers (S.J.F., S.T.) to confirm eligibility for inclusion. Studies included (i) randomized controlled trials (RCTs) that compared either the timing of tracheostomy, or tracheostomy and prolonged intubation, in patients with acute SCI, and (ii) cohort studies that included acute SCI patients receiving early tracheostomy or late tracheostomy/prolonged intubation while admitted in the ICU. There were no exclusion criteria based on the level of SCI used in the studies. Differences between the two reviewers (S.J.F. and S.T.) regarding eligible studies were resolved in consultation with a third reviewer (V.A.M.).

Data Abstraction and Quality Assessment
Data were independently extracted from included studies by two reviewers (S.J.F and S.T.) using a standardized data collection form (Supplemental Appendix 3, http://links.lww.com/ TA/C132). The Newcastle-Ottawa Scale (NOS) was used to assess the quality of included studies (Supplemental Appendix 4, http://links.lww.com/TA/C132). 21 All of the studies, with the exception of 1 case series, were cohort studies and thus the NOS was used as it is one of two tools recommended by the Cochrane Handbook to assess the quality of nonrandomized studies of interventions (Supplemental Appendix 5, 6, http://links.lww.com/TA/C132). 22

Outcomes
The primary outcome was short-term mortality, defined as mortality in the ICU or hospital. Secondary outcomes included long-term mortality (defined as death at hospital discharge, 6 months or 1-2 years following the acute illness), duration of MV, ICU length of stay (LOS), hospital LOS, duration of sedation, incidence of ventilator-associated pneumonia (VAP), rate of tracheostomy procedures performed and tracheostomyassociated complications (

Quantitative Data Synthesis
Meta-analysis was performed using Review Manager 5.4. The qualitative terms of "early" and "late" tracheostomy as defined by the researchers of each study were used in the analysis. Dichotomous data were analyzed using the DerSimonian and Laird random effects model to produce the effect measure as a risk ratio (RR). Continuous data were analyzed using an inverse variance random effects model and reported as the mean difference. A 95% study confidence interval (CI) was used for the analysis of all outcomes. Heterogeneity was assessed using the I 2 statistic, the χ 2 test for homogeneity, and visual inspection of the forest plots. A z test of interaction was performed for all subgroup comparisons, which tests the null hypothesis that the treatment effects in each subgroup are the same.

Literature Search
The database search yielded 3,098 citations. One study was found by gray literature search, three were retrieved from the reference lists of the included studies, and one abstract was included following a search of conference proceedings. In total, 17 studies with 2,804 patients met our inclusion criteria and were included in this systematic review (Fig. 1).

Study Characteristics and Methodological Quality
The characteristics of included studies are summarized in Table 1. Studies differed in their definitions of early and late tracheostomy (Supplemental Appendix 8, http://links.lww.com/ TA/C132), although the majority used a range of 7 days or less (from either injury, intubation, or surgery) for early tracheostomy. [23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38] In one study, early and late tracheostomy were defined as 7 days or less and longer than 7 days, respectively, but the time point from which tracheostomy was measured was not specified. 33 Two studies used a range of 10 days or less and longer than 10 days. 26,31 Two studies did not report the specific timing of tracheostomy. 34,38 Patient characteristics from the included studies are reported in Supplemental Appendix 9, http:// links.lww.com/TA/C132. In one study of 344 SCI patients, 72 patients also had concomitant traumatic brain injury. 33 One study investigated pediatric patients and was thus excluded. 30

Quantitative Data Synthesis Primary Outcome
A summary of the study results is included in Supplemental Appendix 10, http://links.lww.com/TA/C132. Early tracheostomy was not found to be associated with short-term mortality (RR, 0.84; 95% CI, 0.39-1.79; p = 0.65; 10 studies; n = 2,072; 125 events; I 2 = 52%; Fig. 2, Table 2). Flanagan et al. 27 also measured 90-day mortality with a mortality rate of 6.3% in the early tracheostomy group and 3.5% in the late tracheostomy group. In addition to hospital mortality, Jeon et al. 31 also reported ICU mortality (2.6% and 4.7% for the early and late tracheostomy groups, respectively).

Qualitative Assessment of Additional Secondary Outcomes
Bellamy et al. 24 reported over 20 years experience at a single trauma center of respiratory complications in SCI patients with quadriplegia. Twenty-eight tracheostomies performed within 3 days of injury were associated with 39 pulmonary complications and 14 deaths, while four cases of tracheostomy performed after 3 days were associated with 24 pulmonary complications and one case of death. Mortality in this study was measured at 1 year and thus considered a long-term outcome; other than the study performed by Babu et al., 23 which measured mortality at 1 year for the total study population, this was the only study to report long-term mortality. Only one study reported duration of sedation and found that there was no significant difference for patients who underwent early versus late tracheostomy (14.4 ± 10.4 days vs. 10.5 ± 7.1 days, respectively, p = 0.283). 28 The same study also found that timing of tracheostomy did not affect time to initiation of oral nutrition. 28 Vitaz et al. 37 found that implementation of a clinical pathway, including the placement of a tracheostomy approximately 4 days following injury, was associated with a decreased number of both total decubitus ulcers and stage III ulcers (25% and 0%, respectively, in the clinical pathway group compared with 54% and 14%, respectively, in the control group). In a study investigating the impact of performing tracheostomy prior to anterior cervical fusion, three patients who underwent tracheostomy within 7 days of their injury experienced a DVT, while two patients who underwent tracheostomy after 7 days experienced a DVT. 39 One patient in the late tracheostomy also experienced postoperative decubitus ulcers. In 29 patients with traumatic cervical SCI, late (>24 hours after injury) tracheostomy was associated with decreased time to decannulation compared with early (<24 hours after injury) tracheostomy (35.0 (14-46) days vs. 42.0 (23-104) days, respectively). 40 In contrast, Flanagan et al. 27 found that tracheostomy within 7 days of intubation was associated with fewer days to decannulation compared with late tracheostomy (53.0 ± 28.1 vs. 74.3 ± 45.8 days, p < 0.05) when ASIA Impairment Scale and level of neurological injury were controlled for.

Patient and Surgical Factors Associated With Timing of Tracheostomy
Analysis of SCI level and timing of tracheostomy found earlier tracheostomy was more likely performed in patients with a thoracic SCI compared with cervical SCI (RR, 1.56; 95% CI, 1.10-2.21; p = 0.01; two studies; n = 367; 172 events; I 2 = 0%). Evaluating specific SCI levels, there was no difference in timing of tracheostomy in patients with a SCI at or below C5, compared Leelapattana

DISCUSSION
In this systematic review and meta-analysis of patients with acute cervical or thoracic traumatic SCI, we found that early tracheostomy, as compared with late tracheostomy, is not associated with improvements in short-term mortality; however, it is associated with a decreased duration of MV, ICU LOS, and hospital LOS. Early tracheostomy was also associated with a reduced incidence of ventilator-associated pneumonia and tracheostomyrelated complications.
There is an extensive array of literature investigating the timing of tracheostomy in general critically ill populations. Although multiple cohort studies have shown that early tracheostomy may reduce the duration of MV and LOS, shorten the duration of sedation, and lower the incidence of VAP, larger RCTs have established that a strategy of routinely performing early tracheostomies confers no survival benefit and may result in excess procedures. [41][42][43][44][45] Several recent systematic reviews also found that early tracheostomy (within 7-10 days) does not reduce mortality, in addition to finding no effect on the duration of MV or intensive care stay in a general critical care population. [46][47][48] However, meta-analyses found conflicting results regarding the incidence of VAP and duration of sedation. [46][47][48] The indications for endotracheal intubation, MV, and the need for a tracheostomy vary considerably between the heterogeneous mix of critically ill patients included in these studies. Early tracheostomy may help specific subgroups of critically ill populations. For example, patients with acute brain injury typically require airway protection for depressed airway reflexes, rather than respiratory failure. A recent meta-analysis of RCTs, including only patients with severe  Random effects meta-analysis on short-term mortality, expressed as the RR. The blue box represents the point estimate of the study result, the black horizontal line represents the 95% confidence interval of the study result, and the black diamond represents the mean point estimate and mean confidence interval of all the studies. Flanagan et al. 27 measured mortality at admission (ICU). Galeiras et al. 28 measured mortality during admission. Ganuza et al., 6 Kornblith et al., 33 Romero et al., 36 and Wu et al., 38 did not specify the time at which mortality was measured. Guirgis et al. 29 measured ICU mortality. Khan et al. 32 measured in-hospital mortality. Jeon measured in-hospital mortality. 31 Lozano measured in-hospital mortality. 35 ET, early tracheostomy, LT, late tracheostomy. acute brain injury, showed that early tracheostomy results in decreased long-term mortality, duration of MV and ICU LOS. 49 The ongoing respiratory care needs for the homogeneous critically ill population with acute SCI are unique. They require a tracheostomy for the provision of prolonged MV due to the high incidence of respiratory complications, including atelectasis, pneumonia, and ventilatory failure following an SCI. 1 They also require an airway conduit to maintain pulmonary hygiene and suctioning due to the accumulation of secretions from the combined loss of sympathetic innervation, resulting in increased bronchial tone and mucous secretions, and expiratory musculature and ability to cough. 50 Our meta-analysis included data from only critically ill patients with cervical and high thoracic SCI. This patient population may have unique and competing considerations regarding tracheostomy that impact the timing of tracheostomy, as well as subsequent outcomes. These patients often required prolonged MV or respiratory care for pulmonary hygiene, which may not only favor early tracheostomy but also have specific surgical and anatomic considerations, which impact the ability to perform this procedure promptly because of concerns over surgical site infection.
Multiple factors may contribute to the lack of consensus regarding the timing of tracheostomy in patients with acute SCI. The necessity of cervical spine fixation surgery in cases of SCI has historically resulted in delayed tracheostomy because of the perceived risk of cross-contamination between the two incision sites. 39,51 However, we found that early tracheostomy before, or just after, anterior cervical spine fixation surgery did not result in an increased rate of tracheostomy-associated complications, including wound infection (stoma cellulitis and cervical site wound infection) (Supplemental Appendix 15, http://links.lww.com/TA/ C132). 39,[52][53][54][55] In addition, there have been limited published data regarding optimal timing of tracheostomy in SCI patients, with single studies reporting on various different outcomes; several studies have only briefly explored tracheostomy timing within the broader context of identifying factors that predict the need for tracheostomy. 56,57 In the most recent clinical practice guidelines for management of SCI published in 2005, there is no recommendation of the optimal timing of tracheostomy in patients expected to require prolonged MV. Thus, the timing of tracheostomy remains highly variable, with it often being delayed until ventilator weaning and extubation have been attempted (although a primary tracheostomy as opposed to a secondary tracheostomy following extubation failure may decrease ICU mortality and LOS). 58 The interpretation of these findings must consider the heterogeneity between studies. Clinical heterogeneity likely exists because of (1) patient selection of ASIA Impairment Scale and illness severity scores (e.g., acute physiology and chronic health evaluation II or sequential organ failure assessment), and age, which is known to be associated with duration of MV and ICU stay in SCI patients, and (2) variability in cointerventions between studies. Important cointerventions in SCI patients include protocols or algorithms for liberation from MV. The use of mechanical insufflationexsufflation in the liberation process, venous thromboembolism  prophylaxis, and adherence to repositioning protocols to prevent the development of decubitus ulcers, which are known to be associated with duration of MVand ICU/hospital LOS and mortality in SCI patients. We were not able to complete a sensitivity analysis to adjust for these factors in the included studies to determine whether such cointerventions impacted on the pooled estimate of effects on duration of MV and ICU/hospital LOS.
This systematic review synthesizes the current data regarding optimal timing of tracheostomy in acute SCI patients. Strengths of this review include the comprehensive study protocol, rigorous methodology, and transparent reporting process. Data were individually collected by two reviewers to limit bias, and the quality of studies was analyzed using a validated quality assessment tool. Limitations of this study include the heterogeneity between studies, the inclusion of small single-center studies, mixed cervical and thoracic level SCI populations, and lack of a comparator "no tracheostomy" cohort in the included studies (which may have helped to evaluate the presence of a bias with regard to the SCI patients' stability and overall clinical status). In addition, the studies in this systematic review included both percutaneous and surgical techniques, as well as different timings of early tracheostomy (all within 10 days of intubation). However, it should be noted that the effect of percutaneous versus open would likely not bias the estimates as most of the duration of MV and ICU/ hospital LOS would be driven by compromised ventilatory mechanics and VAP development. In addition, in terms of the variability in the point of reference from which duration of tracheostomy was measured, we postulate that because of the high C-spine injury, the day of injury is likely also the day of MV initiation as breathing would have typically been affected instantly or over hours because of decreased ventilation efficacy, increased work of breathing, and secretion burden. [59][60][61] Future research could look more specifically at QOL or other patient-reported outcome measures following tracheostomy in patients with SCI. One study included in this systematic review investigated QOL outcomes and found that the number of days until initiation of oral nutrition was not significantly different between the early and late tracheostomy groups. 28 A prior study examining dysphagia in SCI patients found that those patients without dysphagia experienced a mean orotracheal intubation duration prior to tracheostomy of 10.0 days versus 16.9 days for those with dysphagia. 62 It also found that those patients with dysphagia experienced a higher rate of VAP compared with those without dysphagia (58% vs. 9%). In another study included in this review, one case of vocal cord dysfunction was noted. 30 In addition, researchers at the Johns Hopkins Hospital developed a validated QOL questionnaire for mechanically ventilated ICU patients and found that those who underwent early tracheostomy compared with late tracheostomy had higher scores on the questionnaire. 63 Further investigation into QOL measures, and factors that impact those measures including the ability to phonate and swallow, should be conducted. In addition, the long-term effects of early versus late tracheostomy remains unclear, with only one study found that investigated long-term mortality. 24

CONCLUSION
This systematic review suggests early tracheostomy (within 7 days of injury, intubation, or surgery) in acute SCI patients reduces MV duration, ICU and hospital LOS, VAP, and tracheostomy-related complications. However, early tracheostomy was not associated with a decrease in short-term mortality. The impact of tracheostomy timing on long-term outcomes in SCI patients, including mortality, patient comfort, and QOL, warrants further study. Randomized controlled trials are necessary to establish the optimal timing of tracheostomy, understand patient selection considering the injury level and severity, and inform evidence-based guidelines for critically ill patients with acute SCI.