- Journal List
- Crit Care Explor
- v.6(8); 2024 Aug
- PMC11319316
As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsem*nt of, or agreement with, the contents by NLM or the National Institutes of Health.
Learn more: PMC Disclaimer | PMC Copyright Notice
Crit Care Explor. 2024 Aug; 6(8): e1145.
Published online 2024 Aug 9. doi:10.1097/CCE.0000000000001145
PMCID: PMC11319316
PMID: 39120085
Vassilis G. Giannakoulis, MD,1 Georgios Psychogios, MD, PhD,2 Christina Routsi, MD, PhD,1 Ioanna Dimopoulou, MD, PhD,1 and Ilias I. Siempos, MD, DSc
1,3
Author information Copyright and License information PMC Disclaimer
Abstract
OBJECTIVES:
Optimal timing of tracheostomy in severe traumatic brain injury (TBI) is unknown due to lack of clinical trials. We emulated a target trial to estimate the effect of early vs. delayed tracheostomy strategy on functional outcome of patients with severe TBI.
DESIGN:
Target trial emulation using 1:1 balanced risk-set matching.
SETTING:
North American hospitals participating in the TBI Hypertonic Saline randomized controlled trial of the Resuscitation Outcomes Consortium.
PATIENTS:
The prematching population consisted of patients with TBI and admission Glasgow Coma Scale less than or equal to 8, who were alive and on mechanical ventilation on the fourth day following trial enrollment, and stayed in the ICU for at least 5 days. Patients with absolute indication for tracheostomy and patients who died during the first 28 days with a decision to withdraw care were excluded.
INTERVENTIONS:
We matched patients who received tracheostomy at a certain timepoint (early group) with patients who had not received tracheostomy at the same timepoint but were at-risk of tracheostomy in the future (delayed group). The primary outcome was a poor 6-month functional outcome, defined as Glasgow Outcome Scale-Extended less than or equal to 4.
MEASUREMENTS AND MAIN RESULTS:
Out of 1282 patients available for analysis, 275 comprised the prematching population, with 75 pairs being created postmatching. Median time of tracheostomy differed significantly in the early vs. the delayed group (7.0 d [6.0–10.0 d] vs. 12.0 d [9.8–18.3 d]; p < 0.001). Only 40% of patients in the delayed group received tracheostomy. There was no statistically significant difference between groups regarding poor 6-month functional outcome (early: 68.0% vs. delayed: 72.0%; p = 0.593).
CONCLUSIONS:
In a target trial emulation, early as opposed to delayed tracheostomy strategy was not associated with differences in 6-month functional outcome following severe TBI. Considering the limitations of target trial emulations, delaying tracheostomy through a “watchful waiting” approach may be appropriate.
Keywords: causal inference, head injury, immortal time bias, neurologic outcome, prolonged mechanical ventilation
KEY POINTS
Question: What is the effect of early vs. delayed tracheostomy strategy on 6-month functional outcome of patients with severe traumatic brain injury (TBI)?
Findings: By performing a target trial emulation analysis using individual patient data from 1282 patients with severe TBI enrolled in the TBI Hypertonic Saline trial of the Resuscitation Outcomes Consortium, we found that early tracheostomy as opposed to delayed tracheostomy strategy was not associated with differences in 6-month functional outcome.
Meaning: Delaying tracheostomy through a “watchful waiting” approach may be a reasonable strategy in severe TBI.
Optimal timing of tracheostomy in severe traumatic brain injury (TBI) is unknown. Clinical trials are lacking, and the subject poses significant challenges for observational studies due to immortal time bias and commitment-of-care bias. On the one hand, when a retrospective study categorizes patients into an early as opposed to a delayed tracheostomy group, this introduces an immortal time bias toward the delayed group, as a patient should survive enough to receive delayed tracheostomy (1, 2). On the other hand, patients receiving an early tracheostomy are exposed to an early “commitment-of-care” bias implying better prognosis, whereas patients in delayed tracheostomy potentially suffer a “grave-prognosis” penalty, which may confound results (3). In this context, the timing of tracheostomy in a nonrandomized study may simply reflect the attitudes of caregivers and legal surrogates. Additionally, a traditional retrospective analysis comparing early vs. delayed tracheostomy may not consider that on an intention-to-treat basis, a significant proportion of patients may avoid tracheostomy altogether, if tracheostomy is delayed (4). Finally, the timeframe for early tracheostomy varies in the literature, which may contribute to significant heterogeneity among reported studies (3–7). Some authors have even suggested treating timing of tracheostomy as a continuous rather than a dichotomous variable (8).
Due to the above-mentioned significant limitations of retrospective studies, causal inference in optimal timing of tracheostomy in severe TBI may mostly be achieved through a randomized controlled trial or, alternatively, target trial emulations (9, 10). Target trial emulations leverage observational data to mimic several features (including causal inference) of hypothetical randomized controlled trials and are especially useful when clinical trials are absent, impractical, or not feasible (9, 10). Also, functionality rather than mortality has been advocated as primary outcome for TBI research (11). Considering the above-mentioned, we obtained data from a large randomized controlled trial of severe TBI (12), and we subsequently emulated a target trial to estimate the effect of early vs. delayed tracheostomy strategy on 6-month functional outcome of patients with severe TBI.
MATERIALS AND METHODS
Study Setting and Design
We used individual patient-level data from the Traumatic Brain Injury Hypertonic Saline randomized controlled trial conducted by the Resuscitation Outcomes Consortium (12). Briefly, the parent trial was a double-blind, three-group, randomized controlled clinical trial comparing three different strategies for resuscitation fluid to patients with severe TBI (not meeting criteria for hypovolemic shock) in the out-of-hospital setting: hypertonic saline vs. hypertonic saline/dextran vs. normal saline. The trial found no differences between the groups on 6-month functional outcome or survival. The lack of effectiveness between the three strategies potentially makes trial data especially suitable for secondary research purposes, as confounders due to treatment strategies are not introduced. Full details of the trial have been published (12). Using such data, we performed a target trial emulation analysis, as described by Hernán et al (9) and Hernán and Robins (10). Supplemental Table 1 (http://links.lww.com/CCX/B389) presents the hypothetical target trial characteristics as opposed to the similarities and differences of our emulation trial analysis. We were granted access to trial data from the Biologic Specimen and Data Repository Information Coordinating Center of the National Heart, Lung, and Blood Institute following the submission of a protocol (13). The Institutional Review Board of Evangelismos Hospital waived the need of informed consent and approved of the current study, with protocol title “Early versus delayed tracheostomy in critically ill patients with severe traumatic brain injury” (protocol number 479/2023-10-25, approval date: October 24, 2023). The study was conducted in accordance with the Helsinki Declaration of 1975.
Exposure
We created two study groups (namely, “early tracheostomy” group and “delayed tracheostomy” group) using 1:1 balanced risk-set matching (14). Balanced risk-set matching provides the opportunity to emulate a clinical trial environment by matching a patient that received tracheostomy at a certain timepoint (early group) with a similar patient who has not yet received tracheostomy at that same point but is still alive and at-risk for tracheostomy and may (or may not) receive tracheostomy in the future (delayed group) (1, 2). This method does not use a prespecified cutoff time to define early tracheostomy, but rather treats it as a continuum, as previously suggested (8). Similar to the Tracheostomy Management (TracMan) trial, on an intention-to-treat basis, several patients in the delayed group may not receive tracheostomy (if they either undergo successful extubation or die) (4). Also, through balanced risk-set matching, the two study groups can be balanced for several important demographic and clinical confounders; this fact allows for a straightforward analysis of the two study groups, similar to a clinical trial (14, 15).
Study Participants
The baseline population consisted of patients with TBI and available data on tracheostomy, with an admission Glasgow Coma Scale (GCS) less than or equal to 8, who were alive and on mechanical ventilation on the fourth day following trial enrollment, stayed in the ICU for at least 5 days and did not receive tracheostomy after ICU stay. Patients who received tracheostomy up to first day following trial enrollment were excluded because they were considered to have an absolute indication for tracheostomy, such as severe facial trauma. Patients who died during the first 28 days and had a decision for care withdrawal before death were excluded, to ensure that all patients are exposed to commitment-of-care, irrespective of the tracheostomy decision. Patients with no available data on 6-month functional outcome and on matching covariates were excluded.
Outcome Exposure
The primary outcome was poor 6-month functional outcome, as assessed by Glasgow Outcome Scale-Extended (GOS-E) (16). Poor 6-month functional outcome was defined as GOS-E less than or equal to 4, as previously (12, 17). We also assessed an alternate outcome for 6-month functional outcome by using the Disability Rating Scale (DRS), rather than the GOS-E, at 6 months (18).
Secondary outcomes were occurrence of acute respiratory distress syndrome (ARDS) and pneumonia during the first 28 days following trial enrollment, 60-day mortality, ventilator-free days, ICU length of stay, hospital length of stay, and a combined outcome of mortality or persistent vegetative state at 6 months.
Statistical Analysis
Using 1:1 balanced risk-set matching (14), we balanced the two study groups (namely, “early tracheostomy” group and “delayed tracheostomy” group) for the following variables: age, admission hemoglobin, Marshall head CT classification, pupillary reactivity in the emergency department, New Injury Severity Score (NISS) (19), number of injuries, median GCS motor component score up to day 5, and usage of vasopressors up to day 5. We selected the above-mentioned variables based on the relevant work of the TBI- International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) investigators, as previously (17, 20). Balanced risk-set matching aims to minimize the difference of covariates at various prespecified time points. So, at time t = xi, a patient who received tracheostomy will be matched (using the specified covariates) to a patient who has yet to have a tracheostomy (“at-risk”) at that specific timepoint (14). Missing values on matching covariates and 6-month functional outcome of the prematching population were considered to be missing completely at random (MCAR), and indeed after performing Little’s MCAR test, we have found a p value of 0.893, which strengthened our hypothesis (21). We subsequently compared the demographic and baseline clinical variables of the two study groups. To statistically confirm successful matching, we calculated absolute standardized differences in a traditional analysis approach of the prematching population (treating timing of tracheostomy as dichotomous variable) and the balanced risk-set matching approach (treating timing of tracheostomy as a continuous variable) (22). As the groups were not predefined in the prematching population, we used a 10-day cutoff to define early tracheostomy in the prematching population, based on the post hoc observation that 75% of patients in the early group of the postmatching population received tracheostomy before 10 days. As we confirmed no imbalances, we performed a straightforward, unadjusted comparison of clinical outcomes between the two groups. Specifically, for continuous variables, we used the nonparametric Mann-Whitney U or Kruskal-Wallis H tests, whereas for categorical variables we used the chi-square or the Fisher exact test, as appropriate. A stacked bar chart was used to visually depict 6-month GOS-E categories. We performed balanced risk-set matching through the R Statistics software (R Foundation for Statistical Computing, Vienna, Austria) using the package “rsmatch” (15), whereas we performed all other analyses through IBM SPSS statistics, Version 28 (IBM Corp., Armonk, NY).
RESULTS
Out of 1282 patients available for analysis, 275 were eligible for matching and comprised the prematching population (Fig. Fig.11). Following 1:1 balanced risk-set matching, we successfully matched 150 patients (i.e., 75 patients in the early group and 75 patients in the delayed group). Those 150 patients comprised the final target trial emulation population. Table Table11 presenting absolute standardized differences in a traditional analysis approach of the prematching population as opposed to the balanced risk-set matching approach confirmed successful matching of the two groups.
TABLE 1.
Calculation of Absolute Standardized Differences in a Traditional Analysis of the Prematching Population (Treating Timing of Tracheostomy As Dichotomous Variable) and the Balanced Risk-Set Matching Approach (Treating Timing of Tracheostomy As a Continuous Variable)
| Variable | Early (≤ 10 d) (n = 100) | Late (> 10 d) or No (n = 175) | ASD | Early Strategy (n = 75) | Delayed Strategy (n = 75) | ASD |
|---|---|---|---|---|---|---|
| Age | 39.96 (15.72) | 34.01 (15.80) | 0.187 | 33.03 (14.23) | 33.89 (15.49) | 0.058 |
| Female sex | 16 (16.0) | 40 (22.9) | 0.174 | 14 (18.7) | 16 (21.3) | 0.067 |
| Race | 0.351 | 0.257 | ||||
| White | 55 (55.0) | 90 (51.4) | 40 (53.3) | 43 (57.3) | ||
| Black | 10 (10.0) | 7 (4.0) | 9 (12.0) | 4 (5.3) | ||
| Asian | 2 (2.0) | 7 (4.0) | 2 (2.7) | 2 (2.7) | ||
| Other | 5 (5.0) | 4 (2.3) | 3 (4.0) | 2 (2.7) | ||
| Unknown | 28 (28.0) | 67 (38.3) | 21 (28.0) | 24 (32.0) | ||
| 911 call to ED admission, min | 58.43 (30.61) | 63.97 (29.67) | 0.184 | 59.68 (25.33) | 66.33 (39.46) | 0.200 |
| Admission GCS score | 3.89 (1.57) | 3.98 (1.65) | 0.054 | 4.15 (1.73) | 4.07 (1.67) | 0.047 |
| Admission hemoglobin, g/dL | 12.54 (2.01) | 12.37 (2.20) | 0.080 | 12.80 (1.77) | 12.67 (1.63) | 0.073 |
| Marshall head CT classification > 2 | 37 (37.0) | 64 (36.6) | 0.009 | 24 (32.0) | 24 (32.0) | < 0.001 |
| Pupillary reactivity in the ED | 0.299 | < 0.001 | ||||
| Both reactive | 55 (55.0) | 119 (68.0) | 51 (68.0) | 51 (68.0) | ||
| One reactive | 12 (12.0) | 10 (5.7) | 4 (5.3) | 4 (5.3) | ||
| Both unreactive | 33 (33.0) | 46 (26.3) | 20 (26.7) | 20 (26.7) | ||
| New Injury Severity Score | 44.64 (14.75) | 42.98 (13.87) | 0.116 | 42.49 (12.63) | 42.47 (12.94) | 0.009 |
| Number of injuries | 7.34 (2.96) | 7.30 (3.12) | 0.012 | 7.07 (2.96) | 7.17 (3.07) | 0.035 |
| Highest ED heart rate, beats/min | 112.99 (24.11) | 111.26 (23.39) | 0.073 | 115.36 (24.87) | 110.55 (21.44) | 0.207 |
| Lowest ED systolic blood pressure, mm Hg | 109.33 (25.61) | 108.31 (23.82) | 0.041 | 111.67 (23.36) | 110.67 (23.43) | 0.043 |
| Median GCS motor score up to day 5 | 4.16 (1.41) | 4.07 (1.67) | 0.061 | 4.22 (1.38) | 4.23 (1.37) | 0.010 |
| Any usage of vasopressors up to day 5 | 12 (12.0) | 40 (22.9) | 0.289 | 9 (12.0) | 9 (12.0) | < 0.001 |
Open in a separate window
ASD = absolute standardized difference, ED = emergency department, GCS = Glasgow Coma Scale.
Results are presented as mean (sd) or n (%).
Figure 1.
Patient flow diagram.
In the final target trial emulation population, out of the 75 patients in the early group, 75 patients (100%) received tracheostomy; median time of tracheostomy was 7.0 days (6.0–10.0 d). Out of the 75 patients in the delayed (“at-risk”) group, 30 patients (40%) received tracheostomy; median time of tracheostomy was 12.0 days (9.8–18.3 d). The difference in median time of tracheostomy between the early and delayed groups was statistically significant (p < 0.001). There were no differences between the two groups regarding the variables selected for balance; namely, age, admission hemoglobin, Marshall head CT classification, pupillary reactivity, NISS, number of injuries, median GCS motor score up to day 5, and usage of vasopressors up to day 5 (Table Table22).
TABLE 2.
Baseline Characteristics of the 150 Matched Patients With Severe Traumatic Brain Injury Included in the Early Versus Delayed Tracheostomy Group
| Variable | Early (n = 75) | Delayed (n = 75) | p |
|---|---|---|---|
| Age | 27.0 (22.0–44.0) | 29.0 (21.0–44.0) | 0.873 |
| Female sex | 14 (18.7) | 16 (21.3) | 0.683 |
| Race | 0.650 | ||
| White | 40 (53.3) | 43 (57.3) | |
| Black | 9 (12.0) | 4 (5.3) | |
| Asian | 2 (2.7) | 2 (2.7) | |
| Other | 3 (4.0) | 2 (2.7) | |
| Unknown | 21 (28.0) | 24 (32.0) | |
| 911 call to ED admission, min | 58.0 (38.3–76.0) | 54.4 (40.1–76.0) | 0.718 |
| Admission GCS score | 3.0 (3.0–6.0) | 3.0 (3.0–5.0) | 0.745 |
| Admission hemoglobin, g/dL | 12.9 (11.5–13.8) | 13.0 (11.6–13.7) | 0.817 |
| Marshall head CT classification > 2 | 24 (32.0) | 24 (32.0) | 1.000 |
| Pupillary reactivity in the ED | 1.000 | ||
| Both reactive | 51 (68.0) | 51 (68.0) | |
| One reactive | 4 (5.3) | 4 (5.3) | |
| Both unreactive | 20 (26.7) | 20 (26.7) | |
| New Injury Severity Score | 41.0 (34.0–50.0) | 43.0 (34.0–50.0) | 0.914 |
| Number of injuries | 7.0 (5.0–9.0) | 7.0 (5.0–9.0) | 0.998 |
| Highest ED heart rate, beats/min | 110.0 (96.0–136.0) | 112.0 (95.0–124.0) | 0.319 |
| Lowest ED systolic blood pressure, mm Hg | 118.0 (95.0–130.0) | 110.0 (95.0–127.0) | 0.603 |
| Median GCS motor score up to day 5 | 4.5 (3.0–5.5) | 4.5 (3.0–5.5) | 0.979 |
| Any usage of vasopressors up to day 5 | 9 (12.0) | 9 (12.0) | 1.000 |
Open in a separate window
ED = emergency department, GCS = Glasgow Coma Scale.
Results are presented as median (interquartile range) or n (%).
There was no statistically significant difference between groups regarding poor 6-month functional outcome, defined as GOS-E less than or equal to 4 (68.0% in early vs. 72.0% in delayed; p = 0.593). Figure Figure22 presents a stacked bar chart of each GOS-E category at 6 months in the early vs. delayed group. Consistently, there was no difference between groups in the alternate outcome using DRS, instead of GOS-E, to assess 6-month functional outcome (5.0 [2.0–7.0] vs. 5.0 [3.0–8.0] in late; p = 0.662) (Table Table33). This was also the case for secondary outcomes, such as occurrence of ARDS, occurrence of pneumonia, ventilator-free days, ICU length of stay and hospital length of stay, and the combined outcome of mortality or persistent vegetative state at 6 months (Table (Table3).3). Finally, Supplemental Table 2 (http://links.lww.com/CCX/B389) presents baseline characteristics and outcomes of patients in the delayed (“at-risk”) tracheostomy group stratified according to whether they received or not tracheostomy.
TABLE 3.
Outcomes of the 150 Matched Patients With Severe Traumatic Brain Injury Included in the Early Versus Delayed Tracheostomy Group
| Outcome | Early (n = 75) | Delayed (n = 75) | p |
|---|---|---|---|
| 6-mo Glasgow Outcome Scale-Extended ≤ 4 | 51 (68.0) | 54 (72.0) | 0.593 |
| 6-mo Disability Rating Scale | 5.0 (2.0–7.0) | 5.0 (3.0–8.0) | 0.662 |
| Occurrence of acute respiratory distress syndromea | 11 (14.7) | 13 (17.3) | 0.656 |
| Occurrence of pneumoniaa | 43 (57.3) | 37 (49.3) | 0.326 |
| 60-d mortality | 3 (4.0) | 1 (1.3) | 0.620 |
| Ventilator-free days | 17.0 (11.0–19.0) | 17.0 (12.0–21.0) | 0.244 |
| ICU length of stay, db | 17.0 (12.0–23.5) | 15.0 (11.3–23.0) | 0.370 |
| Hospital length of stay, db | 36.1 (25.1–49.6) | 30.8 (21.7–47.6) | 0.151 |
| Combined outcome of mortality or persistent vegetative state at 6 mo | 6 (8.0) | 4 (5.3) | 0.513 |
Open in a separate window
aWithin 28 d following trial enrollment.
bAmong survivors at 6 mo.
Results are presented as median (interquartile range) or n (%).
Figure 2.
Stacked bar chart of each Glasgow Outcome Scale-Extended (GOS-E) category at 6 mo in the early vs. delayed tracheostomy group. There was no statistically significant difference between groups regarding poor 6-mo functional outcome defined as GOS-E less than or equal to 4 (early group: 68.0% vs. delayed group: 72.0%; p = 0.593; black dotted line).
DISCUSSION
By taking advantage of high-quality data from a large randomized controlled trial on severe TBI (12), we emulated a target trial to estimate the effect of early vs. delayed tracheostomy strategy on functional outcome of critically ill patients with severe TBI (admission GCS ≤ 8). Specifically, we matched 75 patients who received tracheostomy at a certain timepoint (early group) with 75 patients of similar characteristics and clinical course who had not yet received tracheostomy at that same timepoint but were at-risk of receiving tracheostomy in the future (delayed group). Median time of tracheostomy was 7.0 days (6.0–10.0 d) in the early and 12.0 days (9.8–18.3 d) in the delayed group. Only 40% of patients in the delayed (“at-risk”) tracheostomy approach actually received tracheostomy. We found that there was no statistically significant difference between groups regarding poor 6-month functional outcome (68.0% in early vs. 72.0% in delayed; p = 0.593).
There are no randomized controlled trials exploring the association between long-term functional outcome and the timing of tracheostomy specifically in patients with severe TBI. On the one hand, reports based on observational data on severe TBI seem to favor an early tracheostomy approach (3, 23, 24). On the other hand, reports with a high level of evidence (including a randomized controlled trial and a meta-analysis of over 17,000 patients) on another form of acute brain injury, namely, severe ischemic or hemorrhagic stroke, support an absence of benefit in the early tracheostomy approach (5, 25). Our results support the latter, and considering that target trial emulations such as ours aim to mimic several features of reports with a high level of evidence, they may be relevant and reliable.
Through the current target trial emulation analysis, we aimed to circumvent significant limitations of previous retrospective studies, mainly immortal time bias and the inclusion of nonintervention cases (a consequence of delaying tracheostomy). First, immortal time bias reportedly introduces heterogeneity among studies by artificially affecting or potentially reversing survival in favor of delayed tracheostomy in some previous reports (1, 3, 8, 26–28). Second, the inclusion of nonintervention cases is also essential, as on an intention-to-treat basis in real-time clinical practice, delaying tracheostomy may mean not intervening altogether, as documented in the TracMan trial (4). As a matter of fact, in our analysis, only 40% of patients in the delayed group actually received tracheostomy. At the same time, we ensured that all patients, irrespective of the decision to receive tracheostomy, were exposed to commitment-of-care for a reasonable timeframe, as in a previous study the decision to not tracheostomize was associated with higher frequency for withdrawal of treatment and a potential “grave-prognosis” penalty (3). Therefore, our target trial emulation may follow a pragmatic approach that corresponds better to the studied patient population.
In comparison to previous retrospective studies, we did not create a rigid (and often arbitrary) cutoff to define early tracheostomy (4–7, 29). Our approach aimed to answer the question “should tracheostomy be performed now, or should it be delayed for later?,” with “now” being defined flexibly, as early tracheostomy occurred at a median day of 7.0 days (6.0–10.0 d). Importantly, recent scientific literature supports our point of view in treating timing of tracheostomy as a continuous variable. Specifically, the investigators of a Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) study aimed to convey the message of abandoning the dichotomized approach on tracheostomy timing (3, 8).
The results of our target trial emulation analysis may have implications for clinical practice. Although there was no statistically significant difference in 6-month functional outcome between the two groups, one cannot ignore that through delaying tracheostomy, two thirds of patients in the delayed tracheostomy approach were not subject to the invasive procedure. At the same time, this approach did not seem to lead to higher healthcare resource utilization, as ICU length of stay and hospital length of stay were comparable between the two groups. Overall, our results may support a “watchful waiting” approach to tracheostomy.
Our target trial emulation may also inform the design of future clinical trials on the subject. We observed that 68% of patients in the early tracheostomy strategy and 72% of patients in the delayed tracheostomy strategy presented worse functional outcome at 6 months. To find the minimum number required for adequate study power, we used an alpha of 0.05, a power of 80% and an enrollment ratio of 1 (i.e., equal enrollment between groups) (30). We have found that the minimum sample size to identify this difference is 4118 patients (2059 in each group). Accordingly, we observed that 8.0% of patients in the early tracheostomy group and 5.3% of patients in the delayed tracheostomy group presented death or persistent vegetative state at 6 months. Using similar criteria to the above example, we calculated a minimum of 2672 patients (1336 in each group) to identify a difference in the hard combined outcome of mortality or persistent vegetative state. Overall, considering both power calculations, a clinical trial or a target trial emulation would require 4118 patients with severe TBI. These numbers may be considered extremely high for severe TBI, thus rendering a clinical trial not feasible, as even the large parent trial of the present analysis enrolled a total of 1282 patients from 114 North American emergency medical services agencies within the Resuscitation Outcomes Consortium over a period of 3 years. Finally, while for 6-month functional outcome the hypothetical 4% difference is in favor of early tracheostomy strategy, for death or persistent vegetative state at 6 months there is a 2.7% difference in favor of delayed tracheostomy strategy. This conflicting finding may raise the pretrial probability that indeed there is no difference between the two strategies, as our emulation observed.
Our study has limitations. First, it was subject to the inherent limitations of a post hoc analysis, such as missing data and potential residual confounding. However, it took advantage of high-quality data (including long-term follow-up) from a large randomized controlled trial (12), the postmatching population had no missing data, and the methodological approach was sophisticated. Second, the postmatching study population might seem small and it did not allow for further categorization of the patients, although the parent trial was large. However, this was the result of the strict methodology of a target trial emulation analysis. This type of analysis is an established way to circumvent several limitations of observational studies, namely, immortal time bias, timing dichotomization, inclusion of nonintervention cases, and commitment-of-care bias. Third, we cannot exclude the possibility of imbalances or residual confounding on unavailable covariates between the compared groups leading to selection bias, especially regarding respiratory condition. For example, the indication for tracheostomy (which could directly provide insight on the specific condition of the patient) and the respiratory muscle conditions of the patients (which may have allowed some patients to ventilate without tracheostomy) were not available. Overall, while target trial emulation does help circumvent immortal time bias, it was still left at the discretion of the clinician on the timing of tracheostomy (confounding by indication). As such, there are numerous factors, particularly related to tracheostomy, that would factor into the decision by the treating clinician, and that would never be captured by the available data (such as secretion management). However, during the matching process, balancing was performed on traditionally perceived important variables affecting outcome following TBI, based on the meticulous work of TBI-IMPACT investigators (20). Fourth, parameters such as patient comfort or long-term pulmonary condition (which tracheostomy potentially affects) in either approach were not assessed. Still, we assessed an important, patient-centered outcome, namely, poor functionality following 6 months of severe TBI.
CONCLUSIONS
In a target trial emulation analysis using data from a high-quality randomized controlled trial on severe TBI (admission GCS ≤ 8), we found that early as opposed to delayed tracheostomy was not associated with statistically significant differences in 6-month functional outcome. Considering the limitations of target trial emulations and given that only 40% of patients in the delayed tracheostomy group actually received the procedure, delaying tracheostomy through a “watchful waiting” approach may be appropriate.
ACKNOWLEDGMENTS
We acknowledge the incredible work by the Resuscitation Outcomes Consortium researchers, without which this study would not have been possible.
Supplementary Material
Click here to view.(342K, pdf)
Footnotes
This study was supported by grants to Dr. Siempos from the Hellenic Thoracic Society (2023) and from the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the “2nd Call for H.F.R.I. Research Projects to support Post-Doctoral Researchers” (Project Number: 80-1/15.10.2020). This study was prepared using research materials from the Traumatic Brain Injury Hypertonic Saline randomized controlled trial conducted by the Resuscitation Outcomes Consortium.
Research materials were obtained from the Biologic Specimen and Data Repository Information Coordinating Center of the National Heart, Lung, and Blood Institute (NHLBI) and the article does not necessarily reflect the opinions or views of the researchers who performed this trial or the NHLBI.
The authors have disclosed that they do not have any potential conflicts of interest.
Drs. Giannakoulis and Siempos had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Giannakoulis and Siempos were involved in conceptualization, methodology, and visualization. Dr. Giannakoulis was involved in data curation, formal analysis, and writing the original draft. Dr. Siempos was involved in funding acquisition and supervision. Dr. Dimopoulou was involved in project administration and resources. Drs. Giannakoulis, Psychogios, Routsi, Dimopoulou, and Siempos were involved in investigation. Drs. Psychogios, Routsi, Dimopoulou, and Siempos were involved in reviewing and editing the writing.
Data which this secondary analysis was based on are available through the Biologic Specimen and Data Repository Information Coordinating Center of the National Heart, Lung, and Blood Institute (https://biolincc.nhlbi.nih.gov/home/).
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccejournal).
REFERENCES
1. Hoshino T, Yonekura H, Nakamori Y, et al.: Early versus late tracheostomy in patients with traumatic brain injury.Intensive Care Med2020; 46:1286–1287 [PMC free article] [PubMed] [Google Scholar]
2. Andersen LW, Grossestreuer AV, Donnino MW: “Resuscitation time bias”-a unique challenge for observational cardiac arrest research.Resuscitation2018; 125:79–82 [PMC free article] [PubMed] [Google Scholar]
3. Robba C, Galimberti S, Graziano F, et al.; CENTER-TBI ICU Participants and Investigators: Tracheostomy practice and timing in traumatic brain-injured patients: A CENTER-TBI study.Intensive Care Med2020; 46:983–994 [PMC free article] [PubMed] [Google Scholar]
4. Young D, Harrison DA, Cuthbertson BH, et al.; TracMan Collaborators: Effect of early vs late tracheostomy placement on survival in patients receiving mechanical ventilation: The TracMan randomized trial.JAMA2013; 309:2121–2129 [PubMed] [Google Scholar]
5. Bösel J, Niesen WD, Salih F, et al.; SETPOINT2 and the IGNITE Study Groups: Effect of early vs standard approach to tracheostomy on functional outcome at 6 months among patients with severe stroke receiving mechanical ventilation: The SETPOINT2 randomized clinical trial.JAMA2022; 327:1899–1909 [PMC free article] [PubMed] [Google Scholar]
6. Siempos II, Ntaidou TK, Filippidis FT, et al.: Effect of early versus late or no tracheostomy on mortality and pneumonia of critically ill patients receiving mechanical ventilation: A systematic review and meta-analysis.Lancet Respir Med2015; 3:150–158 [PubMed] [Google Scholar]
7. Bösel J, Schiller P, Hook Y, et al.: Stroke-related early tracheostomy versus prolonged orotracheal intubation in neurocritical care trial (SETPOINT): A randomized pilot trial.Stroke2013; 44:21–28 [PubMed] [Google Scholar]
8. Galimberti S, Robba C, Graziano F, et al.: Tracheostomy in traumatic brain injured: Solving the conundrum of the immortal time bias.Intensive Care Med2020; 46:1288–1289 [PubMed] [Google Scholar]
9. Hernán MA, Wang W, Leaf DE: Target trial emulation: A framework for causal inference from observational data.JAMA2022; 328:2446–2447 [PubMed] [Google Scholar]
10. Hernán MA, Robins JM: Using big data to emulate a target trial when a randomized trial is not available.Am J Epidemiol2016; 183:758–764 [PMC free article] [PubMed] [Google Scholar]
11. Honeybul S, Ho KM, Gillett GR: Long-term outcome following decompressive craniectomy: An inconvenient truth?Curr Opin Crit Care2018; 24:97–104 [PubMed] [Google Scholar]
12. Bulger EM, May S, Brasel KJ, et al.; ROC Investigators: Out-of-hospital hypertonic resuscitation following severe traumatic brain injury: A randomized controlled trial.JAMA2010; 304:1455–1464 [PMC free article] [PubMed] [Google Scholar]
13. Coady SA, Mensah GA, Wagner EL, et al.: Use of the National Heart, Lung, and Blood Institute data repository.N Engl J Med2017; 376:1849–1858 [PMC free article] [PubMed] [Google Scholar]
14. Li YP, Propert KJ, Rosenbaum PR: Balanced risk set matching.J Am Stat Assoc2001; 96:870–882 [Google Scholar]
15. Kent S, Paukner M: rsmatch: Matching Methods for Time-Varying Observational Studies. R package version 0.2.0.2023. Available at: https://skent259.github.io/rsmatch/. Accessed October 18, 2023
16. Wilson JT, Pettigrew LE, Teasdale GM: Structured interviews for the Glasgow Outcome Scale and the extended Glasgow Outcome Scale: Guidelines for their use.J Neurotrauma1998; 15:573–585 [PubMed] [Google Scholar]
17. Vrettou CS, Giannakoulis VG, Gallos P, et al.: Effect of different early oxygenation levels on clinical outcomes of patients presenting in the emergency department with severe traumatic brain injury.Ann Emerg Med2023; 81:273–281 [PubMed] [Google Scholar]
18. Rappaport M, Hall KM, Hopkins K, et al.: Disability rating scale for severe head trauma: Coma to community.Arch Phys Med Rehabil1982; 63:118–123 [PubMed] [Google Scholar]
19. Osler T, Baker SP, Long W: A modification of the injury severity score that both improves accuracy and simplifies scoring.J Trauma1997; 43:922–925; discussion 925–926 [PubMed] [Google Scholar]
20. Steyerberg EW, Mushkudiani N, Perel P, et al.: Predicting outcome after traumatic brain injury: Development and international validation of prognostic scores based on admission characteristics.PLoS Med2008; 5:e165; discussion e165 [PMC free article] [PubMed] [Google Scholar]
21. Little RJA: A test of missing completely at random for multivariate data with missing values.J Am Stat Assoc1988; 83:1198–1202 [Google Scholar]
22. Gupta S, Wang W, Hayek SS, et al.; STOP-COVID Investigators: Association between early treatment with tocilizumab and mortality among critically ill patients with COVID-19.JAMA Intern Med2021; 181:41–51 [PMC free article] [PubMed] [Google Scholar]
23. Alali AS, Scales DC, Fowler RA, et al.: Tracheostomy timing in traumatic brain injury: A propensity-matched cohort study.J Trauma Acute Care Surg2014; 76:70–76; discussion 76–78 [PubMed] [Google Scholar]
24. Khalili H, Paydar S, Safari R, et al.: Experience with traumatic brain injury: Is early tracheostomy associated with better prognosis?World Neurosurg2017; 103:88–93 [PubMed] [Google Scholar]
25. Premraj L, Camarda C, White N, et al.: Tracheostomy timing and outcome in critically ill patients with stroke: A meta-analysis and meta-regression.Crit Care2023; 27:132. [PMC free article] [PubMed] [Google Scholar]
26. Mubashir T, Lai H, Oduguwa E, et al.: Effect of tracheostomy timing on outcomes in patients with traumatic brain injury.Proc (Bayl Univ Med Cent)2022; 35:621–628 [PMC free article] [PubMed] [Google Scholar]
27. Rizk EB, Patel AS, Stetter CM, et al.: Impact of tracheostomy timing on outcome after severe head injury.Neurocrit Care2011; 15:481–489 [PubMed] [Google Scholar]
28. Azari Jafari A, Mirmoeeni S, Momtaz D, et al.: Early versus late tracheostomy in patients with traumatic brain injury: A US nationwide analysis.Neurocrit Care2024; 40:551–561 [PubMed] [Google Scholar]
29. Kwak PE, Connors JR, Benedict PA, et al.: Early outcomes from early tracheostomy for patients with COVID-19.JAMA Otolaryngol Head Neck Surg2021; 147:239–244 [PMC free article] [PubMed] [Google Scholar]
30. Kane SP: Sample Size Calculator. ClinCalc. 2019. Available at: https://clincalc.com/stats/samplesize.aspx. Accessed May 11, 2024 [Google Scholar]
Articles from Critical Care Explorations are provided here courtesy of Wolters Kluwer Health
