Electrical impedance tomography-derived flow index during spontaneous breathing trial stratifies the risk of reintubation within 48 h after extubation

Abstract

Background

Spontaneous breathing trial (SBT) is the standard test of readiness for liberation from invasive mechanical ventilation, but some patients who successfully complete SBT still experience early post-extubation failure. Inspiratory effort during SBT has emerged as an important physiological determinant of weaning tolerance and extubation outcome because it reflects the balance between ventilatory load and respiratory muscle capacity, yet its assessment often requires airway occlusions, dedicated ventilator algorithms, or intermittent measurements. We therefore evaluated whether an electrical impedance tomography (EIT)-derived flow index (EFI) measured during SBT was associated with SBT outcome and with reintubation within 48 h after extubation.

Methods

In this prospective multicenter observational study, mechanically ventilated adults receiving invasive ventilation for at least 48 h underwent a standardized 30-min pressure support SBT with continuous EIT monitoring. EFI was calculated breath by breath from EIT-derived inspiratory flow-time profiles. The primary endpoint was SBT outcome. A prespecified key secondary endpoint was reintubation within 48 h after planned extubation among patients who successfully completed the SBT and underwent extubation according to routine care. EFI discrimination for SBT failure was assessed using receiver operating characteristic analysis. For the post-extubation analysis, 48-h freedom from reintubation was evaluated using Kaplan–Meier analysis and Cox proportional hazards models.

Results

A total of 150 patients were enrolled, of whom 107 successfully completed the SBT and underwent planned extubation. EFI showed excellent discrimination for the primary endpoint of SBT outcome (AUC 0.980) and provided early physiological warning of evolving SBT intolerance. Reintubation within 48 h occurred in 7 of 107 extubated patients. Using the EFI threshold of 1.333 derived from the primary endpoint analysis, patients with EFI < 1.333 had significantly lower 48-h freedom from reintubation than those with EFI ≥ 1.333, and all reintubation events occurred in the EFI < 1.333 group. In univariable Cox analyses, EFI, RSBI, MIP, and P0.1 were each associated with reintubation within 48 h after extubation. Because of severe multicollinearity among these indices and the very small number of events, a combined multivariable model was not pursued.

Conclusions

EFI measured during SBT was strongly associated with SBT outcome and showed an exploratory association with reintubation within 48 h after extubation. Beyond identifying SBT intolerance, EFI may help identify patients with residual physiological vulnerability after a successful SBT; however, this secondary finding requires cautious interpretation and external validation.

Trial registration

ClinicalTrials.gov NCT06876792. Registered on March 10, 2025.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-026-06098-3.

Introduction

Spontaneous breathing trial (SBT) is a cornerstone for assessing readiness to liberate patients from invasive mechanical ventilation in the intensive care unit (ICU) [1, 2]. Conventional approaches, including pressure-support ventilation or T-piece trials, evaluate the ability to sustain spontaneous breathing by monitoring gas exchange, respiratory rate, tidal volume, and hemodynamic stability [3–5]. However, these global parameters may not capture early physiological stress that precedes overt intolerance—particularly in patients who are vulnerable to deterioration after extubation [6, 7]. As a result, some patients who meet conventional SBT success criteria may still experience early post-extubation failure requiring reintubation, which is associated with worse outcomes [8].

Identifying such patients before extubation remains challenging [9].

Beyond breathing pattern and gas exchange, inspiratory effort during SBT has gained attention because it reflects the balance between ventilatory load and respiratory muscle capacity. Effort-related indices measured during SBT, such as airway occlusion pressure-derived metrics, have been reported to associate with extubation outcomes [10, 11]. Nevertheless, many effort assessments require airway occlusions, dedicated ventilator algorithms, or intermittent measurements, limiting feasibility for continuous, interruption-free bedside monitoring.

Electrical impedance tomography (EIT) is a noninvasive bedside imaging technique that continuously monitors regional ventilation and its temporal characteristics [12]. EIT-based analyses have been used to describe phenotypes with increased respiratory drive/effort and to provide dynamic physiological information beyond conventional weaning indices [13–15]. These capabilities make EIT attractive for real-time assessment during SBT [16], with the potential to identify patients likely to tolerate the trial and to stratify risk of early post-extubation failure.

We developed an EIT-derived flow index (EFI) to characterize the temporal shape of inspiration during the spontaneous breathing trial (SBT) from EIT-derived inspiratory flow behavior. In this prospective multicenter study across three tertiary ICUs, EFI was measured during a standardized 30-min pre-extubation pressure support SBT. We aimed to determine whether EFI measured during SBT was associated with SBT outcome, the primary endpoint of the study, and with reintubation within 48 h after planned extubation among extubated patients, a prespecified key secondary endpoint. We also evaluated whether EFI could provide an early physiological signal of evolving SBT intolerance during the trial, alongside conventional weaning indices including RSBI, MIP, and P0.1.

Methods

Study design and setting

This prospective multicenter observational study was conducted between December 1, 2024 and December 31, 2025 in three tertiary ICUs in China: Ruijin Hospital (Shanghai), Fujian Provincial Hospital (Fuzhou), and Peking Union Medical College Hospital (Beijing). The protocol was approved by the institutional ethics committees of all participating centers (Ruijin IRB No. 2025030; Fujian Provincial IRB No. K202502002; Peking Union IRB No. SK1859). Written informed consent was obtained from patients or their legal representatives. The study was registered at ClinicalTrials.gov (NCT06876792; registered on March 10, 2025). All participating centers used identical inclusion and exclusion criteria, a standardized spontaneous breathing trial (SBT) protocol, and the same data collection procedures (Fig. S1).

Participants

Adults (≥ 18 years) receiving invasive mechanical ventilation for at least 48 h were eligible when they met local readiness criteria for an SBT. Inclusion required: (1) acute respiratory failure requiring endotracheal intubation; (2) hemodynamic stability without escalation of vasoactive support during the preceding 24 h; and (3) oxygenation compatible with extubation practice (FiO₂ ≤ 0.5 and PEEP ≤ 8 cmH₂O). Exclusion criteria were neuromuscular disease impairing spontaneous breathing, high cervical spinal cord injury, deep sedation (RASS ≤  − 3), inadequate EIT signal quality, or tracheostomy at screening [1, 11, 16, 17].

Spontaneous breathing trial protocol and measurements

All patients underwent a standardized 30-min pressure support SBT with pressure support of 8 cmH₂O, PEEP of 5 cmH₂O, and FiO₂ ≤ 0.5 [1, 11, 16, 17]. Measurements were collected at prespecified time points (Fig. S1). Before the SBT (T0), conventional weaning indices, including RSBI, MIP, P0.1, as well as vital signs and arterial blood gases, were recorded, and continuous EIT monitoring was initiated [12, 18]. During the SBT (T1), EIT was recorded continuously and EFI was calculated breath by breath and summarized minute by minute. At the end of the SBT or at early termination (T2), baseline measurements were repeated and SBT outcome was adjudicated according to prespecified failure criteria.

Patients who successfully completed the SBT were returned to their pre-SBT ventilator settings for a 1-h recovery period before planned extubation according to routine clinical care. They were then followed for 48 h after extubation. No EIT data were collected after extubation; therefore, analyses of post-extubation outcomes were based exclusively on physiological data measured during the SBT.

EIT acquisition and calculation of EFI

Continuous EIT was recorded throughout the SBT using a thoracic electrode belt positioned at a standardized mid-thoracic level in accordance with current recommendations [14, 15]. The global EIT impedance-time signal (ΔZ) was extracted continuously and used as a surrogate of tidal volume change over time.

EFI was calculated using an approach adapted from the previously validated Flow Index framework, which quantifies the concavity of the inspiratory flow-time profile [19] (Fig. S2). Briefly, the global ΔZ signal was processed breath by breath. To reduce high-frequency noise and minimize amplification of signal irregularities during differentiation, the impedance signal underwent automated preprocessing before analysis. Individual respiratory cycles were then identified from the global ΔZ waveform, and the inspiratory phase of each breath was segmented from the onset of inspiratory impedance increase to the transition toward expiration. A flow-like waveform proportional to inspiratory flow was obtained by time differentiation of the inspiratory ΔZ signal.

For each breath, the descending portion of the inspiratory waveform, after the initial rise and before cycling to expiration, was identified using an automated trimming algorithm (Fig. S2). Breaths affected by obvious motion artifact, signal distortion, unstable cycle detection, or inadequate fitting of the inspiratory segment were excluded from analysis. The selected segment was fitted with a nonlinear model, in which the exponent c quantified the concavity of inspiratory flow decay. This parameter was defined as the breath-level global EFI. For patient-level analysis, EFI values were summarized over the analyzed recording period using the prespecified study algorithm.

Alert time and early-warning analysis

EFI alert time was defined as the earliest time during the SBT when EFI first fell below the prespecified cutoff (EFI < 1.333) and remained below that cutoff for at least two consecutive breaths. For conventional indices (RSBI, MIP, and P0.1), alert times were derived using prespecified thresholds for early risk identification only and were not used as formal SBT failure criteria. Early-warning performance was assessed among patients with SBT failure as the time from first alert to trial termination and was compared across indices.

Outcomes

The primary endpoint was SBT outcome (pass/fail). SBT success was defined as completion of the 30-min SBT without meeting predefined failure criteria. SBT failure (early termination) was adjudicated according to prespecified criteria consistent with published recommendations [11, 17], including any of the following: respiratory rate > 35 breaths/min for at least 5 min; SpO₂ < 90% (or PaO₂ < 60 mmHg) despite FiO₂ ≥ 0.5; heart rate > 140 beats/min or a sustained change > 20%; systolic blood pressure > 180 or < 90 mmHg or need for escalating vasoactive support; new arrhythmia; marked agitation or diaphoresis; or clinician-determined intolerance requiring termination.

A prespecified key secondary endpoint was reintubation within 48 h after planned extubation among patients who passed the SBT and underwent extubation according to routine clinical care. Additional secondary outcomes included ventilator-free days at day 7 (VFD-7) and ICU mortality. VFD-7 was defined as the number of days alive and free from invasive ventilation within 7 days after extubation.

Statistical analysis

Continuous variables are presented as mean ± standard deviation or median (interquartile range), as appropriate, and categorical variables as number (percentage). Baseline characteristics were summarized for the overall study population and, for post-extubation analyses, for the pooled extubated cohort according to EFI groups (EFI < 1.333 vs EFI ≥ 1.333). Group comparisons were performed using Student’s t test or Mann–Whitney U test for continuous variables and the chi-square test or Fisher’s exact test for categorical variables.

The discriminatory performance of EFI and conventional weaning indices for the primary endpoint of SBT outcome was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC) with 95% confidence intervals (CIs). Early-warning time was calculated among patients with SBT failure as the interval from first threshold crossing to SBT termination.

For the prespecified key secondary endpoint, patients were categorized according to the EFI cutoff of 1.333, which had been derived from the primary endpoint analysis and was carried forward to the post-extubation analysis without re-derivation. This represented internal reuse of the threshold within the same study cohort and was not externally validated. Freedom from reintubation within 48 h after extubation was analyzed using Kaplan–Meier curves and compared between EFI < 1.333 and EFI ≥ 1.333 groups using the log-rank test. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% CIs for reintubation within 48 h. As a post hoc descriptive analysis, we also estimated the observed power for the between-group comparison of 48-h reintubation rates using a two-sided comparison of two independent proportions based on the observed event rates in the two EFI groups; this analysis was interpreted cautiously because of the sparse-event structure and the absence of events in one subgroup. For comparison with conventional weaning indices, RSBI, MIP, and P0.1 were dichotomized using prespecified thresholds (RSBI 68.33 breaths/min/L, MIP 20.71 cmH2O, and P0.1 − 3.043 cmH₂O). Multicollinearity among EFI, RSBI, MIP, and P0.1 was assessed using variance inflation factors (VIFs). Because severe multicollinearity was present when these indices were entered simultaneously, and because the number of reintubation events was very small, a combined multivariable model including all four indices was not pursued. Univariable Cox models were therefore constructed to evaluate their associations with 48-h reintubation.

Exploratory physiological analyses included pairwise Spearman correlations between EFI and conventional weaning indices. These analyses were intended to assess the relationship between EFI and conventional respiratory burden-related measurements during SBT and were interpreted as exploratory.

All statistical tests were two-sided, and p < 0.05 was considered statistically significant. Statistical analyses were performed using R (version 4.3.2) and SPSS (version 27).

Results

Study population

A total of 150 mechanically ventilated patients underwent the standardized 30-min pressure support spontaneous breathing trial (SBT) with concurrent EIT monitoring. Among them, 107 patients successfully completed the SBT and underwent planned extubation; these patients constituted the pooled extubated cohort for post-extubation outcome analysis. Reintubation within 48 h after extubation occurred in 7 of 107 patients. Baseline characteristics of the overall study population and of the pooled extubated cohort according to EFI group (EFI < 1.333 vs EFI ≥ 1.333) are summarized in Table 1 and Supplementary Table S1.

Table 1.

Baseline characteristics of the overall study population

Variable	Overall (n = 150)	SBT Success (n = 107)	SBT Failure (n = 43)	P value
Demographics
Age, years	65 (56–73)	64 (55–71)	68 (59–75)	0.042
Male sex, n (%)	92 (61.3)	65 (60.7)	27 (62.8)	0.815
BMI, kg/m2	23.5 (21.2–25.8)	23.4 (21.0– 25.6)	23.8 (21.5– 26.1)	0.563
APACHE II score	18 (14–23)	17 (13–21)	21 (16–26)	< 0.001
Clinical characteristics
Reason for intubation, n (%)				0.685
Pneumonia	72 (48.0)	50 (46.7)	22 (51.2)
Post-surgery	38 (25.3)	28 (26.2)	10 (23.3)
ARDS	22 (14.7)	16 (15.0)	6 (14.0)
COPD exacerbation	10 (6.7)	8 (7.5)	2 (4.7)
Other	8 (5.3)	5 (4.7)	3 (7.0)
Comorbidities, n (%)
Hypertension	65 (43.3)	45 (42.1)	20 (46.5)	0.621
Diabetes mellitus	35 (23.3)	24 (22.4)	11 (25.6)	0.678
Duration of MV prior to SBT, days	6 (4–9)	5 (3–8)	8 (5–12)	< 0.001
Respiratory parameters at baseline (T0)
PaO₂/FiO₂, mmHg	235 (198–285)	242 (205–292)	218 (185– 265)	0.012
PaCO₂, mmHg	42 (38–48)	41 (37–46)	45 (40–52)	0.03
RSBI, breaths·min⁻1·L⁻1	52 (38–72)	46 (35–62)	78 (60–95)	< 0.001
MIP, cmH₂O	28 (20–36)	32 (24–40)	18 (14–24)	< 0.001
P0.1, cmH₂O	− 2.4 (− 3.8 to − 1.5)	− 1.9 (− 2.8 to  − 1.2)	− 4.2 (− 5.5 to  − 3.1)	< 0.001
EIT-derived parameters
EFI (Breath-by-breath mean)	1.45 (1.12–1.78)	1.62 (1.35– 1.92)	0.98 (0.82– 1.15)	< 0.001

Data are presented as mean ± SD, median (IQR), or n (%), as appropriate. The table summarizes baseline demographic, clinical, and respiratory characteristics of enrolled patients before or at the start of the spontaneous breathing trial. P values compare patients with SBT success and SBT failure

Data are presented as median (IQR) or n (%). P values compare patients with SBT success and SBT failure

EFI, Electrical impedance tomography-derived flow index; RSBI, rapid shallow breathing index; MIP, maximal inspiratory pressure; P0.1, airway occlusion pressure at 100 ms; SBT, spontaneous breathing trial

BMI, body mass index; APACHE II, Acute Physiology and Chronic Health Evaluation II; MV, mechanical ventilation; RSBI, rapid shallow breathing index; MIP, maximal inspiratory pressure; P0.1, airway occlusion pressure at 100 ms; EFI, Electrical impedance tomography-derived flow index

Primary endpoint: EFI and SBT outcome

EFI showed excellent discriminatory performance for the primary endpoint of SBT outcome, with an area under the receiver operating characteristic curve of 0.980 in the pooled analysis (Supplementary Fig. S3). An EFI cutoff of 1.333 was derived from this primary endpoint analysis to discriminate SBT success from failure and was subsequently carried forward to the post-extubation analysis without re-derivation. Among patients who failed the SBT, EFI abnormalities fell below the threshold before actual trial termination, providing an early warning of impending SBT intolerance (Supplementary Table S2).

Prespecified key secondary endpoint: EFI and reintubation within 48 h after extubation

Using this prespecified cutoff, patients in the pooled extubated cohort were classified into the EFI < 1.333 group (n = 43) and the EFI ≥ 1.333 group (n = 64). All reintubation events within 48 h occurred in the EFI < 1.333 group, whereas no patient in the EFI ≥ 1.333 group required reintubation. Kaplan–Meier analysis demonstrated significantly lower 48-h freedom from reintubation in patients with EFI < 1.333 than in those with EFI ≥ 1.333 (Fig. 1, log-rank p < 0.001). In a post hoc descriptive power calculation using a two-sided comparison of two independent proportions based on the observed event rates in the two EFI groups (7/43 vs 0/64), the estimated observed power was 84%. In univariable Cox proportional hazards analyses, EFI, RSBI, MIP, and P0.1 were each associated with reintubation within 48 h after extubation (Table 2; Fig. 2). EFI ≥ 1.333 versus < 1.333 was associated with a lower hazard of reintubation (HR 0.25, 95% CI 0.11–0.57; p < 0.001). The corresponding estimates were HR 0.29 (95% CI 0.15–0.55; p < 0.001) for RSBI < 68.33 vs ≥ 68.33, HR 0.38 (95% CI 0.20–0.72; p < 0.001) for MIP ≥ 20.71 versus < 20.71 cmH₂O, and HR 0.55 (95% CI 0.34–0.89; p = 0.013) for P0.1 ≥  − 3.043 vs <  − 3.043 cmH₂O. Collinearity diagnostics showed severe multicollinearity among EFI, RSBI, MIP, and P0.1 when these indices were entered simultaneously (VIFs: 57.7, 30.7, 154.0, and 5.6, respectively). Therefore, a combined multivariable model including all four indices was not pursued, and the post-extubation findings should be interpreted as exploratory given the very small number of events.

Fig. 1.

Kaplan–Meier curves of 48-h freedom from reintubation after extubation stratified by EFI cutoff group in the pooled extubated cohort. Patients were categorized according to the prespecified EFI threshold as EFI < 1.333 or EFI ≥ 1.333. Tick marks indicate censoring. The P value was calculated using the log-rank test. The analysis included patients who successfully completed the spontaneous breathing trial and underwent planned extubation

Table 2.

Univariable association of EFI and conventional weaning indices with reintubation within 48 h after extubation in the pooled extubated cohort

Variable	Threshold (Cutoff)	Univariable HR (95% CI)	P value
EFI	≥ 1.333 vs < 1.333	0.25 (0.11–0.57)	< 0.001
MIP	≥ 20.71 vs < 20.71 cmH₂O	0.38 (0.20–0.72)	< 0.001
RSBI	< 68.33 vs ≥ 68.33 breaths·min⁻1·L⁻1	0.29 (0.15–0.55)	< 0.001
P0.1	≥  − 3.043 vs <  − 3.043 cmH₂O	0.55 (0.34–0.89)	0.013

Hazard ratios (HRs) are reported for the category meeting the cut-off compared with the reference category. EFI ≥ 1.333, MIP ≥ 20.71 cmH₂O, RSBI < 68.33 breaths·min⁻¹·L⁻¹, and P0.1 ≥  − 3.043 cmH₂O were coded as exposed categories. HR < 1 indicates a lower hazard of reintubation within 48 h after extubation. The pooled extubated cohort included patients who successfully completed the spontaneous breathing trial and underwent planned extubation. EFI, Electrical impedance tomography-derived flow index; HR, hazard ratio; CI, confidence interval; RSBI, rapid shallow breathing index; MIP, maximal inspiratory pressure; P0.1, airway occlusion pressure at 100 ms

Hazard ratios (HRs) are reported for the category meeting the prespecified cutoff compared with the reference category not meeting the cutoff. HR < 1 indicates a lower hazard of reintubation within 48 h in this exploratory analysis

EFI, Electrical impedance tomography-derived flow index; MIP, maximal inspiratory pressure; RSBI, rapid shallow breathing index; P0.1, airway occlusion pressure at 100 ms; HR, hazard ratio; CI, confidence interval

Fig. 2.

Univariable Cox proportional hazards analysis of EFI and conventional weaning indices for reintubation within 48 h after extubation. Forest plot showing hazard ratios and 95% confidence intervals for EFI, RSBI, MIP, and P0.1 in the pooled extubated cohort. Hazard ratios < 1 indicate a lower hazard of reintubation for the category meeting the prespecified threshold. The analysis included patients who successfully completed the spontaneous breathing trial and underwent planned extubation. The EFI threshold used in this post-extubation analysis was derived from the primary endpoint analysis of spontaneous breathing trial outcome and was not re-derived within the extubated cohort. A combined multivariable model including all four indices was not pursued because of severe multicollinearity among EFI, RSBI, MIP, and P0.1

Exploratory physiological analyses

In exploratory analyses, EFI showed substantial physiological overlap with conventional weaning indices, particularly RSBI (Supplementary Fig. S4).

Discussion

In this prospective multicenter study, EFI measured during spontaneous breathing trial (SBT) was strongly associated with the primary endpoint of SBT outcome and showed an exploratory association with the prespecified key secondary endpoint of reintubation within 48 h after extubation. EFI showed excellent discriminatory performance for identifying SBT failure and provided an early physiological signal of evolving SBT intolerance (Supplementary Fig. S3; Supplementary Table S2).In addition, among patients who successfully completed SBT and underwent planned extubation, EFI was associated with the observed occurrence of early post-extubation failure: patients with EFI below the prespecified threshold had substantially lower 48-h freedom from reintubation, and all reintubation events occurred in the EFI < 1.333 group. Taken together, these findings suggest that EFI may be useful for monitoring physiological tolerance during SBT and may also help identify patients with residual physiological vulnerability after a successful SBT. However, the post-extubation findings should be interpreted cautiously and viewed as hypothesis-generating.

The association between EFI and reintubation within 48 h after extubation is particularly relevant from a clinical perspective. Clinically, the most challenging cases are not those with obvious SBT failure, but patients who pass SBT yet later require reintubation [6, 7]—suggesting that a binary pass/fail classification may miss residual vulnerability at extubation. Our exploratory analyses also suggested substantial physiological overlap between EFI and conventional weaning indices, particularly RSBI, supporting the interpretation that EFI reflects overlapping rather than clearly independent physiological information. In this context, EFI may help identify patients who pass SBT but remain physiologically borderline, reflecting a possible mismatch between respiratory load and the patient’s remaining neuromuscular and ventilatory reserve [10].

In this context, EFI may help identify patients who formally pass the SBT but remain physiologically borderline. Because EFI is derived continuously from EIT-based inspiratory flow behavior during the trial, it may capture subtle but persistent abnormalities in inspiratory mechanics, breathing burden, or compensatory inspiratory dynamics that are not fully reflected by intermittent conventional measurements [12, 18, 19]. Such patients may therefore satisfy standard SBT completion criteria while still requiring substantial residual support to maintain stability. Once the endotracheal tube is removed and ventilatory assistance is further reduced or withdrawn, this latent imbalance may become clinically manifest, resulting in early post-extubation failure. The separation of Kaplan–Meier curves between EFI groups (EFI < 1.333 vs EFI ≥ 1.333), together with the observation that all reintubation events occurred in the EFI < 1.333 group, suggests that EFI may mark an otherwise easily overlooked subgroup with residual physiological vulnerability. However, given the very small number of events and the absence of events in one group, this observation requires cautious interpretation and external validation before any clinical application.

EFI also appears to capture physiologically relevant deterioration during the SBT itself. Conventional indices such as RSBI, MIP, and P0.1 are informative but are usually assessed intermittently and mainly reflect breathing behavior at discrete time points [1, 10, 17]. By contrast, EFI is derived continuously from EIT-based inspiratory flow behavior throughout the SBT [12, 18, 19]. This allows EFI to capture evolving changes in breathing pattern burden and compensatory inspiratory dynamics that may not be fully reflected by spot measurements, which may help explain its excellent discrimination for the primary endpoint and its ability to provide an early physiological signal of impending SBT intolerance (Supplementary Fig. S3; Supplementary Table S2). At the same time, EFI should not be interpreted as a static surrogate for any single weaning index. Rather, EFI shares physiological overlap with conventional measures while also capturing a distinct and dynamic aspect of inspiratory behavior during SBT. In this regard, EFI may complement rather than replace established weaning indices.

An important point is not simply whether EFI is related to conventional weaning indices such as RSBI, MIP, and P0.1, but what its potential contribution may be beyond these established measures. In our exploratory analyses, EFI showed substantial physiological overlap with conventional weaning indices, particularly RSBI, rather than clearly distinct information. We therefore do not interpret EFI as a replacement for conventional indices or as proof of incremental predictive superiority. Instead, its potential value lies in how it is acquired and what it may capture physiologically over time. Unlike conventional indices, which are generally assessed intermittently at isolated time points, EFI is derived continuously, breath by breath, from noninvasive EIT monitoring during the SBT. This approach may allow EFI to reflect evolving inspiratory dynamics and physiological burden during the trial in a way that is not fully captured by snapshot measurements. Whether this continuous physiological characterization provides clinically meaningful incremental value beyond conventional indices remains uncertain and requires dedicated validation in larger prospective cohorts.

This potential clinical role is also noteworthy because EFI is derived from continuous, noninvasive EIT monitoring and does not rely on repeated airway occlusions or isolated point measurements. Beyond pre-extubation assessment, this raises the possibility that EFI or related EIT-derived inspiratory pattern metrics could eventually be evaluated during transitional phases of respiratory support, including after extubation in patients receiving noninvasive ventilatory assistance. In particular, future studies could evaluate whether EFI remains informative after extubation in patients receiving prophylactic or rescue noninvasive ventilatory support, where continuous assessment of inspiratory burden might help refine monitoring intensity or support titration. Although our study did not assess EFI under noninvasive support and no conclusions can be drawn for that setting, the ability to track inspiratory burden continuously and noninvasively may represent a promising direction for future studies of post-extubation monitoring and support titration.

Several limitations should be acknowledged. First, the number of reintubation events within 48 h was very small (7 events), which limited the robustness of statistical inference, increased the risk of unstable effect estimates and model instability, and precluded robust conclusions regarding post-extubation risk stratification. Although the post hoc descriptive power estimate based on the observed event rates was in the moderate-to-high range, such an estimate is largely driven by the observed sample split and should be interpreted cautiously in the presence of sparse events and a zero-event subgroup. In addition, the absence of events in the EFI ≥ 1.333 group suggests possible sample-specific separation, further limiting the reproducibility and generalizability of this observation. Second, severe multicollinearity among EFI, RSBI, MIP, and P0.1, together with the sparse-event structure and the absence of events in one EFI-defined subgroup, precluded stable adjusted modeling even with reduced or penalized approaches, and we therefore did not pursue a combined multivariable Cox model.Third, the EFI cutoff of 1.333 was derived from the primary endpoint analysis and then reused in the post-extubation analysis within the same study cohort, without external validation. This internal reuse may introduce optimism bias and may overestimate post-extubation risk stratification performance. Fourth, EFI was derived from EIT data obtained during a standardized 30-min pressure support SBT, and its generalizability to other weaning strategies, ventilator settings, or non-standardized clinical workflows remains uncertain [1, 2, 4, 5, 17]. Fifth, although EIT provides unique physiological information, its availability and the expertise required for acquisition and interpretation may limit immediate implementation in some ICUs [18, 20, 21, 22]. Sixth, the mechanistic interpretation of EFI remains indirect. EFI is an EIT-derived implementation inspired by the previously described Flow Index framework, but it is not equivalent to the original flow-based index. Although EFI showed exploratory associations with conventional weaning physiology, dedicated validation against gold-standard measures of inspiratory effort, such as esophageal pressure-derived indices or Pmus, is still needed to better define its physiological meaning. Seventh, detailed post-extubation respiratory support data were not systematically analyzed in the present study, and such management may have influenced reintubation risk. Finally, because this was an observational study, our findings do not establish that EFI-guided interventions improve extubation outcomes.

Conclusion

In this multicenter prospective study, EFI measured during SBT was strongly associated with SBT outcome and showed an exploratory association with reintubation within 48 h after extubation. Beyond identifying SBT intolerance, EFI may help identify patients with residual physiological vulnerability after a successful SBT. These findings support further evaluation of EFI as a complementary bedside physiological tool for weaning assessment, but external validation is required before EFI can be considered for post-extubation risk stratification in clinical decision-making.

Author contributions

RZ,JX and JL designed the work. RZ,JW, RT,HH,YL,DW and DH collected and analyzed the data. RZ and JX wrote the manuscript. JL, HQ and ZZ critically revised the manuscript. All authors agree to be accountable for all aspects of the work and to ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All authors read and approved the final manuscript.

Funding

Shanghai Municipal Health Commission (20224Y0029). Shanghai Natural Science Foundation Project (23ZR143980).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

No animal studies are presented in this manuscript. No potentially identifiable human images or data are presented in this study. The study was approved by the institutional ethics committees of all centers (Ruijin IRB No. [2025030], Fujian Provincial IRB No. [K202502002], Peking Union IRB No. [SK1859]). Written informed consent was obtained from all participants or their legal representatives prior to inclusion. The trial was registered at ClinicalTrials.gov (Identifier: NCT06876792; registered on March 10, 2025).

Consent for publication

In accordance with legislation, written informed consent was obtained from the patient. Patients who were unable to provide consent prior to randomization due to orotracheal intubation or other medical conditions were informed as soon as conditions permitted.

Competing interests

The authors declare no competing interests.