Mediastinal Masses, Anesthetic Interventions, and Airway Compression in Adults: A Prospective Observational Study

This study was approved by the Institutional Review Boards of both institutions (Brigham and Women’s Hospital; University of Virginia Health System). Patients with radiographically large mediastinal masses and/or evidence of moderate to severe airway compression observed via chest computed tomography and who were scheduled for procedures with general anesthesia were recruited between 2016 and 2020, after written informed consent was obtained. Exclusion criteria included: pregnancy; less than 18 yr old; prisoner status; and refusal by patient, surgeon, or anesthesiologist assigned to the case. A data analysis and statistical plan was written after the data was accessed.

After venous and radial arterial catheter insertion and mild sedation, patients were positioned in the operating room with their backs raised to comfort. All patients expressed a preference for 30- to 45-degree elevation of the head of the bed for ease of breathing. Airways were topically anesthetized with aerosolized 2% lidocaine and sedation was titrated, if desired, to maintain a verbally interactive state. Awake fiberoptic placement of an 8.5-mm (inner diameter) endotracheal tube or supraglottic airway device was performed, followed by application of supplementary topical lidocaine to the glottis or distal airway as needed. The video bronchoscope (EVIS EXERA III BF-1TH190; Olympus Medical Systems Corp., Japan) with high-definition resolution and 6.0-mm external diameter was then positioned with a view of the most stenotic region of the trachea or mainstem bronchus.

Continuous video recording was then performed through each phase of a staged induction: (1) awake with spontaneous ventilation; (2) anesthetized with spontaneous ventilation; (3) anesthetized with positive pressure ventilation; and (4) anesthetized with positive pressure ventilation and neuromuscular blockade. The airway device was connected to the anesthesia circuit by a swivel bronchoscope adaptor with 2 to 6 l/min oxygen (inspiratory oxygen fraction = 1.0) flows without added positive end-expiratory pressure (PEEP). Induction was performed by sevoflurane inhalation and confirmed by processed electroencephalogram (bispectral index [BIS Medtronic-Covidien, Ireland] less than 60), loss of lid reflex, and loss of response to verbal or physical stimulus. Positive pressure ventilation was applied thereafter via manual bag ventilation by a dedicated assistant instructed to match the same pattern (tidal volume [V_T]), inspiratory/expiratory ratio, respiratory rate [RR], and flow- vs. time-pattern) as that occurring during general anesthesia with spontaneous ventilation. Neuromuscular blockade was induced with either succinylcholine or a nondepolarizing agent and confirmed by twitch monitor, followed by a recording period as previously discussed.

To ensure safety, protocol dictated that in the event of airway closure, near-closure, inadequate ventilation, or hemodynamic instability, the recording protocol would terminate, and appropriate rescue maneuvers would be initiated. An a priori plan devised to treat worsening tracheal collapse was developed and included (1) bronchoscopically-guided intubation distal to the obstruction to stent the airway; and (2) availability of a rigid bronchoscope or extracorporeal support at the discretion of the anesthesiologist or surgeon.

Raw videos were reviewed by at least two experienced bronchoscopists, and clips were extracted representing each phase of the staged induction for each patient. The video segments for each phase were chosen with attention to selection of segments that were at least 20 s in duration and provided the clearest view of the most compressed portion of the airway, with minimal secretions, suctioning, or blurriness.

Scoring of the airway caliber as assessed by anterior–posterior diameter, relative to baseline (awake, spontaneous ventilation) was made separately by seven experienced bronchoscopists. Judges viewed clips one at a time in random, in scrambled order for each patient, blinded to which patient or phase of the staged induction they were viewing, and with the baseline clip displayed as a continuous loop for side-by-side comparison on identical size computer screens. Scoring was based on gross assessments across multiple respiratory cycles of anterior–posterior diameter changes in the region of maximal compression, disregarding background motion from inspiration/expiration, heartbeats, and bronchoscope movement. Such background motion could not be eliminated, and since it was not possible to mark which specific video frames corresponded to which portion of the respiratory cycle, fine distinctions based on respiratory cycle could not be evaluated. Judges were provided with a drawing depicting 0%, 25%, and 50% changes in anterior–posterior diameter for reference.

For analytic purposes, airway anterior–posterior dimension changes were converted to central airway patency scores as follows: unchanged = 0; 25 to 50% larger = +1; more than 50% larger = +2; 25 to 50% smaller = −1; more than 50% smaller = −2. Central airway patency scores (−2 to +2), derived from scoring of airway anterior–posterior diameters are presented as means with 95% CIs at each stage of anesthetic intervention. Anterior–posterior diameter changes greater than 50% were considered to be clinically significant. Comparison of central airway patency scores was performed using a linear mixed model which included intercepts for each subject/anesthetic intervention stage combination and separate random intercepts for each rater. Interrater reliability was estimated using the interclass correlation coefficient. Anesthetic stage was included as a fixed effect. Interrater reliability was estimated using the interclass correlation coefficient. Interrater reliability was estimated using the interclass correlation coefficient. No statistical power calculation was conducted; the sample size was based on the available data. All hypothesis testing was two-tailed (R version 4.1.0; R Foundation for Statistical Computing, Austria) and the statistical criterion for significance was P < 0.05.

Twenty-one patients were recruited and studied between 2016 and 2020. Results from four were lost due to inadequate or irretrievable video bronchoscopy data. No patients were excluded or had their studies interrupted due to airway collapse, difficulties with ventilation, or hemodynamic instability. Video bronchoscopy images for two patients (no. 8 and no. 15) were incomplete (a total of three time points were missing), but the stages with satisfactory video data were included in the aggregate data analysis. It should be noted that similar images and observations from patient no. 1 were previously reported.⁹  Patient and mass characteristics of the 17 subjects are shown in tables 1 and 2. The presence or absence of orthopnea was not reliably recorded in patients’ records, and therefore was omitted from table 1. However, six patients had documented severe orthopnea (no. 1, no. 5, no. 6, no. 14, no. 16, and no. 17), and all patients were positioned to stated breathing preference in a semisitting position for this study. Computed tomographic scan–based cross-sectional areas of the central airways at the most compressed region were determined by the method of Cotton–Myer^10,11  as follows: less than or equal to 50% = I; 51 to 70% = II; more than 70% = III; and complete occlusion = IV.

Mean (95% CI) airway patency scores for the different phases of a staged induction are shown in figure 1A. Compared to baseline, the transition to general anesthesia was neither associated with a statistically significant (0 [95% CI, 0 to 0]; P = 0.953) change in the mean airway patency score nor a clinically significant decrease (mean estimate greater than 50%) in anterior–posterior diameter in any single patient. The mean airway patency score increased significantly and sequentially—first with the initiation of positive pressure ventilation (0 [95% CI, 0 to 1]; P = 0.024) and then, as increase persisted, with neuromuscular blockade (1 [95% CI, 0 to 1]; P < 0.001)—relative to that of the awake, spontaneous ventilation state. Distribution of anterior–posterior airway evaluations by phase are presented in figure 1B. Airway patency scores for different phases of a staged induction for individual subjects are shown in figure 2. The type-2 intraclass correlation coefficient, which was used to assess agreement among raters, was estimated to be 0.52 (95% CI, 0.42 to 0.64). Although this intraclass correlation coefficient value is generally indicative of moderate reliability, it may be influenced by the discrete nature of the data, and the intraclass correlation coefficient may to some extent obscure the reliability of the agreement. For example, across all combinations of patients and anesthetic phases, 70% of ratings match the median rating for the patient and phase, only 1% of ratings differed from the median by two categories, and no ratings were more than two categories from the median.

Representative still images for two subjects (no. 3 and no. 10) are presented in figures 3 and 4, respectively. Computed tomographic images, virtual bronchoscopic images, and still bronchoscopic images for all 17 subjects are provided in Supplemental Digital Content (1 [http://links.lww.com/ALN/C701], 2 [http://links.lww.com/ALN/C702], 3 [http://links.lww.com/ALN/C703], 4 [http://links.lww.com/ALN/C704], 5 [http://links.lww.com/ALN/C705], 6 [http://links.lww.com/ALN/C706], 7 [http://links.lww.com/ALN/C707], 8 [http://links.lww.com/ALN/C708], 9 [http://links.lww.com/ALN/C709], 10 [http://links.lww.com/ALN/C710], 11 [http://links.lww.com/ALN/C711], 12 [http://links.lww.com/ALN/C712], 13 [http://links.lww.com/ALN/C713], 14 [http://links.lww.com/ALN/C714], 15 [http://links.lww.com/ALN/C715], 16 [http://links.lww.com/ALN/C716], and 17 [http://links.lww.com/ALN/C717]). The video bronchoscopy recordings for all 17 subjects at each phase of the staged inductions are also provided (Supplemental Digital Content 18 [http://links.lww.com/ALN/C718], 19 [http://links.lww.com/ALN/C719], 20 [http://links.lww.com/ALN/C720], 21 [http://links.lww.com/ALN/C721], 22 [http://links.lww.com/ALN/C722], 23 [http://links.lww.com/ALN/C723], 24 [http://links.lww.com/ALN/C724], 25 [http://links.lww.com/ALN/C725], 26 [http://links.lww.com/ALN/C726], 27 [http://links.lww.com/ALN/C727], 28 [http://links.lww.com/ALN/C728], 29 [http://links.lww.com/ALN/C729], 30 [http://links.lww.com/ALN/C730], 31 [http://links.lww.com/ALN/C731], 32 [http://links.lww.com/ALN/C732], 33 [http://links.lww.com/ALN/C733], and 34 [http://links.lww.com/ALN/C734]).

This prospective observational study provides direct visual evidence of the dynamic behavior of extrinsically compressed central airways in 17 adult patients during each phase of a staged induction. Our most important finding was that no statistically significant decrease in anterior–posterior diameter was observed with the introduction of general anesthesia, positive pressure ventilation, or neuromuscular blockade compared to the awake, spontaneous breathing state. Airway patency scores increased with the introduction of positive pressure ventilation and this change persisted with neuromuscular blockade. No patient suffered airway collapse or required hemodynamic or ventilatory rescue interventions.

These observations challenge prevailing physiologic concepts regarding positive pressure ventilation and use of neuromuscular blocking agents. The number and variety of mediastinal masses in this study lend weight to this finding (table 2) and have important mechanistic, and potential clinical implications. We emphasize that only adults were included in this study and that patterns of ventilation (V_T, RR, inspiratory-to-expiratory ratio, and flow- vs. time-pattern) were similar during the different phases of the staged induction. Hence, our conclusions may not apply to children or to situations where ventilation parameters are changed.

All patients met inclusion criteria of at least moderate airway compression based on preoperative computed tomographic scan–determined airway cross-sectional area grades of II (50% to 70% reduction in cross-sectional area) or III (more than a 70% reduction in cross-sectional area) by the Cotton–Myer method^10,11  (table 2).

Observations of airway caliber changes during the different phases of a staged induction focused on anterior–posterior diameter rather than airway cross-sectional area for two reasons. First, because extrinsic airway compression from anterior mediastinal masses takes place primarily in the anterior–posterior direction, it logically represents the more sensitive and clinically meaningful dimension. Second, there is evidence of greater reliability using anterior–posterior diameter to assess central airway compression compared to cross-sectional area.^13,14 

We employed subjective assessments of airway changes for several reasons. No standard accepted technique exists for quantitative measurements of airway stenosis by bronchoscopy.¹²  Objective techniques to adjust for the bronchoscope’s wide-angle lens distortions,^15–18  or for 3- to 2-dimensional conversions,¹⁹  are complex and have not been widely validated or embraced.¹²  By analogy, a 2-dimensional picture of railroad tracks stretching to the horizon depicts converging rails, but subjectively, the brain more accurately perceives the separation between rails as a constant quantity.²⁰  In a survey of 118 interventional pulmonologists, 91% reported using subjective visual estimation alone.¹²  Interobserver variability in our study was satisfactory, and the magnitude of observed changes in anterior–posterior diameter was below what would be considered clinically significant.

Prevailing assumptions that spontaneous ventilation and intact muscle tone help preserve central airway patency extrinsically compressed by a mediastinal mass derive entirely from physiologic arguments and case reports. The vast majority of case reports involve children. All are subject to selection bias and other inherent limitations of case reports. Most predate readily available fiberoptic bronchoscopes and lack objective confirmation of central airway collapse. While some appear to support an apparent temporal association between ventilatory crises and induction, transition to positive pressure ventilation, or paralysis (or resolution with emergence and return of spontaneous ventilation),^1,21–24  that association is poor or reversed in others,^1,25–27  and often the precise timing of events is difficult to determine. More compressible, small diameter pediatric airways likely behave differently than those of the adult.

In adults with mediastinal masses, case reports of central airway obstruction are rare and do not reflect a temporal correlation with initiation of positive pressure ventilation or neuromuscular blockade. Gardner described bronchoscopically-verified airway collapse despite maintenance of spontaneous ventilation.²⁸  Kafrouni described the mediastinal mass–related inability to ventilate in an adult during spontaneous ventilation that improved with positive pressure ventilation.²⁹  Bechard retrospectively reviewed 105 anesthetics in adults with mediastinal masses, of which 97 involved general anesthesia with positive pressure ventilation and paralysis for some or all of those cases. No instances of airway collapse were observed, but seven cases of life-threatening ventilatory complications were, all of which occurred postoperatively during awake, spontaneous ventilation.³⁰ 

Current physiologic explanations for mediastinal mass–related airway collapse during anesthesia have been thoroughly reviewed elsewhere.^1,6,7  Transitioning to the anesthetized state, with its associated decrease in functional residual capacity,³¹  would be expected to exacerbate crowding and mass effects on compressible airways. Associated airway smooth muscle relaxation may also reduce resistance to compression.^1,32  We observed a minor, nonsignificant effect of induction in our study, possibly because the patients remained in the semisitting position; the supine position further reduces functional residual capacity and may have changed our results.

Substituting positive pressure for negative pressure ventilation increases mediastinal pressures³³  and may reverse the distending transpulmonary pressure gradient of spontaneous ventilation. However, the net transpulmonary pressure gradient is the sum of all incident forces. Increased intraluminal pressure may compensate for lost negative pleural pressure, and favor airway caliber preservation or even expansion. Air trapping distal to the stenotic central airway may exacerbate compressive forces. Thus, we contend that the pattern of ventilation, and its net effect on transpulmonary forces determines airway caliber changes rather than the positive or negative mode of ventilation per se. During positive pressure ventilation the competing effects of pneumatic stenting and dynamic hyperinflation may interact in a complex way over time, possibly explaining delayed airway collapse. We attempted to mimic the pattern of spontaneous ventilation throughout in order to limit the number of changes between phases. We speculate that exacerbation of airway compression may occur if the pattern of ventilation is altered with increased V_T or inspiratory-to-expiratory ratios predisposed to air trapping distal to stenotic central airways (although this remains unproven). More severely stenotic airways would be more prone to air trapping.

The independent effect of neuromuscular blockade is difficult to separate from the preceding phase of positive pressure ventilation. It appeared that the effect of positive pressure ventilation persisted into the paralysis phase without change. We also observed a previously described⁹  diminution in respiratory variation during the paralysis phase which remains unexplained.

Although we observed respiratory cycle–related movement, changes in airway caliber due to the respiratory cycle appeared to be relatively small were not technically feasible to score, and were beyond the aims and scope of this study to measure. Airway patency scoring was based on the maximal degree of airway compression irrespective of respiratory phase.

As with positive pressure during inspiration, PEEP may act as a “pneumatic stent” during the expiratory phase, and has been demonstrated to improve ventilation to lung regions previously excluded by extrinsic mainstem bronchial compression from an anterior mediastinal mass.³⁴  However, the effects of PEEP with obstruction are complex and difficult to predict.³⁵  Although extrinsic PEEP was not used in our study, some degree of intrinsic PEEP cannot be excluded. During expiration in the absence of PEEP, one would expect the effects of spontaneous and positive pressure ventilation on airway obstruction to be similar since expiration is passive in both modes.

If verified, the forgoing observations and mechanistic hypothesis will impact the rationale behind clinical practice recommendations and to a lesser extent, may influence practice. As it is more likely that dynamic hyperinflation will occur with positive pressure ventilation, there should remain a preference for spontaneous ventilation. However, our findings indicate that at least brief periods of similarly patterned positive pressure ventilation may safely be employed if, for example, desaturation or apnea occurs during an attempted induction of spontaneous ventilation. The staged induction protocol recommended by leading texts for patients at intermediate/indeterminant risk³⁶  calls for return to spontaneous breathing if ventilation difficulty occurs with positive pressure. Our findings do not reassure that such a change will reverse a collapsing airway and we would instead propose advancement of a stenting endotracheal tube. For patients deemed high risk, we advocate early intubation to provide for monitoring airway compression and if necessary, for advancing a stenting conduit.

We acknowledge several limitations. First, the number of patients was small due to the rarity of the syndrome. Second, few patients had critical airway compression and all had secured upper airways. Third, despite efforts to maintain consistent respiratory patterns across phases, breath-to-breath variation was inherent and unavoidable. Fine discrimination between end-inspiration and end-expiration airway diameters was not technically feasible. Fourth, the duration of observation for each phase was relatively brief and we are unable to confirm that prolonged positive pressure ventilation or neuromuscular blockade would have changed our results. Fifth, assessments of anterior–posterior diameter changes were based on subjective assessments as previously explained. However, the precise magnitude of airway diameter changes is less important than the finding that the direction of change with positive pressure ventilation is in opposition to predictions based on prevailing assumptions.

In conclusion, we observed no worsening of central airway compression in adults with large mediastinal masses when induction, positive pressure ventilation, and paralysis were introduced in a staged manner. This suggests a need to reexamine presumed mechanisms of airway collapse during anesthesia in such patients. Our observations may only apply to the specific staged protocol described (including semisitting posture and maintaining ventilation patterns tolerated during spontaneous breathing). We emphasize that our findings do not prove that conventional (nonstaged) inductions are safe for this population, which is also at risk of hemodynamic collapse.

The authors wish to acknowledge Ryan E. Lesh, M.D. (Department of Anesthesiology, University of Virginia, Charlottesville, Virginia), for participating in the scoring of video bronchoscopy images. The authors also wish to acknowledge Steven Kyriakidis (Department of Materials Management, Brigham and Women’s Hospital, Boston, Massachusetts) for assistance with video editing and James Rathmell, M.D. (Department of Anesthesiology, Perioperative and Pain Medicine, Brigham & Women’s Hospital), for assistance with figure 2.

Support was provided solely from institutional and/or departmental sources.

Dr. McMurry reports the following financial relationships: Biocore, LLC (Charlottesville, Virginia); Diffusion Pharmaceuticals (Charlottesville, Virginia); Bon Secours Health System (Richmond, Virginia); and U.S. Centers for Disease Control and Prevention, National Institute of Occupational Safety and Health (Spokane, Washington).The other authors declare no competing interests.