Predictors and Clinical Outcomes from Failed Laryngeal Mask Airway Unique™: A Study of 15,795 Patients

• Failure of supraglottic airway is not rare, but little about the risk factors for failure and its impact on respiration and patient outcomes.

• Laryngeal mask airway failure occurred in 1.1% of an adult surgical population of more than 15,000 patients, resulting in adverse respiratory events and unplanned hospitalization. Mask ventilation was difficult in 5.6% of patients with laryngeal mask airway failure.

SINCE the invention of the laryngeal mask airway (LMA) in 1981 and the subsequent introduction to clinical practice in 1988, its use during general anesthesia has gained popularity in the operating room with well-documented success.1,–,3However, just as with general anesthesia with tracheal intubation, LMAs are not without patient risks. The fourth National Audit Project of the Royal College of Anesthetists and Difficult Airway Society4reviewed complications of airway management that led to death, brain damage, the need for an emergency surgical airway, unanticipated intensive care unit admission, or prolongation of intensive care unit stay.4,5The estimated incidence of these life-threatening complications with supraglottic airway device use for anesthesia was 1 in 46,174 (95% CI; 1 in 34,684 to 1 in 69,051). In comparison, adverse respiratory events, such as significant airway obstruction and laryngospasm during Classic LMA ™ (LMA North America, Inc., San Diego, CA) use, are seen more frequently,6,7with reported rates of 0.15–7%. The incidence of Classic LMA ™ failure ranges from 0.19 to 4.7%.6,–,8However, there is a paucity of data on independent risk factors for supraglottic airway device failure reflecting a gap in current knowledge. Within our institution, a standardized LMA, the LMA Unique ™ (uLMA ™; LMA North America, Inc., San Diego, CA) is used to manage planned intraoperative cases. There are no previous large observational studies of the uLMA ™ in the literature, and data on failure rates are derived from small trials comparing the uLMA ™ with other airway devices. The uLMA ™ failure rates in these studies range from 0 to 2.5%.9,10This variability in reported rates of adverse respiratory events and LMA failure supports the need for examining these rates in a large observational study setting and identifying independent risk factors for the failure. Clinical decisions regarding the choice of uLMA ™versus  endotracheal tube may be informed by knowledge of these risk factors.

To further characterize perioperative risk factors for uLMA ™ device failure requiring an acute rescue tracheal intubation, we conducted this retrospective observational study at our quaternary care facility. We hypothesized that patient factors and intraoperative characteristics exist that place patients at increased risk of uLMA ™ failure.

After obtaining approval from the Institutional Review Board (University of Michigan, Ann Arbor, Michigan), we reviewed the prospective perioperative electronic clinical information available within the anesthesiology unit of our hospital system. Because no care interventions were mandated, signed patient consent was waived. Through this computerized database (Centricity; General Electric Healthcare, Waukesha, WI), preoperative, intraoperative, and postoperative data are documented routinely by anesthesiology residents, attending staff, and certified registered nurse anesthetists. As is mandatory in our institution, structured electronic preoperative history and physical notes are recorded for each surgical patient. All clinical elements (e.g. , cardiac symptoms, history of sleep apnea) are stored as discrete database elements. In addition, a structured, predefined pick-list is used by the clinician to enter information to make the database readily searchable for specific clinical findings (appendix). Within each intraoperative record, patient data such as mask ventilation grade, intubation ease, and vital signs are uniquely stored as binary or continuous data. Through a consistent recording of these data, a searchable database for observational research has been established and recently used in multiple large-scale clinical studies.11,12 

From this prospectively populated computerized database, a search query was performed to obtain all relevant data for this study. Inclusion criteria were adult (≥18 yr) patients who underwent general anesthesia with a planned uLMA ™ from January 1, 2006, to October 30, 2009. This period was chosen because intraoperative data relevant to this study were collected using the Anesthesia Information Management System during the period. The study included use of uLMA ™ in patients in both ambulatory and nonambulatory settings. Exclusion criteria were children, instances in which the laryngoscopy was performed before uLMA ™ placement, and when the uLMA ™ removal was related to a change in surgical plan.

For each patient included in the study, a number of variables were evaluated based on a thorough literature review of all frequently used perioperative assessment tools shown to be associated with a difficult airway and upper airway obstruction. Airway-related variables included modified Mallampati class 3–4,12,–,14reduced thyromental distance estimated less than 6 cm,11,12reduced mouth opening estimated less than 3 cm,13inability to protrude lower incisors anterior to upper incisors,11,13qualitatively assessed thick neck,12,15beard presence,13clinically estimated reduced cervical spine mobility,14,15and clinically assessed poor dentition defined as edentulous, having dentures, or having missing, loose, or broken teeth.13,15Preoperatively recorded patient characteristics were analyzed for each patient, including age, body mass index, sex, and obstructive sleep apnea.11,–,13Age and sex previously have been shown to be associated with differences in upper airway resistance.12,16Recent upper respiratory tract infections, asthma, chronic obstructive pulmonary disease, and current smoking also were assessed both individually and collapsed as one variable because hyperreactive airways have been shown to be associated with an increased incidence of adverse respiratory events with LMA anesthesia.3,17 

Because anesthesia provider experience has been shown to be associated with an increased incidence of tracheal intubation failure,18this variable was included in the analysis: novice caregivers were defined as interns, nonanesthesia rotators, and clinical anesthesia year-1 residents, who typically have less than 1 yr of clinical anesthesia experience. Experienced caregivers included all clinical anesthesia year-2 or year-3 residents, anesthesiology fellows, anesthesiology attending physicians, and certified registered nurse anesthetists. Intraoperative characteristics related to surgical positioning were evaluated because they have been studied previously with LMA anesthesia19; these characteristics included nonsupine patient positioning and any case in which the surgical table was rotated. Nonsupine positions included lithotomy, lateral, sitting, and prone. Finally, on the basis that mask ventilation and uLMA ™ are both considered to be rescue techniques to be used in the event of each other's failure, we evaluated the relationship between uLMA ™ failure and difficult mask ventilation, defined as mask ventilation that was inadequate, unstable, impossible, or required two providers.20Difficult mask ventilation was included in a separate univariate analysis because not all patients in the study underwent an assessment of mask ventilation with uLMA ™ placement. The presenting features of the airway events leading to uLMA ™ failures were extracted by individual chart review. Data on uLMA ™ size (relative to patient sex and weight), airway event type (inability to ventilate because of leak or airway obstruction), presence of adverse respiratory event (significant desaturation, hypercapnia, or increased peak inspiratory pressures), and prelaryngoscopy airway event management (reinsertion of uLMA ™ or use of succinylcholine) were collected. For the purposes of defining a significant adverse respiratory event, the following variables were used. Desaturation was defined as SpO²⩽85%, hypercapnia was defined as EtCO²≥50 mmHg, and increased peak inspiratory pressure was defined as peak inspiratory pressure ≥25 cm H²O on two consecutive readings, each 1 min apart, with uLMA ™ in place.

The primary outcome was uLMA ™ failure, defined as any acute airway event occurring between insertion of uLMA ™ and completion of surgical procedure that required uLMA ™ removal and rescue endotracheal tube placement. This definition was chosen as the primary outcome measure because it encompassed all possible characteristics of a clinically significant airway event (from inadequate ventilation to severe desaturation, hypercapnia, and airway obstruction) associated with a failed uLMA ™ for which an acute airway intervention was clinically indicated. Cases involving the primary outcome were identified initially in our database by noting any case in which a direct laryngoscopy was performed after uLMA ™ placement. These case records were then manually reviewed by two investigators (MM and SKR) to exclude cases in which the uLMA ™ replacement was performed in response to a change in surgical management (fig. 1). The secondary outcome measures of this study were the incidence of difficult mask ventilation in patients with failed uLMA ™ and frequency of unplanned hospital admission. Unplanned admissions were defined as patients originally planned to undergo ambulatory procedures who required admission to the hospital.

Our data analysis was carried out with SPSS® version 15.0 (SPSS Inc., Chicago, IL). Univariate analyses comparing patients with and without uLMA ™ failure were conducted for each data field variable, with P  values calculated using Pearson chi-square or Fisher exact tests for categorical variables and Student t  tests for continuous variables.

Because no previous adjusted models were available for LMA failure, all studied variables were included in a nonparsimonious logistic regression model to evaluate the strength of any univariate association and reduce the risk of missing important variables. Collinearity diagnostics and Pearson correlations were conducted on all pairs of variables to assess for independence. Condition indices more than 30 were used to identify covariates that are highly correlated with one another before building the logistic regression model.21The Omnibus test was used to evaluate the goodness of fit by the presence of statistically significant differences between the explained and unexplained variance within the model.22The resultant chi-square statistic value is a measure of the relationship between observed and expected frequencies. A P  value of <0.05 in this test denotes that the null hypothesis is rejected.22The predictive value of the resulting regression model was then evaluated using a receiver operating characteristic area under the curve.23The area under the curve represents the fractions of outcomes, both positive and negative, that are accurately predicted by the model. All variables deemed to be significant in the logistic regression model (P <0.05) were established as independent predictors. Adjusted odds ratios with 95% CI were then used to describe the individual effect size of each independent variable on uLMA ™ failure.24Finally, the frequency of unplanned admissions and difficult mask ventilation were evaluated between patients with and without uLMA ™ failure.

Sample size estimation was performed to define the limits of data analysis. Based on uLMA ™ failure rates as great as 2.5%, a sample size of 5,000 patients (125 failed uLMA ™ events) would allow evaluation of at least 12 covariates in a logistic regression model with minimal risk of over-fitting.25 

Of the 15,795 adult uLMA ™ cases performed between 2006 and 2009 at a single quaternary care facility, 313 cases were identified in which a laryngoscopy was performed during an uLMA ™ anesthetic. Through a subsequent manual review, 143 cases were excluded from the uLMA ™ failure group because tracheal intubation was performed in response to changes in the surgical plan. Thus, 170 (1.1%) patients experienced the primary outcome of a failed uLMA ™, defined as an acute airway event necessitating uLMA ™ removal and subsequent tracheal intubation.

On univariate analysis, the following variables had a significantly higher frequency in patients with uLMA ™ failure, as described in table 1: advanced age, increased body mass index, male sex, reduced thyromental distance, thick neck, poor dentition, smoking, and surgical table rotation. Hyperreactive airway, sleep apnea, increased modified Mallampati score, reduced mouth opening, reduced neck movement, limited jaw protrusion, beard presence, novice caregiver, and nonsupine patient positioning were not associated with uLMA ™ failure on univariate analyses.

Event analyses of failed uLMA ™ cases revealed the following findings (table 2). The event occurred early (induction to incision) in 104 cases (61.1%, 95% CI 53.7–68.2%), during table rotation in 27 cases (15.9%, 95% CI 10.7–22.3%), and during maintenance of anesthesia in the remaining 39 cases. Although 47.7% of cases involved use of uLMA ™ outside the recommended size for body weight range, there were no difference in the frequency of failed uLMA ™ relating to choice of size by patient sex or body weight (table 1). There was no documented attempt to reinsert an uLMA ™ in 61%, one attempt at reinsertion in 25%, and two or more attempts in the remaining cases. Among the 27 instances of uLMA ™ failure relating to table rotation, surgical procedures involving the head and neck were observed in nine cases. Inadequate ventilation secondary to leak was observed in 72 patients (42.3%, 95% CI 35.1–50.2%). Significant adverse respiratory events manifesting as desaturation, hypercapnia, or increased peak inspiratory pressures were seen in 106 patients (62.3%, 95% CI 54.6–69.7%). Severe desaturation (SpO²less than 85%, more than 1 min) was seen in 22.4% of failed uLMA ™ cases. Gastric contents were observed in the uLMA ™ in three patients. In eight instances, laryngospasm was denoted in the anesthetic record as the cause of airway obstruction, requiring active treatment with succinylcholine. An additional 44 patients had airway obstruction, stridor, increased peak inspiratory pressures, bronchospasm, coughing, bucking, hiccups, grunting or phonating.

On multivariate analysis, collinearity diagnostics did not demonstrate any condition indices greater than 30. Four independent predictors of uLMA ™ failure were identified on logistic regression analysis: surgical table rotation (adjusted odds ratio 5.00, 95% CI 3.12–8.02), male sex (adjusted odds ratio 1.74, 95% CI 1.23–2.45), poor dentition (adjusted odds ratio 1.58, 95% CI 1.00–2.49), and increased body mass index (adjusted odds ratio 1.06 per unit body mass index increase, 95% CI 1.03–1.09) (table 3). The model was evaluated using the omnibus test of model coefficients, which demonstrated a chi-square value of 106 with 17 degrees of freedom and a P  value of <0.001. Receiver operating characteristic curve analysis demonstrated an area under the curve of 0.708 (95% CI 0.667–0.748).

Of the 15,795 adult uLMA ™ cases studied, mask ventilation was attempted in 1,089 cases (6.9% of study patients), as shown in figure 2. Of 1,005 control group patients, the overall incidence of difficult mask ventilation in patients without a failed uLMA ™ was 1.9% (19 of 1,005), whereas the incidence of difficult mask ventilation in patients with a failed uLMA ™ was 5 of 84 (5.6%, P  value = 0.013, unadjusted odds ratio 3.3, 95% CI 1.24–8.73).

Finally, it was noted that 13,170 (83.4%) of the uLMA ™ cases performed were conducted in the ambulatory setting. There was a lower frequency of uLMA ™ failure in ambulatory anesthesia compared with in-patient settings (0.99 vs.  1.48%, P  value = 0.03). Of the cases involving a failed uLMA ™, 131 of 170 were performed in the ambulatory setting, 18 (13.7%, 95% CI 8.4–20.8%) of which resulted in an unplanned hospital admission. In addition, 2 of the 170 failed uLMA ™ cases (one inpatient and one outpatient) subsequently required unplanned intensive care unit admission for persistent hypoxemia.

The quality of data output for this study was verified through a method of systematic sampling (every tenth case) and manual chart audits for data completeness and accuracy. For all data fields, the completion rate was noted to be 99% or greater for all but two variables, body mass index (98.9% completion rate) and modified Mallampati score (87.5%). Missing data analysis demonstrated a significantly lower rate of uLMA ™ failure in patients with missing modified Mallampati score (0.55 vs.  1.15%; P  value = 0.02).

The incidence of uLMA ™ failure was 1.1% among 15,795 adult patients undergoing general anesthesia. We identified four independent predictors of uLMA ™ failure: intraoperative surgical table rotation, male sex, poor dentition, and increased body mass index. Failed uLMA ™ was associated with an approximately 3-fold increased incidence of difficult mask ventilation among patients with attempted mask ventilation. Unplanned hospital admission was seen in 13.7% of ambulatory patients with uLMA ™ failure, and 5.6% of admissions needed intensive care for hypoxemia.

The incidence of failed uLMA ™ in our study is consistent with previously reported Classic LMA ™ failure rates in large observational studies8and surveys of 0.1–4.7%.6This relatively low failure rate supports the safety of uLMA ™ for intraoperative airway management. Life-threatening complications relating to supraglottic airway device use are estimated to be extremely rare occurrences.4In our study, all of the uLMA ™ failures were associated with airway events. Although a significant fraction (approximately 40%) of uLMA ™ failures in our study presented as inability to ventilate relating to leaks, almost two thirds of uLMA ™ failures were associated with an adverse respiratory event, manifesting as desaturation, hypercapnia, or increased peak inspiratory pressures. The higher frequency of these adverse respiratory events may reflect the influence of high prevalence of obesity in the study population.

Consistent with previous studies on difficult airway, increased body mass index was demonstrated to be an independent predictor of uLMA ™ failure. However, this finding also contrasts with current literature in support of LMA usage in comparison with tracheal intubation in obese patients.26Despite this finding, the Classic LMA ™ has been demonstrated as a valuable tool in managing an unanticipated difficult airway in obese patients.1Our study also demonstrates an independent increase in uLMA ™ failure risk for male patients, independent of uLMA ™ size and body weight. This finding is consistent with the increased incidence of increased upper airway resistance in men compared with women, leading to upper airway narrowing, obstruction, and obstructive sleep apnea.16,27,28Of the airway measures evaluated in our study, absence of normal dentition was identified as an independent predictor of uLMA ™ failure, and because all but one of these patients had missing rather than damaged teeth, this association is likely attributable to the reduced oropharyngeal support for the uLMA ™. None of the other evaluated airway variables were associated significantly with the primary outcome on adjusted analyses. Of the anesthetic technique risk factors assessed in our study, it was interesting to note that no significant association between anesthesia provider experience and uLMA ™ failure risk existed. This finding may suggest the relative technical ease of supraglottic airway device placement and management, as supported by literature documenting uLMA ™ success even among early clinical trials.9,10 

Of the surgical risk factors assessed in our study, nonsupine patient positioning was not determined to predict uLMA ™ failure, consistent with a review of previous studies demonstrating the feasibility of uLMA ™ anesthesia in nonsupine positioned patients.19,29Intraoperative surgical table rotation was the most significant risk factor independently associated with uLMA ™ failure. The results may reflect displacement of the uLMA ™ position during untwisting of the circuit connection, during reattachment of the circuit to the uLMA ™, rotation of the table with uLMA ™ connected, or intraoperative dislodgement of the uLMA ™. In addition, surgical procedures on the head and neck were seen in one-third of uLMA ™ failures related to table rotation, suggesting that surgical factors may have played a small but important role.

When considering the medical setting in which uLMA ™ cases were performed, it was noted that most took place in an ambulatory setting, although the uLMA ™ failure rates across inpatient settings were marginally greater. The high (13.7%) incidence of unplanned hospital admission for patients after uLMA ™ failure may signify the need to improve preoperative prediction of this complication, lending support to the rationale for this study.

The univariate relationship between failed uLMA ™ and difficult mask ventilation could not be further evaluated by adjusted techniques because of the small proportion of patients with difficult mask ventilation included in the study. Despite this, the 3-fold overall increase in incidence of difficult mask ventilation in patients with failed uLMA ™ is of concern. This finding could suggest a common tendency to airway closure in these patients or simply be a reflection of the high incidence of laryngospasm associated with uLMA ™ failure. In either case, the study findings are provocative and add to the knowledge on the relationship between failed uLMA ™ and difficult mask ventilation.

It should be noted that data obtained from this study have several limitations. First, because this study was observational in nature, no specific care protocol was enforced regarding the decision to use, manage, or remove an uLMA ™ and replace it with a more secure airway. Consequently, our primary outcome was dependent on the clinical judgment of the attending anesthesiologist rather than any objective measure, and we cannot be certain that the criteria for uLMA ™ removal and subsequent tracheal intubation were similar for all providers involved. We have addressed this limitation by using a rigorous manual review of the chart to separate surgical causes from airway event-driven uLMA ™ failures. We did not record how the uLMA ™ was secured; it usually is taped. The size of uLMA ™ relative to patient sex and body weight was outside the recommended range in a significant number of patients, and this may have influenced the risk of uLMA ™ failure. However, we were unable to demonstrate a weight- or gender-based relationship to uLMA ™ failure in these patients. In rare cases, ProSeal^TM  LMAs (LMA North America, Inc., San Diego, CA) or other types of supraglottic airway devices may have been used for research purposes. When these are used for research purposes, the departmental expectation is a notation to that effect in the medical record. None of the patients who experienced uLMA ™ failure had a notation that an alternate study LMA had been placed.

Although such events are rare, it should be noted that an unknown number of adverse respiratory events under uLMA ™ anesthesia may have been managed by deepening the plane of anesthesia or by administration of muscle relaxants with subsequent clinical improvement, rather than uLMA ™ replacement with a endotracheal tube. These cases were not concluded to be uLMA ™ failures for purposes of this study, although they represent potential risks to patients. Next, because of the paucity of current data available regarding the preoperative and intraoperative characteristics of a failed uLMA ™, variables included in this study were limited to our own hypotheses of potential risk factors. In addition, we limited the analyzed variables to prevent regression model over-fitting.25It should be noted that an unmeasured selection bias existed for the patient population, which may have influenced the primary choice of uLMA ™ over a endotracheal tube. Depth of anesthesia may have been an important factor that was not assessed in this study. We routinely do not collect data on processed electroencephalogram, so weight-based dosage analysis is likely to have errors relating to individual variability in determining thresholds for anesthesia and laryngeal reflex suppression.

This study was subject to the limitations inherent within the electronic perioperative medical record from which our data were obtained. Because this electronic medical record is used primarily to manage clinical care, data outside the scope of this purpose are unable to be obtained. Within our medical record system, no rigorous validation of data entry exists, so this study is limited by the reliability of the data entry from which it has been derived. However, it should be noted that a robust data audit process was involved in verification of the accuracy of the data used in this study. Unplanned hospital admissions may have been for other indications that were not apparent in the medical record. Despite this limitation, the unplanned admission rates seen with uLMA ™ failure were significantly higher than expected rates at our institution (∼0.5%). The prediction model has modest discrimination, and additional work may need to be done to enhance the model's accuracy in the future. The validity of the model needs to be tested in an independent population to confirm generalizability of findings. Finally, because data from this study were drawn from a single quaternary care center, caution should be observed when applying results to patients nationally or internationally because anesthesia care delivery processes are variable across these regions. It is quite possible that other hospitals do not have the same rate of uLMA ™ (or other supraglottic airway device) displacement when the operating table is rotated, and the 5-fold independent increase in risk raises the possibility of a system-level error in uLMA ™ management during table rotation. In addition, it is potentially the one risk factor (unlike male gender, increased body mass index, and poor dentition) that the anesthesiologist can actively modify by increased attention to detail.

Despite these limitations, our data shed light on a largely unstudied clinical issue commonly encountered by anesthesia providers. Through identifying risk factors for intraoperative uLMA ™ failure in a large patient sample, this study allows for a clinician's decision to use or refrain from using an uLMA ™ to be drawn in part from evidence, rather than solely clinical intuition. In addition, our study has identified high-risk patient and operative characteristics that can serve to focus efforts of future prospective, randomized trials testing intraoperative therapies and management techniques.

In conclusion, we report that uLMA ™ failures occur approximately once in every 93 anesthetic inductions and are independently associated with surgical table rotation, male sex, poor dentition, and increased body mass index. There are clinically relevant consequences of uLMA ™ failure, as evidenced by the high rate of adverse respiratory events and the need for unplanned hospital admissions. Difficult mask ventilation was encountered in 5.6% of patients with uLMA ™ failure, representing a 3-fold increase from patients with successful uLMA ™.