Previous studies have failed to detect high body mass index (BMI) as a risk factor for difficult tracheal intubation (DTI). BMI was investigated as a risk factor for DTI in patients planned for direct laryngoscopy.
A cohort of 91,332 consecutive patients planned for intubation by direct laryngoscopy was retrieved from the Danish Anesthesia Database. A four-point scale to grade the tracheal intubation was used. Age, sex, American Society of Anesthesiologists physical status classification, priority of surgery, history of previous DTI, modified Mallampati-score, use of neuromuscular blocker, and BMI were retrieved. Logistic regression to assess whether BMI was associated with DTI was performed.
The frequency of DTI was 5.2% (95% confidence interval [CI] 5.0-5.3). In multivariate analyses adjusted for other significant covariates, BMI of 35 or more was a risk for DTI with an odds ratio of 1.34 (95% CI 1.19-1.51, P < 0.0001). As a stand alone test, BMI of 35 or more predicted DTI with a sensitivity of 7.5% (95% CI 7.3-7.7%) and with a predictive value of a positive test of 6.4% (95% CI 6.3-6.6%). BMI as a continuous covariate was a risk for failed intubation with an odds ratio of 1.031 (95% CI 1.002-1.061, P < 0.04).
High BMI is a weak but statistically significant predictor of difficult and failed intubation and may be more appropriate than weight in multivariate models of prediction of DTI.
DIFFICULT tracheal intubation (DTI) is feared among anesthesiologists due to the increased risk of perioperative morbidity and mortality.1–4The ability to predict a DTI allows anesthesiologists to take precautions to reduce the risk.5Several studies have investigated the possible association between obesity and DTI, but their results are ambiguous. Some studies did not demonstrate obesity as an independent risk of DTI,6–8but others seemed to show that obese patients are at risk of DTI9,13. These studies may have failed to detect or reject obesity as a risk factor for DTI due to small patient numbers and subsequent lack of power. Obesity is a worldwide, constantly growing problem. It is therefore appropriate to evaluate whether obesity confers a risk of DTI and failed tracheal intubation (FTI).
There is a lack of consensus about how to measure obesity as a risk factor for DTI. Two studies used weight (kg),9,13and others used body mass index (BMI; kg · m−2).7,8,10,12The clinical cut-off value defining obesity by BMI has been applied from that used for other medical events and may not be appropriate for DTI. One study did not demonstrate that morbidly obese patients were more at risk of DTI than those with moderate obesity.12In contrast other studies have suggested that the risk of DTI increases with weight.9,13The aim of this study is to assess whether obesity measured by BMI is associated with DTI, independent of other risk factors registered in the Danish Anesthesia Database (DAD). We also wish to assess whether the risk of DTI is greater in patients with high BMI. We will evaluate the different levels of BMI used to categorize obesity. We will compare BMI and weight to decide if there are differences in their association with a DTI or FTI. We will evaluate the accuracy of obesity as a stand-alone clinical test to predict a DTI.
Materials and Methods
Fourteen Danish anesthesia departments in 2005 and 25 in 2006–07 prospectively and consecutively recorded patients undergoing surgery and anesthesia in the DAD. The DAD contains specific quantitative anesthetic and surgical indicators describing the perioperative period. This information is registered immediately after each anesthesia by the anesthesiologist. The departments (appendix I) were connected via the Internet to a central server.
The Danish Data Protection Agency, Copenhagen, Denmark (journal-number 750.16–5) approved the registration in the DAD. The Ethics Committees for Biomedical Research, Glostrup, Denmark (reference KA-06751), and the Steering Committee of the DAD approved this study and provided access to the data.
We retrieved 129,925 records of tracheal intubations of patients undergoing general or combined anesthesia from January 1, 2005, to September 30, 2007, (fig. 1). We excluded patients aged less than 15 yr and those primarily scheduled for fiberoptic intubation. There were no records of the reason for these patients to be allocated for fiberoptic intubation. Thus, these patients may be allocated for other reason than anticipated difficult tracheal intubation. We included 91,332 patients tracheally intubated 110,382 times. Only the last record of 13,135 patients scheduled for intubation by direct laryngoscopy more than once was included. The final cohort includes 91,332 patients, each represented by only one session of attempted tracheal intubation by direct laryngoscopy. Except for cardiothoracic surgery, all types of surgery are represented in the DAD.
There is no Danish recommendation for the evaluation and handling of the airway in patients undergoing tracheal intubation. The anesthesia departments may differ in their recommendations for the evaluation and handling of the airway.
There is no international consensus defining a difficult intubation. Difficult laryngoscopy is often used as a surrogate outcome for difficult intubation,14but others suggest a specific definition of DTI.5,15We registered an intubation score (DTI score, table 1) for all patients primarily attempted for tracheal intubated by direct laryngoscopy.
The DTI score, age, sex, priority of surgery, weight, height, American Society of Anesthesiologists physical status classification, modified Mallampati score,16use of neuromuscular blocking agents (NMBA), and history of previous difficult intubation (PDI) were used as covariates in the assessments.
Every patient is registered with a unique identifying number from the centralized civil register. This unique identifier contains information regarding the patient’s sex and date of birth. This identifies each patient during the statistical analysis, allowing exclusion of duplicates and patients anesthetized and registered more than once during the observation period. A total of 13,135 patients were anesthetized more than once (fig. 1). For these patients, PDI was categorized as Yes or No on the basis of DTI score of the penultimate intubation. Patients anesthetized and registered only once, with absence of information of a possible PDI, were categorized as Unknown (lack of clinical information in contrary to statistical missing value). Priority of surgery was defined as nonscheduled if a patient was anesthetized without being planned for surgery the previous day. Height and weight were registered on the basis of preoperative measures at the surgical wards or as reported by the patients. If records of height or weight were omitted, they were categorized as a missing value. BMI was calculated as weight · height−2(kg · m−2). An automatic validation of weight and height is incorporated in the DAD. A warning appeared during registration if the calculated BMI exceeded 35 or was below 17, to emphasize that the weight and height entries should be reconsidered. We manually performed an additional validation before the statistical analysis, in which only height of 125–230 cm and height of 30–250 kg were accepted; otherwise, the registration was categorized as missing. If the Mallampati class was registered as unknown, the registration was categorized as a missing value. For the analyses, the Mallampati score was dichotomized by combining classes I with II and classes III with IV. The use of NMBA was categorized as no relaxation or relaxation. If relaxation was used, it was not possible to distinguish between relaxation used for intubation, during anesthesia or both. The DAD did not record the type of equipment used for intubation.
We attempted a power and sample size calculation in the protocol.17With a type I error risk set to α= 0.05, the type II error risk of β= 0.10, a relative risk increase of obesity of 0.30, a frequency of DTI of 0.058,18and a multivariate correlation coefficient, R2= 0.917, the number of patients needed was 25,618. R2was unknown and estimated from a linear regression of BMI dependent on height and weight. This assumption is restrictive as height and weight are closely associated; therefore, the true number of patients needed for investigation may be lower.
We performed univariate regression analyses to evaluate the associations between the covariates and DTI. In the primary analysis, weight and height were replaced by BMI. A subsequent multivariate logistic regression analysis was performed including all significant covariates from the univariate analyses. Backward stepwise regression was performed to identify a final model. Only interactions of the first order between BMI and all the other covariates were explored. Hereafter, BMI was categorized into six intervals, and a univariate logistic regression analysis was performed to determine if the risk of DTI increased with BMI. We also explored the combination of some of the intervals and the changing of their borders. After a final fitting of the categorized BMI, it was included in a new multivariate logistic regression analysis. A model control was performed. The receiver operating characteristic curve tested the final fitted model, and the area under the curve was estimated.
On the basis of the final model, we evaluated dichotomized BMI intervals as a stand-alone predictor of DTI. The accuracy of the predictor was described by sensitivity, specificity, predictive value of a positive test, predictive value of a negative test, positive likelihood ratio, and negative likelihood ratio.
The prevalence and pattern of missing values among all covariates were described. Afterwards, multiple imputations for missing values were performed.19SPSS v. 15.0 and AMOS v. 7.0 (SPSS Inc., Chicago, IL) and NORM v 2.03 by Joseph L. Schafer, M.Sc., Ph.D. (Associate Professor, Department of Statistics and The Methodology Center, The Pennsylvania State University, University Park, Pennsylvania), were used for the analyses. The number of imputations (m = 10) was calculated to reach 99% efficiency.20Ten complete datasets were analyzed as described above, and the estimates were pooled for overall estimates.20A similar analysis was performed for the original dataset with listwise deletion patients with missing data—a complete case analysis. The pooled estimates and their corresponding complete case analysis estimates, including 95% confidence intervals (CI), were compared. If there were any noticeable differences between complete case analysis and analysis of the pooled estimates of the multiple imputations, both results are presented. Otherwise, only the pooled estimates of the multiple imputations are presented. P < 0.05 was regarded as statistically significant. This study has been presented according to the Strengthening the Reporting of Observational Studies in Epidemiology statement.21
The overall proportion of patients with DTI was 5.2% (95% CI 5.0–5.3). The proportion of patients with DTI in 2005, 2006, and 2007 were 5.8%, 4.9% and 5.1%, respectively. FTI occurred in 141 patients with a frequency of 0.15% (95% CI 0.13–0.18). The characteristics of the patients are displayed in table 2. In univariate analyses, the covariates American Society of Anesthesiologists physical status classification, BMI, age, sex, priority of surgery, Mallampati score, use of NMBA, and PDI were all associated with DTI with statistical significance (P < 0.0001).
The results of the univariate analysis of BMI stratified in six categories are listed in table 3. The odds ratio (OR) for DTI increased with BMI. Based on level of significance and the OR, it seemed reasonable to combine the six intervals into three: less than 25; 25–34; 35 or more. A univariate analysis with BMI less than 25 used as reference demonstrated OR for DTI of 1.42 (95% CI 1.26–1.59, P < 0.0001) for BMI of 35 or more and OR of 1.24 (95% CI 1.17–1.32, P < 0.0001) for BMI 25–34. Further analysis did not demonstrate other reasonable alternative intervals for the categorization of BMI. A multivariate analysis, including all significant covariates from the univariate analyses, identified all covariates except the American Society of Anesthesiologists physical status classification to be independent risk factors of DTI. Adjusted for all other significant covariates, BMI of 35 or more and BMI 25–34 remained statistically significant risk factors of DTI with an OR of 1.34 (95% CI 1.19–1.51, P < 0.0001) and 1.11 (95% CI 1.04–1.18, P < 0.0016), respectively (table 4).
Controlling the multivariate model with the Hosmer and Lemeshow goodness-of-fit test (P < 0.3) indicated that the statistical model was reasonable fitted. Adding age2as an extra covariate did not improve the model. In a receiver operating characteristic curve analysis of the final model, the area under the curve was 0.65 (95% CI 0.64–0.66, P < 0.0001).
Both weight and height were statistically significant in a univariate analysis associated with DTI. Performing another multivariate analysis including the same statistically significant covariates from the univariate analyses included previously, but replacing BMI with weight and height, demonstrated only weight as an independent risk factor of DTI with OR of 1.004 (95% CI 1.002–1.006, P < 0.0001). A goodness-of-fit test (P < 0.01) indicated that this model was poorly fitted. We used the same stratified weight as El-Ganzouri et al. 9A multivariate analysis with weight less than 90 kg as reference demonstrated OR for DTI of 1.28 (95% CI 1.12-1.46, P < 0.0002) for weight of 110 kg or more and OR of 1.10 (95% CI 1.02–1.20, P < 0.02) for 90–110 kg.
Table 5illustrates the agreement of the two categorical covariates, BMI and weight, with a weighted kappa value of 0.44 (95% CI 0.43–0.44, P < 0.0001). The distribution of the patients in this table is asymmetric; more patients are located below than above the diagonal of the table. This indicates that patients categorized with a low or a moderate risk of DTI according to their weight may be categorized with a moderate or a high risk of DTI according to their BMI. On the other hand, patients categorized with low BMI had no or much lower probability of being categorized with moderate or high risk of DTI according to their weight. Thus, despite weight being included in the calculation of BMI, the risk profile for DTI for a specific patient might differ considerably when the two categorical risk factors are compared.
We performed a multivariate logistic regression analysis to determine if it was possible to include both BMI and weight. In the first step of the analysis, we included BMI, weight, height, and all other covariates that were statistically significant in univariate analyses. According to nonsignificant and high P values as well as nonsignificant changes of deviances, we excluded first weight and subsequently height from the model, leaving BMI as the only independent significant risk factor for DTI. Repeating the analysis with both the categorical BMI and weight confirmed the result. Therefore, height and weight may be confounders for BMI in the prediction of DTI.
Based on the cut-off values defining the three intervals of BMI and weight, the following dichotomous stand-alone tests were assessed (table 6). BMI of 35 or more as the only predictor had a sensitivity of 7% and a specificity of 94%. The predictive value of positive test was 6%. Using weight of 110 kg or more gave a predictive value of positive test of 7%, a sensitivity of 6%, and a specificity of 95%.
In both univariate and multivariate analyses, age, sex, Mallampati score, PDI, NMBA, weight, and BMI were significantly associated with FTI. In a multivariate analysis, BMI as a continuous covariate remained an independent risk factor of FTI (OR 1.031, 95% CI 1.002–1.061, P < 0.04); as a categorical covariate, BMI was not significantly associated with FTI. If BMI was replaced with weight, similar multivariate analyses demonstrated weight as an independent continuous covariate (OR 1.012, 95% CI 1.003–1.021, P < 0.01). As categorical covariate, only weight of 90–109 kg was significantly associated with FTI (OR 2.44, 95% CI 1.67–3.56, P < 0.0001), whereas weight of 110 kg or more was not (P < 0.119).
With the same methodological approach used for DTI, we performed post hoc analyses of the association between BMI of 35 or more and planned fiberoptic intubation. A new study cohort including patients primarily scheduled for fiberoptic intubation was retrieved. We performed logistic regression with Yes and No values for the dependent dichotomous covariate primarily scheduled fiberoptic intubation. In the univariate analysis, the OR was 1.31 (95% CI 1.12–1.53, P < 0.001); in the multivariate analysis (table 7), the OR was 1.17 (95% CI 0.99–1.39, P = 0.0588) for a patient with a BMI of 35 or more being scheduled to fiberoptic intubation.
We found that 5.2% of the patients had DTI, confirming the estimate in a previous meta-analysis.18In this large DAD cohort, obesity was associated with an OR of 1.42 for DTI with BMI of 35 or more and an OR of 1.24 with BMI of 25–35. This is much less than the estimates found by Shiga et al. , where obese patients had a three times higher risk of DTI than lean patients.18We also found that Mallampati score of III and IV and a history of PDI were predictors of DTI, with ORs of 3.7 and 6.3, respectively. Furthermore, male sex, scheduled surgery, and absence of NMBA were identified as risk factors of DTI with ORs of 1.35, 1.34, and 1.59, respectively. These OR were similar to that of obesity. The impact of obesity, in itself, on the risk of DTI therefore may seem weak.
The risk of DTI increased with degree of obesity, which concurs with earlier studies.9,13However, in our study the most reasonable cut-off values stratifying BMI differed from previous studies.7–10,12,22Our data suggest that a BMI of 35 is the cut-off value that relevantly divides overweight and obese patients into two groups with different risks of DTI. BMI of 35 or more may be a better clinical cut-off than BMI of 30 or more or BMI of 40 or more, as previously suggested.
We found that weight was also an independent risk factor for DTI. Compared to El-Ganzouri et al. 9our estimates of the OR were considerably lower, although we used the same stratification of weight. This difference may be explained by their use of Cormack and Lehanes classification of the glottic view as a surrogate parameter for DTI. In our final multivariate analysis, where BMI was compared with weight, only BMI qualified as a statistically significant risk factor. This indicates that BMI rather than weight may be the measure of choice for describing obesity as a risk for DTI.
As sole predictors of DTI, the accuracy of both BMI and weight assessed as dichotomous tests performed poorly. Previous studies have also suggested that the value of screening tests for DTI is limited when a single test is used.18,23Therefore, combining different predictors may improve diagnostic accuracy, and a multivariate test including numerous known risk factors for DTI may allow the use of BMI as a measure of obesity.
In our results, the low ORs of BMI and weight may indicate a limited importance of obesity as a risk of DTI. However, obesity has been identified as a risk of difficult mask ventilation.22,24,25Kheterpal et al. investigated 22,660 attempts of mask ventilation where BMI of 30 or more was identified as an independent risk factor for the combination of difficult mask ventilation and difficult intubation. Mask ventilation is an important rescue technique in a situation with DTI or FTI; therefore, the knowledge of obesity being a risk factor for DTI simultaneously with difficult mask ventilation is important. The airway management of obese patients may also be associated with accelerated oxygenic desaturation26and difficult emergency tracheotomy.27
Our analyses indicated that both age and male sex were risk factors of DTI. This may be due to sex being a confounder of parameters not observed, and the status of age is in agreement with Kheterpal et al. , who identified age of 57 yr or more as a risk of difficult mask ventilation.22
The present study is based on a large cohort of prospectively and consecutively collected data representing everyday experience from Danish clinical practice. Presently, half of Danish anesthesia departments register in the DAD. The number of registered patients is approximately half of all the patients in Denmark undergoing anesthesia. The DAD requires all registered indicators to be subjected to relevant rules of validation. This minimizes subsequent problems of missing and invalid data, which is supported by the fact that there was no noticeable difference between the estimates from the complete case analyses and the pooled estimates from the multiple imputations. The interface of DAD to register the airway evaluation, plan, and management was the same for all the registrations as well as the validation and user manual for the DAD. This confers a high external validity to our results. Our study included more than 91,000 registered intubations, thereby exceeding the total number of intubations included in the latest meta-analyses dealing with prediction of DTI.18,23This large number of patients enabled us to detect or reject even weak associations with great power and strengthened the precision of the estimates by narrowing their confidence intervals. However, we cannot ensure that controlled and uniform conditions were met and applied in all the patient encounters due to a heterogeneous population of patients and reporters and a lack of a national recommendation for airway management. This may reduce the internal validity of our data.
Obesity may create anatomical difficulties for the intubation caused by the decreased mobility and enlargement of structures in the throat and around the neck. Therefore, it seems rational to hypothesize that obesity in terms of BMI may be independently associated with DTI. However, the number of risk factors that may be considered for difficult intubation used in our study was limited. Including more predictors of DTI in our investigation may have changed the result, revealing BMI as a confounder for other and more closely related risk factors for DTI. This is suggested by the receiver operating characteristic curve analysis in our final model revealing an area under the curve of 0.65, indicating ample room for model improvement. That gender was independently associated with DTI may exemplify a likelihood that unknown confounding variables were missing in our analyses. The neck circumference may be a better and more relevant predictor than BMI, but again the current literature does not provide an adequate answer to this question.7,28–30Including other additional risk factors for DTI as the thyromental distance, ability of mouth opening, range of neck movement, or ability to prognath in a future analysis may remove obesity as an independent risk factor for DTI. Although these covariates are considered well-known risk factors for DTI, they have not succeeded as a stand-alone test or in combination to predict DTI with sufficient accuracy.9,18,31In our analysis, however, high BMI was identified as a risk factor for DTI. At the obese patient’s preanesthetic visit, the anesthetist should therefore take the precaution of including a very careful airway examination.
The importance of relaxation is supported by our results, as the lack of NMBA use for intubation was identified as a risk of DTI and FTI. It is a limitation of the study that there was no record of the educational level or years of experience of the persons performing the intubations. Those with least experience may have the most episodes of difficult intubations. No information regarding different height of the pillow and the position of head and neck of the obese patients during intubation was registered. An elevated compared with a sniffing position may improve the conditions for intubation of obese patients.32This information might have changed the results of our study if we had been able to include it in our analyses.
Confounding by indication is known to introduce bias when dealing with forecasts of DTI in any nonrandomized study involving interventions.33The airway management of anticipated DTI is likely to differ from that of unanticipated DTI. Our results could be biased by numerous variables not recorded in DAD. These unknown confounding variables may be important for the airway handling depending on the BMI. As an example, a more experienced physician may be allocated for the task; therefore, the obese patient may have been successfully tracheally intubated. It may be another confounder that a patient may be scheduled for a fiberoptic intubation because of obesity, leaving this particular patient ineligible for the analysis. This finding may be supported in part by our post hoc analyses, where the risk of being scheduled for fiberoptic intubation for patients with BMI of 35 or more was evaluated. Thus, the association was significant in our univariate assessment; whereas the association was marginally nonsignificant in our multivariate model. These results are encumbered with uncertainty because we cannot identify if a patient was initially allocated for fiberoptic intubation because of an anticipated difficult intubation or for other reasons. It is likely that an experienced anesthetist with the best airway management skills will be allocated for a patient judged to be of high priority. In our study, nonscheduled surgery may be associated with high-priority surgery, which may explain why it was associated with a decreased risk of DTI when compared with scheduled surgery.
This study adds to the numerous studies dealing with the prediction of DTI. It adds to the description of the total risk profile of patients with DTI. In our large cohort, increasing obesity was demonstrated as a risk factor for DTI independent of other risk factors registered in DAD. The impact of BMI of 35 or more on DTI was limited compared to other known risk factors such as PDI and the modified Mallampati score. BMI appears to be a better measure than weight to describe obesity as a risk for DTI. Obesity measured by BMI cannot in itself identify patients at risk of DTI. A multifactorial test including previously validated risk factors as well as obesity in terms of BMI of 35 or more may improve prediction of DTI.
The authors thank all anesthesia departments reporting to the Danish Anesthesia Database ( appendix 1), and to Jane Craknell, B.A., M.Sc., Coordinator and Managing Editor of Cochrane Anesthesia Group, Rigshospitalet, Copenhagen Ø, Denmark, for linguistic reviewing the manuscript.
Appendix 1. Contributing Departments.
Department of Anesthesia, Abdominal Center, Copenhagen University Hospital, Rigshospitalet, Copenhagen.
Department of Anesthesia, Head and Orto Center, Copenhagen University Hospital, Rigshospitalet, Copenhagen.
Department of Anesthesia, Juliane Marie Center, Copenhagen University Hospital, Rigshospitalet, Copenhagen.
Department of Anesthesia, Neuro Center, Copenhagen University Hospital, Rigshospitalet, Copenhagen.
Department of Anesthesia and Surgery, Bispebjerg Hospital, Copenhagen University Hospital, Copenhagen.
Department of Anesthesiology, Hvidovre, Hospital, Copenhagen University Hospital, Hvidovre.
Department of Anesthesia and Surgery, Amager Hospital, Copenhagen University Hospital, Copenhagen.
Department of Anesthesia, Frederiksberg Hospital, Copenhagen University Hospital, Frederiksberg.
Department of Anesthesia and Surgery, Glostrup Hospital, Copenhagen University Hospital, Glostrup.
Department of Anesthesia and Intensive Care, Herlev Hospital, Copenhagen University Hospital, Herlev.
Department of Anesthesiology, Næstved Hospital, Næstved.
Department of Anesthesiology, Nykøbing Falster Hospital, Nykøbing.
Department of Anesthesiology, Bornholm's Hospital, Rønne.
Department of Anesthesiology, Horsens Hospital, Horsens.
Department of Anesthesiology, Vejle Hospital, Vejle.
Department of Anesthesiology, Kolding Hospital, Kolding.
Department of Anesthesiology, Brædstrup Hospital, Brædstrup.
Department of Anesthesiology, Regionshospital Holstebro, Holstebro.
Department of Anesthesiology, Regionshospital Herning, Herning.
Department of Anesthesiology, Regionshospital Silkeborg, Silkeborg.
Department of Anesthesiology, Århus Sygehus, Århus University Hospital, Århus.
Department of Anesthesiology, Regionshospital Randers, Randers.
Department of Anesthesiology, Odder of Århus Hospital, Odder.
Department of Anesthesiology, Skejby Hospital, Århus University Hospital, Århus.
Department of Anesthesiology, Thy-Mors Hospital, Thisted.