In the United States, out-of-hospital endoscopy procedures are prevalent and continue to increase in number. In 2012, the Centers for Disease Control reported 15 million colonoscopies performed in the US, with the vast majority occurring in out-of-operating-room locations.1 In 2017, a private analytics company estimated an approximate 51.5 million gastrointestinal endoscopies being performed across the United States in that year alone. With factors including the updated recommendation for earlier colorectal cancer screening, these numbers are only expected to increase.2

It is also widely recognized that the rates of morbid obesity and super morbid obesity continue to rise rapidly.3 The association of morbid obesity with significant physiologic changes including redundant airway tissue, increased baseline oxygen consumption, ventilation-perfusion mismatching, atelectasis, right to left shunt caused by airway closure, and decreases in functional residual volume and partial pressure of oxygen narrow the margin of safety for deep sedation.4,5,6,7

While out-of-hospital locations maximize efficiency with rapid case turnover, they often offer limited resources for rescue airway management (i.e., limited personnel, lack of anesthesia machine, limited airway tools). Therefore, to provide safe and effective deep sedation during endoscopy procedures, thoughtful risk stratification of patients is essential in the decision to perform anesthesia at an outpatient location. The lack of consistency among providers and institutions regarding body mass index (BMI) restrictions for out-of-hospital locations highlights the need for appropriate and evidence-based recommendations, while considering the resources and limitations at each individual procedure site.

It has remained unclear at what threshold morbid obesity independently increases complications among patients undergoing out-of-hospital deep sedation. In 2006, the rates of hypotension, hypoxia, delayed discharge, and unplanned hospital admission among morbidly obese patients were similar to those among non-obese patients, but numbers were small.8 Other studies examining obesity complication rates have small sample sizes, low mortality rates, or focus on the presence of other comorbidities.9

Our objective was to quantify the severity and duration of hypoxemic events in morbidly obese patients compared with non-obese patients during outpatient endoscopy procedures performed under deep sedation provided by solo anesthesiologists and supervised certified registered nurse anesthetists (CRNA). We also sought to determine the relationship of intravenous fentanyl on hypoxemia. We define hypoxemia as oxygen saturation (SpO2) of less than 90% by pulse oximetry for greater than one minute. We define severe hypoxemia as SpO2 less than 85% or greater than one minute. The goal of this large-scale retrospective analysis is to provide evidence-based institutional guidance for risk stratification of otherwise healthy morbidly obese patients for the off-site, out of hospital setting. We hypothesized that patients with increased BMI were more likely to experience hypoxemia than those with normal BMI during outpatient endoscopy procedures.

Methods

This was a retrospective cohort study of adult patients who underwent colonoscopy or esophagogastroduodenoscopy (EGD) under deep sedation provided by either solo anesthesiologists or supervised CRNAs between June 2015 and June 2016. The database of included patients was constructed using discharge demographic data from a single institution in an urban setting that includes six freestanding ambulatory gastroenterology centres. Only adult patients (18 yr and older) were included. Our institutional review board has deemed this project non-human subject research and does not require review of these projects.

From the preoperative medical comorbidities, the following were extracted by text search from the problem list, which is included in every patient electronic health record and can be populated as discrete fields by any documenting provider: coronary artery disease, asthma, chronic obstructive pulmonary disease, gastroesophageal reflux disease, obstructive sleep apnea, diabetes mellitus, and smoking status. The problem list was searched using these precise words as discrete terms can be entered to define the given comorbidity (i.e., chronic obstructive pulmonary disease is listed as “chronic obstructive pulmonary disease” or “chronic obstructive pulmonary disease (COPD)” and not only “COPD”). Body mass index was reported in kg·m−2. BMI was calculated by the electronic health record based on the most recent available weight in kilograms and height in metres. For elective procedures, height and weight were measured and entered into the electronic health record for each patient upon admission to the presurgical holding area on the day of the procedure.

Intraoperative vital signs were queried from the electronic health record (EPIC, Verona, WI, USA), which included SpO2 as measured by the pulse oximeter with a resolution of one reading per minute. Intraoperative fentanyl administration was also queried. Only adult outpatient gastroenterological procedures were included. Cancelled cases were excluded. Records that did not contain a BMI or contained a physiologically unlikely BMI (> 100 kg·m−2 or < 10 kg·m−2) were excluded.

Patients were pre-selected for American Society of Anesthesiologists (ASA) physical status score (I or II) and absence of obstructive sleep apnea (OSA) requiring continuous positive airway pressure or bilevel positive airway pressure devices. Recent documentation of cardiology assessment was required in the presence of congestive heart failure, coronary artery disease (CAD), coronary stents, arrhythmia, or anticoagulation use. ASA score was defined based on the designation made by the attending anesthesiologist on the day of the procedure.

ASA physical status scores of three patients including those with a BMI greater than 40 kg·m−2, presence of OSA, poorly controlled hypertension, alcohol dependence or abuse, implanted pacemakers, history (> three months) of CAD/stents [9] were permitted at the discretion of the anesthesiologist. A core group of 20 attending anesthesiologists practiced solo. A single site utilized 1:3 CRNA coverage and accepted patients with BMI < 45 kg·m−2. All patients received nasal cannula oxygen therapy and primarily propofol based total intravenous anesthesia. Nasal oxygen flow rate was set at the discretion of the provider and this data did not automatically import to the electronic medical record and was not included in our study. Propofol was dosed to a level of deep sedation with purposeful response to repeated or painful stimulation.10

Body mass indexes were separated into World Health Organization classification.11 All SpO2 measurements were recorded from the time of attaching the pulse oximeter (Nellcor, Boulder, CO) to the patient until the pulse oximeter was removed upon departure from the procedure area. Recorded oxygen saturations of less than 60% were considered aberrant and excluded from analysis, based on a method utilized in previous studies.12 Desaturations less than one minute duration were considered artifacts and likely clinically irrelevant to conclusions. Based on previous studies, hypoxemia was defined as at least one recorded measurement of SpO2 less than 90%. Severe hypoxemia was defined as at least one recorded measurement of SpO2 less than 85%. Because SpO2 is recorded in our anesthesia information management systems at a resolution of one reading per minute, total time of hypoxemia (in minutes) was defined as the sum of SpO2 measurements less than 90%.

Out-of-operating-room location was defined as any procedural site that is not a fully operational operating room. Off-site location was defined as any procedure location that was not a fully functional operating room and geographically separate from the main hospital. An off-site ambulatory surgery centre (ASC) was designated an Article 28 facility regulated by The Joint Commission on Accreditation of Healthcare Organizations. Office-based was defined as an off-site location regulated by American Association for Accreditation of Ambulatory Surgery Facilities, Inc. Of our six procedure locations, four were ASC and two were office-based.

All descriptive values are presented as mean (standard deviation), median [interquartile range (IQR)], or frequency (percentage) as deemed appropriate. Normality was assessed using qq-plots and histograms. Normal continuous data were compared using t test or one-way analysis of variance, non-parametric continuous data were compared using Wilcoxon rank sum test or Kruskal–Wallis test, and categorical data were compared by Chi square or Fisher’s exact test.

Logistic regression analyses were performed with hypoxemia, severe hypoxemia, and prolonged hypoxemia (greater than five minutes of hypoxemia) as the binary outcome variable in a separate analysis. Categories of BMI were dummy-coded independent variables, using the normal range of BMI (18.5–24.9 kg·m−2) as the reference category. Unadjusted analysis was performed, as well as risk-adjusted analysis accounting for the following confounders: age, sex, ASA physical status score, type of procedure, smoking status, COPD, asthma, use of intravenous fentanyl, and length of procedure, with the exposure being the interaction between categories of BMI, and OSA or intravenous fentanyl administration. Increasing age is associated with reduced oxygen tension.5 ASA physical status score is a general indicator of global patient status. Chronic obstructive pulmonary disease, smoking status, and asthma may be associated with reduction of intraoperative oxygenation due to relationship with lung function. Use of intravenous fentanyl was included because of its effect as a respiratory depressant, and the interaction was included to explore the relationship between increasing BMI and use of intravenous fentanyl. Both unadjusted and adjusted odds ratio (AORs) were determined from the logistic regression coefficients, and 95% confidence intervals (CI) are reported as well.

Visual assessment of restricted cubic spline plots suggested the increased probability of hypoxemia, severe hypoxemia, and prolonged hypoxemia at a given BMI. Adjusted analyses were used for these plots, with BMI as a continuous variable and consistent default values across all three plots for the other covariates so that plot shapes could be contrasted.

Sensitivity analyses

Several sensitivity analyses were performed to further explore the data and the relationships to contributing factors. First, an analysis was performed which included increasing durations of hypoxemic time (≥ one minute, ≥ two minutes, ≥ three minutes, ≥ four minutes and ≥ five minutes). Both unadjusted and adjusted analyses were done, just as in the primary analysis, controlling for the same variables. Second, instead of eliminating missing values (approximately 6% of data), an analysis was performed in which missing values were imputed using multiple imputation (“mice” package in R). The primary analysis was then conducted on the imputed data. Third, an analysis was performed which only included variables available prior to the start of the procedure, specifically fentanyl use and length of the procedure. Fourth, an analysis was performed in which procedure type (EGD and/or colonoscopy) was included as an interaction term with BMI class.

A P < 0.05 was considered statistically significant. All statistical operations were performed using R statistical software (version 3.1.1, R Foundation for Statistical Computing, Vienna, Austria).

Results

After application of exclusion criteria, 11,595 patients were identified (Figure). Characteristics (age, sex, medical comorbidities) of patients that were excluded for having an obviously erroneous or missing BMI in the health record were assessed, and found to be similar to those that had an available BMI, indicating a representative sample. Baseline patient characteristics are shown in Table 1, and patient characteristics separated by BMI classification are shown in Table 2.

Figure
figure 1

Flowchart of patients included in the study cohort

Full size image
Table 1 Patient characteristics
Full size table
Table 2 Characteristics of patients separated by body mass index classification
Full size table

Hypoxemia occurred in 1,625 patients (14%), severe hypoxemia in 685 patients (5.9%), and prolonged hypoxemia in 76 patients (0.66%) (Table 3).

Table 3 Intraoperative hypoxemia in all patients and separated by body mass index classification
Full size table

When comparing incidence of hypoxemia across ranges of BMI, 600 patients (13%) within the normal BMI range experienced hypoxemia, compared with 314 (18%) of class I obesity patients, 159 (27%) of class II obesity patients, and 24 (19%) of class III obesity patients.

Of normal BMI patients, 215 (5.5%) experienced severe hypoxemia, compared with 127 (7.2%) of class I obesity patients, 72 (12%) of class II obesity patients, and 11 (8.5%) of class III obesity patients. Prolonged hypoxemia occurred in 23 (0.58%) of patients with normal BMI, and in 34 (0.88%) of all obese patients.

Adjusted restricted cubic spline plots suggested the increase in probability of hypoxemia rose most sharply with BMI greater than 25 kg·m−2, the probability of severe hypoxemia with BMI greater than 30 kg·m−2, and the probability of prolonged hypoxemia with BMI greater than 30 kg·m−2.

Both unadjusted and adjusted analyses showed increased odds of hypoxemia with increasing BMI classification until class III obesity (Table 4). Adjusted odds ratio of any occurrence of intraoperative hypoxemia were 1.61 (95% CI, 1.35 to 1.90; P < 0.001) in the class I obesity group and 2.61 (95% CI, 2.05 to 3.30; P < 0.001) in the class II obesity group compared with patients with normal BMI.

Table 4 Unadjusted and adjusted odds ratios of developing hypoxemia, severe hypoxemia, and greater than five minutes of hypoxemia for different ranges of body mass index
Full size table

Fentanyl use

Fentanyl was used in 1,249 (10.8%) of cases. It was used in 962 (16%) cases with an EGD, and 690 (8.4%) cases with a colonoscopy. In cases in which fentanyl was used, the median [IQR] dose was 100 [50–100] µg. Intravenous fentanyl was associated with intraoperative hypoxemia (AOR, 1.67; 95% CI, 1.33 to 2.09; P < 0.001), though the interaction between fentanyl use and BMI was not significant for any category of BMI. Adjusted odds ratio of experiencing severe hypoxemia were significant in the class I obesity group (AOR, 1.47; 95% CI, 1.13 to 1.89; P = 0.003), and the class II obesity group (AOR, 2.59; 95% CI, 1.86 to 3.57; P < 0.001). Use of intravenous fentanyl was associated with intraoperative severe hypoxemia (AOR, 1.79; 95% CI, 1.30 to 2.43; P < 0.001), though the interaction between fentanyl use and BMI was not significant for any category of BMI. Adjusted odds ratio of experiencing prolonged hypoxemia increased with each category of BMI from 1.97 (95% CI, 1.08 to 3.69) in the overweight group to 9.20 (95% CI, 4.74 to 18.03) in patients with class III obesity.

Sensitivity analyses

The first sensitivity analysis included increasing durations of hypoxemic time to further characterize the relationship between BMI and hypoxemia. The results of this analysis are shown in Table 5.

Table 5 Unadjusted and adjusted odds ratios of developing increasing durations of hypoxemia
Full size table

The second sensitivity analysis was performed on a data set with imputed data instead of eliminated missing data. In the adjusted analysis, BMI was not associated with hypoxemia in the overweight group (P = 0.139), but the analysis did suggest an association with class I obesity (OR, 1.1; 95% CI, 1.3 to 1.8; P < 0.001), class II obesity (OR, 1.5; 95% CI, 1.9 to 3.0; P < 0.001), and class III obesity (OR, 2.0; 95% CI, 1.2 to 3.2; P = 0.005).

The third sensitivity analysis was performed without certain intraprocedural variables, specifically fentanyl use and procedure length. In the adjusted analysis, BMI was not associated with hypoxemia in the overweight group (P = 0.08), but an association was suggested between hypoxemia and class I obesity (OR, 1.5; 95% CI, 1.3 to 1.8; P < 0.001), class II obesity (OR, 2.7; 95% CI, 2.2 to 3.4; P < 0.001), and class III obesity (OR, 1.8; 95% CI, 1.1 to 2.8; P = 0.02).

In the fourth sensitivity analysis, the interaction term between procedure type and BMI class was not statistically significant for any class of obesity. There were significant relationships between hypoxemia and EGD (OR, 2.3; 95% CI, 1.8 to 2.9; P < 0.001) and hypoxemia and colonoscopy (OR, 0.74; 95% CI, 0.58 to 0.94; P = 0.01) in the sensitivity multivariable analysis.

Discussion

This study represents a large cohort of patients examined for intraprocedure hypoxemia during elective endoscopy using deep sedation by anesthesia providers. Other studies have directed study of associated comorbidities including OSA, asthma, smoking, hypertension, congestive heart failure, CAD, and include data from non-anesthesia provider directed sedation.8,13 We found that even modest obesity confers an independent risk of hypoxemia. The incidence of severe hypoxemia increased nearly six-fold in obese patients and 8.5-fold in class III obese patients when compared with those of normal BMI. We also found that while fentanyl use was associated with both severe and prolonged hypoxemia, this relationship did not differ across BMI categories.

Frequently, the question arises as to what maximum BMI is acceptable for ambulatory procedures involving deep sedation. Production pressure influences the decision to perform elective anesthesia on patients with increasing comorbidities. Body mass index limits thus far are derived from opinion rather than published literature as major adverse cardiac events are extremely rare in outpatient endoscopic procedures. In our centre, an urban academic medical centre comprised of six free-standing ambulatory centres, ASA I and II are allowed (ASA III at the discretion of the anesthesiologist), and one centre allows BMI up to 45 kg·m−2. Even with a dedicated anesthesia provider monitoring and administering deep sedation, hypoxemic events occur at an increased rate among obese patients. Odds ratios show a doubling of likelihood of hypoxia starting at BMI 30 kg·m−2. Even at five of our six examined locations, which limit BMI to less than 40 kg·m−2, we observed a trend of increasing likelihood of hypoxemic events as BMI increases. These results are similar to an exploration of hypoxemic events in obese patients having general anesthesia.14

We therefore advocate that caution should be observed in patients with a BMI greater than 30 kg·m−2. In addition, many of the interventions often employed to minimize hypoxemia during general anesthesia, such as positive pressure ventilation and recruitment maneuvers, are often not readily available without an airway device.15,16,17 Patients with a constellation of comorbidities known to elevate risk including OSA, asthma, CAD, smoking, and obesity represent the highest risk for elective endoscopic and colonoscopic procedures and deserve the greatest consideration for risk assessment and risk mitigation.18,19

Fentanyl use has been advocated to decrease overall propofol dosage requirement and decrease airway reflexes for upper endoscopy. Among the six endoscopy centres included in this study, only a single centre represents a higher BMI limit (45 kg·m−2) and primarily utilizes supervised CRNAs; however, use of fentanyl is homogeneous across each of the six endoscopic locations. Increase in the likelihood of hypoxemia likely represents a synergistic effect of combined propofol, fentanyl, ± midazolam. Our data, therefore, do not support the use of fentanyl for endoscopic procedures because of the independent increase in the likelihood of hypoxemic events.

Our study is limited by its retrospective design. Because individual patient care was left to the discretion of the anesthesia provider, variability among practice patterns exist. These differences may include variable oxygen flow rates, thresholds for intervention for hypoxemia, and medication choices. Nevertheless, these providers consist of a core group of 20 attending anesthesiologists who frequently cover off-site locations and are known to practice similarly. We therefore feel that these differences are small and unlikely to alter our results. Additionally, we recognize that these data were harvested exclusively from a large academic medical centre in an urban environment, so may not be generalizable to other populations. Finally, because only a single site included patients with a BMI 41–45 kg·m−2, the sample size for these patients is small.

Although there are limitations to this study, there are also several strengths which enhance the existing body of literature. First, this is a large cohort of healthy patients undergoing colonoscopies and EGD under deep sedation by a dedicated anesthesia provider. Additionally, physiologic data were auto-populated into the electronic medical record at frequent intervals and not entered manually, making it less susceptible to manipulation. Frequent interval SpO2 measurements also ensured that short episodes of hypoxemia were detected.

Patient selection must include obesity to appropriately risk stratify patients to locations with adequate resources for safe care.