Introduction

Sedation for endoscopy, especially advanced endoscopic procedures, is recommended.1,2 Sedation has been shown to increase the quality of various procedures1,3,4, while being safe1,5, especially if propofol mono-sedation is used, as recommended by the European Society of Gastrointestinal Endoscopy (ESGE) and others1,2,6. However, the discussion on whether sedation by the endoscopist is safe or should be performed and monitored by an anesthesiologist (possibly under general anesthesia) is ongoing.1,2,7

Very scarce prospective data exists concerning hypoxemia and other sedation-related adverse event rates of endoscopic retrograde cholangiopancreatography (ERCP) performed under non-anesthesiological/nurse-administered propofol sedation (NAPS).8 Studies of various designs and using various definitions have reported respiratory depression (hypoxemia) rates ranging from ~ 0.1% to 45.5%.5,9,10,11 Studies investigating monitored anesthesia care (MAC) performed ERCPs often excluded higher risk patients or the decision to use general anesthesia was not random.9,10,11 Thus, the real rate of desaturations among NAPS-performed ERCPs in everyday practice is not really known.

Another important aspect of high quality endoscopy is documentation and monitoring of adverse events. This is recommended not only to ensure proper quality standards, but also to provide opportunities to learn and improve through situational awareness and feedback.12,13,14,15 However, to our knowledge, no study has been performed investigating endoscopist (and staff) awareness of adverse events. Thus, the aims of our prospective, quality control study were to assess the rate of hypoxemic events and measures taken by endoscopists during unselected NAPS-performed ERCPs, as well as evaluate endoscopist awareness of overall adverse events evaluated by adequate documentation.

Methods

This prospective, observational, quality control study was conducted at the Department of Internal Medicine 2 Gastroenterology & Hepatology, University Hospital of St. Pölten. The department is an academic, tertiary referral center with an interdisciplinary out- and inpatient endoscopy unit attached. Approximately 700 ERCPs are performed per year and there is an on-call ERCP team available during weekends and holidays. ERCPs are mostly performed with non-anesthesiological/nurse-administered propofol sedation (all with low-flow (2L O2/min) oxygen supplementation over nasal cannula). Rarely, ERCP is performed with anesthesia stand-by or under general anesthesia (GA). All patients stay overnight after an ERCP (unless they are transferred to another hospital) and a standardized laboratory is drawn a few hours after the procedure and the next day. Due to the quality control nature of the study, no informed consent was collected from patients, to reduce bias. The study was submitted to the local ethics committee (Ethics Committee of the State of Lower Austria, St. Pölten, RefNr 141,222–1451) and a waiver was obtained.

Patients and data collection

All patients (male and female, any age) undergoing ERCP for any indication during the defined time period (mid of January to May 2023) were included. Patients were also included if they underwent emergency ERCP, also during weekends and holidays, and not excluded in case of more than one procedure. Endoscopy nursing staff completed a predefined data record sheet for every patient including the type of sedation method (NAPS or anesthesia), baseline oxygen flow (at the start of the procedure), whether a desaturation occurred, lowest SpO2 value, and what measures were taken to correct for hypoxemia. Patients were then followed up by our study team by searching the electronic health care records for any postprocedural adverse events. Also, basic epidemiological data and data on comorbidities was collected, the Charlson Comorbidity Index (CCI) was calculated for each patient. Department policy states that all adverse events occurring during or after endoscopy procedures must be recorded (by endoscopists or physician becoming aware of the adverse event) via a dedicated electronic health care record sheet imbedded in the endoscopy report. Whether an adverse event report was made was also recorded by the study team. Endoscopy nursing stuff filled in the data record sheet, but where not specifically instructed about the nature of the study and endoscopists were not informed about the quality control study – both to reduce bias (Hawthorne effect).

The study team regularly collected the data record sheets and matched the recorded ERCPs with the performed ERCPs. All ERCPs performed in the study period were recorded separately by the study team to detect missed ERCPs.

Outcomes and statistical analysis

The main objective of the study was to estimate the incidence of desaturations during NAPS. The definition of desaturation was based on the ESGE definition16, i.e., an SpO2 < 85%. Secondary desaturation-related objectives were the incidence of hypoxemic events, defined as either a desaturation or an intervention for a drop in SpO2 values, exploring risk factors for hypoxemic events, and exploring whether hypoxemic events are associated with worse outcomes.

Furthermore, the aim was to prospectively collect overall incidences of ERCP-related adverse events and to assess awareness of ERCP-related adverse events by correlating (mandatory) documentation of adverse events by endoscopists (and staff) with the actual incidences as detected by the study team. All definitions of ERCP-related adverse events were based on ESGE definitions16. The only modification to the ESGE definition was made concerning bleeding. Bleeding was also counted if any endoscopic intervention (e.g., stenting) was performed to stop bleeding, due to this fulfilling AGREE IIIa criteria for adverse events (which are used at our department for classifying adverse events).17

Proportions and 95% confidence intervals of adverse events were computed. Associations of risk factors were analyzed with t-tests or Mann–Whitney tests, for categorical data the Fisher’s exact test was used. Categorical variables are reported as absolute and relative frequencies and continuous variables as arithmetic mean and standard deviation (SD) (or 95%CI) or median and 25–75% interquartile ranges (IQR), as appropriate. After identification of statistically significant univariable risk factors, multiple logistic regression analysis was performed to assess the association between hypoxemic adverse events and clinical and procedure-related data to identify risk factors. Statistical significance was defined as p ≤ 0.05 and statistically significant results are considered exploratory. All analyses were conducted as 2-sided tests. Adjustments for multiple testing were made in the multiple logistic regression model concerning predictors of hypoxemic events, where adjusted Odds Ratios are derived from. Data were analyzed and graphics were produced in Graphpad Prism 10 for Windows (GraphPad Software, Boston, MA, USA).

Results

During the study period, 232 ERCPs were performed and of those 14 (6%) were performed under general anesthesia (included for the purposes of overall adverse events). Mean age of patients was 67.8 (SD 15.7) years, 51.7% were female, and median length of stay was 5 (IQR 2;11) days (Table 1).

Table 1 Epidemiological and clinical data.
Full size table

Concerning comorbidities, 47.6% of patients were current or former smokers, 38.9% had a history of cancer, 29.2% suffered from diabetes, 25.9% from congestive heart failure, 10.6% from obstructive pulmonary disease, and 10.6% had a history of a myocardial infarction. The median CCI was 4 (IQR 2;7). All-cause 30-day mortality was 5.5%. Among the 218 NAPS-performed ERCPs, median patient ASA-status was 2 (IQR 2;3), median propofol dose 430 (IQR 310;670) mg, median procedure length was 59 (IQR 40;83) min – as recorded by nurses with pre- and post-preparation, not endoscope intubation to extubation), and 51 (23.4%) were emergency procedures. Propofol mono-sedation was used in 98.2%, in 1.8% midazolam was added, and 86.8% of procedures were started on 2 L/min of oxygen flow. A valid desaturation data record sheet was available for 199 (91.3%) of NAPS-procedures.

Sedation-related adverse events

A desaturation occurred in 22.6% (95%CI 16.7–28.5%) of procedures (Fig. 1). The minimum recorded SpO2 of patients who desaturated was 80% (IQR 72;82%), which was significantly lower than of those who did not desaturate (97%, IQR 95;98%; p < 0.0001). A hypoxemic event occurred in 28.1% (95%CI 21.8–34.4%) (Fig. 1), i.e., staffed intervened more often earlier than when the official desaturation definition was met. The most common interventions to correct for hypoxemic events (multiple possible) were increasing O2-flow (26.6%), chin-lift (14.6%), and placing a Wendel-tube (9%), the procedure was terminated in 1% (2 cases). No cases of unplanned intubation occurred. Hypotension, another sedation related adverse, events occurred most often (40.1% of cases). Performing cardio-pulmonary-resuscitation (CPR) was necessary in 1 case (0.4%).

Fig. 1
figure 1

Desaturation and association with mortality. Left side incidences of desaturation (dotted filled) and interventions for hypoxemia (dashed filled); right side mortality of patients without hypoxemic events (unfilled) and with hypoxemic events (dashed filled); means with 95% CIs; * =  p < 0.05.

Full size image

Hypoxemic patients were, on univariable analyses, heavier (p = 0.01), or had higher BMI (kg/m2) respectively (p = 0.009), and were more likely ASA3/4 (p = 0.003), and reported to be snoring more often (p = 0.002) (Fig. 2). However, age (p = 0.44), CCI (p = 0.19), emergency procedure (p = 0.84), procedure length (p = 0.61), propofol dose in mg per kg body weight (p = 0.72), and presence of cholangitis (p = 0.63) were all not associated with hypoxemic events. Also, sex was not associated with desaturations (p = 0.87). On multiple regression analysis, including significant parameters (BMI, ASA-status, snoring) and important clinical parameters (age, sex, CCI, emergency procedure) only BMI (p = 0.04, adjusted OR 1.08 (95%CI 1.00–1.17)) and ASA-status (p = 0.04, adjusted OR 2.55 (95%CI 1.34–4.87)) remained statistically significant. Unfortunately, no clinically meaningful risk score or estimation could be derived from the data. Of note, 30-day overall mortality of patients with hypoxemic events was statistically significantly higher than of those without (10.7% vs. 2.8%, RR 3.83, p = 0.03) (Fig. 1). The majority of causes of death were due to (rapid) overall decline due to oncological background disease (pancreatic cancer n = 3, cholangiocellular carcinoma n = 2, gallbladder cancer n = 1), and otherwise cholangitis in the context of oncological disease (n = 3) and sepsis (n = 3).

Fig. 2
figure 2

Parameters statistically significantly (p < 0.05) associated with interventions for hypoxemia in univariable analysis; upper left graph weight in kg of patients with hypoxemia (filled dots), and without (grey dots) with mean and 95%CI; upper right graph BMI (kg/m2) of patients with hypoxemia (filled dots), and without (grey dots) with mean and 95%CI; lower left graph proportion of patients with hypoxemia according to ASA status (note: very patients with ASA 1); lower right graph proportion of patients with hypoxemia according to snoring as reported by the patient (note: few patients with unknown status).

Full size image

Other adverse events

The most common other adverse event was intraprocedural hypertension, occurring in 11.2% (95%CI 7.1–15.3%) (Table 2). Also, intra- or postprocedural bleeding occurred in 9.1% (95%CI 5.3–12.7%), however none were fatal or demanded interventional radiological or surgical intervention. Post-ERCP pancreatitis occurred in 3.9% (95%CI 1.3–6.3%), post-ERCP cholangitis in 1.7% (95%CI 0.03–3.4%). No cases of post-ERCP cholecystitis and perforations were recorded. Cardiopulmonary adverse events (any) increased statistically significantly with higher ASA-status (p = 0.012) (Fig. 3).

Table 2 Occurrence and documentation of AEs.
Full size table
Fig. 3
figure 3

Proportion of patients with any cardiorespiratory adverse event according to ASA status (hypotension, hypertension, hypoxemia, cardiopulmonary resuscitation, 30-day all-cause mortality – any one of those), chi-squared test p = 0.012.

Full size image

AE documentation and awareness

Compliance with adverse documentation overall was low, with only 8.5% of adverse events found by the study team having been documented by staff (Table 2).

All cases of CPR (1/1) were documented. Documentation was second highest with post-ERCP pancreatitis (22.2%) and third highest with desaturations (20%). Of all bleeding events, 19% had been officially documented. Surprisingly, only 2.2% of intraprocedural hypotensive events (being the most common adverse event overall) were documented. No documentation was found concerning post-ERCP cholangitis (0%), intraprocedural hypertension (0%), and 30-day all-cause mortality (0%). It should be noted, that no deaths were directly linked to the procedure after individual review of all deceased patients. Nevertheless, all deaths within 30 days of an endoscopic procedures should be recorded, according to department policy.

Discussion

In this prospective, observational, quality control study desaturations occurred in 22.6% and hypoxemic events in 28.1% of NAPS-performed ERCPs, with 99% of procedures successfully completed. Routine awareness of ERCP-related adverse events among endoscopists and staff seems low, with only 8.5% of all adverse events being officially documented.

Sedation during advanced endoscopy procedures is recommended and safe1,2, but there is disagreement as to whether anesthesia monitored care is necessary to ensure adequate patient safety. One of the largest prospective studies of 470 MAC-performed ERCPs (and 58 GA cases) by Berzin et al.9 reported overall sedation-related adverse events in 21%, desaturations in 12.5%, and emergency intubations in 3%. Other trials evaluating MAC (vs. general anesthesia) or sedation regimens reported desaturations in 19.2% of cases and airway maneuvers to correct for hypoxemia 45.5%11 and desaturation rates of 15.4%10, comparing well to our 22.6% and 28.1% rates. A single study investigating NAPS8 reported a desaturation rate of 16.2%. However, care must be taken when directly comparing desaturation rates between studies, since some studies of MAC excluded emergency procedures or unstable airway patients 11 or the decision whether MAC or intubation was indicated was left to the anesthesiologist9, while others used additional monitoring such as capnometry8,10, all possibly influencing hypoxemia rates. No randomized controlled trials exist that have demonstrated superior safety outcomes of MAC over NAPS. Even more so, without strict exclusion criteria hypoxemia rates of 18.9% with GA have been reported.18 Overall, sicker patients, e.g. higher ASA class or specific comorbidities, seem to experience significantly more sedation-related adverse events, regardless of whether the procedure is performed under MAC, and only intubation may significantly reduce the risk of respiratory complications.9,11,18 Interestingly, respiratory AEs in our study were statistically significantly associated with higher all-cause 30-day mortality (10.7% vs. 2.8%, p = 0.03). It is unlikely that mortality was causally linked to desaturations/NAPS – 30-day all-cause mortality in the GA group was 14.3% (of note, low numbers) – but, possibly, desaturations should lead to closer postinterventional monitoring, especially of sick patients. Mortality in this group is likely linked to the comorbidity of this group, while further confounders cannot be excluded. Thus, postprocedural observation, possibly within the context of intermediate care units for example, might be warranted and may possibly improve outcomes. High quality studies directly comparing NAPS with MAC for prevention of sedation-related AEs do not exist and whether transient desaturations or prevention of those (by GA, e.g.) are clinically meaningful has not been proven.18 Given the ongoing shortage of anesthesia capacity and the uncertain clinical significance of respiratory AEs, it is questionable whether an increase in GA endoscopies and the associated reduction in endoscopy unit efficiency is warranted.5,18,19,20

Another important finding is that awareness or consciousness of periprocedural AEs among endoscopists and staff is low. Only 8.5% of adverse events were documented (including other department staff), despite mandatory documentation according to department policy. Feedback is a crucially important aspect of endoscopy training.12,13,14,15 Being aware of the nature and incidences of adverse events provides opportunities to improve and is essential to uphold quality standards. Thus, guidelines recommend keeping track of (overall and individual) complication rates.14,15. However, mental recognition of an adverse event having occurred is a prerequisite for any measures aimed at managing, following up, and reducing complications, as well as providing feedback. We are not aware of other studies evaluating individual documentation and awareness of adverse events in endoscopy, except for surveys evaluating knowledge and use of minimal standard terminology21, which was low. The reasons for low AE reporting in this study can only be speculated upon, as cause-analysis was not the focus of this investigation. Internal discussions with endoscopists indicated that the most commonly cited reasons for not documenting AEs included time constraints, unclear definitions of what constitutes an AE, and lack of awareness regarding the occurrence of late complications. It is well established that even general quality control reporting adherence is low.22,23 Constantly increasing routine clinical work-loads leading to severe time constraints, electronic health record (EHR) systems not designed for clinical usability and feedback, and underrecognition of an adverse having occurred are likely factors.24,25 Endoscopy is regularly described as very or extremely safe but adverse events happen (quite often). Obviously some adverse events were recognized and acted upon (i.e., stent placement for bleeding), however, consciousness of the occurrence of an adverse event includes awareness beyond the immediate periprocedural setting. Recognizing complications and implementing structured support of and feedback for endoscopists, especially trainees, after (critical) adverse incidents provide the chance to learn, cope, and grow, also reducing the risk of burning out in adversity.26 Systematically improving adverse event recognition and reporting is challenging, and likely requires intervention at a higher level than individual endoscopists or endoscopy units.24 Large language models reducing manual documentation efforts could be useful in this regard, as has been shown, but would require integration into the EHR.27

A limitation of our study is that whether the procedure was performed with NAPS or general anesthesia was not random. However, the quality control nature of our study precluded randomization. This is also a strength, since all patients and procedures were included, thus our data are more generalizable and not selective. Also, a waiver was obtained by the ethics committee, i.e., informed consent was not collected for this quality control study. However, after intense discussion it was concluded that it would not be possible, post-hoc, to exclude the possibility that high-risk groups (e.g., emergency procedure patients, very sick patients,…) would be underrepresented if subjects could only be included with informed consent, severely limiting validity and generalizability.

In conclusion, sedation-related adverse, especially desaturations, are common among NAPS-performed ERCPs, but comparable to reported rates in the literature of MAC-performed ERCPs. Most procedures are completed successfully and clinically relevant, immediate consequences of such temporary desaturations are uncertain. Unfortunately, endoscopists’ awareness of adverse events seems low as measured by documentation of complications.