We conducted a single-centre retrospective cohort study in a tertiary referral university hospital. Electronic medical records of 823 adult patients were reviewed based on sample-size calculation. The included patients were ≥18 years old, had American Society of Anesthesiologists (ASA) physical status classification scores I–III, and were undergoing elective outpatient GIEPs managed by anaesthetists using propofol TCI sedation during September 2018. According to the recommendations of the Anesthesiology Department, all gastrointestinal endoscopies are performed using TCI propofol sedation except in the cases that require general anaesthesia due to individual or procedural characteristics or contraindication to propofol. Patients scheduled for endoscopic retrograde cholangiopancreatography, enteroscopy, and procedures performed under planned tracheal intubation were excluded in both cohorts.
Ethical approval was provided by the Ethics Committee of the Hospital Italiano de Buenos Aires, Buenos Aires, Argentina (Chairperson: Dr. Augusto Pérez, Ethical Committee N° 3885) on September 24, 2018. The requirement for written informed consent was waived due to the retrospective nature of the study.
The patient variables recorded included sex, age, ASA physical status classification score, weight, height, and significant comorbidities (diabetes mellitus, arterial hypertension, dyslipidaemia, current/former tobacco smoking, coronary artery, cerebrovascular, and peripheral vascular diseases; congestive heart failure, chronic obstructive pulmonary disease, and obstructive sleep apnea). The procedures comprised diagnostic and therapeutic esophagogastroduodenoscopy (EGD), colonoscopy, or both. Procedure-related data recorded included anaesthetic strategy (propofol TCI sedation), total propofol dose (mg), use of opioid drugs (fentanyl), and total fentanyl dose (mcg); type of procedure (EGD, colonoscopy, or combined procedure), and procedure duration (minutes; including time for monitoring and nasal cannula placement, operating room (OR) checklist, and the procedure itself). Intra-procedural adverse events recorded were oxygen saturation (SaO2) <90% and <95%, arterial hypotension (systolic blood pressure <90 mmHg), vasoactive drug administration, significant bradycardia (heart rate <50 beats/minute or chronotropic drug requirement), unplanned supraglottic device (SGD) insertion or endotracheal intubation, and need for advanced cardiac life support (ACLS). Post-procedural data recorded consisted of SaO2 <90% and <95% events, hypotension, nausea and/or vomiting episodes or administration of antiemetic drugs; time to post anaesthetic care unit (PACU) discharge (minutes), unplanned post procedural hospitalisation, and need for ACLS.
The main data sources comprised electronic medical records, including procedure subsections with attached anaesthetic and post-anaesthetic charts (PDF documents). Anaesthetic charts are automatically generated during procedures by electronic capture of real-time vital-sign measurements from multiparameter monitors and graphically displayed. Drug administration and comments on adverse events are manually entered by anaesthesiologists. Upon case completion, a PDF document is generated and available for review in the procedure subsection. During the post-procedure period at the PACU, vital-sign measurements and drug administration are entered by PACU nurses, and a post-anaesthetic chart is generated upon discharge.
All patients scheduled for colonoscopy received bowel cleansing preparation (Pico-Sulphate - Picoprep) before the procedure according to standardised hospital protocols. Vital signs were monitored (pulse oximeter, noninvasive blood pressure cuff, and five lead electrocardiography) once inside the operating room and after intravenous cannulation. Supplementary oxygen was delivered through a nasal cannula, and propofol was administered using TCI systems at titrated doses aiming for a deep level of sedation . The TCI system included a Perfusor® Space pump (B. Braun Melsungen AG, Melsungen, Germany) controlled by a microprocessor system. After introducing patient data (sex, age, weight, and height), the Schnider or Marsh pharmacokinetic models and initial target concentrations were selected by the anaesthesiologist. During the procedures, automatic adjustments to the infusion rate were generated by perfusor pumps based on the predicted plasma or tissue levels of propofol and the target concentrations set by the operator, guided by clinical judgement.
All patients were transferred to the PACU immediately after completion of the procedure and were monitored for at least 30 minutes before discharge.
Our primary outcome was the incidence of adverse cardiovascular and/or respiratory events among patients undergoing GIEP under propofol TCI sedation. The secondary outcome was the existence of associations between hypotension and oxygen desaturation events and potentially related variables.
Sample size calculation
Regarding the primary aim, we calculated a sample size of 823 patients based on an estimated incidence of 14% for SaO2 <95% events and aiming for a CI with a hemi amplitude of 2.5% . An incidence of 5% was estimated for SaO2 <90% events in this population, predicting approximately 41 of these events among 823 patients. This allowed to build a multivariate model through the logistic regression technique including approximately four potentially related variables.
Descriptive analyses were performed using the mean ± SD for continuous variables and numbers (proportions) for categorical variables. Qualitative variables derived from each group were compared using the χ2 test or Fisher’s exact test in cases involving low expected counts. Student’s t-test was used to analyse normally distributed quantitative data, while the nonparametric Wilcoxon rank-sum test was used to analyse non-normally distributed quantitative data. All statistical analyses were conducted using STATA.13 (StataCorp, College Station, TX).
We conducted two multivariate logistic regression analyses to investigate which variables would be associated with the development of SaO2 <90% and hypotension events during propofol TCI sedation for GIEPs. For both models, variables were included based on significant associations (P<0.1) with the event in the bivariate analysis. The final multivariate model for SaO2 <90% included age, ASA physical status classification score, obesity, EGD procedure, propofol doses, and fentanyl administration; for hypotension events the model included age, ASA physical status classification score, history of arterial hypertension, obesity, colonoscopic procedures, propofol doses, and fentanyl administration.
Finally, we performed linear regression to examine whether propofol doses would decrease when administered together with fentanyl, adjusting for potentially confounding factors such as age, ASA classification status, and body mass index (BMI).