This single-blind, prospective, randomized controlled trial was approved by the Ethics Committee of Nanjing First Hospital (KY20201102-04) and registered in the China Clinical Trial Center (20/10/2021, ChiCTR2100052144) before patients enrollment. All patients signed an informed consent form.

Patients scheduled for propofol-sedated gastroscopy were enrolled in this study. The inclusion criteria were as follows: (1) ≥ 65 years old; (2) ASA classification of I - II; and (3) BMI < 30 kg/m2. The following were the exclusion criteria: (1) emergency endoscopy; (2) upper gastrointestinal obstruction and impaired gastric emptying; (3) coagulation disorders or upper respiratory bleeding, severe cardiovascular, pulmonary, liver, or kidney disease; (4) infections or tumours of the mouth, nose, or pharynx; (5) a history of difficult intubation, severe sleep apnea syndrome [respiratory/pause hypoventilation index (AHI) > 40]; (6) allergy to propofol, eggs, soy or egg whites; and (7) unaccompanied or unattended individuals.

The research equipment included the HFNC (AIRVO 2 provided by Fisher & Paykel, Panmure, New Zealand) and a regular nasal cannula.

Using computer randomization software (SPSS 24.0), each patient was randomly categorized into a nasal cannula oxygen group (Group C), 100% FiO2 in HFNC group (Group H100), and 50% FiO2 in HFNC group (Group H50) in a 1:1:1 ratio. The grouping results were secured in a sealed opaque envelope, which was opened only by the researchers before anesthesia induction. Blinded members of the research team conducted postoperative interviews with the patients. A member of the study team worked on registering and assigning patients, who were unaware of the randomization grouping, and the anesthesiologists involved in the study were trained and qualified in the use of HFNC.

The patients fasted for 8 h and ceased drinking for 2 h before the gastrointestinal endoscopy procedure and took no more than 50 ml of mucosal cleanser 30 min before the procedure. The specific preoperative preparation requirements were in accordance with the relevant guidelines for gastrointestinal endoscopic procedures.

Riphaus demonstrated that bolus and continuous propofol infusions provide similar good controllability of propofol sedation. Therefore, an IV push technique was adopted rather than an infusion pump [16]. Following arrival to the operating room, patients were routinely monitored with electrocardiography, respiration, blood pressure, and SpO2. Then, they were instructed to lie on their left side. In Group C, 8 L/min pure oxygen was used via a nasal cannula for calm breathing, while 30 L/min oxygen was used via HFNC in both Groups H100 and H50. The FiO2 values of Groups H100 and H50 were 100% and 50%, respectively. Each group underwent a minute of calm breathing with adequate denitrogenation. The initial load of propofol was slowly administered intravenously at a dose of 1.5–2.5 mg/kg. After the patients’ Ramsay sedation score reached 4, the flow was increased to 60 L/min in Groups H100 and H50 while keeping FiO2 constant. The gastroenterologist began the endoscopic procedure. If the consultation time was slightly longer or the stimulation of the operation was stronger, additional propofol 0.2–0.5 mg/kg was injected intravenously. An extra dose based on the patient’s signs, such as deepening respiration, increased heart rate, and even paradoxical response, was used and included in the record of the propofol dosage. No benzodiazepines or opioids were used. When treating the patient, maintaining a good level of anesthesia and sedation was essential to ensuring unconsciousness.

During the peri-examination period, hypoxia (SpO2 ≤ 92%) was treated via the following protocols: (1) stimulate the patient; (2) stop the medication; (3) increase the oxygen flow rate from 8 to 10 L/min in Group C; adjust the FiO2 to 100% in Group H50; (4) open the airway by lifting the jaw; (5) exit the gastroscope and parallel mask ventilation; and (6) perform endotracheal intubation for mechanical ventilation. During the examination, ephedrine 5–10 mg was administered intravenously if hypotension was found (systolic pressure below 80 mmHg for more than 1 min); atropine 0.25–0.50 mg was administered if the heart rate was less than 50 beats per minute. If necessary, the drug administration was repeated.

The incidence of hypoxia during the peri-examination period (SpO2 ≤ 92%) was the primary endpoint. Kelly AM et al. suggested that SpO2 ≤ 92% was a valid indicator for screening for systemic hypoxia, although hypoxia is defined as SpO2 < 90%. Based on previous research [17, 18], the inclusion of elderly patients and 50% FiO2 in Groups C and H50, our study defined SpO2 ≤ 92% as hypoxia [19].

The following were secondary endpoints: (1) hypoxia-related indicators: incidence of SpO2 < 90%; incidence of severe hypoxia (SpO2 ≤ 85%); incidence of prolonged hypoxia (SpO2 ≤ 92% for 1 min); SpO2 1 min after anesthesia induction; (2) interventions used to treat hypoxia events; (3) patient’s overall condition and airway assessment; (4) general information about gastroscopy(duration, dose of propofol and wake-up time); (5) adverse events related to HFNC ventilation at 5 and 30 min postoperatively(airway injury or any barotrauma including pneumothorax, subcutaneous emphysema, etc.); and (6) any sedation-related adverse events(paradoxical response, nausea/vomit, reflux, airway obstruction, or choking).

The primary endpoint of this study was the incidence of SpO2 ≤ 92% during the peri-examination period. According to the preliminary test results, the incidence of SpO2 ≤ 92% during the peri-examination period was 33%, 15%, and 20% in Groups C, H100, and H50, respectively. According to PASS 11.0 (NCSS, LLC., Kaysville, UT, USA), with an error of 0.05 (two-tailed) and a power of 0.80, 294 patients were needed. Due to attrition, a total of 369 patients was finally identified (123 patients in each group).

SPSS software (version 24.0; SPSS, Inc., Chicago, IL, USA) was used for the statistical analysis. We performed the Shapiro-Wilk test and Levene’s test on all continuous variables. According to the test results, the data are expressed as the means ± standard deviations or medians [interquartile ranges]. The normally distributed measurement data (dose of propofol) were compared between groups by one-way ANOVA, and the LSD method was further used for multiple comparison. Measurement data with skewed distribution (duration, wake-up time, median 1 min after induction of anesthesia SpO2) were compared between groups using the rank sum test, and Dunn’s method was used for multiple comparisons. We compared the incidence of hypoxia and sedation-related adverse events between groups with a chi-square test or Fisher’s exact test, and the P value of the multiple comparison was corrected by the Bonferroni method. Statistical significance was set at P < 0.05.