In this issue of the Korean Journal of Helicobacter and Upper Gastrointestinal Research, the authors take on precisely this problem [2]. Their prospective study of 98 consecutive ESD patients asks a simple but long-overdue question: can continuous capnography warn us of a hypoxic event before it happens, and if so, what do we do with that warning? The authors propose that these events can be detected early through the instability of end-tidal carbon dioxide (EtCO₂) instead of its overall levels; they further suggest that nasal high-flow oxygenation (NHFO) enables rescue without requiring changes in sedation or interfering with the procedure.

Thanks to its simplicity, low cost, and reliability, pulse oximetry remains the standard respiratory monitor in every endoscopy suite. However, it is a lagging indicator. A patient on supplemental oxygen can hypoventilate—or even stop breathing—for quite a while before peripheral capillary oxygen saturation (SpO₂) levels finally drop. By the time the alarm sounds, the airway has often collapsed, forcing the clinician to react to a crisis rather than prevent one. Most experienced endoscopists have had that moment: suctioning, jaw-thrusting, and bagging all at once because the first warning they got was a saturation reading in the 80s.

Capnography can be useful for closing that gap. By measuring exhaled CO₂ with every breath, it provides a real-time reflection of alveolar ventilation. The clinical impact was clearly demonstrated in Qadeer et al.’s [3] landmark trial, where adding capnography to standard monitoring during deep sedation for endoscopic retrograde cholangiopancreatography and endoscopic ultrasonography cut hypoxia rates nearly in half. A broad meta-analysis from Saunders et al. [4], which showed a consistent 50% reduction in hypoxia risk across various procedural sedations shows similar results. These data have been translated into practice guidelines; the ASGE (American Society for Gastrointestinal Endoscopy) and its multi-society sedation guideline recommend capnography for endoscopy performed under deep sedation [5].

Unlike general anesthesia, patients undergoing procedural sedation often belch, talk, or breathe through their mouths. These actions cause the capnography waveform to fluctuate erratically, triggering frequent nuisance alarms that can frustrate the operator and ultimately lead to decreased vigilance.

The shift toward variability-based metrics rests on two practical observations. First, waveform-based capnographic assessment has been shown to outperform pulse oximetry and visual inspection for early detection of respiratory depression [5], whereas trials relying on absolute EtCO₂ thresholds have produced inconsistent benefit [6]. Second, because artifact-laden data points were excluded a priori, the resulting standard deviation reflects variability of a relatively clean ventilatory signal in which sustained hypoventilation produces cumulative drift before desaturation.

To mitigate the various artifacts and noise encountered during endoscopy, the authors defined EtCO₂ instability based on its fluctuations rather than absolute values. Specifically, they utilized the standard deviation of EtCO₂ during the initial 30 minutes of the procedure as a key metric. Multivariate analysis revealed that this instability was independently associated with hypoxia (odds ratio 2.417, p=0.002). Receiver operating characteristic analysis further yielded a cutoff value of 4.33, with a specificity of 87.1%, a sensitivity of 64.3%, and an area under the curve of 0.806. This represents a respectable predictive performance from a parameter that is both simple and easy to compute. The robustness of this metric is supported by the study design: the 30-minute window is anchored to the induction and early maintenance phase of sedation, where respiratory depression most likely occurs. Oxygen delivery was standardized to 2 L/min via nasal cannula, and CO₂ insufflation was deliberately avoided to eliminate exogenous CO₂ as a confounder. Approximately 17.7% of EtCO₂ data points contaminated by belching or oral breathing were excluded prospectively using pre-specified waveform criteria mitigating the influence of transient movement artifacts.

The risk factors for EtCO₂ instability, including Mallampati scores and baseline SpO₂, are well-established components of pre-procedural assessment. This metric does not describe a new physiological phenomenon; rather, it quantifies existing clinical observations. This consistency suggests high potential for clinical adoption.

In the event of hypoxia, rescue measures must be initiated without delay. However, conventional interventions including administration of sedative antidotes, cessation of the procedure, and bag-mask ventilation inevitably disrupt the ongoing dissection and may preclude its successful completion. In this study, the authors demonstrated that NHFO therapy can restore SpO₂ without compromising procedural continuity or causing additional complications. These findings are consistent with the existing literature: Lin et al. [7] showed in a randomized trial that high-flow nasal cannula oxygenation significantly reduced both hypoxia and subclinical respiratory depression during propofol-sedated endoscopy, and the multicenter ODEPHI trial by Nay et al. [8] confirmed a similar benefit among patients at elevated risk of hypoxemia undergoing gastrointestinal endoscopy. The present study extends these observations to a setting involving substantially longer sedation durations, as required for ESD.

In this study, NHFO was applied only as a rescue therapy. The authors also discuss prophylactic use in high-risk patients, supported by the ODEPHI trial [8], though cost-effectiveness limits routine adoption. The hypoxia predictors reported here (high Mallampati score, oral breathing, low baseline SpO₂, obstructive lung disease) help identify candidates for upfront NHFO, while patients begun on standard oxygen can be monitored with the EtCO₂ instability cutoff and escalated if needed.

However, a few points remain to be refined before widespread adoption: the single-center cohort is relatively small, artifact exclusion criteria would benefit from further standardization, the relative definition of hypoxia differs from the absolute thresholds commonly used elsewhere, and whether the 30-minute observation window is optimal for longer procedures is still an open question. From a practical standpoint, capnography monitoring during endoscopic sedation is not currently covered under the Korean National Health Insurance system, and NHFO equipment is not universally available across community hospitals, posing additional barriers to routine implementation. Nonetheless, selective application of these tools in patients with identified risk factors—such as high Mallampati scores, habitual oral breathing, or low baseline SpO₂—may offer a meaningful safety benefit for high-risk procedures.

In conclusion, this study offers a clinically meaningful reframing of capnographic monitoring by shifting the focus from absolute EtCO₂ values to their variability, coupled with a practical rescue strategy that preserves procedural continuity. While further multi-center validation and resolution of reimbursement barriers are needed, these findings represent a promising step toward safer sedation practices in therapeutic endoscopy.