Why it matters
Most patients arrive in the PACU breathing supplemental oxygen, and most are monitored with a pulse oximeter. The oximeter is trusted as the monitor that will announce respiratory trouble. On supplemental oxygen it cannot do that job for hypoventilation: a patient can retain carbon dioxide for a long time while the saturation stays reassuring.
How often that happens has now been measured, and the answer is not the one this page assumed. A blinded multicenter trial recorded capnography and oximetry alongside standard monitoring in adult recovery rooms, on an average of 3.8 L/min of supplemental oxygen, and analyzed 172 patients chung-2020-capnography. Applying its most severe thresholds, apnea appeared in 113 patients and bradypnea in 53, while measured hypercapnia appeared in only 2 chung-2020-capnography. So the common finding on oxygen is not a rising carbon dioxide; it is periodic apnea and a slow respiratory rate, and 78% of patients triggered at least one severe notification during standard monitoring chung-2020-capnography. That is the point of this day stated from the other side: the events were there, and the monitor in use was not showing them.
The consequence of getting this wrong is not a missed alarm. It is an alarm that is structurally impossible on the monitor you are watching.
Mechanism
Pulse oximetry measures arterial hemoglobin saturation, which sits on the flat upper part of the oxyhemoglobin dissociation curve whenever alveolar oxygen is generous. Hypoventilation raises alveolar carbon dioxide and lowers alveolar oxygen. On room air, the fall in alveolar oxygen is enough to move saturation off the flat part of the curve, so the oximeter reads the change. With supplemental oxygen, the alveolar oxygen tension starts high enough that the same rise in carbon dioxide leaves saturation on the flat part, and the oximeter reads nothing.
That is the whole mechanism: oxygen does not treat hypoventilation, it hides its only oximetric sign. Carbon dioxide rises exactly as it would have; the monitor no longer reports it.
Evidence
Fu 2004 is a two-phase study in surgical patients, not volunteers fu-2004. Phase 1, in the operating room, was a prospective patient-controlled clinical trial of 45 patients fu-2004. Phase 2, in the PACU, was a prospective randomized clinical trial of 288 patients fu-2004.
In phase 1, hypoventilation was defined as a 50% reduction in minute ventilation fu-2004. Patients breathed room air (FiO2 0.21, n=25) for up to 5 min or until SpO2 fell below 90%, or oxygen at FiO2 0.25 (n=10) or 0.30 (n=10) for up to 10 min fu-2004. SpO2 declined only in the room-air group; there was no decline at all at FiO2 0.25 or 0.30 fu-2004.
In phase 2, PACU patients were randomized to room air (n=155) or supplemental oxygen (n=133), with SpO2 recorded every minute for up to 40 min fu-2004. Desaturation, defined as SpO2 below 90%, occurred in 9.0% of the room-air group and 2.3% of the supplemental-oxygen group, P=0.02, roughly fourfold higher on room air fu-2004.
What this does not show
Phase 2 found desaturation more common on room air: 9.0% versus 2.3% fu-2004. Read quickly, that says room air is dangerous and oxygen is protective. That reading inverts the study. Both groups were recovering from anesthesia and both hypoventilated; the room-air group desaturated because room air lets the oximeter see hypoventilation, and the oxygen group did not desaturate because oxygen removes the signal. The direction of the result is the point, not a contradiction of it.
The study does not show that room air is the right choice for PACU patients, and it does not show that supplemental oxygen harms them. It shows that on supplemental oxygen the pulse oximeter is not a ventilation monitor. Whether routine oxygen changes any patient outcome is a separate question this source does not answer. The ward observation that respiratory depression episodes were 2.7 times more frequent on supplemental oxygen while desaturations did not differ is the same mechanism seen with capnography, and it comes from the ward, not the PACU doufas-2023.
The oximeter reads wrong in a second way, and this day carries it as its boundary. In methemoglobinemia the pulse oximeter overestimates the true saturation in proportion to the methemoglobin fraction until about 35%, where it plateaus at 84 to 86% and does not fall further even as the true saturation continues to drop; that plateau was measured in anesthetized dogs, and the design vocabulary has no term for an animal study barker-1989. In a summary of 242 published episodes related to local anesthetics, a saturation of 90% or below with an arterial oxygen tension of 70 mmHg or more was present in 91.8%, the gap between the oximeter and co-oximetry ranged from minus 6.2% to 44.7%, a single spray of benzocaine was enough to induce it, coma was reported at methemoglobin fractions of 32.2% in children and 29.1% in adults, and rebound to 59.9% up to 18 h after methylene blue was described with mucosal benzocaine; the author’s conclusion is that benzocaine should no longer be used guay-2009. The treatment has a label: methylene blue 1 mg/kg intravenously over 5 to 30 min, repeated once after an hour if the fraction stays above 30% or symptoms persist; it is contraindicated in G6PD deficiency, where it risks hemolysis and may not work, it carries a boxed warning for serotonin syndrome with serotonergic drugs and opioids, and it can itself lower the oximeter reading fda-methylene-blue-label-2025.
At the bedside
Decide, for each patient on oxygen, what is actually monitoring ventilation. If the answer is the oximeter, nothing is. Look at the patient: respiratory rate, depth, pattern, and level of consciousness are the ventilation monitors the oximeter is standing in for. Where capnography is available, it measures the thing you care about.
Treat a normal saturation on oxygen as information about oxygenation only. Treat a falling saturation on oxygen as late, because the reserve that oxygen provides has already been spent.
When you consider removing supplemental oxygen, be clear about the reason: it restores the oximeter’s sensitivity to hypoventilation. It does not make the patient safer by itself, and this source does not test whether it does.
Sources
Every number above carries its ledger key. Each key below resolves to the source record.
[[fu-2004]] randomized trial 2004 n: Phase 1 n=45 (OR, patient-controlled: FiO2 0.21 n=25, 0.25 n=10, 0.30 n=10); phase 2 n=288 randomised in the PACU (room air 155, supplemental O2 133) paywalled era-limited
Fu ES, Downs JB, Schweiger JW, Miguel RV, Smith RA. Supplemental oxygen impairs detection of hypoventilation by pulse oximetry. Chest. 2004 Nov;126(5):1552-1558.
Two-phase study in surgical patients, not volunteers. Phase 1 (OR, n=45): hypoventilation defined as a 50% reduction in minute ventilation; SpO2 declined only in the room-air group, with no decline at all at FiO2 0.25 or 0.30. Phase 2 (PACU, n=288, randomised room air vs supplemental O2, SpO2 every minute for up to 40 min): desaturation below SpO2 90% in 9.0% on room air vs 2.3% on supplemental O2, P=0.02. The direction is the teaching trap: supplemental oxygen removes the oximeter's hypoventilation signal.
[[doufas-2023]] cohort 2023 n: 202 open
Doufas AG, et al. Incidence of postoperative opioid-induced respiratory depression episodes in patients on room air or supplemental oxygen: a post-hoc analysis of the PRODIGY trial. BMC Anesthesiol. 2023 Oct 4;23(1):332.
Surgical WARD patients (not PACU) with blinded continuous oximetry and capnography: 74 always on room air, 88 intermittent, 40 continuous supplemental oxygen. On supplemental oxygen vs room air: all OIRD episodes IRR 2.7 (95% CI 1.4-5.1), apnea IRR 2.8 (1.5-5.2), bradypnea IRR 3.0 (1.2-7.9); high PRODIGY score IRR 4.5 (2.2-9.6). 'Despite oxygen desaturation events not differing between SO and RA, SO may clinically promote OIRD.' Answers the Day 3 placeholder only by analogy (ward, not PACU).
[[chung-2020-capnography]] cohort 2020 n: 250 enrolled, 172 completed open
Chung F, Wong J, Mestek ML, Niebel KH, Lichtenthal P. Characterization of respiratory compromise and the potential clinical utility of capnography in the post-anesthesia care unit: a blinded observational trial. J Clin Monit Comput. 2020;34:541-551.
ANSWERS THE DAY 3 PLACEHOLDER, AND NOT IN THE DIRECTION THE PLACEHOLDER ASSUMES. Prospective multicentre blinded observational trial in adult PACU; capnography and oximetry recorded blinded alongside standard monitoring. Supplemental oxygen, VERBATIM: 'The average supplemental O2 rate in the PACU was 3.8+/-2.4 L/min, delivered to 81 patients by mask and 82 patients by nasal cannula; 82% of patients were transported out of the PACU with supplemental O2.' Level I thresholds: apnea >=10 s twice in a 15 min epoch; bradypnea <=6 bpm for more than 30 s; hypercapnia EtCO2 >=60 mmHg for more than 30 s. Incidence, VERBATIM: 'Within Level I notifcations, 2 patients had hypercapnia, 48 had hypocapnia, 113 had apnea, 6 had tachypnea, and 53 had bradypnea.' (sic, typo in the published text.) Also VERBATIM: '163 (95%) patients had a Level II notification and 135 (78%) had a Level I notification during standard monitoring.' READ THIS CAREFULLY BEFORE USING IT: on ~3.8 L/min of oxygen, capnography found apnoea in 113 of 172 and bradypnoea in 53 of 172, but measured hypercapnia in only 2 of 172. The day must not say 'hypercapnia is common in the PACU on oxygen'; this study says the opposite. What is common is apnoea and bradypnoea, which the oximeter also does not show.
[[guay-2009]] case series 2009 n: 242 published episodes paywalled
Guay J. Methemoglobinemia related to local anesthetics: a summary of 242 episodes. Anesth Analg. 2009 Mar;108(3):837-845.
Systematic summary of 242 published episodes of local anaesthetic-related methaemoglobinaemia: SpO2 <=90% with PaO2 >=70 mmHg discrepancy in 91.8%; SpO2 minus co-oximetry difference -6.2% to 44.7%; a single spray of benzocaine may induce methaemoglobinaemia; coma reported at MetHb 32.2% (children) and 29.1% (adults); rebound MetHb up to 59.9% up to 18 h after methylene blue with mucosal benzocaine; complications include hypoxic encephalopathy, MI and death; conclusion 'Benzocaine should no longer be used'. Prilocaine adult limit 5.0 mg/kg. PubMed pubtype: Meta-Analysis.
[[barker-1989]] cohort 1989 n: 5 anesthetized dogs paywalled
Barker SJ, Tremper KK, Hyatt J. Effects of methemoglobinemia on pulse oximetry and mixed venous oximetry. Anesthesiology. 1989 Jan;70(1):112-117.
ANIMAL study (design vocabulary has no term; 'cohort' is a placeholder - director to decide). MetHb induced up to 60% by intratracheal benzocaine on FiO2 1.0: SpO2 overestimated SaO2 proportionally until MetHb ~35%, then plateaued at 84-86% and did not fall further; with further desaturation SpO2 changed far less than SaO2 (slopes 0.16-0.32). This is the primary source for the '~85% plateau'.
[[fda-methylene-blue-label-2025]] advisory 2025 n: NA open
Methylene Blue Injection, USP 0.5% [prescribing information]. Amneal Pharmaceuticals LLC. Revised 11/2025. DailyMed set ID 67da2a69-8166-46c7-921c-d6ec47bfeac8.
FDA label: indicated for acquired methemoglobinemia (pediatric and adult); 1 mg/kg IV over 5-30 min, repeat 1 mg/kg after 1 h if MetHb remains >30% or symptoms persist; CONTRAINDICATED in G6PD deficiency (risk of hemolytic anemia) and 'may not be effective in patients with G6PD deficiency'; boxed warning for serotonin syndrome with serotonergic drugs and opioids; 'The presence of methylene blue in the blood may result in an underestimation of the oxygen saturation reading by pulse oximetry.'
Check yourself
Three items. Every option carries an explanation. Progress is not saved.