IV meds (photo from 123RF)

This is the fourth article in a series exploring the impact of pulse oximetry alarm thresholds in hospitalized patients.

In the first article, “Improving the Safety of Post-Surgical Care,” I introduced the concept that, although the current approach to physiologic threshold monitoring (triggering an alarm when oxygen saturation falls below 90%) works well in the OR, it is unreliable on post-surgical floors.

In the second post, “Pulse Oximetry False Alarms on Post-Surgical Floors,” I explored in more depth why the threshold for triggering a pulse oximetry alarm should vary depending on the site of care (OR vs post-surgical floor). The key to appreciating why this is the case is understanding that the clinical conditions that threaten oxygenation on post-surgical floors are different from the type of sudden, life-threatening airway compromise that occur in ORs. Those conditions often have an insidious onset and comprise sepsis, aspiration, congestive heart failure, pulmonary embolus, and two different types of opioid-associated respiratory depression.

In the third post, “Detecting Deadly Post-Surgical Respiratory Dysfunction,” I reviewed the pattern of respiratory compromise that characterizes the conditions not related to opioid use. In this post, I will discuss the respiratory risk of opioids in post-surgical settings.


Type II – CO2 narcosis

The second pattern of respiratory dysfunction (Type II) is called carbon dioxide (CO2) narcosis. Adverse events associated with this pattern are considerable—an estimated 20,000 events per year occur in the USA that includes accidental death and severe anoxic brain injury. As illustrated in the chart below, it is not unusual to trigger a deteriorating, self-propagating process where both opioids and a rising carbon dioxide (PaCO2) contribute to an unstable central depression of patients’ ventilatory drives. The vicious cycle begins with a rise in carbon dioxide (PaCO2) due to neuro-inhibition of the brainstem by opioids.

Type II Respiratory Dysfunction (CO2 Narcosis)

Instead of leveling off, this depressive state continues to advance unchecked either from opioid overdose or unanticipated patient sensitivity or susceptibility. This ultimately leads to carbon dioxide intoxication with PaCO2 reaching levels throughout the body so extraordinarily high that the patient’s own oxygen (PaO2) becomes severely diluted. This leads to inevitable respiratory arrest.

Contrary to common belief, this progressive hypoventilation is not always coupled with a progressively slowing respiratory rate. In fact, the respiratory rate can remain adequate even though progressively shallower breaths are being inhaled.

Catching CO2 Narcosis early would mean detecting it at or close to an elevated carbon dioxide (PaCO2) value of 70 mmHg, essentially twice its normal. If patients are breathing room air during this process, progressive accumulation of carbon dioxide trapped in patients’ blood and lungs begins to crowd out oxygen normally occupying the lung space, limiting its availability. When this starts, astute clinicians will be able to detect a steady downward drifting of oxygen saturation (SPO2) values over minutes to hours if continuous pulse oximetry monitoring is being used. All the while these patients look as if they’re blissfully asleep.


The key to early detection

Anticipating this oxygen dilution process being responsible for these downward saturation drifts is the key to detection and successful rescue. Clinicians should then go immediately to the bedside and try to arouse these patients. They most often look to be remarkably content while soundly sleeping, but actually are severely poisoned with all that’s standing between them and a respiratory arrest being attempts to wake them.

Supplemental oxygen can again be harmful when ordered prophylactically, but for a different reason than that discussed with the Type I pattern. Downward drift of oxygen saturation (SPO2) can only be reliably detected in early CO2 narcosis (PaCO2=70 mmHg) if supplemental oxygen is delivered at no more than a fraction of inspired oxygen (FIO2) of 27% to patients with otherwise normal lungs. (This has been purposefully worked out through multiple simulations using the same mathematical models relied upon by blood gas laboratories.)

One difficulty nurses face is that supplemental oxygen is often ordered given through nasal cannula in flow rates of liters per minute, not FIO2. While rules of thumb have been derived that convert FIO2 values to flow rates, they are often unreliable. This is especially true when patients are given opioids because of two key variables that determine the actual FIO2 that patients inhale into their lungs at any given flow rate:

  • Minute ventilation
  • The percent of O2 actually entrained with each inspiratory effort.

These variables can often change significantly due to the ventilatory depression caused by opioids. For this reason, I recommend either limiting supplemental oxygen flow rates to no more than 1 liter/minute by nasal cannula. If higher flows are ordered, request a respiratory therapist to provide precise FIO2 delivery not to exceed 27%. Unfortunately, many otherwise excellent clinicians have never had the opportunity to learn through mathematical model simulation how vulnerable SP02 drift detection can become using even relatively low flow rates of O2 through nasal cannula.

So to summarize, reasonably early detection of Type II patterns can be achieved using a 90% SPO2 threshold alarm, but only if prophylactic supplemental oxygen is limited to an FIO2 of no more than 27%. (Note: This does not pertain to therapeutic oxygen delivery ordered for patients with known lung disease.)


Type III – Repetitive reductions in airflow

This last pattern of respiratory dysfunction (Type III) is also opioid-related. It was discovered (actually rediscovered) at the early turn of this century by an anesthesiologist, Ann Lofsky, MD.

In 2002, in a special article written for the Anesthesia Patient Safety Foundation Newsletter, Lofsky described how patients with obstructive sleep apnea (OSA) are more prone to airway obstruction from drugs like opioids that reduce the contractibility of the pharyngeal musculature. This promotes airway collapse and obstruction that is not always amenable to routine reversal with the opioid antagonist, naloxone. Death can follow quickly if the patient is not being observed directly and physically rescued.

She cited a case published by SI Samuels and W Rabinov in 1986 (Anesth-Analg 1986;65:1222-4) where this exact sequence happened to a patient in an intra-operative setting, which fortunately led to a successful resuscitation. Life-threatening arousal failure and arousal arrest were reproduced in this patient while severe hypoxemia reached PaO2 levels of 30 mmHg in a follow-up formal sleep study. This gave rise to Lofsky’s speculation that had this intra-operative arrest occurred on the post-surgical floors, it would have likely ended much differently.

Today, direct patient observation by clinical staff on general care floors occurs routinely only 4% of the patient’s total time spent there. Without some form of continuous electronic monitoring, these patients are being put at risk.

The Type III pattern differs from our classic Type II CO2 narcosis pattern because of its association with true sleep breathing disorders, primarily OSA. It involves an arousal failure component that likewise occurs only during sleep as opposed to Type II CO2 narcosis which is associated with a diminished ventilatory drive occurring both while awake and asleep.

Cyclical airflow reductions and apneas associated with OSA are extremely common in postoperative populations receiving parenteral opioids on post-surgical floors. Generally, each repetitive apnea is associated with a brief, transient state of hypoxemia and a small, transient elevation in PaCO2 which contributes to the generation of a neurologically-induced arousal as illustrated in the figure below.

Type III Respiratory Dysfunction (Repetitive Reduction in Airflow)

Medications, like opioids, increase CO2 arousal thresholds. Even a slight additional arousal delay (for example from just one more PCA opioid administration) can permit critically low oxygen saturation (SPO2) levels to be reached.

Likewise, conditions that encourage ongoing cycling hypoxemia, such as chronically depleted oxygen reserves in both venous beds and lung from obesity, can have the same effect by accelerating desaturations to critically low levels even before normal PaCO2 arousal generating thresholds can be reached.

With both types of insults, arterial oxygen saturation in some patients will fall to the point where the brain no longer receives sufficient oxygen for a central arousal to occur. This is called the “Lights Out Saturation” (LOS). If resuscitation is not immediately provided, brain death follows within minutes.

This Type III pattern argues best for why all patients receiving parenteral opioids on post-surgical floors should be continuously monitored during sleep. Yet, as we have already discussed, this can’t work when using oximetry threshold alarms set at 90%.

Type I and II patterns usually take many minutes to hours to evolve before a death occurs, providing staff multiple opportunities to detect trouble and intervene even without continuous monitoring. But with Type III patterns a patient can die in less than 10 unobserved minutes, while sleeping, without any visible or audible warnings unless some capable continuous electronic monitor is being used. Patients not on continuous electronic monitoring are left unobserved by professionals for far longer periods and are, therefore, always at risk because of it.


Key takeaways:

  • Opioids can cause two different types of respiratory dysfunction (Type II and Type III).
  • Type II is CO2 narcosis—it is caused by neuro-inhibition of the brainstem by opioids.
  • Opioids promote airway collapse and obstruction, worsening the underlying problem of people with obstructive sleep apnea.
  • Both types of respiratory dysfunction can lead to death if not recognized and appropriately intervened.


Is there one pulse ox threshold solution for all three patterns?

The answer is yes. In the final post in this series, I will explain how it can be done.

Paul Curry, MD
Dr. Curry received his MD degree in 1971 from the University of Florida. He held many leadership and teaching positions while at St. Mary’s Medical Center in Long Beach, including liaison physician for Long Beach’s nationally recognized paramedic program, advanced trauma, pediatric, and cardiac life support instruction, and teaching responsibilities for rotating UCLA Emergency Medicine Residents. In the late nineties, Dr. Curry teamed with the University of Pennsylvania under a grant from Nellcor (division of Covidien) to study the problems surrounding early recognition and response to patients becoming unstable while on post surgical hospital floors. Over the next four years, Dr. Curry authored 10 nationally accepted abstracts, an additional paper on the subject, and went on to create at Hoag Memorial one of the finest Rapid Response Teams in the country. Most currently, he retired from his Anesthesiology practice in 2014 but continues to serve as honorary medical staff at Hoag while publishing and lecturing nationally. He consults with Lyntek Medical Technologies, working closely with its Founder, CEO, and close friend Dr. Lawrence Lynn, assisting with its PatientStormChaser technology created to provide unsurpassed clinical safety well into the future. Dr. Curry also sits on the Clinical Committee of the Society of Anesthesia and Sleep Medicine.


  1. I read this article and I was truly interested. I have had two major back surgeries and have chronic i back pain. Oxycontin has be en part of my life for years now. Night I have trouble breathing and wake situada panic. Could this be caused by the drug ¿


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.