Tox and Hound – Great! Naloxone worked! Now what?

by Jon Cole

In some ways, you never outgrow your training. Before my medical toxicology fellowship, I completed a residency in emergency medicine. It is my primary specialty and practice, and it’s by far the most common place I encounter most of the patients I see. As opioids have become the primary cause of poisoning deaths, more and more I am forced to ponder a question I struggled with as a resident: what do you do after naloxone works?

I present for you a patient that is likely to be, at this very moment, in any number of emergency departments all across the country:

A young man is found down by passers-by. He was noted to be obtunded and cyanotic. 911 was called, EMS was activated, and the patient was given naloxone. The patient awakens in the ambulance, and upon being placed in their room in the ED, clearly states he is leaving.

Immediately two questions arise:

1. How long should I observe this patient?
2. What do I do if he refuses to stay?

In pondering this clinical scenario above, and the questions it spurred, I will answer both questions in honor of Dr. Ed Panacek, with “Well, it depends . . .” (Dr. Panacek taught the EMBRS course at ACEP I attended several years ago and I learned a lot from him, including how to answer complex questions by starting with “Well, it depends . . . “)

Duration of Observation after Naloxone

Since the FDA approved naloxone in 1971, clinicians have wondered what to do after naloxone reversal.^​1,2​ Largely based upon expert opinion, a 4-6 hour observation period is typically recommended.^​3​ This opinion, however, isn’t completely arbitrary. As naloxone has a relatively short half-life (somewhere between 30 – 90 minutes), there has always been a concern that the half-life of the antidote (naloxone) will be shorter than the half-life of the poison (the opioid), raising concern for recurrent toxicity. Since we rarely know the identity or concentration of the opioid that was used, we can’t rely on pharmacokinetic data (for many of the fentanyl derivatives currently being sold as heroin we don’t even have pharmacokinetic data). Because of this, the safest thing to do is to base observation time upon when we think the naloxone is gone so that there is no “masking” of opioid poisoning by any lingering naloxone. For most medications, 5 half-lives are usually enough to ensure there is no active drug left. Given the 30 – 90 minute half-life of naloxone, a 4 – 6 hour observation period has traditionally been recommended (figure courtesy of &Howard, adapted from Katzung^​4​).

Clearly this conservative estimate creates consternation, both for patients and clinicians. We all have cared for a heroin overdose who sat in a room for 4-6 hours who never displayed a hint of a nap, let alone recurrent opioid poisoning, and as the patient becomes more irritated and the ED waiting room builds, the room the patient is occupying becomes more precious by the minute. As a consequence, multiple studies have attempted to answer the question of “how long do I really have to watch patients after naloxone?” Willman and colleagues, in 2017, published a systematic review to address this question.^​5​ They attempted to answer three questions in the paper, two of which directly address this topic.

1. Does a heroin user revived with naloxone always require ambulance transport? What are the medical risks to a heroin user who refuses ambulance transport after naloxone?
2. If the heroin user is treated in the ED, how long must the patient be observed prior to discharge?

For the first question, the authors found 8 pertinent studies (all retrospective), and admittedly it gets murky here because not all “heroin” overdoses, even in the 1990s, were actually heroin. The authors were able to identify 5,443 patients treated with naloxone and released without being transported to the ED after a presumed opioid overdose and found 4 deaths potentially attributed to rebound opioid toxicity. This means the number needed to treat (with a forced ambulance ride) to save one life would have been 1,361.

I’m not going to weigh in on whether or not 1,360 “unnecessary” ambulance rides are worth saving a life – that’s a question for people far more important than me. What I am going to suggest (as Willman and colleagues do themselves) is that this number is likely a lot lower than 1,360.

First, all these studies were retrospective. Cases could have been missed. Even in prospective studies where mortality is a primary outcome, determining death can be difficult. Investigators in the U.S. attempted to use county medical examiner databases to supplement hospital medical records^​6–9​, but clearly a patient could have refused transport and died in another county. Some Danish^​10​ and Scandinavian^{​11–13​} investigators attempted similar studies, which initially seem more reliable given the completeness of European medical records, and yet as anyone who has been to the E.U. can tell you, it is really easy to cross international borders in Europe. Patients still could have refused transport and died in another country.

Second, using mortality as an outcome can be troublesome. As a study outcome, mortality is appealing. It’s easy to understand and about as definitive as you can get. And yet, it misses some really critical outcomes. I imagine you’ve all seen patients like this:

• Opioid overdose resulting in anoxic brain injury and permanent institutionalization
• Opioid overdose resulting in ARDS and 3-week ICU stay
• Opioid overdose resulting in rhabdomyolysis, acute kidney injury, and long-term hemodialysis

All of these patients, by definition, fall into the “good” outcomes category when a dichotomous study uses only mortality as an outcome. As such, many toxicologists remained skeptical of some of this literature demonstrating “safe” EMS transport refusal after naloxone.

For Willman and colleagues’ second question (how long after ED naloxone?), they identified one prospective study^​14​ and four retrospective studies^{​1,13,15,16​} that could potentially answer the question. The four retrospective studies have all the same limitations; however, the prospective study raises some interesting questions. It’s a derivation study of 573 patients by Christenson et al. conducted at St. Paul’s Hospital in Vancouver, BC, that had a sensitivity of 99% (and specificity of 40%) to predict safe discharge if the patients met all of the following criteria at one hour after naloxone administration (whether naloxone was given prehospital or in the ED):

1. Can mobilize as usual
2. Room air pulse ox >92%
3. Respiratory rate >10 breaths/min and <20 breaths/min
4. Temperature >35.0℃ and <37.5℃
5. Heart rate >50 beats/min and <100 beats/min
6. Glasgow Coma Scale = 15

These 6 criteria became known as “St. Paul’s Early Discharge Rule” (though I agree with those who prefer to call this 6-point list a tool and not a rule). Willman and colleagues concluded that, depending on the patient and the clinical picture, an ambulatory patient with a GCS of 15 and normal vital signs are low risk and may be allowed to leave the ED after one hour. There were two major problems with using St. Paul’s rule at this time – first, it had never been externally validated^​17​, and second, there’s the opioid elephant in the room that heroin isn’t “just heroin” anymore.

While other opioids have been occasionally substituted for heroin in the past (alpha-methyl fentanyl sold as “China White” is a classic example; fentanyl sold as “Tango and Cash” is another)^​18​, these heroin substitution outbreaks were typically self-limited in nature, and heroin was, for the most part, actually heroin (plus other sundried excipients . . . ) Then, in the early part of this decade, we started seeing other opioids routinely sold as heroin. The first of these that I recall was in 2013 when acetyl fentanyl was sold as heroin in Rhode Island^​19​, but shortly thereafter case after case emerged of opioids being substituted for one another.^{​20–23​} Heroin is now commonly adulterated by or completely replaced with fentanyl and fentanyl analogues. Adulterated heroin is now so common that heroin deaths are actually falling, commensurate with the rise in fentanyl-related deaths.^​24​ Knowing the best way to handle an actual heroin overdose from 2-3 decades ago in 2019 is a little like knowing the best VCR/DVD combo player of 2019 – it’s just not relevant anymore.

Enter the HOUR study (Hospital Observation Upon Reversal), a validation study of St. Paul’s Early Discharge Rule published this year.^​25​ Clemency and colleagues performed this prospective observational validation study from 2016-2017 at a busy, urban safety-net hospital with some minor modifications to St. Paul’s Rule – they increased the tolerable pulse ox threshold to >95% on room air and increased the bradycardia threshold to 60 beats/min. Below is an infographic I found useful to help remember the six parameters, courtesy of Dr. Mark Ramzy (@MRamzyDO)

In addition to asking providers to collect St. Paul’s criteria at 1 hour, they also collected a data point on clinical judgment; namely, they asked the provider (e.g., attending, resident or advanced practice provider) if the patient, based on their clinical judgment, was safe for discharge or not. Importantly, this data point was collected after St. Paul’s Rule was applied. They found St. Paul’s Rule performed well, but not as well as it had in the derivation study (a common phenomenon in validation studies), and that the rule was not superior to clinical judgment. Another major difference from the derivation study – the vast majority (>85%) of patients received intranasal (IN) naloxone (at a mean dose of 3.1 mg). There are some other notable differences too, however, this post is not meant to be a complete journal club review of the HOUR study – for a great review check out The Skeptics’ Guide to Emergency Medicine episode on this paper or the visual abstract below courtesy of the SGEM and Dr. Kirsty Challen (@KirstyChallen).

Much has been written about the HOUR study, and for the record, this is a great study that adds significantly to the body of literature on this topic. This is the first large, prospective study involving intranasal naloxone in an era of prevalent illicitly manufactured fentanyl, and yet there are some substantial limitations. First, the opioid exposures were not confirmed. We don’t know which opioid overdoses these data apply to. Epidemiologically, it’s probably fentanyl, but there’s no way to know for sure. Second, over 85% of the patients in this study were reversed with IN naloxone. IN naloxone has a substantially different pharmacokinetic profile than IM or IV, as seen in Figure 1 in this open-access article from McDonald, et al.^​26​ Although IN naloxone has a similar time to peak concentration as IM, at the doses used in the HOUR study IN naloxone results in serum concentrations that remain substantially higher over time than those that result from typical 0.4 mg IM or IV dose.

This calls into question the validity of the HOUR rule in patients who receive a smaller IM or IV naloxone dose – i.e. does the sustained duration of IN naloxone provide additional protection from “retoxicification” in the first few hours after it is given?^​27​ While it’s quite true that pharmacokinetics and pharmacodynamics are very different concepts^​28​, data in the HOUR study are insufficient to know if this is indeed the case. Third, in addition to a lack of confirmation of the specific opioids, we are not provided with the route the patient used their opioids. I believe this is a substantial limitation, largely because patients that either injected or snorted their opioids likely have absorbed all the offending drug, whereas an oral overdose (particularly since opioids themselves delay drug absorption, including other opioids) could continue to absorb drug long after the naloxone has been metabolized and cause delayed poisoning. The HOUR data suggest this indeed might be a concern. Of the 10 cases missed by a combination of a negative HOUR score plus provider judgment where an adverse events still occurred, four were oral overdoses, including two of the cases with the most severe adverse events (one case requiring BiPAP for pulmonary edema and another case requiring additional naloxone – both were purported to be caused by methadone). Fourth, as large as this study is, it is still a single-center study of a site that clearly sees a lot of opioid overdoses. We do not know if these data are generalizable to another clinical environment that might have a different prevalence of disease or a different familiarity with opioid overdose. And last, as promising as these data are, even if universally true, are they good enough? For a vulnerable population where death is a definitively possible outcome, is a sensitivity of 84% truly enough to change practice from our usual observation period (whatever that may be in your part of the world)?

Here’s how I use the HOUR study. I use the 6 criteria as a cross-check against my own clinical judgment. My hope is it augments my gestalt, like a checklist for airline pilots. It reminds me to re-check some key data points before I discharge a patient. There’s rationale to use it in this fashion too – clinical judgment in the HOUR study was assessed after the 6-point tool was collected, so it’s likely performance of clinical judgment in the study was affected by the collection of the data itself (a concept called incorporation bias). Are there opioid overdose patients I discharge after one hour of observation? Yes. But there’s more to it than what’s in the HOUR study. What route was the exposure? I generally think IV and IN exposures are more predictable. How reliable is the history, and do they have co-ingestions that are concerning? If there is another respiratory depressant, like alcohol or a benzodiazepine on board, I’m more cautious. What is their underlying health? Do they have chronic diseases like COPD, sleep apnea, or heart failure that raise the risk of respiratory depression? How risky is their discharge? Do they truly understand the risks of recurrent toxicity? Are they going home with someone who can call for help, and/or administer take-home naloxone? What time of day is it? Discharges in the middle of the night make me nervous. Caring companions are far more likely to be asleep at that hour, making home observation less ideal, and diseases like sleep apnea can compound the risk of an early discharge in the middle of the night. Do I have take-home naloxone to give them? And last, is there something to be gained by keeping them longer? Are there other resources we can connect them to, such as addiction counseling or outpatient treatment? Sometimes the extra few hours in the ED gets the patient a lot more than just a period of observation. While the HOUR study is promising and helps me quantify risk in some patients, I’m not yet ready to make it my standard approach for all patients without taking into account a lot of other factors.

What if they refuse to stay?

So the second big question after “how long?” is “what if they won’t stay?” As much as “It depends. . .” is the answer to the first big question of “how long?”, it’s really the answer to the second question of “what if they say they’re leaving?” I’d like to think we all strive to practice evidence-based medicine. EBM is a complex concept that involves not just the most relevant scientific evidence and your clinical judgment, but also the patient’s values and preferences. As demonstrated nicely by the figure below, EBM is the intersection of all 3 of these concepts. We don’t practice medicine in a vacuum.

So what happens then when you are treating a patient whose values and preferences come into direct conflict with your clinical judgment and your assessment of the relevant scientific evidence? Standard de-escalation techniques, reflective listening, and a genuine display of concern can go a long way. I’m surprised how often a patient will agree to stay if I just say to them “For your safety, I can’t in good conscience let you leave.” Some of this can also be prevented by judicious dosing of naloxone. If naloxone is titrated IV using small doses with the purpose of restoring only normal respirations and not to the point of a “full wake-up,” this conflict can be avoided. It also, by the way, might make discharge more expedient – if only small doses of naloxone are used to restore normal respiratory drive and the patient later wakes on their own, the clinical scenario is more like alcohol than opioids – i.e., once the patient fully awakens on their own you can be certain the naloxone is no longer having an effect and discharge may be safer (nod to @lapizity for the opioid/alcohol simile).

But what about those patients who are literally walking out the door as you are disagreeing with them? Here’s where the “it depends. . .” part really comes into play. In the U.S., it varies substantially from state to state whether or not such patients are holdable. I’ve held a medical license in two states (Minnesota and Wisconsin), and the two are radically different from one another. This is an oversimplification, but in Minnesota, there is a statute for emergency holds that allows a peace or health officer (e.g. physicians) to hold a patient for “emergency care and treatment” if they meet one criteria from both column A and column B in the chart below. It reads, “if the examiner is of opinion, for stated reasons, that the person is (something from A) and (something from B) if not immediately detained…”

…then the patient can be held for up to 72 hours. This statute, under the right circumstances, could be used to hold a patient who wants to leave despite the physician’s recommendation to stay. In contrast, in Wisconsin, no such statute exists that would allow an emergency physician to place a similar hold. Instead, this authority rests with law enforcement.^​30​ When I worked in Wisconsin if a patient left the ED and I thought they needed to be held for their safety, we called the local police who would find the patient, make their own assessment, and then (sometimes) return them to the hospital. This post is by no means meant to capture all the nuances of placing a patient on an emergency hold, but rather it is meant to demonstrate how different the process can be depending on where you practice; and from speaking with colleagues all around the country, rules around involuntarily holding a patient vary significantly.

This is a long-winded way of saying that wherever you work, knowing the institutional policies as well as local law on matters like this is imperative. There are many times where medicine and law don’t see eye to eye, and this is one of them. The kinetics of opioids and naloxone behave how they are going to behave, regardless of state statutes. Knowing the local laws and how to navigate them before the conflict is imminent is your best strategy to ensure your patient stays safe and you stay out of legal jeopardy. In states where emergency holds are impractical or impossible, it is critical to have other strategies to ensure patient safety, including take-home naloxone or engaging with friends and family to arrange observation at home. As one of our most experienced hospital attorneys once told me, “the best legal defense is the good practice of medicine.”

While knowing the laws and processes for an emergency hold (if one is possible where you practice) is critical, fortunately most opioid overdose cases don’t end in legal conflict. Usually, genuine concern, some catering to the patient’s preferences and values (where we safely can), good clinical judgment, and application of the latest, best available evidence allow a safe disposition, whether that’s an HOUR, or a few more.