The number of opioid overdose deaths in the United States has continued to increase for more than two decades (1) with modeling studies predicting up to 1.2 million additional fatalities over this decade (2, 3). Illicitly manufactured synthetic opioids (“synthetics”) such as fentanyl have been linked to >90% of the more than 80,000 opioid overdose deaths reported for the 12 months ending in September 2023 (1). Counterfeit pills (produced to resemble medications including methylphenidate, oxycodone, and alprazolam) often containing lethal quantities of synthetics have flooded the United States over the past 5 years (4, 5), and the ready availability of these pills is now the face of what is commonly referred to as the ‘4^th wave’ of the opioid epidemic (6). The Drug Enforcement Administration estimated 78.4 million counterfeit pills were seized in 2023, with 70% containing a potentially lethal dose of synthetic opioid (7). It is likely the number of counterfeit pills seized by law enforcement agencies represents only a fraction of the total number smuggled into the United States. Both the sheer quantities and ready availability of these counterfeit pills on the ‘gray market’ often puts unwitting users at a very high risk of a fatal opioid overdose.

As the initial treatment of opioid overdose shifted from the emergency department (8) to first responders (e.g., police, fire department, friends and family of overdose victims), intranasal (IN) naloxone has become the standard rescue agent in a community setting (9). First approved by the FDA in November 2015, IN naloxone (4 mg) can be administered with little or no training, is absorbed as rapidly as an intramuscular (IM) injection (10) and eliminates the potential for needlestick injury. While multiple factors (such as the type and quantity of opioid, route of administration, presence of other drugs, and interval between overdose and intervention) ultimately determine a patient’s prognosis (11), both clinical and preclinical evidence indicate that higher doses of naloxone are needed to reverse a synthetic opioid overdose than are typically used by first responders. Multiple clinical studies have reported high doses of parenteral naloxone (in some cases followed by a naloxone infusion) are required for rescue, with recommendations of up to 12–15 mg if a synthetic like fentanyl is involved (12–14). Moreover, preclinical studies examining respiratory depression in the absence of the multiple factors that complicate interpretation of a clinical overdose have demonstrated that up to ten-fold higher doses of naloxone are needed to reverse respiratory depression produced by fentanyl compared to an opium-based alkaloid like morphine (15) despite comparable affinities at μ-opioid receptors (16). More recent studies (17) have provided key molecular insights which may contribute to this apparent paradox.

Mechanistic modeling studies present an unbiased, alternative approach to estimate the effectiveness of μ-opioid receptor antagonists in a first-responder setting. Model-based approaches have been successfully applied in drug development and can help address information gaps when it is challenging to conduct clinical trials. A translational mechanistic model recently published by Mann et al. (18) was developed to predict the extent of respiratory depression and incidence of cardiac arrest triggered by a synthetic opioid overdose in the absence of antagonist treatment and following administration of IM naloxone. This model was recently implemented to evaluate the efficacy of IN naloxone given as single or multiple doses (19).

In the present study, we leverage and expand this translational model to compare the efficacy of IN nalmefene to IN naloxone, the latter generally considered the “gold standard” for first responders to reverse an opioid overdose. Nalmefene is a more potent μ-opioid receptor antagonist than naloxone and was recently approved for the treatment of opioid overdose (induced by natural or synthetic opioids) in the United States in the form of an IN formulation delivering 3 mg nalmefene hydrochloride (HCl) equivalent to 2.7 mg nalmefene free base.

There is a compelling body of evidence, both preclinical (15–17) and clinical (12–14, 31), that higher doses of naloxone are needed to reverse a synthetic opioid overdose. Consistent with this evidence, the State Unintentional Drug Overdose Reporting System (SUDORS) database indicates that some form of naloxone had been administered in more than 20% of opioid-related overdose deaths in 2022; in some jurisdictions, more than 35% of decedents had received naloxone (32). Nonetheless, there is resistance to using either higher doses of naloxone or more potent and rapid onset agents like nalmefene (33–35). Much of this resistance is driven by the potential for more severe withdrawal in patients with a pre-existing opioid use disorder and justified by reports that “standard doses of naloxone” needed to reverse an overdose have not risen significantly (36–38) despite the demonstrable rise in synthetic opioids in the illicit drug supply (39). However, conclusions based on anecdotal observations, case studies, and chart reviews suffer from the inherent heterogeneity of an opioid overdose, variability in reporting outcomes, and study quality. For example, multiple reports used to support the position that naloxone doses have not increased attempted to standardize IN, IV, and IM naloxone dosing as “naloxone IV equivalents” based on bioavailability (36–38). While the attempt to standardize naloxone dosing based on plasma exposure is understandable, it does not account for differences in early plasma concentrations across routes of administration (18, 19, 40), which is essential for the successful reversal of a synthetic opioid overdose (11, 18, 19). These reports may also introduce a significant reporting bias, with reported EMS rescue rates of ≥99% (38, 41) which underestimate opioid overdose deaths reported in the SUDORS database. A translational model of opioid overdose offers an unbiased approach to estimating the effectiveness of naloxone in the face of a potentially lethal opioid overdose.

Among the most compelling evidence that standard IM doses of naloxone (2 mg/2 mL) favored by many first responders would result in a significant loss of life following an IV synthetic opioid overdose is the translational model developed by Mann et al. (18). These authors evaluated the effectiveness of IM naloxone in reversing hypoxia-induced cardiac arrest produced by fentanyl and carfentanil. Fentanyl doses of 1.63 mg and 2.97 mg were modeled because these doses result in plasma concentrations similar to the mean and one standard deviation above the mean plasma concentrations, respectively, measured post-mortem in a study of approximately 500 fatal fentanyl overdoses (18). In the Mann model, an IV bolus of 1.63 mg fentanyl produced cardiac arrest in approximately 52% of 2000 virtual chronic opioid users in the absence of antagonist treatment. Following IM administration of naloxone (2 mg/2 mL), the incidence of cardiac arrest was reduced to 30%, corresponding to a rescue rate of 42%. The Mann model predicts that IM naloxone will be even less effective at reducing the incidence of cardiac arrest following administration of either 2.97 mg fentanyl (rescue rate of 30%) or the ultrapotent synthetic opioid, carfentanil (rescue rates of 18% to 38% depending on the carfentanil dose: 0.022 mg or 0.012 mg, respectively) ( Table 3 ).

Because of the widespread use of IN naloxone in a community setting (9), we implemented and expanded the Mann model to compare the effectiveness of IN nalmefene to standard doses of IN naloxone HCl (4 mg), the latter generally considered the “gold standard” for first responders (22). The model predicts large and clinically significant reductions in the incidence of cardiac arrest following rescue with IN nalmefene compared to IN naloxone under each of four conditions (i.e., medium and high doses of fentanyl and carfentanil) and for each population investigated (chronic opioid users and opioid naïve individuals) ( Tables 3 , 4 ). Following a simulated IV overdose with 1.63 mg fentanyl producing cardiac arrest in approximately 52% of patients in the absence of antagonist treatment, the incidence of cardiac arrest following IN nalmefene was 2.2% (rescue rate of >95%) compared to 19.2% following IN naloxone (rescue rate of 63%). Large differences in the effectiveness of IN nalmefene compared to IN naloxone were also evident in overdoses simulating a higher fentanyl dose as well as both the medium and high doses of carfentanil.

The “medium” and “high” doses of fentanyl used in the simulations were based on data from approximately 500 cases of fatal fentanyl overdoses, the great majority with a history of opioid use and/or chronic pain (18). Tolerance to the respiratory effects of opioids is known to develop following repeated use (42), and the incidence of cardiac arrest in this population would be predicted to be lower than those in an opioid naïve population. Implementing the model in opioid naïve individuals confirmed this prediction, with a higher incidence of cardiac arrest following both fentanyl and carfentanil administration compared to simulations in an opioid-tolerant population. Nonetheless, the differences in effectiveness between nalmefene and naloxone were as robust in this population as in simulated chronic opioid users. This is critical because the increased availability of synthetic opioids in counterfeit pills (7) has fundamentally changed the demographics of overdose, with increased numbers of opioid naïve individuals, especially adolescents (43) exposed to potentially lethal doses of synthetic opioids.

Because our simulations demonstrated large reductions in the incidence of cardiac arrest after a single dose of IN nalmefene compared to a single dose of IN naloxone, we examined the rates of cardiac arrest following administration of multiple doses of naloxone. Simultaneous dosing of two, three and four doses of naloxone (total amount of 8, 12, and 16 mg) produced dose-related reductions in the incidence of cardiac arrest. However, at both medium and high doses of fentanyl or carfentanil, four doses of IN naloxone (16 mg) administered simultaneously were needed to reduce the incidence of cardiac arrest to values approaching a single dose of nalmefene ( Tables 3 , 4 ). The simulation results for IN naloxone were consistent with recent data by Strauss et al. (19) who used the same translational model as in the present work despite different data used for their IN naloxone pharmacokinetic model. In their simulations, Strauss et al. evaluated the use of either ‘rapid’ or ‘standard’ dosing regimens of IN naloxone and showed that four doses of naloxone given at 2.5-minute intervals would not produce a “meaningfully lower” incidence in cardiac arrest compared to a single IN naloxone dose. They also showed that the simultaneous delivery of two doses of IN naloxone produced additional reductions in the incidence of cardiac arrest compared to one dose of IN naloxone, which drove our decision to evaluate simultaneous (and not sequential) doses of naloxone for comparison with IN nalmefene.

In the Mann model, μ-opioid receptor antagonists were administered one minute after ventilation was decreased to 40% of baseline to mimic a delay between recognizing respiratory depression and administering the reversal agent. In practice, recognizing the signs of an opioid overdose requires training, and there is very often a longer delay in administering the reversal agent. Therefore, with the assumption that an opioid overdose would be immediately recognized and treated, our model predictions likely overestimate the effectiveness of both μ-opioid receptor antagonists in a field setting. Strauss et al. (19) reported that even a brief (3–10 min) delay in administration of IN naloxone could result in a dramatic loss in the effectiveness of IN naloxone. Also, while fentanyl remains the predominant synthetic found in DEA seizures (11), the illicit drug supply varies over time and by region. Thus, the quantities of fentanyl used in the Mann model may not reflect either the actual quantities or composition of opioids (including other opioids such as heroin) in a real-world scenario.

Other limitations of this work are inherent to the translational model itself. Mann et al. used an extensive set of data to develop, calibrate and validate their model, including in-vitro binding experiments, pharmacokinetic data, clinical studies of opioid effects on ventilation in chronic opioid users and opioid naïve subjects, and animal studies with severe hypoxia-induced cardiac arrest which could not ethically be conducted in humans (18). As pointed out by the authors, although some datapoints in their validation datasets had high variability, the ratio of mean values between model predictions and observations had a median of 0.95 (interquartile range, 0.90–1.01) across validation datasets, demonstrating a very good performance of the model. One model limitation is the lack of robust pharmacokinetic model for IV carfentanil. Mann et al. (18) had to adjust the pharmacokinetic model for IV fentanyl to predict carfentanil pharmacokinetics and match the carfentanil half-life of 45 minutes reported in the literature. Another limitation is the use of the same pharmacodynamic parameters for all synthetic opioids to describe their effects on ventilatory drives based on receptor occupancy, ignoring possible differences in terms of G protein activation and tolerance mechanisms. Nonetheless, Mann et al. were able to successfully predict the respiratory effects of fentanyl, remifentanil and alfentanil in clinical laboratory studies (18). With respect to the model expansion with IN naloxone and IN nalmefene data, robust pharmacokinetic models were developed from pharmacokinetic data collected in three studies (22, 23). Since the Mann model did not include binding parameters for nalmefene, those were derived from published data (24) using a scaling approach. Additional validation showed that the model was able to closely reproduce the effects of IN nalmefene and IN naloxone in reversing remifentanil-induced respiratory depression in healthy volunteers ( Figure 2 ).

Overall, the difference in the apparent effectiveness between IN nalmefene and IN naloxone is consistent with preclinical and clinical evidence including: a higher affinity of nalmefene at μ-opioid receptors (24, 44), higher plasma concentrations of nalmefene delivered at the critical early time points (e.g., 5 min) after dosing (10, 22, 23), and a more rapid onset of action as demonstrated in a clinical model of opioid-induced respiratory depression (23). The model may be further used to assess the effects of other synthetic opioids, including novel synthetic opioids (NSOs) such as the benzimidazoles (e.g., etonitazene and isotonitazene). Because this latter class of synthetic opioids was never approved for human use, additional clinical pharmacology studies (including pharmacokinetic and medical toxicology data [e.g. plasma concentrations from overdose victims]) would be required to adequately interrogate the model.

Nalmefene has a significantly longer plasma half-life (t_1/2 7.1–11 h) (22, 45) than naloxone (t_1/2 ~2 h) (10) as well as higher affinity at μ-opioid receptors (24) which have the potential risk to produce a more severe and longer precipitated withdrawal in individuals with an opioid use disorder. A double-blind, randomized clinical study (46) comparing IV doses of naloxone (2 mg) and nalmefene (1 or 2 mg) in 176 patients admitted to nine emergency departments with suspected narcotic overdose reported that both opioid antagonists produced rapid and robust reversals of respiratory depression in patients with a confirmed opioid overdose. Adverse events were noted in opioid-positive patients in all three treatment arms, but the overall difference among treatment arms was not significant (p > 0.27) and no significant overall time-treatment interactions (measured out to 240 min post-dosing) emerged; no statistical differences in withdrawal outcomes were seen between treatment groups (46). While it is difficult to extrapolate the results of a study conducted in an emergency department to the use of IN naloxone by first responders in a community setting, it is noteworthy that the overall incidence of adverse events following IV administration of these opioid antagonists (46) is lower than in a contemporary study using information provided by first responders and community-based organizations following rescue with 4 mg IN naloxone (47). Furthermore, the higher plasma naloxone concentrations which may be needed to reverse a synthetic opioid overdose also carries an increased risk of precipitated withdrawal. Ironically, multiple doses of IN naloxone administered using a standard regimen (i.e., a dose administered every 2–5 minutes if there is no response) increases the risk of sustained precipitated withdrawal without conferring significant therapeutic advantage compared to a single IN dose (19).

Whether the risks of underdosing with an opioid antagonist outweigh the risks associated with precipitated withdrawal remains a matter of debate. Although the majority of opioid overdoses are not fatal (11), every overdose is potentially lethal in the absence of intervention and the primary goal of opioid antagonist treatment is to prevent death. While withdrawal symptoms precipitated by an opioid antagonist can be unpleasant and distressing, they are medically manageable and rarely life threatening (8). As such, they should not limit the use of more effective reversal strategies, especially when the odds of a synthetic opioid overdose are high (18, 48, 49). On the other hand, there are concerns that precipitated withdrawal could pose acute risks to the patient and medical personnel and could trigger further opioid use to counter the effects of the opioid antagonist (33, 35, 41, 50, 51). Additionally, the knowledge and avoidance of precipitated withdrawal could reduce both use and acceptance of the reversal agent among patients with an opioid use disorder (33, 35, 41, 50). Nonetheless, a recent study (52) of 1152 patients entering treatment for opioid use disorder across 49 addiction treatment facilities reported that most respondents had either no preference (48.4%) or preferred a higher dose (35.9%) reversal agent if they were to experience another overdose. These findings indicate that while the concerns about precipitated withdrawal are valid, the majority of people who have had a lived experience accept the need for effective reversal agents for themselves and others within a community setting (52).

In conclusion, simulations using a validated translational model of opioid overdose demonstrate that a single dose of IN nalmefene produces clinically meaningful reductions in the incidence of cardiac arrest compared to naloxone tools (IN, IM) frequently used by first-responders across a variety of scenarios involving fentanyl and carfentanil. These findings are consistent with converging lines of pharmacological evidence that the rapid delivery of high concentrations of a potent reversal agent favors a successful rescue in the era of synthetic opioids.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by WCG IRB, 1019 39th Ave. SE, Suite 120, Puyallup, WA 98374 and IntegReview Institutional Review Board, 3815 S Capital of Texas Hwy, Austin, TX 78704. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

CL: Conceptualization, Supervision, Writing – original draft. PP: Supervision, Writing – review & editing. ND: Formal analysis, Writing – review & editing. IG-G: Formal analysis, Writing – review & editing. PS: Conceptualization, Supervision, Writing – original draft.