All adverse events, including those consequents to reversal medications, span a range of severities from trivial to life-threatening. We a priori restricted our analysis to side effects that required pharmacologic treatment and were thus both noticed by the responsible clinicians and deemed sufficiently severe to treat. Because some medications, including epinephrine, atropine, methylprednisolone, and diphenhydramine, are given for various reasons, we further required 3 independent anesthesiologists to adjudicate each event. Glycopyrrolate, ephedrine, and salbutamol were considered a priori diagnostic for bradycardia and bronchospasm. The events we report are thus all presumably clinically meaningful.
Using our definitions, the overall incidence of clinically important adverse events was 3.4% in patients given sugammadex and 3.0% in those given neostigmine. The difference of 0.4% is based on the higher incidence of bradycardia (2.4% vs 2.2%) and bronchospasm (0.98% vs 0.78%). Both side effects are transient and easy to treat with appropriate drugs. The number of patients who needed to be given neostigmine rather than sugammadex to avoid an episode of bradycardia was 500 and similarly was 500 for avoiding bronchospasm. The number-needed-to-harm for both complications combined was thus 250 patients. Few clinicians would consider a number-needed-to-harm of 250 patients for minor complications, much less 500 patients, a reasonable basis for selecting one drug or the other. In contrast, cardiac arrest and anaphylaxis are life-threatening. Cardiac arrest (0.01% with neostigmine versus 0.02% with sugammadex) and anaphylaxis (0.02% vs 0.004%) were both rare. There were no impressive differences in the incidence of either side effect with the 2 drugs; although with only 6 cases of anaphylaxis and 14 cases of cardiac arrests, our comparisons are underpowered despite including 89,000 surgical cases.
Our results contradict 2 recent meta-analyses of small randomized trials, both of which reported that sugammadex is associated with significantly fewer adverse events. One reported significantly fewer composite adverse events in patients given sugammadex than neostigmine (RR, 0.60; 95% CI, 0.49–0.74; 28 trials; n = 2298). The other reported adverse events in 11% (78 of 684) of patients given sugammadex versus 21% (133 of 630) in those given neostigmine (OR, 0.47; 95% CI, 0.34–0.66; 13 studies; n = 1384). Severe adverse events included anxiety, depression and marked fatigue, acute lung injury, severe hypoxemia, bradycardia, and postoperative abdominal pain and were observed in 0.15% of patients given sugammadex and 1.27% of patients given neostigmine (OR, 0.33; 95% CI, 0.09–1.19; P = .091). Cumulatively, both meta-analyses included <3800 patients spread across 41 studies, with sample sizes ranging from 22 to 1198 adults, and included various clinical settings and patient characteristics. Both meta-analyses are seriously underpowered for serious but rare side effects such as anaphylaxis and cardiac arrest, and by variability and inconsistencies in definitions of the outcomes across trials. It is also unclear how anxiety, depression, and fatigue are mechanistically related to reversal agent selection.
Among the side effects we evaluated, bradycardia was by far the most common consequence of sugammadex and neostigmine administration, accounting for about two-thirds of the adverse events. The anticholinesterase neostigmine is routinely coadministered with a muscarinic receptor antagonist such as glycopyrrolate or atropine to prevent bradycardia that would otherwise occur with elevated plasma acetylcholine concentrations. This clinical practice usually prevents severe bradycardia and did so in about 98% of our cases. The incidence of clinically meaningful bradycardia was 2.2% in patients given neostigmine, which is consistent with a previously reported incidence of 2.9% in a retrospective analysis of >73,000 cases.
Sugammadex also causes bradycardia, although the mechanism remains unclear. The incidence of bradycardia in patients given sugammadex was 2.4% and, therefore, nearly identical to patients given neostigmine (2.2%). Results of our analysis differ from a recent meta-analysis that reported bradycardia in 50 of 597 patients (8.3%) given neostigmine and 4 of 621 patients (0.6%) given sugammadex (RR, 0.16; 95% CI, 0.07–0.34; 11 studies; n = 1218). The difference between the meta-analysis and our results is based on differing definitions of bradycardia. We only considered bradycardia events requiring a pharmaceutical intervention instead of using a defined heart rate threshold. Why previous publications did not report any bradycardia after sugammadex administration remains unclear.
There are anecdotal reports of cardiac arrests after administration of sugammadex or neostigmine.[16,17] We identified 14 cardiac arrests, 2 among 16,480 patients given sugammadex (0.01%) versus 12 among 73,273 patients given neostigmine (0.02%), none of whom died during their initial hospitalization. Our analysis extends previous work by suggesting that the odds of cardiac arrest are similar in patients given sugammadex and neostigmine; although with so few events, we have a little power for identifying even substantial differences.
Anaphylaxis is a major clinical concern and is independently associated with morbidity and mortality. We identified 6 cases of anaphylaxis, 3 of which occurred in patients given sugammadex (0.02%) and 3 of which followed neostigmine (0.004%). None of these 6 cases progressed to cardiac arrest, and none died during their initial hospitalization.
Our results are consistent with a large Japanese multicenter retrospective analysis that included nearly 50,000 surgical patients in which the overall incidence of laboratory-confirmed sugammadex-related anaphylaxis was 0.02% (6 of 29,962 patients). Interestingly, the investigators did not identify any anaphylaxis attributable to neostigmine (0 of 3157 patients). However, the study was seriously underpowered for neostigmine-related anaphylaxis which occurred only once per 25,000 patients in our cohort.
A postmarketing surveillance study of the US Food and Drug Administration (FDA) reported an incidence of sugammadex-related anaphylaxis of 0.024%. Another Japanese single-institutional study reported an incidence of 0.039%. The reported incidence of anaphylaxis to sugammadex of 0.02% in our study is, therefore, consistent with these results.
The observed incidence of anaphylaxis in our patients was about 5 times more common with sugammadex (1 of 5000 cases) than with neostigmine (1 of 25,000 cases). However, there were so few anaphylaxis events that our estimates are imprecise, and the relative risk with each remains uncertain. Consequently, anaphylaxis is so rare after sugammadex or neostigmine that is should not be a basis for selecting one drug or the other.
Bronchospasm is a known side effect of neostigmine's antimuscarinic mechanism and is aggravated by excessive respiratory secretions. Bronchospasm is usually temporally limited and easily treated with salbutamol or inhaled epinephrine but can also be the initial sign of an allergic or anaphylactic reaction. The incidence of bronchospasm was comparable with each reversal agent, with an incidence of 0.98% (161 of 16,480) in patients given sugammadex and 0.78% (570 of 73,273) in patients given neostigmine. There are several anecdotal reports of bronchospasm after neostigmine, but none after sugammadex that has been given to many fewer patients.
Respiratory adverse events following administration of sugammadex and neostigmine were evaluated in a recent meta-analysis that included 6 randomized trials. The overall incidence of respiratory adverse events was 1.4% (5 of 345) in patients given sugammadex and 4.5% (15 of 335) in patients given neostigmine (OR, 0.36; 95% CI, 0.14–0.95; P = .0386). However, none of the underlying studies specifically reported bronchospasm.
A strength of our approach was electronically identifying potential cases, followed by individual review and assessment by an independent adjudication committee. Individual review allowed us to identify the most appropriate clinical diagnosis. For example, epinephrine was given to 140 patients, each of whom was individually evaluated by at least 3 independent adjudicators. Ultimately, the adjudication committee allocated these 140 cases to 13 cardiac arrests, 22 episodes of bronchospasm, 6 of anaphylactic events, and 97 cases of bradycardia, with 2 others being obvious misclassifications.
Retrospective analyses are subject to reporting bias and random errors, such as misclassification due to incorrect data input. To reduce misclassification errors and overreporting, a substantial fraction of the adverse events that were electronically identified was individually reviewed by the adjudication committee. Four hundred fifty-nine potential events, including all potential events of anaphylaxis and cardiac arrests, were individually reviewed and <3% (11 events) were considered due to incorrect data input. Because bronchospasm and bradycardia events were not reviewed, misclassification is more likely for these common outcomes. However, there is a little reason to expect misclassification to differ depending on the reversal agent.
Clinicians might have selectively given one reversal drug or the other. We, therefore, used the IPTW approach based on the known and measured confounder to balance baseline and demographic characteristics. Because we included pediatric and adult patients in our cohort, we did not consider reversal agent doses. However, both are usually given on an mg·kg–1 basis, and there was a relatively narrow range of doses.
Suspected causative agents of anaphylaxis are often gauged by assessing the time of administration of each drug in relation to symptom onset. This approach does not work well during anesthetic induction because various medications are typically given over short periods of time. However, reversal medications are often the only medication given during emergence from anesthesia. We restricted our analysis to the period between administration of neuromuscular blockade reversal agents and departure from the operating room. While late-onset side effects might initially appear in the postanesthesia care unit (PACU) or surgical ward, such long delays seem unlikely—especially for the more severe side effects.
Analyses of large datasets often yield statistically significant associations that are not clinically meaningful. We, therefore, a priori defined 1.2 as the smallest OR that we considered clinically meaningful in this setting. Using observed event rate of 3% in neostigmine patients with noninferiority bound of OR ≤1.2, the number-needed-to-harm from adjusted OR was 165 (102–383). A final limitation is that we did not consider minor potential side effects such as nausea and vomiting.
The incidence of severe adverse events—anaphylaxis and cardiac arrest—was so rare that even with >89,000 cases, we had little precision for estimating the incidence, much less power for a valid comparison between neostigmine and sugammadex. In contrast, there were many episodes of bradycardia and bronchospasm, both of which are transient side effects that are easy to treat and not especially serious. Our composite was, therefore, largely driven by bradycardia and bronchospasm, and from a practical perspective was essentially a composite of just these 2 side effects.
Neuromuscular blockade reversal with neostigmine was associated with slightly fewer adverse events than sugammadex (3.0% vs 3.4%). Neostigmine was found to be statistically superior to sugammadex, but the composite was mostly driven by minor, temporary, and relatively unimportant complications like bradycardia and bronchospasm. This difference of just 0.4% corresponds to a number-needed-to-harm of 250 patients (95% CI, 137–769). That is, 250 patients would need to be given neostigmine rather than sugammadex to avoid 1 episode of minor complications like bradycardia or bronchospasm. We, therefore, conclude that sugammadex and neostigmine are comparably safe.
AIDS = acquired immune deficiency syndrome; ASA = American Society of Anesthesiologists; ASD = absolute standardized difference; CI = confidence interval; FDA = food and drug administration; GEE = generalized estimating equation; GI = gastrointestinal; ICD, International Classification of Diseases; IPTW = inverse probability of treatment weighting; IV = intravenous; NA = not applicable; OR = odds ratio; PACU = postanesthesia care unit; RR = risk ratio; SD = standard deviation
Anesth Analg. 2022;134(5):1043-1053. © 2022 International Anesthesia Research Society