With Cleveland Clinic Foundation Review Board approval and waived consent (identifier 20–062, January 22, 2020, primary investigator Kurt Ruetzler, MD), we conducted a retrospective cohort analysis of patients having surgical procedures with general anesthesia who were given rocuronium or vecuronium at Cleveland Clinic, Main Campus between June 2016 and December 2019. We considered patients who were given either neostigmine or sugammadex for reversal of neuromuscular blockade at the end of surgery. Sugammadex and neostigmine were available in our institution throughout the entire study period. We excluded patients with incomplete intraoperative medication records or time stamps.
We collected demographic information (age, sex, weight, and height), relevant medical history (cardiac, renal, hepatic, and neurological), and perioperative data, including American Society of Anesthesiologists (ASA) physical status score, length of anesthesia, type of surgery, medications administered during and after surgery, vital signs, and diagnosis codes.
Important adverse events potentially associated with administered reversal medication were defined by clinical deterioration during or shortly after reversal medication administration that prompted treatment by the responsible anesthesia clinicians. Our primary outcome was a composite of bradycardia, cardiac arrest anaphylaxis, and bronchospasm occurring between administration of the reversal agent and departure from the operation room. Specifically, anesthesia records were electronically screened for the administration of any medication or intervention specified in Table 1 during this period. Neostigmine is routinely coadministered with intravenous (IV) atropine or IV glycopyrrolate to counteract the muscarinic effects, and we, therefore, only considered any additional dose of either drug after the initial dose was given. Because pediatric patients were included, we did not require any particular dose for drug interventions.
Administration of the second dose of IV glycopyrrolate and IV ephedrine was a priori assumed to be given for bradycardia. Administration of salbutamol was a priori assumed to be given due to bronchospasm. All other cases suspected of having a clinically important adverse event as defined in Table 1 were evaluated by an adjudication committee.
The primary adjudication committee included 4 anesthesiology attendees (S.C., K.M., S.K., and A.T.) and 1 cheif adjudicator (K.R.). Adjudicators were asked to evaluate cases suspected of having clinically meaningful adverse events potentially associated with whichever reversal medication was given. For each suspect case, administrative and medical information was accessible to the adjudicators, but they were blinded to the patient's diagnosis codes, administered reversal medication, and to other adjudicators' comments. Each adjudicator was responsible for: (1) understanding and accepting our definitions for clinically meaningful adverse events and the adjudication guidelines; (2) carefully reviewing each suspect case; and (3) documenting the assessment of the diagnosis for suspect cases. Potential diagnosis included bradycardia, cardiac arrest, anaphylaxis, bronchospasm, or unrelated/unknown.
Each suspect case was reviewed by at least 3 adjudicators. The chief adjudicator (K.R.) then reviewed each assessment. Cases with consistent assessments by all assigned adjudicators were considered diagnosed. Cases with inconsistent assessment were individually reviewed by the chief adjudicator (K.R.). A diagnosis was considered established if the assessment of the chief adjudicator matched assessments from at least 2 other adjudicators. When there was greater disagreement, the chief adjudicator had the option of consulting other clinicians if deemed necessary. Finally, the adjudication committee (4 adjudicators and the chief adjudicator) met in person and individually reviewed each unresolved case to reach final consensus decisions.
Our primary goal was to estimate the average treatment effect at the population level and then conduct noninferiority testing at the composite level. Some components of the composite are rare, such as cardiac arrest and anaphylaxis, and might be excluded with a propensity score matching method. We, therefore, used inverse probability of treatment weighting (IPTW) that allowed us to include all observations while simultaneously controlling observed potential confounding variables, such as age and sex. Specifically, we first estimated the probability of receiving sugammadex (ie, the propensity score) for each patient using logistic regression, with sugammadex as the outcome and all of the prespecified potential confounding variables listed in Table 2 as the explanatory variables.
We then calculated "stabilized" propensity score weights as follows: the weights for the sugammadex group were calculated as the proportion of patients who received sugammadex divided by the propensity score; the weights for the neostigmine group were calculated as the proportion of patients who received neostigmine divided by 1 – the propensity score.[12,13] To avoid extreme weights unduly influencing results, we truncated at the 1st and 99th weighting percentiles. Successful control for confounding was assessed by comparing the 2 groups on all the potential confounding variables used to construct the propensity score using absolute standardized difference (ASD) after weighting each observation. Any covariable with an ASD >0.10 would be entered into the models to reduce potential confounding. Histogram plots of estimated propensity scores were used to evaluate overlap (ie, common support) and similar distribution (ie, balance) across the groups.
IPTW was used to adjust for available confounding variables (Table 2) in the comparison of sugammadex and neostigmine patients on outcome variables. Sugammadex would be considered noninferior to neostigmine (or vice versa) if the OR for incurring the collapsed composite outcome was no more than a predefined delta of 1.2.
Noninferiority was assessed using a 1-tailed t test, with the log-transformed noninferiority delta as well as estimated log-OR and its standard error from a generalized estimating equation (GEE) to count within-subject correlation resulting from multiple surgeries per patient. If noninferiority was found in either direction, a superiority test would be performed using a 1-tailed t test. The significant level was 0.025 for noninferiority and superiority tests in each direction.
The GEE model included the inverse propensity score weighting and adjusted for any covariable still imbalanced after the weighting. A sensitivity analysis excluding pediatric and cardiac surgeries was also conducted.
Sample Size Considerations. We planned to use all available surgical cases for the study duration who met the inclusion/exclusion criteria for our study and did not conduct an a priori power analysis to guide sample size estimation. Based on a preliminary query from our database, the incidence of the composite outcome was 3%, among 16,480 cases in the sugammadex group and 73,273 in the neostigmine group. With an upper bound of noninferiority boundary of OR at 1.2, this sample would provide 98% power to show noninferiority at a 1-sided alpha level of 0.025 comparing patients given sugammadex versus neostigmine, assuming the odds were similar in each group. SAS statistical software version 9.4 was used for all analyses.
Anesth Analg. 2022;134(5):1043-1053. © 2022 International Anesthesia Research Society