Continuous data were presented as medians and interquartile ranges due to skew; binary primary and secondary outcomes were summarized by frequencies and percentages for each matched group. Some continuous variables were transformed consistent with published clinical standards (body mass index) or clinically meaningful categories that incorporate realities of clinical documentation accuracy (estimated blood loss, time from last dose to reversal of extubation) as described in Supplemental Digital Content 2 and 3 . Unadjusted differences between patients receiving sugammadex versus neostigmine were assessed using conditional logistic regression to account for the matching.
To assess the independent association between administration of sugammadex versus neostigmine reversal and the primary composite pulmonary complication, separate multivariable conditional logistic regression models were developed. Additional variables not used in matching were assessed for residual confounding using absolute standardized differences. Any covariate with a standardized difference > 0.10 was included in the multivariable analysis. In addition, surgical body region/invasiveness (16 distinct categorical variables) and NMBA (rocuronium alone, vecuronium alone, or both) were included. Adjusted odds ratios with 95% confidence intervals were reported for all models. Model discrimination and calibration were assessed using standard logistic regression methods, since current statistical software is unable to calculate diagnostic measures that account for the matched design.30 ,31 (link) All statistical analyses were performed using SAS® version 9.4 (SAS Institute), and hypothesis testing was two-sided.
Using an approximated formula and assuming a conservative estimate for the sample proportions and 95% confidence, to achieve a margin of error of ±1%, we would need a study sample size of approximately 9,600. For the defined study period, we expected to observe greater than 30,000 patients receiving sugammadex.
To assess the independent association between administration of sugammadex versus neostigmine reversal and the primary composite pulmonary complication, separate multivariable conditional logistic regression models were developed. Additional variables not used in matching were assessed for residual confounding using absolute standardized differences. Any covariate with a standardized difference > 0.10 was included in the multivariable analysis. In addition, surgical body region/invasiveness (16 distinct categorical variables) and NMBA (rocuronium alone, vecuronium alone, or both) were included. Adjusted odds ratios with 95% confidence intervals were reported for all models. Model discrimination and calibration were assessed using standard logistic regression methods, since current statistical software is unable to calculate diagnostic measures that account for the matched design.30 ,31 (link) All statistical analyses were performed using SAS® version 9.4 (SAS Institute), and hypothesis testing was two-sided.
Using an approximated formula and assuming a conservative estimate for the sample proportions and 95% confidence, to achieve a margin of error of ±1%, we would need a study sample size of approximately 9,600. For the defined study period, we expected to observe greater than 30,000 patients receiving sugammadex.
