Effect of Intraoperative Neuromonitoring on Rates of Recurrent Laryngeal Nerve Injury during Thyroidectomy: A Doubly Robust Approach.
Kelvin Memeh1, Tanaz Vaghaiwalla2, Xavier Keutgen1, Peter Angelos1
1Surgery, University of Chicago, Chicago, Illinois, United States, 2Surgery, University of Tennessee Medical Center - Knoxville, Knoxville, Tennessee, United States
Introduction: The effect of intraoperative neuromonitoring (IONM) on rates of recurrent laryngeal nerve (RLN) injury during thyroidectomy remains controversial. Several studies on this topic apply the Ordinary Least-Square (OLS) methods to adjust for confounding in their effect estimation. However, OLS estimate might be inconsistent and biased in the setting of a rare outcome of interest such as RLN injury. The current study employs a doubly robust method to reduce the selection bias on measured covariates in the estimation of the effect of IONM on RLN injury rates.
Methods: A pooled thyroidectomy dataset was created by linking the 2016 - 2019 NSQIP General Participant User File (PUF) with the corresponding Targeted-Thyroidectomy PUF using the unique case numbers. The primary outcome was RLN injury rates, and the secondary outcomes were operating time and postoperative length of stay. A doubly robust estimator approach was used as follows: First, the propensity score(PS) of each patient having IONM during thyroidectomy was computed using preoperative covariates. Then an inverse probability weighting was performed using the PS, thus, creating a PS-weighted cohort - balanced on the preoperative covariates. Next, a regression-adjusted outcome model was fitted to estimate the effect of IONM on the outcomes of interest in the PS-weighted cohort using perioperative prognostic factors. Sensitivity analysis of the impact of unmeasured confounder(s) on the estimates was reported as E-value.
Results: 24,370 patients were evaluated, 17,514 met inclusion criteria , out of which 12,130 (70%) had IONM during thyroidectomy, with RLN injury occurring in 1,125 (6.4%) cases. In unadjusted analysis, cases with IONM were slightly younger (52 vs 52.1yrs, p =0.83), white (73.8 vs 68.1%, p < 0.001), non-Hispanic (95.7 vs 93.5%, p < 0.001), obese ( 50.7 vs 43%, p < 0.001). They are also more likely to use vessel-sealant device (79.9% vs 62.4%, p < 0.001), have benign disease (85.6 vs 82.6%, p < 0.001), have neck dissection(21.7 vs 16.4%, p < 0.001), and have prior neck surgery( 5.8 vs 4.5%, p = 0.01). In the propensity-adjusted analysis, the use of IONM conferred significant absolute reduction in the overall rate of RLN injury [-1.76% points (CI, -2.66 to - 0.85% points), RR 0.77 (CI, 0.69 to 0.86), p < 0.001] , and length of stay [- 2 hours (CI, - 3.86 to - 0.2 hours) p = 0.03)]. However, IONM use was associated with an increase in operating time [17.19mins (CI, 15.09 to 19.29 mins, p < 0.0001 )]. The number needed to treat to prevent one RLN injury was 57 cases.
Sensitivity analysis revealed that it would take an unmeasured confounder associated with both IONM and RLN injury by a RR of 1.92-fold each, to explain away the observed effect of IONM on RLN injury.
Conclusion: In a balanced cohort of patients undergoing thyroidectomy from multiple sites and by multiple surgeons participating in NSQIP, the use of IONM conferred significant protection against RLN injury.
Back to 2022 Abstracts