IONM indicates intraoperative nerve monitoring; LOS, loss of IONM signal; RLN, recurrent laryngeal nerve.
IONM indicates intraoperative nerve monitoring; RLN, recurrent laryngeal nerve.
The IONM RLN injury rate was varied as a percentage of no IONM (visual identification only) RLN injury rate. IONM indicates intraoperative nerve monitoring; RLN, recurrent laryngeal nerve.
The RLN injury rates were varied for both IONM arms as the rate in the no IONM arm (visual identification only) was varied. IONM, intraoperative nerve monitoring; RLN, recurrent laryngeal nerve.
eFigure. Detailed decision tree used to model IONM use in total thyroidectomy.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rocke DJ, Goldstein DP, de Almeida JR. A Cost-Utility Analysis of Recurrent Laryngeal Nerve Monitoring in the Setting of Total Thyroidectomy. JAMA Otolaryngol Head Neck Surg. 2016;142(12):1199–1205. doi:10.1001/jamaoto.2016.2860
Copyright 2016 American Medical Association. All Rights Reserved.
Is intraoperative nerve monitoring (IONM) of the recurrent laryngeal nerve cost-effective for total thyroidectomy?
In this cost-utility analysis, visual identification is more cost-effective than IONM in nearly all cases, including most high-risk cases, for total thyroidectomy. In selected high-risk cases, IONM becomes cost-effective if a surgeon can use it to decrease nerve injury rate by at least 50.4% over visual identification alone.
Visual identification alone is more cost-effective than IONM for most surgeons in the setting of total thyroidectomy.
Intraoperative nerve monitoring (IONM) of the recurrent laryngeal nerve (RLN) is used as a tool to decrease the rate of nerve injury, although study findings are divergent on IONM efficacy. The cost-effectiveness of this approach to total thyroidectomy has not been well studied.
To determine whether IONM is a cost-effective intervention in the setting of total thyroidectomy.
Design and Setting
This study creates a decision-tree model of total thyroidectomy to analyze, from a societal perspective, the cost-effectiveness of universal IONM (ie, use in every case) vs selective IONM (ie, high-risk cases including reoperative cases, substernal or toxic goiters, and cases with known cancer) vs no IONM (visual identification only). Parameters for the model were derived from review of the literature, and deterministic and probabilistic analyses were performed to test the model’s robustness. All analyses were performed from the model; there were no human participants.
Modeled total thyroidectomy with and without IONM of the RLN.
Main Outcomes and Measures
Cost-effectiveness of universal IONM vs selective IONM vs visual identification only of the RLN.
Visual identification of the RLN led to a cost savings of $179.40 and $683.20 per patient, and an improvement of 0.001 and 0.004 quality-adjusted life-years, over selective IONM and universal IONM, respectively. Visual identification was the most cost-effective approach, despite variations in costs and utilities in both deterministic and probabilistic sensitivity analyses. In a 1-way sensitivity analysis, decreasing the probability of RLN injury with IONM made selective IONM more cost-effective. When the rate of RLN injury for visual identification was kept constant (at 3.86%), selective IONM became the most cost-effective approach when its RLN injury rate dropped below 1.9%. As the rate of RLN injury with IONM dropped below 50.4% of the visual identification RLN injury rate, selective IONM became the most cost-effective approach.
Conclusions and Relevance
Visual identification of the RLN is more cost-effective than any use of IONM. If a clinician can, with use of IONM, decrease the rate of RLN injury by 50.4% or more compared with visual identification, selective use of IONM in high-risk cases is most cost-effective.
One of the most clinically significant complications of thyroid surgery is an injury to the recurrent laryngeal nerve (RLN). Paralysis rates from surgery on the thyroid gland have fallen substantially over the past few decades, likely owing to routine identification of the RLN.1 Even with routine identification of the RLN, however, nerve injury rates are reported as high as 8%, and RLN injury is one of the more common reasons for malpractice claims against surgeons.2
Intraoperative nerve monitoring (IONM) has been championed as a tool to reduce the risk of RLN injury. Findings of studies looking at the question of efficacy of IONM have been divergent,3-8 but 2 meta-analyses have shown no significant difference in injury rates between IONM and visual identification alone.1,9
The implications of routine nerve monitoring in a resource-constrained health care system are not small. One study examined the implications of IONM in terms of operating room time,10 but to our knowledge, there is no study that examines the cost-effectiveness of this procedure.
The present study looks at the cost-effectiveness of IONM from a societal perspective using a decision-tree model of total thyroidectomy surgery. We compare universal IONM (ie, use in all total thyroidectomy cases), selective IONM (ie, use in high-risk total thyroidectomies only), and no IONM (visual identification of the RLN only). This issue has relevance especially because payers and patients seek to reward high-quality care that is cost-conscious.
We used a decision-tree model to examine the question of which approach to total thyroidectomy is most cost-effective. Figure 1 shows a simplified version of the decision tree; for a detailed version, see the eFigure in the Supplement. A 10-year time frame was used, and a societal perspective was taken. The model has 3 main arms: (1) universal IONM (use in all total thyroidectomy cases), (2) selective IONM (use in high-risk total thyroidectomies only), and (3) no IONM (visual identification of the RLN only). We defined high-risk cases as reoperative cases, substernal or toxic goiters, and cases with known cancer.1 The model attempts to take into account all potential complications and outcomes that would result from a decision to undergo a total thyroidectomy.
In the universal IONM arm, the model incorporates the possibilities of postoperative hematoma, death from hematoma, loss of IONM signal (LOS) intraoperatively (including true- and false-negative LOS), false-positive IONM signal, surgery termination due to LOS, temporary RLN injury, permanent RLN injury, and subsequent treatment for any permanent nerve injury. The no IONM arm looks at similar outcomes, but there is no intraoperative decision based on IONM signal. The selective IONM arm incorporates the possibilities of both the universal IONM and no IONM arms.
Probability parameter estimates in the base case were derived from a review of the literature.1,11-15 When multiple estimates were available, we generally used estimates from studies with the largest sample size to ensure reliability of estimates. Nerve injury rates used for the present analysis were obtained from a previous meta-analysis.1 The probabilities of other complications, including rates of hematoma, death from hematoma, surgery termination after LOS, medialization of paralyzed vocal folds, and hypothyroidism after a hemithyroidectomy, were determined from a variety of sources (Table).1,11-22 Where no data were available to guide parameters, estimates were made based on expert opinion.
Costs were derived from the literature and from Medicare reimbursement data for relevant Current Procedural Terminology (CPT) codes (Table).10,15-19 Costs included the cost of hemithyroidectomy, total thyroidectomy, completion thyroidectomy, tracheotomy, medialization, hematoma evacuation, hypocalcemia, hypothyroidism, and disposable IONM equipment. Consistent with the recommendations for cost-effectiveness analysis by the panel of Cost-Effectiveness in Health and Medicine,23 the capital cost of purchasing an intraoperative nerve monitor was not included. Patient loss of income and the cost of follow-up were included. All cost parameters were inflated to present-day value using inflation rates from the US Department of Labor, Bureau of Labor Statistics Consumer Price Index24 for medical care. Costs were varied in the sensitivity analysis, and all costs incurred after year 1 of the model were discounted at 3% per annum in the base-case analysis. All costs were measured in US dollars.
Utilities and durations spent in a given health state for the base-case model were also derived from a review of the literature where possible and estimated from expert opinion otherwise (Table).11,14,20-22 These utilities were derived for all of the treatment outcomes and complications. For each arm, utilities were multiplied by duration spent in any given health state to determine quality-adjusted life-years (QALY) for each treatment arm.
Cost-effectiveness was calculated for each of the treatment arms. The incremental cost was calculated from the difference in cost among the 3 arms, and the incremental effectiveness was calculated in the same manner. The incremental cost-effectiveness ratio was calculated by dividing the incremental cost by the incremental effectiveness. Acceptability curves were generated for the base-case analysis.
Deterministic 1-way sensitivity analysis was performed for several relevant variables. These variables included the costs of completion thyroidectomy and IONM; the probabilities of unilateral RLN injury and false LOS when using IONM; and the utilities of hypoparathyroidism, hypothyroidism, unilateral vocal cord paralysis, and bilateral vocal cord paralysis. These parameters and those remaining were varied in a probabilistic sensitivity analysis and analyzed using second-order Monte Carlo simulation with 1000 simulations. Distributions were created, in general, using 90% of the base-case value as the low value and 110% as the high value. Some variables, including rates of RLN injury in the different arms of the model, were varied based on expert opinion. Sampling of these parameters was performed from a triangular distribution because there was not sufficient published data to determine the type of continuous distribution for these variables.25 Two-way sensitivity analysis was also performed varying the rate of RLN injury with universal IONM and the rate of RLN injury with no IONM. All analyses were performed using TreeAge Pro software (TreeAge Software Inc, 2015).
The model included several assumptions. We assumed that the rates of complications—other than injury to one or both RLNs—were equivalent regardless of study arm because we could find no literature to support any differences in complication rates. We could also find no literature to support any difference in recurrence or reoperative rate whether IONM was used or not, so we assumed no difference here as well. There were no data on what percentage of patients with aborted total thyroidectomy due to LOS eventually went on to have a completion lobectomy. We assumed that the reasons for total thyroidectomy, as opposed to thyroid lobectomy, were compelling. Thus, we assumed that 60% of patients would undergo completion thyroidectomy after aborting mid-procedure due to LOS when the RLN injury was permanent. We assumed that 99% would have a completion thyroidectomy if the injury was temporary. We also assumed that the utility of patients with a paralyzed vocal fold who underwent medialization laryngoplasty was 1, that all laryngoplasties would occur at 1 year after injury, that a temporary RLN injury would take 9 months to fully resolve, and that all patients with bilateral vocal cord injuries underwent tracheotomy.
In the base-case scenario, the cost-effectiveness in both the universal IONM arm and the selective IONM arm was lower than in the no IONM arm. The cost in the no IONM arm over a 10-year horizon was $13 417.58. The costs in the universal IONM and selective IONM arms were $14 100.77 and $13 596.99, respectively. The no IONM arm realized a cost savings of $179.40 per patient over the selective IONM arm and $683.20 per patient over the universal IONM arm. The quality-adjusted life expectancy in the no IONM arm was 9.889 QALY over a 10-year time horizon. The universal IONM and selective IONM arms yielded 9.888 QALY and 9.885 QALY, respectively. The no IONM arm thus showed incremental effectiveness of 0.001 and 0.004 QALY compared with the selective IONM and universal IONM arms, respectively. Assuming a societal willingness to pay $50 000 per QALY, there is an 80.2% likelihood that the no IONM arm is the most cost-effective of the 3 RLN monitoring techniques (Figure 2).
One-way sensitivity analysis was performed on several cost parameters, including the costs of hemithyroidectomy, total thyroidectomy, completion thyroidectomy, and disposable IONM components. Variation in these cost parameters did not change the cost-effectiveness outcomes for the 3 arms. The no IONM treatment arm proved to be the most cost-effective, regardless of cost variability.
Probability parameters were also varied, including the rate of unilateral RLN injury and the rate of false LOS. In this analysis, the probabilities of RLN injury in the high-risk selective IONM group and in the low-risk no IONM group were expressed as percentages of the RLN injury rate in the selective IONM group and no IONM group, respectively. The assumption was that the rate of RLN injury with selective IONM would not be independent of the RLN injury rate with no IONM. In other words, using selective IONM would offer some proportional benefit or decrement over no IONM. As the rate of RLN injury decreased with the use of selective IONM (while the no IONM RLN injury rate was held stable), using selective IONM became more cost-effective. At a threshold of 1.9% RLN injury rate with selective IONM, using IONM selectively was the most cost-effective strategy. Decreasing the rate of RLN injury in the selective IONM arm, calculated as a percentage of the injury rate in the no IONM arm, made selective IONM become more cost-effective. Threshold analysis showed that when the RLN injury rate with selective IONM was less than 50.4% of the rate of injury with no IONM, selective IONM was the most cost-effective approach (Figure 3). Varying the rate of false LOS for selective IONM did not change the cost-effectiveness of no IONM.
Two-way sensitivity analysis was also performed varying the rates of RLN injury for both IONM groups and the no IONM group. The findings are graphically depicted in Figure 4. Again, the no IONM arm proved to be the most cost-effective approach in most cases.
Utility parameters were also varied. The utilities of bilateral vocal cord paralysis, hemithyroidectomy, and total thyroidectomy were varied. Variability in these parameters did not affect the cost-effectiveness calculation. The no IONM arm remained the most cost-effective arm regardless of the utility of these parameters.
Probabilistic sensitivity analysis was also performed varying the model parameters. Assuming a societal willingness to pay $50 000/QALY, over a 10-year time horizon 80.5% of simulations showed no IONM to be the most cost-effective strategy. Selective IONM was the most cost-effective in 16.9% of simulations.
In the setting of total thyroidectomy, visual identification of the RLN (no IONM) is more cost-effective than using IONM. The cost savings, while relatively small per patient, are more substantial on a societal level. In 2011, there were 72 344 total thyroidectomies, which notably was nearly a 60% increase from 2006.26 If IONM were used in a third of these cases, a no IONM strategy would save our health care system more than $16 million per year.
The cost savings from the no IONM strategy is a result of avoiding the per-case costs of IONM and what is often an unnecessary delay in completion of the total thyroidectomy. This delay also accounts for the small increase in QALYs with the no IONM strategy. LOS after completing the first hemithyroidectomy will often result in aborting the remainder of the procedure. Since there is a small but real rate of false LOS with IONM, and since the risk of injury to the contralateral RLN is small, the vast majority of patients would likely benefit from not aborting the procedure. Delay in performing the completion procedure leads to a costly second procedure in most cases with additional loss of income for patients. This is not to suggest that the decision to abort the surgery after LOS is necessarily flawed. Our analysis was performed from a societal perspective, and the decisions of whether to use IONM and whether to abort after LOS are often made on an individual level. Certainly the consequences of bilateral cord paralysis are catastrophic, and in an individual patient, a surgeon may decide that even a low risk is unacceptable. However, from a societal perspective the incidence of this complication is too low to justify use of IONM in the base-case cost-effectiveness analysis.
The sensitivity analysis in this study suggests that unless a surgeon can decrease the likelihood of RLN injury by 50.4% with IONM, visual identification alone is the most cost-effective approach. This portion of the analysis is particularly important because an individual surgeon’s rate of RLN injury with visual identification and with IONM are not likely to be independent of one another. In other words, if there were an improvement to be had from IONM, it would be a proportional improvement from the rate of injury with visual identification. This finding also illustrates that the magnitude of improvement necessary to achieve cost-effectiveness is quite substantial, and this may be difficult to achieve in the context of low nerve paralysis rates with the visual identification strategy in general. Moreover, even if this substantial improvement is present, the most cost-effective approach is selective use of IONM in high-risk cases only.
This study certainly has limitations. We assumed that the risk of bilateral cord paralysis is the square of the rate of injury to a single nerve. This may not be the case if anatomical features that lead to injury to one nerve are present bilaterally or if certain nerves are more susceptible to injury and that susceptibility is a bilateral phenomenon. There are no data to suggest this that we are aware of. It is also possible that the utility of bilateral cord paralysis is even worse than we estimated, given the propensity of surgeons to abort the remainder of a thyroidectomy after LOS on the initial hemithyroidectomy. However, the sensitivity analysis examines variation in the rate of nerve injury and the utility of bilateral cord paralysis.
We also did not attempt to quantify the utility of IONM to the individual surgeon in terms of confidence in RLN identification or confidence in RLN function after completing a hemithyroidectomy. This is certainly a valid consideration although difficult to quantify. The widespread use of IONM, despite a lack of evidence for efficacy, suggests that there is some utility to surgeons outside of diminishing the RLN injury rate. Furthermore, most cost-effectiveness analyses rely on probability and utility estimates, and the validity of the findings of the present study are only as good as the estimates used in our model.
This study also does not examine the use of IONM in thyroid lobectomy. In 2011, there were nearly 58 000 lobectomies.26 However, the justification for use of IONM in these cases is likely to be even less persuasive because there is no potential benefit from avoidance of bilateral RLN injury. Thus, a visual identification strategy in these cases is likely to lead to even further cost savings.
Injury to the RLN is a clinically significant complication of thyroid surgery, but IONM at best offers a relatively small improvement in RLN injury rates. This study demonstrates that from a societal cost-effectiveness standpoint, the improvement in RLN injury rate needs to be quite substantial—a decrease of over 50%—before IONM makes sense, and then only in high-risk cases. While the cost savings per patient are not high, the cost savings from a United States health care standpoint are substantial, given the large number of total thyroidectomies per year.
Corresponding Author: Daniel J. Rocke, MD, JD, Department of Surgery, Division of Head and Neck Surgery & Communication Sciences, Duke University Medical Center, PO Box 3805, Durham, NC 27707 (firstname.lastname@example.org).
Accepted for Publication: July 31, 2016.
Published Online: October 13, 2016. doi:10.1001/jamaoto.2016.2860
Author Contributions: Drs Rocke and de Almeida had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Rocke, Goldstein.
Drafting of the manuscript: Rocke, de Almeida.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Rocke, de Almeida.
Administrative, technical, or material support: All authors.
Study supervision: Rocke, de Almeida.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Previous Presentation: This study was presented at the American Head & Neck Society Ninth International Conference on Head and Neck Cancer; July 16-20, 2016; Seattle, Washington.
Create a personal account or sign in to: