Protective Effects of Intraoperative Nerve Monitoring (IONM) for Recurrent Laryngeal Nerve Injury in Thyroidectomy: Meta-analysis

Recurrent laryngeal nerve (RLN) injury is an intractable complication of thyroidectomy. Intraoperative nerve monitoring (IONM) was designed to prevent RLN injury. However, the results concerning the protective effect of IONM on RLN injury are still controversial. We searched all eligible databases from 1980 to 2017. Meta-analysis was performed to evaluate the effect of IONM on RLN injury. Sensitivity analysis was also conducted to check the stability of our results. There were 34 studies included in the analysis. Overall analysis found a significant decrease in total injury (RR = 0.68, 95%CI: 0.55 to 0.83), transient injury (RR = 0.71, 95%CI: 0.57 to 0.88), and permanent injury (RD = −0.0026, 95%CI: −0.0039 to −0.0012) with IONM. Subgroup analysis found IONM played a preventive role of total, transient and permanent injury in patients undergoing bilateral thyroidectomy. IONM also reduced the incidence of total and transient injury for malignancy cases. Operations with IONM were associated with fewer total and transient RLN injuries in operation volume < 300 NARs per year and fewer total and permanent RLN injuries in operation volume ≥ 300 NARs per year. The application of IONM could reduce the RLN injury of thyroidectomy. Particularly, we recommend routine IONM for use in bilateral operations and malignancy operations.

Recurrent laryngeal nerve (RLN) injury is the most severe complication of thyroid surgery, leading to transient or permanent voice changes, which is one of the most common causes for medical litigation 1 . According to reported data, the incidence of transient RLN injury and permanent RLN injury in thyroidectomy are 2-11% and 0.6-1.6%, respectively [2][3][4] . In certain situations, the risk of RLN injury is relatively high, including thyroid surgery for malignant tumors, history of neck operation, and surgery of toxic goiter or substernal goiter 5,6 .
Routine visual identification of RLN is applied in thyroid surgery to prevent RLN injury, which serves as a conventional measure to reduce vocal cord paralysis. Dionigi et al. found that the causes of RLN injury included transection, clamping, ligatures, suction, traction, thermal injuries and physical compression 7 . Intraoperative nerve monitoring systems (IONM) use the electromyographic signal of vocal muscle movement to reflect the function of the RLN 8,9 . IONM increases the identification rate of RLN, reduces the time of IONM identification and predicts the postoperative function of the vocal cords 10,11 . IONM can be divided into 2 generations, non-continuous intraoperative nerve monitoring (NCIONM) and continuous intraoperative nerve monitoring (CIONM). NCIONM is still the most common method of nerve monitoring, and CIONM is a recent surgical trend that could prevent RLN injury though real-time monitoring.
Although IONM is designed to reduce the incidence of RLN injury, the effect of IONM on RLN injury prevention in thyroid surgery is still controversial 12,13 . Some results found that IONM could reduce transient RLN injury, while other results revealed non-significant help of IONM in RLN injury prevention 4,14,15 . Several meta-analyses have been conducted to discuss this topic 12,13,[16][17][18] . Most of these studies found that IONM was not superior to visual identification in preventing permanent RLN injury 12,18 . Some results indicated that IONM could decrease RLN injury incidence in high risk thyroid surgery, particularly in cancer operations or in those with a previous history of neck operation 17 . For operative sites, incidence of bilateral injury under IONM was relatively higher (0.2%), and staged thyroidectomy was recommended when one-side RLN injury occurred to prevent further bilateral RLN injury 19,20 .
During the last 2 years, many large clinical trials emerged on this topic, which made it necessary to update the data for further conclusion 10,11,[21][22][23][24][25][26] . Considering the existing controversial results and new data, we conducted this meta-analysis to clarify the effects of IONM in thyroidectomy. Subgroup analysis was performed to identify the effects of IONM in bilateral operation, malignancy operation, reoperation group and operation volume.

Materials and Methods
Search strategy and selection criteria. We performed a comprehensive literature search in MEDLINE (PubMed), BIOSIS Previews (ISI Web of Knowledge) and Cochrane library from January 1980 to July 2017. We also manually added relevant articles by reviewing the references. Search terms included: "Neuromonitoring" or "nerve monitoring", "thyroidectomy" or "thyroid surgery", and "recurrent laryngeal nerve" or "RLN". Studies included were confined to randomized controlled trials, case control studies and cohort studies, with convertible effects data of IONM in thyroidectomy. If there were several papers extracting from the same population, we would only include the most recent publication. We used the criterion of 6 months to distinguish transient RLN injury from permanent RLN injury, which was consistent with relevant studies 10,27 . Data extraction and quality assessment. Data were extracted by two authors (BL Bai and WZ Chen) independently. Detailed information was recorded on first author, year of publication, title, location, study type, diagnosis, inclusion and exclusion criteria, allocation method, study duration, device used, outcome measurement, definition of permanent injury, number of case and control groups, the event of transient and permanent RLN injury, the population type, multicenter research or not, main conclusion and Newcastle-Ottawa Scale (NOS) score. Nerve at risk (NAR) was calculated based on number of lobectomy (1 NAR for each side), subtotal thyroidectomy (2 NARs) and total thyroidectomy (2 NARs). In the section of operation volume, we omitted multicenter research studies for the unavailability of hospital volume data and would discuss the effects of IONM under open surgical operation because of insufficient sample size of endoscopic surgery. The Newcastle-Ottawa Scale for Quality was used to assess the included cohort studies 28 . According to the Cochrane Collaboration, risk ratio (RR), risk difference (RD) and confidence interval (CI) were estimated from the median of the posterior distribution and calculated by the random-effects model with inverse   variance weight [30][31][32] . If zero events occurred in both experimental and control group, the pooled RD estimates and 95% CIs were applied. Otherwise, RR was applied. RR below one or RD below zero indicated a benefit of the experimental intervention. We estimated 95%CI from the 2.5th and 97.5th percentiles of the posterior distribution, and calculated two-sided p values from the posterior distribution. If the 95%CI did not contain the valid value 1, the result was considered to have significant difference. Publication bias was assessed using Begg's test and Egger's regression asymmetry test 33 . The heterogeneity between trials was estimated from the median between-trial variance (τ²) observed in the posterior.
Availability of supporting data. All data are fully available without restriction.
Publication bias and sensitivity analysis. Begg's and Egger's test were conducted to assess the publication bias in this meta-analysis. None of the tests was statistically significant in total injury (Begg's p = 0.97 and Egger's p = 0.14), transient injury (Begg's p = 0.79 and Egger's p = 0.15) and permanent injury (Begg's p = 0.012 and Egger's p = 0.19). Meta-analysis was recalculated after omitting every record included in the analysis to check the sensitivity of our overall results. As shown in Figs 2, 3 and 4, the results of total, transient and permanent injury were robust after omitting any record.

Discussion
IONM had received widespread acceptance in the field of thyroid surgery for its role in RLN identification. However, the protective effect of IONM on RLN injury remained controversial. Zheng et al. proposed that IONM could reduce total and transient RLN injury in thyroidectomy 13 , but the results of transient injury had been proven to be hypersensitive 13 . Lombardi et al. gave a negative conclusion about the protective effects of IONM on permanent injury 18 . Wong et al. demonstrated that high-risk subpopulations might benefit from IONM, especially in reoperation and malignancy subgroups 17 . Several large studies were recently published in the past year; Some studies found preventive effects of IONM on total and transient injury 23,24,26 , while others declared an insignificant result on this topic 10,11,22,25 . To identify the protective effect of IONM on RLN function and explore more subgroups which could benefit from IONM, we conducted a meta-analysis that included the most recent developments in this field. Additionally, our meta-analysis was the first to investigate the protective effect of IONM on the topic of bilateral operation and operation volume and revealed the preventive effect of IONM on permanent injury in these subgroups.
Our meta-analysis found preventive effects of IONM on total, transient and permanent RLN injury. The result was still robust after sensitivity analysis. Previous meta-analyses found a decrease of total and transient injury in IONM cases, but for permanent injury the result was not statistically significant 12,13 . The negative results might result from small data sample size (NARs less than 40,000). A total of 59,380 NAR involved in this analysis was enough to reveal IONM's effect on permanent injury. Lombardi et al. mentioned the influence of permanent injury definition and measurement 18 . In this study, both the permanent injury definition (over 6 months) and laryngoscopy measurement were included and did not influence the final results (Appendix Table 3). Based on existing results, we recommended IONM as a routine measurement to be used in thyroidectomy.
In subgroup analyses, the study revealed preventive effects of IONM on total and transient RLN injury in bilateral operation and malignancy operation subgroups. However, for permanent injury, a similar effect was only found in bilateral operation cases. Bilateral thyroidectomy requires relative large operation extent and long operation duration. Previous studies revealed RLN injury incidence in bilateral operation was 1.2-2.3% 2,5 . In this study, pooled incidences of total, transient and permanent injury in bilateral operation were 2.3%, 1.6% and 0.61%. Our results supported the recommendation of IONM application in bilateral operation. Above all, bilateral operation might induce bilateral RLN injury, which could lead to airway obstruction. To avoid this fatal complication, someone proposed routine IONM application in bilateral operation and a staged thyroidectomy would be needed for cases with intraoperative RLN signal loss 19 .
Previous meta-analyses only showed positive results in transient injury for malignancy operation. Our analysis included 3 new studies 11,21,25 and 1 study with updated data 22 . The results demonstrated a statistically significant decrease of total and transient injury in malignancy operation cases with IONM. Malignancy operation might contain potential risk of malignancy invasion of RLN and lymph node dissection, which increase the risk of RLN injury. IONM might reduce the risk by real-time monitor of RLN function.
Reoperation of the thyroid has higher risks than normal thyroidectomy, for local conglutination and anatomical changes can result in RLN injury. Local conglutination makes the identification of RLN difficult and the anatomical position of RLN might change due to a previous operation. Wong et al. found a decreased incidence of total injury in the reoperation group with IONM. However, our study demonstrated a negative result in the reoperation group with IONM, with 2 new studies 21,45 and 1 study with updated data 22 included. The effect of IONM in this subgroup still needs to be identified. Further large size randomized trials need to be conducted to clarify this debate.
The impact of operation volume was also emphasized in several studies. In a German retrospective multicenter trial, a large operation volume medical center had a lower rate of permanent RLN injury, and thus IONM might be more helpful in small operation volume centers than in large operation volume centers 5 . However, in our subgroup analysis, we found IONM could help both small and large operation volume medical centers. IONM reduced total and transient RLN injury in small operation volume centers, and decreased total and permanent RLN injury in large operation volume centers. With the aid of IONM, inexperienced surgeons could avoid hazardous surgical situations near or on the RLN with real-time identification of the RLN 54 .
There were several limitations in our study. First, there were still several studies didn't declare the outcome measurement or follow up for enough duration. The robust test had been passed including all the studies with laryngoscopy measurement and more than 6 months of follow-up. Second, there still existed confounders and heterogeneity in our analysis. Some studies included cases that involved intentional RLN transection. The influence of lymph node dissection could also not be eliminated. The sensitivity analysis, subgroup analysis, and random-effects model was applied to address this.

Conclusions
This meta-analysis indicated that IONM could reduce the incidence of total, transient and permanent RLN injury compared with conventional visual identification. We also recommended IONM in bilateral operations and malignancy operations. IONM could help surgeons perform a better thyroidectomy regardless of the medical center's operation volume. The benefit of IONM in reoperation cases needs to be further explored. We hope our findings can be assessed by large prospective, randomized clinical trials in the future with a standardized IONM application and outcome measurement.