Prognostic value of subthreshold biphasic electromyographic (EMG) waveforms in Type I loss of signal during thyroid intraoperative neuromonitoring

Authors: Yong-Il Cheon, Ha-Nee Kwon, Byung-Joo Lee, Sung-Chan Shin

Abstract

Background

During intraoperative neuromonitoring (IONM), an evoked amplitude below 100 µV is typically regarded as loss of signal (LOS). However, after adjusting the event threshold, previously undetectable electromyographic (EMG) waveforms may become visible. This study aimed to evaluate the diagnostic and clinical significance of such subthreshold EMG waveforms.

Methods

A total of 231 patients (335 nerves at risk, NARs) who underwent thyroidectomy with IONM using either EMG endotracheal tube electrodes or adhesive skin electrodes were retrospectively reviewed. Evoked EMG parameters, including mean amplitude and latency after stimulation of the recurrent laryngeal nerve (RLN) and vagus nerve (VN), were analyzed. Preoperative and postoperative laryngoscopic examinations were performed to assess vocal fold mobility.

Results

Among the 335 NARs, twelve NARs (3.5%) demonstrated evoked EMG amplitudes below 100 µV and were confirmed as true LOS, all of which were classified as Type I (segmental) LOS. After stepwise lowering the event threshold from 100 µV to 80 and subsequently to 50 µV, biphasic EMG waveforms became detectable in 10 of these 12 cases. Although all 12 nerves exhibited immediate postoperative vocal fold paralysis, the 10 nerves with detectable low-amplitude waveforms showed complete recovery of vocal fold mobility within eight months postoperatively.

Conclusions

In cases of Type I LOS during thyroid surgery, the presence of subthreshold biphasic EMG waveforms may reflect residual neural conduction and is strongly associated with postoperative vocal fold recovery. These findings suggest that optimization of threshold settings can improve the sensitivity of IONM interpretation and assist in intraoperative prognostic assessment.

Introduction

Intraoperative neuromonitoring (IONM) has become an indispensable tool in thyroid and parathyroid surgery to help preserve the recurrent laryngeal nerve (RLN) and external branch of the superior laryngeal nerve (EBSLN) (1). Reliable electromyographic (EMG) signal acquisition is critical for accurate nerve identification, real-time functional assessment, and postoperative prognostication (1). Several types of recording electrodes are commonly used in thyroid IONM to detect EMG signals, including endotracheal tube with surface electrodes (EMG-ETT), needle electrodes, and adhesive skin electrodes, all of which have been reported to be effective (2-4).

According to the International Standards Guideline of IONM, an event threshold value of 100 µV is recommended in monitor setup (1). Evoked EMG amplitude ≥100 µV is generally considered to indicate normal nerve function, whereas an amplitude below 100 µV is automatically regarded by the IONM system as a loss of signal (LOS) and is not displayed on the monitor. In general, LOS suggests possible nerve injury. When LOS is detected intraoperatively, the surgeon should determine, based on a standardized troubleshooting algorithm, whether the observed LOS represents a true or false LOS. The positive predictive value (PPV) of IONM has been reported to be approximately 77–83%, and the negative predictive value (NPV) around 99%. When the standard IONM protocol is properly followed, an event identified as true LOS intraoperatively is highly predictive of postoperative vocal cord paralysis (5-7).

However, biphasic EMG waveforms can still be observed at subthreshold level. To date, no studies have investigated whether vocal cord paralysis occurs in such cases, or, if it does, whether recovery follows. The aim of this study was to analyze the incidence and recovery of vocal cord paralysis in patients who were initially diagnosed with true LOS intraoperatively, but in whom EMG waveforms were subsequently detected at subthreshold level. We present this article in accordance with the STROBE reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-515/rc).

Methods

Subjects

This retrospective study reviewed medical records of 231 consecutive patients, 335 nerves at risk (NARs) who underwent conventional thyroid lobectomy or total thyroidectomy and/or central lymph node dissection with IONM using EMG-ETT or skin electrodes at Department of Otorhinolaryngology-Head and Neck Surgery, Pusan National University Hospital from September 2019 to August 2021. Patients with preexisting vocal cord palsy, those undergoing revision neck surgery, lateral neck dissection and patients undergoing parathyroid surgery were excluded. Demographics and clinicopathologic data were collected and analyzed. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Pusan National University Hospital (No. 2509-017-154) and individual consent for this retrospective analysis was waived.

General anesthesia and IONM setup

After standard general anesthesia induction, anesthesia was maintained with sevoflurane. A single intubating dose of a neuromuscular blocking agent was administered only during induction to facilitate endotracheal intubation, using a conventional low dose commonly employed in IONM (rocuronium 0.6 mg/kg). No additional neuromuscular blockade was administered after induction, and no reversal agents were used during the procedure. Patients were placed with Rose position and IONM was performed using either adhesive skin electrode (DSE3125, Medtronic Xomed Inc., Jacksonville, FL, USA) or EMG-ETT (Medtronic Xomed Inc., Jacksonville, FL, USA). The NIM-Neuro 3.0 system (Medtronic Xomed Inc., Jacksonville, FL, USA) was used for IONM. A video laryngoscope was employed to better visualize the glottic view and verify the proper electrode-cord position after patient’s neck positioning if EMG-ETT was used for IONM.

Stimulation duration was set as 100 µs with a frequency of 4 Hz. The initial event threshold was set at 100 µV. All set-up and monitoring procedures complied with the standards outlined by the International Neural Monitoring Study Group (INMSG) guidelines (1). Intermittent IONM (I-IONM) was used in this study, whereas continuous IONM (C-IONM) was not employed in this study due to additional operative complexity and cost.

IONM procedure & threshold adjustment protocol

Neural mapping was performed at a stimulation intensity of 3 mA for both the RLN and vagus nerve (VN), while nerve identification was conducted using 1 mA for the RLN and 3 mA for the VN after visualization. All procedures were monitored according to the standard four-step IONM protocol (V1–R1–R2–V2) proposed by Chiang* et al. *(8), which includes sequential stimulation of the VN before (V1) and after (V2) thyroid dissection, and the RLN before (R1) and after (R2) its dissection. When intraoperative R2 signal loss was observed, the contralateral V2 signal was checked to exclude false LOS. If a true LOS was confirmed under the standard IONM setting with an event threshold of 100 µV, a stepwise threshold adjustment protocol was applied to assess the presence of residual EMG activity. The event threshold was initially reduced from 100 to 80 µV. If no reproducible biphasic EMG waveform was detected at this level, the threshold was further lowered to 50 µV. At each threshold level, stimulation was repeated to confirm waveform reproducibility and exclude random noise. Threshold values below 50 µV were not evaluated, as lower thresholds were associated with excessive background noise that interfered with reliable waveform discrimination and signal interpretation.

In patients initially planned for total thyroidectomy, the occurrence of true LOS in the first side did not routinely mandate conversion to staged surgery or hemithyroidectomy. Instead, when LOS was confirmed, the contralateral thyroid dissection was performed with heightened caution, including meticulous RLN identification and strict avoidance of excessive traction or thermal injury. The decision to proceed with contralateral surgery was individualized, rather than adhering rigidly to an automatic staged surgery recommendation.

Type and mechanism of injury

After confirmation of true LOS, nerve injuries were classified as Type I (segmental) or Type II (global) according to the INMSG guideline, based solely on intraoperative electrophysiological findings (1). Type I LOS was defined as a segmental loss of EMG response with preserved distal stimulation, whereas Type II LOS was defined as a global loss of EMG response along the entire nerve. These LOS classifications were used to describe intraoperative signal patterns and were not intended to directly correspond to histopathological nerve injury classifications such as neuropraxia, axonotmesis, or neurotmesis.

Postoperative laryngoscopy and follow up strategy

All patients underwent preoperative and postoperative evaluations at 2 weeks, 3 months, and 8 months after surgery, including laryngoscopic examination, acoustic analysis [multi-dimensional voice program (MDVP), speaking fundamental frequency (SFF), and voice range profile (VRP)], perceptual analysis (GRABS), and thyroidectomy-related voice questionnaire (TVQ). In addition, all patients with vocal fold paralysis received no medications or other procedures to facilitate recovery, except for voice therapy. Recovery was defined as restoration of vocal fold mobility confirmed by laryngoscopic examination. Permanent vocal fold paralysis was defined as the absence of vocal fold motion persisting beyond 6 months postoperatively, consistent with commonly used clinical criteria. Accordingly, laryngoscopic evaluation at 8 months postoperatively was used as the final assessment to determine recovery or persistence of paralysis.

Statistical analysis

Statistical analyses were performed using Statistical Package for the Social Sciences version 27 for Windows (IBM Corp., Armonk, New York, USA). All data are presented as mean ± standard deviation for continuous variables. Fisher’s exact test and independent t-tests were used to compare categorical and continuous variables, respectively. A P value <0.05 was considered statistically significant.

Results

A total of 231 patients (335 NARs) undergoing thyroid surgery using thyroid IONM were analyzed. The cohort comprised 192 females and 39 males, with a mean age of 51.3 years. Thyroid lobectomy was performed in 127 patients, and total thyroidectomy in 104 patients. Central neck dissection was performed in 186 patients, and lateral neck dissection in 4 patients. All enrolled patients demonstrated normal glottic function in preoperative laryngoscopy examination. The final histopathological diagnoses included 173 (74.8%) papillary thyroid carcinomas, 48 (20.7%) follicular adenomas, and 10 (4.3%) nodular hyperplasia (Table 1).

Mean EMG data for each of the four IONM steps (V1–R1–R2–V2)

A total of 335 NARs, including 335 VNs and 335 RLNs, were assessed. The EMG data, including the mean evoked amplitudes and latencies for each of the four IONM steps (V1–R1–R2–V2) are summarized in Table 2.

Incidence of true LOS and verification of EMG waveform after event threshold adjustment

True LOS was detected in 12 of 335 NARs (3.5%). After event threshold adjustment (from 100 to 50 or 80 µV), biphasic EMG waveforms were identified in 10 of these cases (Figure 1), whereas 2 cases showed no detectable biphasic waveform even after adjustment. Efforts were made to determine the type and mechanism of nerve injury among the 12 true LOS cases. All cases of true LOS were classified as Type I injury. No case of nerve transaction was identified, and most injuries were presumed to result from thermal or traction mechanisms. The details of injury types and mechanisms are summarized in Table 3.

Assessment of vocal cord mobility

Postoperative laryngoscopic examinations verified vocal fold paralysis in all 12 NARs that exhibited true LOS intraoperatively. No therapeutic intervention other than voice therapy was provided for vocal fold paralysis. Follow-up evaluations, including laryngoscopy and voice assessments, were conducted in 2 weeks, 3 months, and 8 months postoperatively. Among the 10 cases in which subthreshold biphasic EMG waveforms were identified during surgery, restoration of vocal fold mobility was verified by laryngoscopy within 8 months postoperatively. In contrast, none of the 2 nerves without detectable subthreshold waveforms recovered (P=0.045).

Discussion

IONM has become an essential adjunct in contemporary thyroid surgery, contributing to improved intraoperative decision-making and postoperative functional outcomes related to RLN preservation (1). Nevertheless, interpretation of intraoperative LOS remains challenging, as LOS encompasses a spectrum of electrophysiological phenomena rather than a uniform indicator of irreversible nerve injury (8-10). The present study specifically focused on a subset of true LOS cases in which reproducible biphasic EMG waveforms reappeared after event threshold adjustment and evaluated the clinical significance of these subthreshold signals.

The principal finding of this study is that detection of reproducible subthreshold biphasic EMG waveforms was strongly associated with postoperative recovery of vocal fold mobility. Although all nerves classified as true LOS demonstrated immediate postoperative vocal fold paralysis, recovery occurred exclusively in cases where low-amplitude waveforms became detectable after lowering the event threshold. Importantly, these findings do not imply preservation of normal neural integrity, but rather suggest the presence of residual neural conduction or partial functional continuity that is not captured under standard monitoring settings (5,10).

Current IONM guidelines define LOS using an absolute amplitude threshold of 100 µV, primarily to minimize background noise and false-positive signals (1). However, increasing evidence indicates that interpretation of intraoperative EMG changes based solely on an absolute cutoff may oversimplify a fundamentally dynamic electrophysiological process (5,10). Previous studies have demonstrated that the concept of relative threshold, which means relative changes in EMG amplitude, particularly reductions or recoveries of approximately 50% from baseline, correlate more closely with postoperative vocal fold outcomes than absolute amplitude values alone (5,10). Consistent with this paradigm, the present study demonstrates that signals below 100 µV—traditionally interpreted as complete LOS—may still carry prognostic information when reproducible waveforms are identified after threshold optimization.

The relevance of this finding becomes particularly evident when considering recording modality–dependent baseline amplitudes. Adhesive skin electrodes, while noninvasive and convenient, are known to yield lower baseline EMG amplitudes and higher noise levels compared with EMG endotracheal tube or needle electrodes (2,11). In our cohort, the mean baseline (V1) amplitude recorded with skin electrodes was approximately 241 µV, indicating that a fixed 100 µV threshold corresponds to nearly a 60% reduction from baseline. Under such conditions, rigid application of an absolute threshold may result in classification of partial or incomplete signal loss as complete LOS (2,11). Detection of biphasic waveforms at a threshold of 50 µV in this study approximates a relative threshold of roughly 20% of baseline amplitude, supporting the notion that some “complete” LOS events under an absolute framework may more accurately represent incomplete functional disruption.

In this context, transcartilage surface recording electrodes have been proposed as an alternative modality that may partially overcome the inherent limitations of low-voltage recording systems (12). Previous experimental and clinical studies have demonstrated that transcartilage electrodes can provide EMG amplitudes comparable to or even exceeding those obtained with EMG endotracheal tube electrodes, while avoiding signal instability related to tube malposition. By improving signal-to-noise characteristics and baseline amplitude, transcartilage recording may reduce the likelihood that clinically meaningful residual neural activity is masked by a rigid absolute threshold, thereby enhancing the interpretability of both absolute and relative EMG changes. Although transcartilage electrodes were not utilized in the present study, their favorable electrophysiological profile suggests that the applicability and clinical relevance of absolute versus relative threshold concepts may vary according to recording modality and warrants further investigation.

It is also important to clarify that LOS classification (Type I* vs. Type II) reflects intraoperative electrophysiological signal patterns rather than histopathological nerve injury severity (1,9). In this cohort, all confirmed LOS events were classified as Type I (segmental) LOS and these findings differ from that reported in some previous studies, in which global LOS accounts for a substantial proportion of signal loss events (13,14). Several factors may explain this discrepancy. First, no cases of complete nerve transection or gross structural disruption were encountered in our cohort. Presumed mechanisms of injury in our cohort were predominantly focal traction- or thermal-related, which typically produce localized conduction disturbances rather than diffuse neural failure. These findings can explain a discrepancy between our cohort and other studies. It should be emphasized that LOS classification (Type I vs. *Type II) reflects intraoperative electrophysiological signal patterns rather than histopathological nerve injury severity and should not be directly equated with neuropraxia, axonotmesis, or neurotmesis (1,9). Accordingly, the prognostic implications of subthreshold biphasic EMG waveform detection demonstrated in this study are applicable only to Type I LOS and should not be extrapolated to Type II LOS, in which diffuse conduction failure. Future studies involving larger cohorts and a broader spectrum of nerve injury severity are warranted to determine whether the prognostic value of subthreshold waveform detection differs between Type I and Type II LOS.

Previous cohort studies have consistently demonstrated that Type I LOS is associated with a higher incidence of postoperative vocal fold paralysis compared with Type II LOS. In contrast, Type II LOS is more often caused by traction-related injury and is generally considered to reflect a reversible conduction block with favorable functional recovery. Importantly, however, Type I LOS represents a heterogeneous entity, and postoperative outcomes vary substantially according to the underlying mechanism of injury. While complete nerve transection is associated with poor recovery, localized non-transection injuries such as traction, thermal injury, or focal compression may demonstrate substantial functional recovery (15).

Consistent with this concept, it has been reported that when macroscopic structural integrity of the RLN is preserved, postoperative vocal fold paralysis recovers in the majority of cases, with recovery rates approaching 95% (16). In the absence of complete nerve transection, both Type I (segmental) and Type II (global) loss-of-signal events have been shown to demonstrate functional recovery within several months (15). In this context, our findings do not contradict the established literature but rather suggest that, within Type I LOS, detection of reproducible subthreshold biphasic EMG waveforms after threshold optimization may help identify a subgroup with preserved residual neural conduction and a higher likelihood of postoperative recovery.

From a physiological standpoint, low-amplitude EMG responses may arise from partial preservation of functional motor units despite transient conduction block or focal demyelination (17). Under standard monitoring conditions, such responses may fall below the detection threshold and be interpreted as absent. Threshold adjustment allows visualization of these residual signals, thereby refining intraoperative interpretation without altering the underlying neural state. This approach should therefore be regarded as an interpretive adjunct, rather than a means of redefining neural integrity or overriding established LOS criteria (1,5).

This study was conducted using intermittent IONM (I-IONM), which inherently differs from continuous IONM (C-IONM) in its ability to detect evolving nerve stress in real time. C-IONM has been shown to identify impending traction-related injury earlier, potentially preventing progression to LOS (5,10). Consequently, some LOS events observed in this cohort may have been avoidable with continuous monitoring. However, the aim of the present study was not to assess LOS prevention, but rather to evaluate the prognostic significance of EMG findings after true LOS had already occurred. In this context, the ability to identify subthreshold biphasic waveforms may remain clinically relevant even in C-IONM settings, particularly when partial signal recovery or low-amplitude responses are observed (5,10).

Several limitations merit consideration. This was a retrospective, single-institution study with a relatively small number of true LOS cases, reflecting the low incidence of such events in experienced surgical practice. All EMG data were obtained using a single monitoring platform, and results may not be directly generalizable to systems with different hardware or signal-processing algorithms. Additionally, the inherent characteristics of adhesive skin electrodes complicate relative threshold analysis, as amplitudes may rapidly approach the noise floor (2,14). In this setting, manual adjustment of absolute thresholds to identify reproducible waveforms may be particularly valuable. Finally, although postoperative follow-up extended to eight months, delayed recovery beyond this period cannot be entirely excluded. However, permanent vocal fold paralysis is commonly defined as the absence of recovery beyond six months postoperatively, and previous studies have shown that most functional recovery of RLN injury occurs within the first several months after surgery. In our cohort, no additional recovery was observed beyond six months, suggesting that laryngoscopic evaluation at eight months provides a clinically reasonable and conservative time point for final outcome assessment.

In summary, this study supports a functional and relative interpretation of intraoperative EMG changes, emphasizing that LOS should not be viewed as a strictly binary event defined by a fixed absolute threshold. In cases of Type I LOS, detection of reproducible subthreshold biphasic EMG waveforms after threshold adjustment may indicate residual neural conduction and is associated with postoperative vocal fold recovery. Threshold optimization should therefore be considered a supplementary troubleshooting step in IONM interpretation, while avoiding overinterpretation and maintaining postoperative laryngoscopic confirmation as the definitive standard. Future multicenter, prospective studies are warranted to confirm these findings and to determine optimal threshold parameters for different electrode types and clinical settings.

Conclusions

In conclusion, this study demonstrates that in cases of Type I signal loss during thyroidectomy, the detection of biphasic EMG waveforms after event threshold adjustment may serve as a strong predictor of postoperative vocal fold recovery. The presence of subthreshold waveforms does not invariably indicate fully preserved neural function but rather suggests residual neural conduction or partial excitability that may allow subsequent recovery in Type I LOS.

Supplementary