Nerve Injury During Thyroid Radiofrequency Ablation: Mechanisms, Prevention, and Practical Management

Abstract

Voice change is an uncommon but clinically important complication of thyroid radiofrequency ablation (RFA), because it may indicate injury to the recurrent laryngeal nerve or, less commonly, the vagus nerve. Early recognition and appropriate intervention are important, because neural dysfunction may remain reversible before the occurrence of irreversible structural nerve injury. Thermal nerve injury during thyroid RFA is not always a single instantaneous event. Rather, it may involve direct thermal exposure of the nerves, leading to acute endoneurial or perineural edema, microvascular compromise, and delayed inflammation or fibrosis. Therefore, a practical approach to the management of voice changes during thyroid RFA should emphasize prevention, frequent intraprocedural voice assessment, immediate cessation of ablation, and prompt bolus injection of cold 5% dextrose in water. After adequate cooling, local dexamethasone injections and a short course of oral corticosteroid therapy may be considered. In conclusion, a mechanism-based framework may improve operator preparedness and reduce the risk of persistent nerve dysfunction.

INTRODUCTION

Ultrasound-guided radiofrequency ablation (RFA) is an established nonsurgical treatment option for symptomatic benign thyroid nodules and is increasingly being used in carefully selected patients with low-risk papillary or recurrent thyroid cancer [1,2,3,4,5,6]. By producing controlled thermal coagulation within the target tissue, RFA can reduce nodule volume, improve compressive or cosmetic symptoms, and provide local tumor control in selected malignant lesions while avoiding many surgery-related complications [7]. With the growing rates of global adoption of this technique, attention has shifted from efficacy alone to standardization of techniques, prevention of complications, and practical management of adverse events [8,9].

Voice change is among the most concerning adverse events during and after thyroid RFA. It is also the most frequently reported major complication, with a pooled incidence of 1.4% (8% in recurrent thyroid cancer) [10,11,12]. It is usually attributed to recurrent laryngeal nerve (RLN) dysfunction, particularly when ablation is performed near the posteromedial thyroid capsule or tracheoesophageal groove [3]. In selected circumstances, injury to the vagus nerve may also contribute to voice changes, particularly when the nerve lies close to the thyroid gland or ablation zone. Although most cases are transient, persistent voice changes can impair communication, quality of life, and occupational function [13].

Existing guidelines and technical reviews have emphasized prevention, careful monitoring, hydrodissection, and conservative ablation near critical structures [2,3,14]. However, the pathophysiological basis underlying acute and delayed management of voice changes has been discussed less frequently. For operators, understanding how thermal nerve injuries develop and evolve is essential for timely recognition and appropriate interventions. This review presents a mechanism-based framework for the occurrence of voice changes during thyroid RFA, with a focus on prevention, early detection, and management.

NERVES AT RISK

The RLN is most commonly implicated in voice changes during or after thyroid RFA. It usually ascends along the tracheoesophageal groove and enters the larynx near the cricothyroid joint. This anatomic course places the RLN at risk during the ablation of nodules or recurrent tumors located in the posteromedial thyroid, central compartment, or postoperative bed. However, its exact course may vary, and the nerve cannot be directly visualized on ultrasound [15,16,17].

The vagus nerve may also contribute to voice changes in selected cases, particularly when it lies close to the thyroid gland or the ablation zone. Other neural structures, including the cervical sympathetic chain and superficial sensory nerves, may also be vulnerable depending on the lesion location and heat propagation, although injury to these structures usually produces nonvocal symptoms.

Clinically, RLN or vagus nerve dysfunction may manifest as hoarseness, weak voice, reduced pitch control, vocal fatigue, aspiration symptoms, or coughing. Since early voice changes may be subtle, repeated patient speech assessment during the procedure is important for early detection of phonation-related nerve dysfunction. Operator awareness should be particularly high when treating lesions adjacent to the danger triangle, which is the region between the trachea and thyroid gland where the RLN is vulnerable, because even a limited expansion of the ablation zone may affect the adjacent neural structures.

PERIPHERAL-NERVE ANATOMY

The peripheral nerves contain axons, myelin sheaths, Schwann cells, and connective tissue layers, including the endoneurium, perineurium, and epineurium. Axons conduct action potentials, whereas myelin sheaths provide electrical insulation and enable saltatory conduction. The connective tissue layers (i.e., the epineurium, perineurium, and endoneurium) provide mechanical support and help maintain the intraneural microenvironment [18].

This anatomy is clinically relevant in the context of thermal injury. The mechanical protection provided by the epineurium and perineurium does not necessarily prevent heat transfer. In addition, because the perineurium forms a diffusion barrier, heat-induced edema may increase the intraneural pressure and impair microcirculation. Thus, thermal nerve dysfunction may result from both the primary thermal injury and secondary edema or ischemia.

The small caliber of the RLN is also clinically important. Unlike a large tissue mass, in which thermal gradients may protect deeper structures, a small nerve can be affected across much of its cross-sectional area when exposed to sufficient heat. This concept may explain why symptoms sometimes appear suddenly rather than gradually during RFA.

THERMAL VULNERABILITY

Neural substructures differ in their susceptibilities to heat. The lipid-rich myelin sheath is relatively vulnerable to mild thermal stress (43℃–45℃), whereas axonal proteins and collagen-rich supporting layers generally require higher temperatures for irreversible structural alteration [19,20,21]. The structural basis for differential thermal vulnerability across neural substructures is illustrated in Figure 1 and summarized in Table 1.

Fig. 1. Peripheral-nerve structures relevant to thermal injury during thyroid radiofrequency ablation. The myelin sheath and axons are functionally critical and heat-vulnerable inner components, whereas the endoneurium, perineurium, and superficial and deep epineurium provide relatively heat-resistant structural support. Therefore, functional disturbances may begin within these inner components before overt damage to the outer connective tissue layers occurs, which explains how voice changes can occur despite apparent structural preservation.

Table 1. Approximate temperature ranges and representative pathophysiologic changes in peripheral-nerve substructures during thermal injury.

Structure	Approximate threshold temperature range	Pathophysiologic implication
Myelin sheath	43℃–45℃	Lipid-rich structure; early myelin disruption and conduction disturbance
Axon	45℃–50℃	Protein coagulation; conduction block and potential axonal injury
Epineurium, perineurium	50℃–60℃	Collagen denaturation; loss of structural integrity and reduced likelihood of reversibility

For clinical interpretation, the key point is not the presence of a single universal temperature threshold but rather that early functional impairment may occur before overt disruption of the outer connective tissue layers. Therefore, a nerve may remain structurally intact while exhibiting conduction failure. This distinction supports early interventions aimed at maintaining the injury within a reversible spectrum.

THRESHOLD-LIKE BEHAVIOR

Voice changes during RFA can occur abruptly. This phenomenon reflects the threshold-like all-or-none clinical behavior of neural conduction. Clinical symptoms may only become apparent after a sufficient proportion of neural conduction is impaired. Before the functional threshold for voice change is reached, subclinical dysfunction may be present and is detectable using electromyography (EMG) or nerve conduction studies (NCS) despite the absence of overt symptoms [19,20,22]. Once the conduction reserve is exceeded, voice changes may become clinically apparent over a short interval [23].

The same threshold-like behavior may also contribute to the apparently abrupt recovery. A patient may remain symptomatic until neural conduction improves sufficiently to restore the functional reserve, after which voice quality may normalize relatively quickly. Thus, rapid recovery does not necessarily indicate a trivial injury but rather may reflect the restoration of conduction above the functional threshold for voice production.

ACUTE AND DELAYED THERMAL INJURY

Thermal nerve injury is a time-dependent process involving acute and delayed phases [19,22]. In the acute phase, heat may primarily affect heat-vulnerable neural components, particularly the lipid-rich myelin sheath, resulting in early conduction disturbances. Importantly, thermal exposure of the nerve does not necessarily end when the RF generator is turned off. Heat stored in the adjacent ablated tissue may continue to be conducted toward the nerve, thereby prolonging the acute injury beyond the moment of energy cessation. This residual heat conduction provides a mechanistic basis for active cooling, which aims to dissipate stored heat and reduce the local temperature below the range associated with ongoing neural injury.

Heat may also increase vascular permeability, leading to endoneurial or perineural edema. Because the nerve fascicle is a confined space, edema may increase the intraneural pressure and cause secondary microvascular compromise [21]. These early mechanisms support the importance of rapid cooling and control of edema and inflammation. If residual heat and inflammatory edema persist, neural dysfunction may develop beyond the immediate procedural period.

In the delayed phase, persistent edema and inflammation may lead to secondary demyelination, ongoing microvascular compromise, and perineural fibrosis [24]. These processes may prolong recovery and explain delayed or incomplete clinical improvement in some patients. This delayed inflammatory phase also provides a biologically plausible rationale for oral corticosteroid therapy following acute management, particularly when persistent edema or delayed inflammatory changes are suspected to contribute to prolonged nerve dysfunction.

INJURY STAGING

In the Seddon classification of peripheral-nerve injuries, neurapraxia represents conduction failure without axonal discontinuity and is potentially reversible, whereas axonotmesis and neurotmesis involve more severe structural injuries that are irreversible or only slowly reversible [25]. For a practical interpretation during thyroid RFA, neurapraxia may be further conceptualized as subclinical or clinical. Subclinical neurapraxia refers to an early stage in which myelin dysfunction, edema, or microvascular compromise may already be present, while nerve conduction remains sufficient to preserve voice function. At this stage, the patient may have no voice symptoms and dysfunction may be detectable only by objective neurophysiological testing, such as EMG or NCS, rather than by subjective voice assessment. In this conceptual model, clinical neurapraxia begins when the conduction reserve is exceeded and voice changes become clinically detectable. This transition may appear abrupt and threshold-like, even though the underlying thermal and edematous injury processes are gradual. Clinical neurapraxia is an early symptomatic stage that may be reversible if additional heat exposure, residual heat conduction, and secondary edema are promptly controlled [23]. If the injury progresses beyond the threshold for irreversible damage, axonotmesis or neurotmesis may occur, resulting in axonal injury, connective tissue disruption, symptom progression, or persistent dysfunction [24]. The threshold-like progression is summarized in Figure 2. This conceptual framework is intended to be a biologically plausible model that integrates experimental peripheral-nerve data and observed clinical behavior; however, it has not been directly validated in thyroid RFA-related RLN injury.

Fig. 2. Threshold-like clinical manifestation and progression of thermal nerve injuries during thyroid radiofrequency ablation. The model illustrates the progression of thermal nerve injuries from subclinical neurapraxia to clinical neurapraxia and, with increasing severity, to axonotmesis or neurotmesis [19,21,22,23,24,25]. In the subclinical stage, while myelin dysfunction, edema, or microvascular compromise may already be present, voice symptoms are absent because of a preserved conduction reserve [19,21,22,23]. Once the functional threshold for voice change is crossed, clinical neurapraxia becomes apparent as voice changes [19,20,22,23]. The neurapraxic stages are considered potentially reversible, whereas progression beyond the threshold for irreversible injury toward axonotmesis or neurotmesis involves axonal or connective tissue injury and may result in symptom progression and incomplete recovery [24,25]. Frequent voice assessments can detect clinical neurapraxia early, whereas prompt cooling can prevent progression beyond the reversible stage [23,28,29]. EMG = electromyography, NCS = nerve conduction study.

The clinical implication is that the goal of management is not simply to treat voice change after it occurs but also to detect clinical neurapraxia early and prevent reversible physiological disturbances from progressing to structural nerve injury.

PREVENTION

Prevention begins before ablation. The operator should evaluate the relationship between the target lesion and the critical nerves, vessels, trachea, and esophagus. Lesions located at the posteromedial thyroid margin, central compartment, or postoperative bed require particular caution. The planned ablation zone should be evaluated three-dimensionally because heat may propagate beyond the visible electrode tip and the immediate hyperechoic zone.

The moving-shot technique, conservative ablation near critical structures, and hydrodissection are key preventive strategies [3,9,26,27]. Near the expected course of the RLN or the vagus nerve, undertreatment of the high-risk conceptual ablation unit may be preferable to aggressive marginal ablation, particularly for benign nodules. Hydrodissection with 5% dextrose in water (D5W) can create physical separation and may also provide thermal buffering [3,28,29,30,31].

Frequent intraprocedural voice assessment is the most practical and widely available method for real-time nerve-function monitoring [13]. Before ablation, the patient’s baseline voice should be confirmed. During ablation, voice can be monitored both through ongoing conversational speech and through simple, reproducible vocal tasks performed at regular intervals, for example, sustaining a vowel sound (such as /a/ or /i/), counting numbers, or repeating a short standardized phrase; the specific task is necessarily language-dependent and can be adapted to the patient. Rather than being tied to a fixed time interval, assessment is best integrated into the procedural workflow, performed between ablation cycles and at each electrode repositioning during the moving-shot technique, and near-continuously when ablating near high-risk regions such as the tracheoesophageal groove, posteromedial thyroid capsule, or expected vagus nerve course. In practice, this assessment is mainly subjective and relies on both patient-reported symptoms and operator recognition of new hoarseness, weak or breathy voice, reduced pitch control, vocal fatigue, cough, or aspiration. To preserve this monitoring capability, deep sedation or general anesthesia should be avoided when feasible because both preclude real-time voice assessment. To the best of our knowledge, no standardized real-time intraprocedural voice grading system has been specifically validated for thyroid RFA. Continuous laryngeal neuromonitoring has been reported as an objective adjunct that may provide real-time evidence of nerve integrity during RFA; however, it has not yet been standardized for routine thyroid RFA and is not widely available [32]. When a voice change is suspected, real-time transcutaneous laryngeal ultrasound may be used as a complementary and objective method to assess vocal-fold movement immediately during the procedure. This technique may be particularly useful for patients with subtle or equivocal voice changes [33]. However, adequate visualization may be limited in older male patients because of dense ossification of the thyroid cartilage. If a voice change is newly detected or remains clinically concerning, RFA should be stopped immediately, followed by prompt cold D5W injection [23,28,29] because early clinical neurapraxia may still be reversible if additional heat exposure and secondary edema are promptly controlled [23].

MANAGEMENT

Once voice changes are detected during intraprocedural assessment, an immediate response is required. RFA should be stopped immediately. The electrode should not be reactivated near the suspected injured nerve until voice function has recovered and the risk of additional injury has been reassessed [23]. If further ablation is necessary, treatment of adjacent high-risk areas should be minimized or avoided.

Cold D5W bolus injections are a key intervention for suspected thermal nerve injury [28,29]. For cooling alone after the cessation of RF energy delivery, normal saline can also help dissipate residual heat. However, D5W is preferred because additional RFA may occasionally be required after recovery of voice function, and D5W can continue to serve as a nonionic hydrodissection medium during subsequent RF energy delivery [31]. D5W may provide more sustained thermal separation than normal saline because it diffuses more slowly through the surrounding tissues. In practice, cold D5W should be promptly injected as a bolus into the perineural space along the expected course of the affected nerve, often near the tracheoesophageal groove, in patients with suspected RLN injury. The injection volume should not be fixed; rather, it should be titrated on the basis of voice recovery, sonographic separation from the ablation zone, and operator assessment to ensure adequate cooling.

After adequate cooling, local dexamethasone injections around the affected nerve may be administered to reduce perineural edema and inflammation. A short course of oral corticosteroid therapy, commonly administered for approximately 1 month by some operators, may also be considered to address persistent edema and delayed inflammatory change [24]. Although these approaches are biologically plausible and consistent with mechanism-based protocols, the independent benefits of corticosteroids beyond cold D5W treatment remain unclear. A practical management framework is summarized in Table 2.

Table 2. Stepwise management framework and evidence-level summary for voice changes during thyroid RFA.

Phase	Action	Pathophysiologic rationale	Clinical considerations	Representative evidence	Evidence level*	Evidence gaps
Prevention	Hydrodissection (D5W), moving-shot technique	Creates physical separation and thermal buffering; limits heat propagation toward nerve	Standard preventive practice	Guidelines [2,3,9,14], retrospective observational study (n = 39) [30], ex-vivo experimental studies [27,31]	Level 4	No controlled comparison isolating each technique
Detection	Frequent intraprocedural voice assessment; complementary real-time laryngeal ultrasound when voice changes are equivocal	Detects early clinical nerve dysfunction before progression	Particularly important near the danger triangle	Guidelines [14]	Level 5	No validated real-time intraprocedural grading system
Acute	Immediate cessation of ablation	Stops active heat generation	Residual heat may persist after energy delivery stops	Guidelines [2,14], mechanism-based reasoning [23]	Level 5	No direct RFA-specific outcome study
	Prompt cold D5W bolus injection	Dissipates residual heat, lowers perineural temperature, and creates separation from the ablation zone	Central active intervention with the strongest clinical support	Guideline [2], retrospective observational studies (n = 17 [28] and 19 [29])	Level 4	Single-center retrospective data; no RCT
Delayed	Local dexamethasone injection after adequate cooling	Reduces perineural edema and inflammation	May be considered after cooling as part of the management protocol	Mechanism-based reasoning [24]	Level 5	Independent benefit not isolated from cold D5W
	Optional short-course oral corticosteroid therapy	May reduce persistent edema and delayed inflammatory change	Duration adjusted by symptom recovery; individualized by comorbidity	Mechanism-based reasoning [24]	Level 5	No RCT; benefit beyond cold D5W alone unestablished

^*Evidence levels were assigned using the Oxford Centre for Evidence-Based Medicine 2011 Levels of Evidence [34], in which Level 1 represents the highest and Level 5 the lowest level of evidence.

RFA = radiofrequency ablation, D5W = 5% dextrose in water, RCT = randomized controlled trial

CLINICAL OUTCOMES AND RECOVERY PATTERNS

Most voice changes during or after thyroid RFA are transient; however, the recovery pattern may vary depending on the severity of the thermal injury, timing of recognition, and promptness of active cooling. Available clinical data suggest that early recognition and immediate perineural intervention, particularly cold D5W bolus injection, are associated with more rapid recovery than delayed or conservative management alone [28,29]. In a retrospective study, patients who received immediate cold D5W injection showed a significantly shorter mean recovery time than those who did not receive the injection (5.2 ± 17.3 days; range, 0–60 days vs. 54.6 ± 17.0 days; range, 3–90 days; P = 0.041). In that study, 83.3% (10/12) of the treated patients achieved recovery during the injection itself [28]. This finding was further supported by the results of a subsequent retrospective cohort study involving 291 RFA sessions, wherein intraprocedural nerve symptoms were managed with an immediate injection of cold D5W. In that cohort, immediate and complete symptom relief was achieved in 82.4% (14/17) of the patients during 78.9% (15/19) of the affected sessions after the injection of a mean volume of 41 mL (range, 3–260 mL) of cold D5W [29]. In contrast, patients with delayed symptom onset detected the following day, who did not receive cold D5W injection, required 1 to 3 months for complete resolution [29]. Permanent vocal cord dysfunction was not reported in any of these studies.

Several factors may be associated with a favorable recovery, including early intraprocedural detection through frequent voice assessment, immediate cessation of RF energy delivery after voice alteration, prompt cold D5W bolus injection into the perineural space, and avoidance of further ablation near the suspected injured nerve. In contrast, delayed or incomplete recovery may be more likely when voice changes are recognized after the procedure, when residual heat is not promptly dissipated, when the lesion is located in a high-risk area, such as the tracheoesophageal groove or postoperative bed, or when fibrotic tissue or adhesions limit effective hydrodissection and cooling. More severe structural nerve injuries such as axonotmesis or neurotmesis may also contribute to prolonged or incomplete recovery. However, because the current evidence base is mainly composed of small, single-center observational studies, direct quantitative comparisons between early and delayed interventions remain limited, and randomized controlled trials are lacking.

CLINICAL LIMITATIONS AND HIGH-RISK SITUATIONS

Appropriate patient selection is essential for safe thyroid RFA. RFA should be performed only when the target lesion is appropriate for minimally invasive treatment and when the expected benefits outweigh the risk of thermal injury to adjacent critical structures. Comprehensive indications for thyroid RFA are provided in existing society guidelines [2,3,14]. From a technical and safety perspective, treatment should not be attempted when the target area cannot be adequately visualized using ultrasound. Conditions such as contralateral vocal cord paralysis, pregnancy, pacemaker or implantable cardiac defibrillator use, clotting or bleeding disorders, and the inability to discontinue anticoagulant or antiplatelet therapy should be regarded as relative contraindications or situations requiring special caution [14].

From the perspective of nerve safety, particular caution is required for lesions located along the posteromedial thyroid capsule, tracheoesophageal groove, central compartment, postoperative bed, or the expected course of the vagus nerve. Recurrent tumors or nodules adjacent to dense postoperative fibrosis may be especially challenging because the nerve course may be distorted, and hydrodissection may be less effective. Particular caution is warranted in patients with pre-existing contralateral vocal cord palsy, since injury to the functioning nerve in these patients can result in bilateral vocal cord paralysis and airway compromise. Additional caution should be exercised in patients with severe anxiety or inability to cooperate during the procedure.

The reliability of intraprocedural voice monitoring may be reduced by poor baseline voice quality, deep sedation or general anesthesia, language barriers, hearing impairments, altered mental status, unconsciousness, severe anxiety, or inability to cooperate, all of which may impair timely detection or reporting of voice changes during ablation. When intraprocedural voice assessment is considered essential for nerve safety, deep sedation or general anesthesia should be avoided whenever feasible.

Management strategies may be less effective when the nerve injury has already progressed beyond reversible neurapraxia, recognition is delayed, residual heat is not adequately dissipated, or anatomical factors prevent effective hydrodissection and cooling. In addition, the independent benefits of corticosteroids remain uncertain, and local dexamethasone injections and oral corticosteroid therapy should be individualized on the basis of patient comorbidities such as poorly controlled diabetes, active infection, or peptic ulcer disease, in which corticosteroid use may be undesirable.

CONCLUSIONS

Voice changes during thyroid RFA are uncommon but clinically important and may reflect a dynamic thermal injury process involving direct heat exposure, residual heat conduction, acute edema, microvascular compromise, and delayed inflammation or fibrosis. The clinically important target is early clinical neurapraxia in which voice changes are detectable; however, the dominant mechanisms may remain reversible. Therefore, a practical management approach emphasizes frequent intraprocedural voice assessments, immediate ablation cessation, and prompt cold D5W bolus injections. After adequate cooling, local dexamethasone injections and a short course of oral corticosteroid therapy may be considered to reduce persistent edema and inflammation. However, their independent benefits remain uncertain. Understanding the acute and delayed mechanisms of thermal nerve injury may improve early recognition and intervention and reduce the risk of persistent voice dysfunction.