Prognostic Indicators of Unilateral Vocal Fold Paralysis

Objectives  To determine the possible prognostic indicators of unilateral vocal fold paralysis (UVFP) and survey the timing and values of preset laryngeal electromyography (LEMG) rules for UVFP.

Design  Cohort study with retrospective data analysis.

Setting  Voice clinic of a tertiary medical center.

Patients  Complete data for 45 patients diagnosed with idiopathic or iatrogenic UVFP. The LEMG was performed between 3 weeks and 6 months from the onset of symptoms.

Main Outcome Measure  At least 6 months after symptom onset and 3 months after LEMG.

Results  Thirteen subjects showed resolved vocal fold motion (29%), and 32 had persistent vocal fold paralysis (71%). According to the LEMG decision rules proposed by Munin et al in 2003, the predictive values for positive results, negative results, sensitivity, specificity, and accuracy of LEMG were 78.9%, 71.4%, 93.8%, 38.5%, and 77.8%, respectively. We found the false-positive rate to be as high as 50% if LEMG was performed less than 2 months after symptom onset, and only 7.7% if LEMG was performed at least 2 months after symptom onset. After excluding 14 LEMG data recorded at less than 2 months, the predictive values for positive results, negative results, sensitivity, specificity, and accuracy of LEMG were 92.3%, 60%, 92.3%, 60.0%, and 87.1%, respectively. The predictive values of positive results and accuracy significantly improved without compromising sensitivity.

Conclusion  This study confirms that LEMG is a clinically useful tool that can offer prognostic information for UVFP especially if it is done at least 2 months after symptom onset.

Unilateral vocal fold paralysis (UVFP) is a common disorder seen in the practice of otolaryngology. Incomplete vocal fold adduction in UVFP may cause formation of a constant glottal gap, which is usually associated with hoarseness of voice and aspiration during swallowing. The question most frequently asked by patients with UVFP is whether laryngeal function will return to normal. Predicting the outcome of UVFP is also important for physicians in making decisions about methods and timing of surgical intervention.

The impairment of laryngeal nerve function can come from heterogeneous conditions. According to Yamada et al,¹ idiopathic paralysis was the most frequent, and iatrogenic paralysis caused by thyroid surgery and endotracheal intubation came next. Havas et al² found that the leading causes of UVFP were iatrogenic and idiopathic. Data for spontaneous recovery of UVFP based on cause have been reported in some articles,^1,3 but statistical analysis of prognosis from these 2 common causes is still lacking.

Laryngeal electromyography (LEMG) was introduced in 1944 by Weddel and Pattle⁴ and advanced substantially in the 1950s by Faaborg-Andersen,⁵ Buchthal,⁶ and others. Information about spontaneous activity, motor unit morphologic characteristics, and motor unit recruitment obtained from LEMG has been used by several authors when evaluating nerve damage in UVFP and in making a prognosis with varying sensitivity and specificity.^7-12 Because there are no standardized guidelines, the clinical role of LEMG differs greatly between medical centers.¹³ Therefore, it is necessary to collect more information to understand the feasibility of previously proposed LEMG rules in predicting the prognosis of UVFP.

Our purpose in this study was to determine which factor or combination of factors, including patient sex and age, laterality of UVFP, cause, symptom duration, or LEMG findings, are most helpful in determining prognosis of UVFP. Furthermore, the relationship between timing and predictive value of LEMG was reviewed. Relevant literature is also reviewed and discussed.

After institutional review board approval, we retrospectively reviewed the medical records of 53 patients with UVFP who received LEMG from November 2004 to March 2007 within a period of between 3 weeks and 6 months after the onset of symptoms. The outcome measurement was done with flexible laryngoscope or videostroboscope at least 3 months after LEMG and at least 6 months after the onset of symptoms. For example, if a patient underwent LEMG 2 months after symptom onset, the outcome measurement was performed 4 months after LEMG (6 months after symptom onset). If a patient underwent LEMG 5 months after symptom onset, the outcome measurement was then performed 3 months after LEMG (8 months after symptom onset). These criteria allowed at least 3 months (range, 3-5 months) of follow-up to observe the predictive ability of LEMG in this time window (Figure).

After 3 patients with untreated skull base tumor, lung cancer, and esophageal cancer with mediastinal lymph node recurrence were excluded from the study, the causes of UVFP for the remaining 50 patients were either idiopathic (including possible viral infection) or iatrogenic (including skull base, neck, or chest surgery or endotracheal intubation). Five patients were lost to follow-up, so the data of 45 patients were included for analysis.

The office-based LEMG was performed in all cases by 1 of us (C.-C.W.), a laryngologist familiar with the anatomy of intrinsic laryngeal muscles. The LEMG findings were all interpreted by 1 of us (M.-H.C.), a neurologist with 15 years’ experience in electrodiagnostic medicine.

All LEMG examinations were performed by the same machine (Cadwell Sierra 6200A; Cadwell Laboratories Inc, Kennewick, Washington) using a computer-based electrodiagnostic system. Signals of LEMG were monitored on a computer screen and by means of a speaker simultaneously. Low-frequency filter settings were set at 10 Hz, and high-frequency filter settings were set at 10 000 Hz. Motor unit recruitment tracings were recorded with sweep speeds at 10 milliseconds per division using a gain of 200 μV per division. A 37-mm monopolar electrode was used in conjunction with both a surface disk reference electrode and ground electrode, which were affixed to superior cervical and clavicle levels respectively.

The procedure has been described in detail elsewhere.¹⁴ Briefly, the cricothyroid (CT) muscle and thyroarytenoid (TA) muscle were both approached transcutaneously without local anesthesia, while only the signals representing recurrent laryngeal nerve function obtained from the TA muscle were used to determine prognosis of UVFP. The TA muscle was approached by inserting a needle through the cricothyroid ligament approximately 0.5 cm from the midline. The needle was then angled superiorly 30° to 45° to an approximate depth of 1 to 2 cm. The position of the needle was validated by asking the patient to repeat a sustained vowel /i/, thus causing a simultaneous sharp and sustained increase in EMG activity. Activation of EMG was significantly lower during expiration and almost nonexistent during inspiration or neck flexion.

The 3 categories of LEMG parameters were (1) spontaneous activity, (2) morphologic characteristics of motor unit potentials (MUPs), and (3) recruitment pattern. First we recorded the MUP morphologic characteristics by asking the patient to phonate the vowel /i/ with low intensity to see if there was any polyphasic MUP (p-MUP) (>4 phases) in minimal voluntary muscle contraction. Then we asked patients to phonate the vowel /i/ with maximum intensity to see the motor unit recruitment in maximum voluntary contraction. The neurologist compared the recruitment pattern of the paralyzed side to the normal side and estimated the percentage of recruitment reduction (RR%). Based on his experience, less than 20% reduction was “nearly normal.” Finally, we asked the patient to keep silent to see if there was any spontaneous activity such as fibrillation potentials (FIBs) or positive sharp waves (PSWs). The data of FIBs, PSWs, and p-MUPs were dichotomized. Since we used monopolar electrodes and transcutaneous puncture techniques of the TA muscle, as were used in the pioneering study by Munin et al,¹² we also followed the steps outlined by those authors to determine overall prognosis: good (negative test result) or poor (positive test). In the earlier study, normal or nearly normal motor unit recruitment and absence of spontaneous activities such as FIBs or PSWs indicated a good prognosis. In our study, if the RR% was less than 20%, we defined the recruitment reduction pattern (RRp) as nearly normal.

All statistical analyses were performed with SPSS software, version 10.0.7 (SPSS Inc, Chicago, Illinois). The outcome measurements of vocal fold motion were dichotomized into persistent UVFP (disease) or resolved vocal fold motion (normal). The relationship of categorical data including sex, laterality, and cause of the UVFP, FIBs, PSWs, p-MUPs, RRp, and LEMG to give outcome measurements of vocal fold motion were analyzed using χ² analyses. Mann-Whitney tests were performed for continuous data such as age, duration after symptom onset, and RR%. P < .05 was considered significant.

We used definitions first outlined by Sittel et al¹¹: (1) The predictive value for a positive test result (positive predictive value [PPV]) was calculated by dividing the number of patients who had positive results and disease outcome by the total number of patients with positive results. This calculation determines the percentage of patients with positive test results (poor prognosis) who actually had a disease outcome (persistent UVFP). (2) Predictive value for a negative result (negative predictive value [NPV]) was calculated by dividing the number of patients who had a negative test result and normal outcome by the total number of patients with a negative test result. This calculation determines the percentage of patients with negative test results (good prognosis) who actually had normal outcomes (resolved vocal fold motion). (3) Sensitivity was calculated by dividing the number of patients who had a positive test result and disease outcome by the total number of patients with disease outcome. (4) Specificity was calculated by dividing the number of patients who had a negative test result and normal outcome by the total number of patients with a normal outcome. And (5) accuracy was calculated by dividing the number of patients with true-positive test results and true-negative test results by the total number of patients.

Furthermore, the LEMG timing after symptom onset resulting in true-positive results and false-positive results was analyzed by the Mann-Whitney test. We could therefore survey the possible influence of LEMG timing on the accuracy of examination results.

Of our 45 subjects, 32 had persistent UVFP (disease outcome) (71%). Thirteen patients had resolved vocal fold motion (normal outcome) (29%), and all of these patients regained a wide range of vocal fold abduction from previously totally immobile diseased vocal folds. Although in 7 patients we still could observe minimally sluggish movement compared with the healthy side intermittently during the inspiration-phonation cycle, they all showed good glottic closure and had a perceptually normal voice. The detailed data of 45 patients are summarized in Table 1.

The statistical analysis for dichotomized data and continuous data is summarized in Table 2 and Table 3, respectively. For variables other than LEMG findings, sex, laterality, and cause were all not significantly correlated with outcome (2-sided Fisher exact test, P = .74, P > .99, and P = .46, respectively) (Table 2). No significant age differences existed between the 2 groups with different outcomes (Mann-Whitney test, P = .13) (Table 3). The mean (SD) symptom duration before LEMG was 2.60 (1.34) months (range, 1-5 months). The mean (SD) symptom duration for 32 patients with persistent UVFP was 2.94 (1.27) months, significantly longer than the 1.77 (1.17) months for the 13 patients in the recovered group (Mann-Whitney test, P = .01) (Table 3).

In LEMG data, FIBs, PSWs, and p-MUPs were not significantly correlated with outcome (2-sided Fisher exact test, P = .21, P = .62, and P = .67, respectively) (Table 2). The mean (SD) RR% was approximately 57% (31.03%) (range, 0%-100%). Although the mean (SD) RR% for 32 patients with persistent UVFP was 62.81% (28.34%), larger than the 42.69% (33.83%) of the recovered group, the difference was not statistically significant (Mann-Whitney test, P = .06) (Table 3). If we dichotomized RRp into nearly normal and obvious reduction (reduction of more than 20%), RRp had a significant correlation with outcome (2-sided Fisher exact test, P = .03) (Table 2).

Taking FIBs into consideration, as proposed by Munin et al,¹² we found that the overall LEMG result was significantly correlated with outcome (2-sided Fisher exact test, P = .02) (Table 2). Thirty of 38 patients with a positive test result had persistent UVFP (Table 2), while PPV was 78.9%. Five of 7 patients with a negative test result had resolved function, and NPV was 71.4%. Thirty of 32 patients with persistent UVFP showed a positive LEMG result, and sensitivity was 93.8%. Five of 13 patients with resolved function had negative LEMG results, and specificity was 38.5%. The outcome of 35 patients could be accurately predicted by LEMG, while the accuracy of LEMG was 77.8%.

There were 8 patients with false-positive results and 30 patients with true-positive results. The mean (SD) duration of symptoms before LEMG for patients with false-positive results was 1.50 (1.07) months (95% confidence interval [CI], 0.61 - 2.39 months), significantly shorter than the 2.90 (1.30) months (95% CI, 2.42-3.38 months) for patients with true-positive results (Mann-Whitney test, P = .01). There was no significant difference in duration from symptom onset to LEMG between patients with false-negative results (3.50 [0.71] months; 95% CI, −2.85 to 9.85 months) and true-negative results (2.20 [1.30] months; 95% CI, 0.58 to 3.82) (Mann-Whitney test, P = .38). Six of 8 patients with false-positive results received LEMG 1 month after symptom onset (Table 1). It seems that early LEMG (before 2 months' symptom duration) led to a higher rate of false-positive results (Table 4). Because in clinical application the LEMG false-positive rate must be reduced to avoid inappropriate surgical intervention, we excluded 14 patients who received LEMG 1 month after symptom onset to see how the value of LEMG in determining prognosis could be improved.

As was the case for the patients with shorter-duration symptoms, the statistical analysis for dichotomized data and continuous data is summarized in Table 2 and Table 3, respectively. For variables except for LEMG findings, sex, laterality, and cause were not significantly correlated with outcome (2-sided Fisher exact test, P > .99, P > .99, and P = .56, respectively). No significant differences for age existed between the 2 groups with different outcomes (Mann-Whitney test, P = .26) (Table 3). Interestingly, there were also no significant differences in symptom duration before LEMG between the 2 groups with different outcomes (Mann-Whitney test, P = .48) (Table 3).

In LEMG data, PSWs and p-MUPs were not significantly correlated with outcome (2-sided Fisher exact test, P > .99 and P > .99, respectively), but FIBs were significantly correlated with outcome (2-sided Fisher exact test, P = .02). As in the group with symptoms of shorter duration, RRp was significantly correlated with outcome (2-sided Fisher exact test, P = .04) (Table 2). The mean (SD) RR% (continuous data) in patients with persistent UVFP was 63.46% (29.39%), significantly larger than that 24.00% (25.10%) of the recovered group. The difference was statistically significant (Mann-Whitney test, P = .01) (Table 3). The overall LEMG results proposed by Munin et al¹² were significantly correlated with outcome (2-sided Fisher exact test, P = .02) (Table 2). Based on data reported in the LEMG rules row of Table 2, we found that the PPV, NPV, sensitivity, specificity, and accuracy were 92.3%, 60.0%, 92.3%, 60.0%, and 87.1%, respectively.

The decision-making algorithm for the management of patients with UVFP has always been complex. In this algorithm, severity of symptoms, the vocal needs of the patient, and prognostic data are usually most important. Accurate prognostic information is important for the physician to determine the type and timing of surgical intervention to be used.

Data on spontaneous recovery of UVFP based on cause have been reported.^1,3 Many researchers have also described the use of LEMG for determining prognosis of UVFP.^7-12 However, the data characteristics differed between studies on cause of UVFP (including patient characteristics), diagnostic criteria used with LEMG, and the electrodes and techniques applied. In addition, the outcome measurement end times varied or were unspecified in some studies (Table 5).^7,8,10-12 Therefore, in the present study, we limited LEMG examinations to patients with idiopathic and iatrogenic causes; performed the LEMG between 3 weeks and 6 months after symptom onset; and measured outcomes at least 6 months after symptom onset and at least 3 months after LEMG testing. By using focused LEMG criteria and statistical significance testing, our results will be more helpful for physicians in determining prognosis and planning treatment.

It is not surprising that sex, age, and laterality of UVFP had no impact on outcome. To our knowledge, no study shows these factors to have a correlation to prognosis. However, whether UVFP with different causes gives a different prognosis needs further investigation because different studies^2,3,15 have used different classifications or definitions of UVFP cause. Numerous studies have documented multiple causes of laryngeal paralysis, often without agreement on the most common cause.^1-3,15-18 According to Yamada et al,¹ idiopathic paralysis was the most common, followed by iatrogenic paralysis caused by thyroid surgery and endotracheal intubations. Other surgical procedures from skull base to chest, such as skull base surgery, cervical spine surgery, vascular surgery of the carotid artery and the aortic arch, and esophageal surgery, also endanger the integrity of the vagus and recurrent laryngeal nerve function.¹

Because there are so many possible causes of UVFP, we classified our patients into those with idiopathic, iatrogenic, or other defined causes, similar to the classification system used by Havas et al.² But patients with possible viral infection were also included in the idiopathic group. In some studies,^3,16-18 UVFP was most commonly caused by other defined causes such as malignancies, particularly lung cancer. But because patients with vocal fold paralysis caused by tumor compression are not expected to recover before complete treatment of the neoplasm, we excluded patients with UVFP in this etiologic category from the present study (3 patients).

Yamada et al¹ found that paralysis caused by endotracheal intubation showed the best prognosis, followed by idiopathic paralysis. But because of the difficulty in differentiating between the causes of postoperative UVFP, such as endotracheal tube compression or direct nerve injury during surgery, we categorized these 2 etiologic groups as iatrogenic.

We found no difference in prognosis between the idiopathic and iatrogenic groups in either category of symptom duration before LEMG.

Similar to Munin et al,¹² we found that symptom duration was significantly related to outcome in our group of 45 patients whose LEMG was performed at least 3 weeks after symptom onset (Table 3). In an animal study by Mu and Yang,¹⁹ recovery from recurrent laryngeal nerve palsy was complete within 3 months after nerve impairment if over half of the nerve fibers of the impaired nerve were maintained intact without degeneration. It is reasonable to assume that the longer symptoms persist, the more severe the initial nerve damage was, and therefore the prognosis should be worse.

In the present study, symptom duration also represented the timing of LEMG. Although Min et al⁹ and Munin et al¹² suggest that LEMG is most predictive if performed at approximately 6 to 7 weeks from the onset of symptoms, we found that if patients presented 2 months after symptom onset, the timing of LEMG had no correlation with outcome, which suggests that LEMG findings will be more important in determining prognosis of UVFP. According to data from our 45 patients who underwent LEMG at least 3 weeks after symptom onset, the tests of 8 patients with false-positive results (poor prognosis with resolved function) were performed a mean (SD) of 1.50 (1.07) months (95% CI, 0.61-2.39) after symptom onset, and the LEMG tests of 30 patients with true-positive results (poor prognosis with persistent vocal fold paralysis) were performed at 2.90 (1.30) months (95% CI, 2.42-3.38) after symptom onset. The difference was significant (Mann-Whitney test, P = .01). Therefore, it seems that LEMG tests performed sooner than 2 months after symptom onset led to more false-positive results (Table 4). Gupta and Bastian⁸ also observed this trend in their 2 patients. Therefore, we should conduct further study into whether LEMG performed at least 2 months after symptom onset could give better information in regard to predictive UVFP prognosis.

Different studies examine the use of different LEMG electrodes. Koufman et al²⁰ believe that monopolar electrodes reduce the chance of sampling error. In Ford's observation,²¹ monopolar electrodes could record a larger sample of muscle activity, and the amplitude values of the bipolar technique are limited by the small test area between the electrodes. In addition, single-fiber EMG requires more sophisticated equipment and is not often used in larynx.²² Compared with a concentric electrode, a monopolar needle is thinner, less expensive, and can record electrical potential in a 360° circumference. The concentric electrode only allows 180° of electrical sampling.²³ Therefore, we used monopolar electrodes for recording LEMG and used overall LEMG prognostic criteria similar to that proposed by Munin et al,¹² who also used monopolar techniques.

Spontaneous activity is the term used to describe electrical activity in severely denervated, unstable, resting muscles. It usually begins 2 to 3 weeks after denervation, which is why our patients received LEMG at least 3 weeks after symptom onset. Different kinds of abnormal spontaneous activity can be observed, such as FIBs and PSWs. The presence of spontaneous activity indicates a poor prognosis for recovery. However, FIBs and PSWs that represent ongoing axonal degeneration had no correlation to outcome in our group of 45 patients who underwent LEMG at least 3 weeks after symptom onset, nor in the study by Munin et al.¹² In Dedo's study²⁴ of the paralyzed larynx in humans and dogs, crushing the nerves resulted in fibrillations that occurred 7 to 14 days after injury. Some of these vocal folds went on to full recovery, thus implying that fibrillations do not necessarily indicate that vocal folds will be permanently paralyzed. In our study, 4 patients with FIBs (patients 3, 23, 29, and 36) and 2 patients with PSWs (patients 23 and 29) recovered function. They all underwent LEMG 1 month after symptom onset (Table 1). If we performed LEMG at least 2 months after symptom onset, FIBs became significantly correlated with outcome (Table 2). This finding again suggests that better timing might improve the value of using FIBs as a diagnostic tool.

Munin et al¹² suspected that decreased detection of FIBs in subjects with persistent vocal fold paralysis might be owing to the small number and size of muscle fibers per motor unit within the laryngeal muscles. In the present study, FIBs were identified in 53% of subjects with persistent vocal fold paralysis (17 of 32), and PSWs were only identified in 9% (3 of 32). Compared with PSWs, FIBs can be more easily detected because of their typical signal sound similar to raindrops falling on a tin roof. Therefore, when LEMG was performed at least 2 months after symptom onset, FIBs were significantly correlated with outcome, while PSWs remained uncorrelated, probably owing to difficulties in detection.

Configuration of the observed motor units is also an important parameter of LEMG to assess. Normal MUPs are triphasic and relatively narrow. In subacute nerve injuries in which some reinnervation has occurred, low-amplitude p-MUPs can be seen. Their presence indicates early reinnervation. As reinnervation continues after axonal loss injury, larger amplitude p-MUPs are noted.²³ Parnes and Satya-Murti⁷ found that the presence of p-MUPs indicates a better prognosis for recovery, with a false-positive rate of 21%. In the present study, p-MUPs were not significantly correlated with outcome in either symptom duration group (Table 2). This finding is similar to that of Munin et al¹² in that other factors such as synkinesis could lead to persistent vocal fold paralysis even if there is evidence of reinnervation.

Recruitment refers to the serial activation of motor units during increased voluntary muscle contraction and reflects the degree of innervation, which is a reflection of the number of active nerve fibers of a given muscle.¹⁴ Although the reduction of recruitment could be estimated by neurologists, we must recognize that objectively measuring phonation is somewhat difficult owing to the subject's difficulties in controlling the effort put into a specific phonation task. However, this measurement is the clinical technique we have adopted thus far. Under such circumstances, we recommended that LEMG data be interpreted by neurologists who are familiar with interpreting EMG. In our study results for the 45 patients with symptom duration of at least 3 weeks before LEMG, the RR% was not significantly different between patients with persistent vocal fold paralysis and patients with recovered function. This finding agreed with the results of Munin et al.¹² However, we hypothesized that the timing of LEMG was the main reason for negative correlation between RR% and outcome. In our results for the 31 patients receiving LEMG at least 2 months after symptom onset, the RR% of patients with persistent UVFP was significantly larger than that of patients with resolved function (Table 3).

While RR% is estimated from continuous data, it is not clinically convenient to use for predicting results. In the study by Munin et al,¹² RR% was dichotomized until absent or showed greatly decreased recruitment vs all other observed patterns. This dichotomization is significantly related to outcome. However, in the preset LEMG steps used by Munin et al, normal or nearly normal recruitment pattern without FIBs determined a good prognosis. Therefore, we dichotomized RRp into normal or nearly normal vs obviously reduced. This dichotomization was significantly related to outcome in both symptom duration groups in our study (Table 2), suggesting that RRp is another important element of LEMG for predicting prognosis of UVFP.

Munin et al¹² proposed that a normal or nearly normal recruitment pattern without FIBs determines a good prognosis, and overall LEMG results were significantly correlated with outcome in their study. This finding was supported by our study (Table 2). We now know that FIBs and RRp are both important prognostic indicators, especially if LEMG was performed at least 2 months after symptom onset. A combination of these 2 indicators should be a valuable clinical tool for predicting the prognosis of UVFP.

Studies have shown variable sensitivity, specificity, and predictive values for recovery of vocal fold function using data from LEMG recordings.^7-12,25 These differences might come from different techniques used in different studies. For example, most of the patients in the study by Parnes and Satya-Murti⁷ received LEMG 6 months after symptom onset, but for those described by Elez and Celik,¹⁰ the mean (SD) duration from symptom onset to LEMG was only 2.7 (1.6) months. Interestingly, the PPV was always better than the NPV (Table 6),^7,8,10-12 indicating that LEMG accuracy is better in predicting a positive test result. Most of the NPV findings were between 60% and 80%. Only Sittel et al¹¹ reported a very low value (12.8%). However, Sittel et al based LEMG prognosis on Seddon's classification of neurapraxia and axonotmesis,²⁶ and this classification is limited because there are no clinically useful laryngeal motor nerve conduction studies available.¹² Therefore, LEMG still has a moderate predictive value for a negative test result according to the most relevant literature. Synkinesis of abductor and adductor muscles, conformational changes within the larynx caused by contracture, or functional weakness caused by partial denervation have been mentioned as possible explanations for vocal fold paralysis in the presence of intact LEMG function.^9,23,27

Data from LEMG studies can be helpful in the management of UVFP. If patients are not willing to wait for the possibility of spontaneous recovery after a positive test result for a poor prognosis, the validity of more definite permanent surgery can then be supported to shorten the process of voice rehabilitation. Koufman et al²⁰ showed that decisions based on LEMG data altered the timing and or type of surgical procedure in 40% of their patients. From this point of view, an ideal test should have no correlation to timing of the test and should have a high PPV and high sensitivity, allowing us to correctly identify those patients who need surgery. In Table 6, we summarize findings that preset LEMG steps at least 2 months after symptom onset can fulfill this requirement. The PPV and sensitivity are both as high as 92.3%. The accuracy using preset rules is better than when using single elements of LEMG alone (Table 7).

There are limitations to this study. Because there are no standardized LEMG guidelines, the interpretation of LEMG signal was based on our experience in this study and all past studies. The effectiveness of the diagnosis technique needs further investigation. In addition, because of strict exclusion and inclusion criteria, the sample size was small. However, our study was still one of the largest series so far and could serve as an important resource (Table 6). Finally, within the study design, LEMG only predicts the results of UVFP for the next 3 to 5 months. We will collect more data to further investigate our hypothesis and support our conclusions from this study.

Data from LEMG findings, including the preset rules first set forth by Munin et al,¹² are significantly correlated with the outcome of UVFP. Normal or nearly normal motor unit recruitment and absence of spontaneous activities such as FIBs or PSWs determine a good prognosis; otherwise, the prognosis is poor. However, early LEMG tests gave higher rates of false-positive results. If LEMG was performed at least 2 months after symptom onset, the aforementioned rules give high sensitivity (92.3%), PPV (92.3%), and accuracy (87.1%). This reveals that LEMG, with the cooperation of an experienced laryngologist and neurologist, is a valuable tool in the workup of patients with UVFP.

Correspondence: Chen-Chi Wang, MD, Department of Otolaryngology, Taichung Veterans General Hospital, No. 160, Sec 3, Taichung Harbor Road, Taichung, Taiwan 40705 (ccwang@vghtc.gov.tw).

Submitted for Publication: October 12, 2006; final revision received July 8, 2007; accepted September 11, 2007.

Author Contributions: Dr C.-C. Wang had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: C.-C. Wang, Chang, C.-P. Wang, and Liu. Acquisition of data: C.-C. Wang and Chang. Analysis and interpretation of data: C.-C. Wang, Chang, C.-P. Wang, and Liu. Drafting of the manuscript: C.-C. Wang. Critical revision of the manuscript for important intellectual content: C.-C. Wang, Chang, C.-P. Wang, and Liu. Statistical analysis: C.-C. Wang. Administrative, technical, and material support: C.-C. Wang, Chang, C.-P. Wang, and Liu. Study supervision: C.-C. Wang and Chang.

Financial Disclosure: None reported.

Additional Contributions: The Biostatistics Task Force of Taichung Veterans General Hospital helped with the statistical analysis of this report.