Botulinum Toxin Type A for Treating Voice Tremor FREE

Essential tremor is a very common movement disorder that has an increased incidence with age.^{1 - 2} Essential tremor most often affects the hands and head, and although voice tremor is not as common, it can be very embarrassing and disabling. Voice tremor can be isolated or associated with tremor of other body parts, and it must be distinguished from spasmodic dysphonia, a dystonic disorder of the larynx.^{3 - 4}

Treatments for essential tremor include oral medications and surgery. Oral medications are partially beneficial for hand and head tremor but not very effective for voice tremor.^{4 - 5} Surgical intervention, such as thalamotomy and thalamic stimulator implan- tation, is recommended for medication-refractory patients.^{6 - 7} However, these treatments are primarily recommended for hand tremor, with some reports showing benefit for voice tremor.⁸ Botulinum toxin type A injections (BoNT-A) (Botox; Allergan, Irvine, Calif) are considered the treatment of choice for laryngeal dystonia,⁹ and they are beneficial for hand and head tremors.^{10 - 14} Three small studies^{15 - 17} and one case report¹⁸ have found some benefit of BoNT-A for voice tremor. We present a randomized dose, randomized objective video, and randomized objective audio assessment trial of BoNT-A injections for the treatment of voice tremor.

Thirteen consecutive subjects with voice tremor without spasmodic dysphonia or head, mouth, jaw, or facial tremor, who were willing to receive BoNT-A injections, were entered into this study between 1993 and 2000, after they signed written informed consent forms approved by the Mayo Clinic Institutional Review Board. Patients with head, mouth, and jaw tremors were excluded because these tremors can result in tremulous-sounding voices. Other exclusion criteria included age less than 18 years, known neuromuscular junction disorders, concurrent treatment with dantrolene sodium or aminoglycosides, concurrent signs of parkinsonism, and previous treatment with BoNT-A. The presence of limb or trunk tremor did not exclude patients. Women were excluded if they were pregnant, nursing, or had an unprotected risk of pregnancy.

At baseline, all patients completed the tremor severity scale (0 [no tremor] to 4 [severe tremor])¹⁹ and a functional disability scale (0 [none] to 4 [severe, unintelligible speech]); the patients were videotaped and underwent videolaryngostroboscopy. Tremor severity was also rated by the physician. Vocal tasks were recorded onto reel-to-reel audiotapes via headset microphones. Videotape and audiotape segments included the patient reading a standardized paragraph as well as performing a sustained “ah.” Patients taking concurrent tremor medications were instructed not to change the dose or frequency of treatment during the duration of the study; also, new medications could not be given.

Botulinum toxin type A was reconstituted with normal saline at a concentration of 2.5 U/0.1 mL and was injected percutaneously under electromyographic guidance with a 1.0-in 27-gauge needle through the cricothyroid membrane.²⁰ At the time of injection, patients were randomized to receive 1.25 U, 2.5 U, or 3.75 U into each vocal cord. These doses were chosen based on previous reports^{12 - 13} and because 2.5 U was the starting dose used in our office at that time for patients with adductor spasmodic dysphonia. Randomization was accomplished by the treating physician randomly choosing the dose from 15 identical treatment kits created prior to study initiation. Neither the patient nor any of the individuals assessing treatment response were made aware of the treatment allocation (assessments were performed by blinded videotape or acoustical review). Patients were reassessed at weeks 2, 4, and 6. They were asked to rate their vocal tremor response to BoNT-A injection on a scale from +4 (marked improvement) to –4 (marked worsening). Adverse events were recorded at each visit and rated as mild, moderate, or severe.

All videotape segments from all the patients (at baseline and the 3 follow-up visits) were randomized; then they were dubbed onto a master tape by an independent video specialist using a randomization list provided by the statistician. These segments were then reviewed and independently rated by 3 blinded investigators. A total of 52 video segments were rated (13 patients, 4 visits per patient). Sample size was determined prior to beginning the study, but the study was terminated after we enrolled only 13 of the planned 15 subjects because of very slow enrollment. The decision to terminate the study was made prior to unblinding the data.

The modulographic method was used for acoustic assessment of tremor.²¹ The analysis produces low-frequency spectrograms of modulations in fundamental frequency (f₀) and intensity parameters. There were a total of 32 measures. For each parameter, there are ordinarily 3 domains of analysis that use different bandwidths of Fourier analysis: (1) wow, from 0.1- to 2-Hz modulations; (2) tremor, from 2- to 10-Hz modulations; and (3) flutter, modulations from 10 to 20 Hz. The modulogram is preferable to the measures used in previous tremor treatment studies^{18 ,22} because a wide range of specific modulation frequencies can be examined distinctly.

Audio recordings were digitized and transferred to a blinded laboratory group for acoustic analysis to assess modulations of f₀ in the tremor domain (2 to 10 Hz). Files were first submitted to amplitude and pitch determination algorithms in the CSpeechSP environment²³ to extract intensity and f₀. Some phonations were essentially aphonic and hence unanalyzable for f₀. Severely tremulous phonations exhibited frequent voice breaks. To construct a continuous record, simple interpolations were used across gaps in f₀ but only when such gaps spanned less than a complete tremor cycle. Special accommodations were also made for very short (2-to 5-second) phonations by omitting wow analysis (which normally requires more than 5 seconds of continuous phonation). This was acceptable because the primary interest in this population was in the tremor domain of 10 to 20 Hz. The extracted acoustic parameters of these phonations were submitted to the modulogram procedure.²¹ Outcomes are summarized by calculating an instability metric as the product of a frequency-specific modulation (percent variation around mean level) times the proportion of phonation time affected by that modulation.

The modulogram supports these selections by displaying a histogram of cumulated magnitudes for each instability tape; typically, the mode of the histogram represents the chief instability in that domain. In this application, the primary interest was in the instability of f₀ in the tremor domain. However, apparently unrelated wow modulations were often observed to leak into the tremor panel. Hand inspection of each modulogram was performed to check for this artifact. Whenever the dominant instability within a tremor display was attributable to leakage from a lower frequency wow and a higher tremor-domain frequency produced a clear mode, that tremor frequency was instead reported. This technique guaranteed that the pretreatment and posttreatment samples were being compared for the same type of distinctly tremulous component.

Changes from baseline were assessed by using the paired t test for the primary outcome measure (patient tremor rating) and secondary outcome measures (patient disability rating, videotape ratings, BoNT-A response rating, and acoustic measures). The dose-response relationships were quantified by using the slope of the linear regression model for the change from baseline to dose. A 95% confidence interval (CI) was calculated for the slope of each regression line to assess the margin of error for the dose-response relationship. The pairwise correlations among the video scores from the 3 raters were quantified using the Pearson correlation coefficient, and the overall interrater reliability was calculated using the intraclass correlation coefficient. All P values were for 2-sided tests.

The mean age was 73 years (range, 54-81 years), with 11 women and 2 men. Five of the 13 patients had concomitant mild-moderate bilateral postural or terminal tremors, and 1 had unilateral arm tremor. Concomitant tremor medications included propranolol (n = 2), primidone (n = 1), and clonazepam (n = 2). Five patients received BoNT-A 1.25 U/cord, 5 received 2.5 U/cord, and 3 received 3.75 U/cord. All patients completed the entire study.

All patients at all dose levels perceived some effect from the BoNT-A injection, mean time to onset being 2.3 days (range, 1-7 days). Mean patient ratings of the tremor severity scale for all dose groups combined (baseline, 3) improved 1.4 points at week 2, 1.6 points at week 4, and 1.7 points at week 6 (Table). The scores improved in all 3 dose groups at all 3 time points.

Patient-rated functional disability (baseline, 2.31) improved 0.2 points at week 2, 1.1 points at week 4, and 1.5 points at week 6, for the combined group data (Table). Patient-rated effect from the BoNT-A injection showed benefit (Table).

The mean ± SD scores for all 52 video segments rated were 2.15 ± 1.07 for rater 1, 2.46 ± 0.78 for rater 2, and 2.62 ± 0.51 for rater 3. The correlation among reviewers was excellent, with a correlation coefficient greater than 0.70 and an interrater reliability of 67% (95% CI, 30 to 88). The independent randomized tremor ratings revealed significant improvement for the combined data at all time points (Table).

The most sensitive acoustic measure was the log-transformed percent modulation. These results (in original modulation units) indicate that at the main tremor frequencies (usually 5-6 Hz), patients exhibited f₀ modulations on the order of 25% of their mean levels. These levels decreased to about half that level at 2 weeks, to about 7% modulation depth at 4 weeks, and to about 10% modulation depth at 6 weeks. Average levels observed in control subjects (previously published data), however, were on the order of 1%.²¹

Figure 1 and Figure 2 display modulograms of one patient’s phonations before and 2 weeks after a 3.75-U injection. Figure 1 omits wow panels to accommodate the short phonation. High-magnitude modulations are evident in the tremor and flutter panels, but the most relevant of these is a 5- to 6-Hz tremor of f₀ (reflected in a clear histogram mode to the right of the f₀ panel). Figure 2 illustrates that in this patient the treatment was very effective 2 weeks postinjection. The f₀ tremor is still evident at 5 to 6 Hz, but relatively high-magnitude instabilities remain at lower frequencies.

The sample was too small to assess a dose-response relationship within the range of doses studied. At week 6, tremor severity scale scores improved by 0.05 points/unit of BoNT-A (95% CI, −0.85 to 0.98), functional disability scale scores changed −0.41 points/unit (95% CI, −1.1 to 0.29), effect of injection scores increased 0.65 points/unit (95% CI, −0.29 to 1.6), and videotape scores changed −0.19 points/unit (95% CI, −0.62 to 0.25).

The main adverse effects at all doses were breathiness and dysphagia. At week 2, 11 of 13 patients had breathiness, and 3 of 13 had dysphagia; all but 1 improved by week 6. Patients rated these adverse effects as mild-severe in intensity, and they were most severe at week 2. Two patients experienced hoarseness, 1 excess phlegm, 1 fatigue, and 1 a cough.

Voice tremor can be a very disabling disorder that responds poorly to oral medications but well to botulinum toxin injections. While the severity of voice tremor averaged 3 of 4 in this study, many of the more severely affected patients with voice tremor were excluded because of concurrent head or jaw tremor, so results cannot be generalized to these patients. Improvement was significant for patient-rated tremor severity and functional disability and for tremor severity as rated by blinded reviewers. Improvement in voice tremor was of a similar degree as that found in patients with adductor spasmodic dysphonia⁹ ; this could extend the indications for botulinum toxin injections in terms of tremor disorders.¹⁴ Because this study did not include a placebo arm, we cannot entirely rule out the possibility that improvement was due to regression to the mean or the physical nature of the injection into the vocal cord.

This study also assessed the interrater reliability for videotape reviews of vocal tremor severity using the tremor severity scale.¹⁹ The correlation and interrater reliability among all 3 raters was excellent. Videotape reviews are therefore an excellent means for determining treatment effects because they can be randomized and independently rated. While we cannot rule out some placebo effect on the subjective ratings, these videotape results support a definite beneficial effect of the botulinum toxin injections.

This was also the first study to use a frequency-specific acoustic measure of tremor severity. The acoustic measures revealed that some voice modulation remains after the treatment, but the tremor-specific frequencies were often affected most directly by the treatment. As described in “Methods,” lower-frequency modulations can dominate an assessment of overall f₀ instability unless this assessment is focused on a specific frequency. The results of our study show that BoNT-A is a promising treatment for voice tremor, and further controlled studies should be pursued.

Correspondence: Charles H. Adler, MD, PhD, Parkinson’s Disease and Movement Disorders Center, Department of Neurology, Mayo Clinic Scottsdale, 13400 E Shea Blvd, Scottsdale, AZ 85259 (cadler@mayo.edu).

Accepted for Publication: February 26, 2004.

Author Contributions:Study concept and design: Adler, Bansberg, Ramig, and Edwards. Acquisition of data: Adler, Bansberg, Witt, Edwards, Krein-Jones, and Caviness. Analysis and interpretation of data: Adler, Bansberg, Hentz, Ramig, and Buder. Drafting of the manuscript: Adler, Hentz, Buder, and Edwards. Critical revision of the manuscript for important intellectual content: Adler, Bansberg, Hentz, Ramig, Buder, Witt, Krein-Jones, and Caviness. Statistical expertise: Hentz and Buder. Obtained funding: Adler. Administrative, technical, and material support: Bansberg, Hentz, and Krein-Jones. Study supervision: Adler, Bansberg, and Buder.

Funding/Support: This study was funded by a grant from the Mayo Foundation, Rochester, Minn.

Acknowledgment: We thank Bruce Pope for creating the randomized videotape; Alan Zinsmeister, PhD, for initial assistance with study design; and Jose Hernandez and Melissa Garafalo for assistance with data analysis. This article is dedicated to the memory of Brian Edwards.