ORIGINAL ARTICLE

ASSESSMENT OF ALARYNGEAL SPEECH USING A SOUND-PRODUCING VOICE PROSTHESIS IN RELATION TO SEX AND PHARYNGOESOPHAGEAL SEGMENT TONICITY

M. van der Torn, MD,1 C. D. L. van Gogh, MD,1 I. M. Verdonck-de Leeuw, PhD,1 J. M. Festen, PhD,1 G. J. Verkerke, PhD,2 H. F. Mahieu, PhD1

1 Department of Otolaryngology/Head & Neck Surgery, Vrije Universiteit Medical Center, P. O. Box 7057, 1007 MB Amsterdam, The Netherlands. E-mail: [email protected] 2 Department of Biomedical Engineering, University of Groningen, Groningen, The Netherlands

Accepted 23 August 2005 Published online 9 February 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hed.20355

Keywords: laryngectomy; artificial larynx; voice prosthesis; Abstract: Background. A pneumatic artificial sound source alaryngeal speech; voice incorporated in a regular tracheoesophageal shunt valve may improve alaryngeal voice quality. Methods. In 20 laryngectomees categorized for sex and pharyngoesophageal segment tonicity, a prototype sound-pro- Total laryngectomy, as surgical treatment for ducing voice prosthesis (SPVP) is evaluated for a brief period and compared with their regular tracheoesophageal shunt locally advanced laryngeal tumors, interferes speech. with all functions of the larynx (ie, phonation, res- Results. Perceptual voice evaluation by an expert listener piration, deglutition, and indirectly olfaction). and acoustical analysis demonstrate a uniform rise of vocal pitch The most persistent problems of patients with lar- when using the SPVP. Female laryngectomees with an atonic yngectomies are related to the loss of voice.1,2 It is pharyngoesophageal segment gain vocal strength with the SPVP. Exerted tracheal pressure and airflow rate are equivalent generally acknowledged that rapid and effective to those required for regular tracheoesophageal shunt valves. voice restoration is critical to the successful reduc- However, communicative suitability and speech intelligibility tion of psychological, social, and economic set- deteriorate by the SPVP for most patients. Tracheal phlegm backs induced by postlaryngectomy aphonia.3,4 clogging the SPVP is a hindrance for most patients. Since laryngectomy has been performed, numer- Conclusions. The SPVP raises vocal pitch. Female laryngec- tomees with an atonic or severely hypotonic pharyngoesopha- ous attempts have been made with varying suc- geal segment can benefit from a stronger voice with the SPVP. cess to obtain or improve the postlaryngectomy VC 2006 Wiley Periodicals, Inc. Head Neck 28: 400–412, 2006 voice by creating a pneumatic artificial source of voice production.5 Esophageal injection voice and electrolarynx devices, however, formed the stand- Correspondence to: H. F. Mahieu ard approaches to alaryngeal voice rehabilitation Contract grant sponsor: This study was supported by grant GGN until tracheoesophageal (TE) puncture incorpo- 55.3712 from the Dutch Technology Foundation STW. rating a silicone shunt valve prosthesis6 evolved VC 2006 Wiley Periodicals, Inc. worldwide as an established technique for post-

400 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 Table 1. Clinical data for the study group; female and male, in order of pharyngoesophageal segment tonicity.

Months PE segment Patient Age Sex postoperative Myotomy Postoperative RT tonus F1 52 F 54 Yes Yes Atonic F2 65 F 31 Yes Yes Atonic F3 68 F 46 None Yes Atonic F4 76 F 47 Yes Yes Atonic F5 69 F 102 Yes None Severely hypotonic F6 77 F 91 None None Slightly hypotonic F7 53 F 38 Yes Yes Slightly hypertonic F8 76 F 49 Yes None Severely hypertonic F9 79 F 114 None None Severely hypertonic M1 52 M 32 Yes None Slightly hypotonic M2 69 M 31 Yes None Slightly hypotonic M3 71 M 115 Yes None Slightly hypotonic M4 74 M 41 Yes None Slightly hypotonic M5 76 M 31 None None Slightly hypotonic M6 56 M 52 Yes None Normotonic M7 63 M 20 None Yes Normotonic M8 76 M 29 Yes None Slightly hypertonic M9 77 M 65 Yes Yes Severely hypertonic M10 78 M 44 Yes Yes Severely hypertonic M11 64 M 14 Yes None Spasmodic

Abbreviations: RT, radiotherapy; PE, pharyngoesophageal; F, female; M, male. laryngectomy voice restoration. The advantages pneumatic sound source to be incorporated in a of this method over esophageal injection voice are regular TE shunt valve was designed.16 Preceding louder phonation and better intelligibility7;in in vivo studies17,18 proved the feasibility of this addition, it usually enables quick and trouble-free voice production and provided us with directions voice acquisition, higher speech rate, and more for the development of an updated series of sound- sustained phrasing because of a larger available producing voice prostheses (SPVPs). These were air reservoir.8,9 evaluated in vitro by aero-acoustic measurements The term ‘‘voice prosthesis’’ is widely used in and detailed high-speed photographic sequences the literature when referring to TE shunt valves, to establish the most promising sound source con- although these devices do not actually produce figuration for clinical use.19 This article describes sound. In both esophageal injection voice and TE the results with the selected SPVP group of 20 shunt voice, the passage of air through the pha- laryngectomized patients. Our objective was to ryngoesophageal (PE) segment sets the closely determine for which group of laryngectomees an approximated mucosal surfaces of this structure SPVP might be beneficial, in particular with into vibration, producing a low-pitched sound, respect to the patient’s PE segment tonicity and which can be used as a substitute voice. If, how- sex. ever, the tonus of the PE segment is too low to attain sufficient mucosal approximation, the resulting voice will be weak and breathy or merely MATERIALS AND METHODS a coarse whisper.10,11 With sufficient approxima- tion, the vibrating mass is often fairly large, which Patients. The patients were nine women and 11 yields a low fundamental frequency (f0). Female men, with a mean age of 69 years (range, 52–79 laryngectomees in particular often have severe years). All underwent total laryngectomy with TE problems accepting their low-pitched alaryngeal 14 to 115 months before this study (mean, 52 12,13 voice. The mean speaking f0 of laryngeal months). Eight patients received radiotherapy af- female voices is 211 Hz (SD, 2.7 semitones),14 ter laryngectomy, nine subjects underwent unilat- which decreases after laryngectomy and current eral or bilateral neck dissections, six patients voice rehabilitation to an unnaturally low mean f0 required pedicled or free flap reconstructions, and of 108 Hz (SD, 28 Hz).15 To improve voice quality 15 patients underwent a primary pharyngeal my- for these two groups of laryngectomees (women otomy. Clinical data are summarized in Table 1. and those with a hypotonic PE segment), a small All patients received a new Groningen ultra-low-

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 401 FIGURE 1. Blueprint of the sound-producing module with bent silicone lip in gray (dimensions in millimeters). resistance (ULR) shunt valve before the tests. Spoken and written informed consent was ob- tained from all patients in this study. The medical ethics committee of the Vrije Universiteit Medical Center, Amsterdam approved the research pro- tocol.

Sound-producing Voice Prosthesis. The new sound source consists of a single bent tapered silicone lip (11.0 3 3.3 mm), which performs self-sustaining oscillations driven by the expired pulmonary air that flows along the outward-striking lip20 FIGURE 2. Prototype of the sound-producing module in a Gro- through the TE shunt valve. The resulting fre- ningen ULR shunt valve. quency of oscillation and sound intensity can be modified by altering the airflow along the lip.16 For female voice frequencies, the optimal lip Procedure and Speech Recordings. Before being thickness is 0.5 mm at the base and 0.3 mm at the recorded with the new sound-producing module free tip; for male voice frequencies, lip thickness is inserted in their TE shunt valve, patients were 0.35 mm at the base and 0.25 mm at the free tip. encouraged to experience and practice this new These lip configurations were selected by in vitro mechanism of alaryngeal voice for approximately benchmarking using two criteria: favorable sound 1 hour. Two block-randomized groups of patients 19 quality and most natural f0 range for each sex. (block size 2) were formed for this crossover trial. To allow easy placement and removal of the One group started all of the described vocal tests sound source for speech evaluation purposes in with their own Groningen ULR shunt valve, our experimental setting, the bent silicone lip was whereas the other group first performed all mea- fitted in a small stainless steel container (Figure surements with the SPVP. All subjects’ voices 1), that can be partly inserted in a patient’s Gro- were recorded in a sound-treated room, using a ningen ULR shunt valve (Figure 2). Thus, the microphone (MKE 212-3, Sennheiser, Germany) SPVP could be evaluated without the need of and a DAT-recorder (DA-7, Casio Computers, Ja- repeatedly replacing the entire shunt valve. The pan). Subjects were asked to read the first para- intention is to eventually integrate the bent sili- graphs of the Dutch prose ‘‘De Vijvervrouw’’ in a cone lip in the design of a regular TE shunt valve. normal conversational manner. Digital recordings

402 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 of approximately 90 seconds were made for each scales, developed by Nieboer22 and later modi- laryngectomee with both types of voice prosthesis. fied,23,24 was adapted for our purpose. This subset These recordings were used for the communica- included the five scales ‘‘high pitch– low pitch’’, tive suitability judgments, the perceptual voice ‘‘weak–powerful’’, ‘‘tense–nontense’’, ‘‘gurgling– evaluation, and to establish speech rate deter- nongurgling’’ and ‘‘melodious–monotonous.’’ The mined as the number of syllables per minute 40 digital recordings were presented in random (spm). In addition, each subject was asked to read order to a voice and speech pathologist experi- aloud a separate list of 13 short everyday Dutch enced with laryngectomized speakers in general sentences with each voice prosthesis. Further- but blinded for the clinical data. more, approximately 10 minutes of spontaneous conversation with the first author was video Intelligibility. Each of the 20 laryngectomees recorded for each laryngectomee with both types read aloud two lists of 13 short everyday Dutch of voice prostheses. sentences, one list while using regular TE shunt voice and one with the new SPVP. The 520 differ- Pharyngoesophageal Segment Tonicity Assess- ent sentences were sampled at 44.1 kHz and 16 ment. Estimation of PE segment tonicity was bits resolution. After each speech recording, an achieved by rating the produced voice using visual additional fixed-level reference tone was recorded. and acoustic cues on a seven-point scale (atonic, From these recordings, speech level and intelligi- severely hypotonic, slightly hypotonic, normo- bility were determined. The long-term average tonic, slightly hypertonic, severely hypertonic, speech level was determined by measuring the av- spasmodic). This was performed by a panel of erage speech level for each of the 40 lists from the three voice and speech pathologists experienced average sound pressure levels (window width 50 with laryngectomized speakers in general but ms) using the recorded tones as reference. Intelli- blinded for the clinical data. After listening and gibility was determined by measuring the speech watching approximately 10 minutes of video reception threshold (SRT) in competing noise (ie, recorded spontaneous speech of each patient with the signal to noise [S/N] ratio at which 50% of the a regular TE shunt valve, judgment was achieved sentences was reproduced without a single error by consensus of the panel. The results are shown by listeners with normal hearing). Because speech in Table 1. from multiple speakers is the most frequent dis- Communicative Suitability. Communicative suit- turbing sound in everyday situations, SRT was ability has been defined as the situation-depend- determined in interfering noise of 60 dB sound ent adequacy of speech as judged by naive listen- pressure level (SPL), with a spectrum equal to the ers. In a previous study, we evaluated and long-term average spectrum of normal speech. An adjusted a communicative suitability rating in- adaptive up-and-down procedure for accurately strument to be used in a meaningful way to assess measuring the SRT with these lists of 13 senten- ces was developed for evaluating hearing impair- functionality of voice after radiotherapy for T1 25 21 ment and later adapted for intelligibility assess- glottic cancer. The 40 digital recordings were 7,26 presented in random order to a panel of 20 stu- ment of alaryngeal speech. Tucker Davis hard- dents, unfamiliar with alaryngeal speech. Listen- ware (System II, Tucker Davis Technologies, ers rated the suitability of each voice sample for Gainesville, FL) and headphones (DT-48, Beyer, use in three speaking situations on an anchored Germany) were used to play out and attenuate the 10-point scale, keeping in mind the grading scale samples in a sound-treated room. The lists were that is commonly used in the Dutch educational presented binaurally to 12 students with normal system (1, extremely poor; 2, very poor; 3, poor; 4, hearing, unfamiliar with alaryngeal speech, each insufficient; 5, just insufficient; 6, just sufficient; listener assessing in random order only 10 7, amply sufficient; 8, good; 9, very good; 10, excel- patients (20 lists) to avoid decreasing attentive- lent). The speaking situations are low demanding ness. (talking about everyday events with a friend), me- dium demanding (asking a stranger for direc- Acoustic and Aerodynamic Assessment. Regis- tions), and highly demanding (giving a lecture to a trations of sustained vowels were made in a newly founded professional association). sound-treated room with the experimental setup shown in Figure 3. A pressure transducer (DT-2- Perceptual Voice Evaluation by a Professional Liste- 14P-0-10L, Modus Instruments, Clinton, MA), ner. A subset of the seven-point bipolar semantic connected to a tracheostoma adapter (BE-6040,

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 403 FIGURE 3. Experimental setup for in vivo registrations of aerodynamic and acoustic parameters.

Inhealth Technologies, Carpinteria, CA) mea- Austin, TX) and digitally processed by a custom- sured tracheal phonatory pressure. A heated built LabVIEW software application. Recordings screen-pneumotach with integrated differential were made for each laryngectomee with both pressure transducer (MS-B and PT36, Erich types of voice prosthesis during several sustained Jaeger GmbH, Ho¨chberg, Germany) was fitted phonations of the vowel /a:/ at comfortable loud- on a facemask (King Systems Corp., Noblesville, ness and pitch. In addition, patients were asked IN) to determine phonatory airflow rate. These to sustain the vowel /a:/ as soft as possible and as transducers were calibrated by means of a Cali- loud as possible. In this way, the dynamic and bration Analyzer (Timeter RT-200, Allied Health- the melodic range were determined. Intratra- care,St.Louis,MO).Aminiatureelectretcon- cheal pressure, airflow rate, and SPL were read denser microphone with a hybrid-integrated off-line from the raw data, whereas the acoustic amplifier (9468, Microtronic, Amsterdam, The signals were converted to WAV files for further Netherlands) was attached to the flowhead for software analysis. For every 100 ms, the funda- voice recording. SPL at 30 cm from the flowhead mental frequency was calculated off-line from on the facemask was registered by a digital the microphone signal by another custombuilt sound survey meter (CEL-231, CEL Instruments, LabVIEW application using a harmonic product Hitchin, England). An electro-glottograph (EG- spectrum algorithm.27 830, Froekjaer-Jensen Electronics, Holte, Den- mark) was connected to the patient through a set Self-assessment by Patients. Each patient only of Ag/AgCl electrodes. An 8F catheter with three had between 2 and 4 hours of experience with the pressure transducers (UniTip 8366-00-9980-D, new SPVP, during which all measurements were UniSensor AG, Attikon, Switzerland) was in- performed. We deemed this period too short for a serted through the nose until the distal sensor patient to estimate the impact of the SPVP on was located at the level of the esophageal flange daily functioning and quality of life by means of of the TE shunt valve. All sensor signals were an extensive questionnaire, like the Voice Handi- preprocessed using National Instruments SCXI cap Index28 or the European Organization for signal-conditioning hardware, sampled at 10 kHz Research and Treatment of Cancer Quality of Life by means of a PC-based 16-bit analog-digital con- Questionnaire-Head & Neck 35.29 Therefore, a verter (PCI-MIO-16xe-10, National Instruments, short questionnaire was designed to assess on

404 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 FIGURE 4. Mean communicative suitability scores for each patient. The origin of the arrow corresponds to the communicative suitabil- ity of the regular tracheoesophageal shunt voice. The tip of the arrow represents the communicative suitability using the sound-produc- ing voice prosthesis. four-point scales the patients’ primary judgment of tracheal phlegm clogging the vibrating silicone lip voice quality attained with a SPVP, as well as their of the SPVP. The tests were resumed after remov- regular TE shunt valve. The seven items of the ing the SPVP from the patient, flushing it with questionnaire were vocal intensity, pitch of voice, water, and reinserting it. availability of the voice, effort required for speak- ing, fluency of the speech, pitch control for intona- Speech Rate. Using the SPVP speech rate tion, and a general impression of voice quality. (mean, 139 spm; SD, 38 spm) was decreased com- pared with regular TE shunt speech (mean, 146 Statistical Evaluation. Results on the various spm; SD, 42 spm), although this difference was test parameters, obtained in both situations, were not significant (t(19) ¼ 2.058; p ¼ .054). The aver- compared and statistically analyzed (SPSS 11.0, age speech rate of 52 laryngeal reference speakers SPSS Inc., Chicago, IL) using Pearson’s correla- from our institute is 228 spm (SD, 30 spm). In tion and paired Student’s t test for parametric female laryngectomees with a severely hypotonic data or Mann–Whitney U test and Wilcoxon or atonic PE segment (patients F1–F5), speech matched-pairs signed-rank test for ordinal data. rate increased on average 1 spm (SD, 19 spm) For the panels of listeners, the mean score was using the SPVP, whereas it decreased 9 spm (SD, used for further analyses after ascertaining the 13 spm) in the other 15 patients. This between- interrater reliability by calculating Cronbach’s group difference was not significant (t(18) ¼ 1.376; alpha. p ¼ .186).

Communicative Suitability. Interrater reliability, RESULTS assessed separately for the three speaking situa- tions by means of Cronbach’s alpha, seemed to be Observations. For 12 patients (60%) the test high: all alpha coefficients exceeded 0.97. For fur- sequence had to be halted at least once because of ther analyses, the mean scores of the 20 listeners

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 405 Table 2. Perceptual evaluation of voice quality by an expert listener for 20 laryngectomees using the sound-producing voice prosthesis compared with their regular tracheoesophageal shunt voice on a bipolar seven-point scale.

Bipolar semantic scales (number of scale values*) Patient Low pitch–high pitch Weak–powerful Tense–nontense Gurgling–nongurgling Monotonous–melodious F1 2 1 F2 4 2 1 5 1 F3 1 2 2 2 F4 2 1 1 F5 2 1 11 4 F6 21 3 1 F7 4 1 6 F8 13 31 F9 3 31 M1 3 1 2 N.A. 2 M2 2 2 1 M3 5 1 3 23 M4 5 1 3 23 M5 2 1 1 2 1 M6 2 4 2 3 1 M7 2 1 M8 221 M9 11 1 2 M10 4 121 M11 4 2 11 Mean 1.35 0.15 0.65 0.74 0.35 Significancey .026 .595 .042 .117 .560

Abbreviation: N.A., not applicable. *A positive number of scale values indicates improvement: higher pitch, more strength, less tense, less gurgling, and more melodious. A negative num- ber of scale values indicates the opposite. Items unaltered by the sound-producing voice prosthesis were left out. ySignificance of differences caused by prosthesis is based on Wilcoxon matched-pairs signed-rank tests. were used. The suitability ratings for each of the scale the rating difference between SPVP voice three speaking situations strongly correlated, and the patient’s regular TE shunt voice. The indicating redundancy (Pearson’s correlation coef- effects of the SPVP on voice quality varied consid- ficients 0.983, 0.923, and 0.969; significance < erably between patients. According to separate .001). Hence, the mean score of the three speaking Wilcoxon matched-pairs signed-rank tests, two situations was used, resulting in one rating for scales significantly differentiated between regu- each speaker with each shunt valve. lar TE shunt voices and SPVP voices: the SPVP Figure 4 shows the shift in communicative voices were rated higher pitched, yet more tense. suitability for each patient using the SPVP com- Female laryngectomees with a severely hypotonic pared with their regular TE shunt voice. The com- or atonic PE segment (patients F1–F5) gained on municative suitability of the 20 regular TE shunt average 1.4 scale values (SD, 1.3) of vocal strength voices (mean, 4.9; SD, 1.7) was significantly by using the SPVP, whereas the other 15 subjects reduced (t(19) ¼ 4.456; p < .001) by the SPVP lost 0.3 scale values (SD, 2.0). This difference, (mean, 4.0; SD, 1.4). Three patients (patients F1, however, was only weakly significant (U ¼ 17.5; F6, and M3) attained a slightly improved commu- p ¼ .075). The effects of the SPVP on the other four nicative suitability with the SPVP. All others scales differed hardly between these two sub- deteriorated. In female patients with a severely groups of laryngectomees. hypotonic or atonic PE segment (patients F1–F5), the mean reduction of communicative suitability Intelligibility and Speech Level. Because each caused by the SPVP was 0.8 (SD, 0.9), whereas it speaker was assessed by only six listeners to avoid was 1.0 (SD, 1.0) in the other 15 laryngectomees. decreasing attentiveness, interrater reliability This difference was not significant (t(18) ¼ 0.453; was assessed separately for the two groups of six p ¼ .656). listeners by means of Cronbach’s alpha. The alpha coefficients were 0.955 and 0.934, which is high. Perceptual Voice Evaluation by Professional Liste- For further analyses, the mean scores of the 12 lis- ner. Table 2 shows for each patient and each teners were used.

406 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 Table 3. Long-term average speech level, speech reception threshold in noise, and percentage of words repeated correctly when played out without interfering noise.

Average speech Signal-to-noise ratio at level (dB SPL) SRT (dB) Percentage of correctly repeated words Patient Regular VP SPVP Regular VP SPVP Regular VP SPVP F1 70.3 78.1 12.5 1.5 F2 67.2 68.3 9.9 2.8 F3 63.0 71.4 39 40 F4 70.0 74.8 þ1.6 þ4.4 F5 74.2 75.6 11.6 3.8 F6 75.3 77.8 12.3 5.7 F7 74.1 74.1 9.4 2.4 F8 79.5 75.9 25 10 F9 74.2 69.3 þ8.8 100 54 M1 77.2 75.9 þ8.1 100 40 M2 78.8 76.6 0.3 þ9.3 M3 69.9 70.9 7.9 1.7 M4 75.9 79.0 8.5 3.7 M5 74.0 73.5 þ0.9 100 29 M6 82.1 80.2 12.9 4.9 M7 75.2 70.1 5.5 þ12.8 M8 78.5 78.7 þ0.5 þ7.4 M9 81.1 78.7 1.1 þ0.3 M10 78.3 72.9 1.8 þ13.5 M11 72.9 81.8 16 12 Mean (SD) 74.6 (4.8) 75.2 (3.8) 6.5 (5.3) þ1.5 (6.7) 63% (41) 31% (17)

Significance t(19) ¼ 0.635; p ¼ .533 t(13) ¼ 6.728; p < .001 t(5) ¼ 2.600; p ¼ .048 Abbreviations: SPL, sound pressure level; SRT, speech reception threshold; VP, voice prosthesis; SPVP, sound-producing voice prosthesis.

Table 3 presents the long-term average speech ity of the speech. Using the SPVP, intelligibility in levels of each patient for both voice sources. For noise of all laryngectomees was reduced compared the entire group of patients, the difference with their regular TE shunt voices. For most between SPVP speech (mean, 74.6 dB SPL) and patients this intelligibility loss was considerable regular TE shunt speech (mean, 75.2 dB SPL) was (>6 dB). A paired Student’s t test showed the SRT not significant (t(19) ¼ 0.635; p ¼ .533). Female lar- differences between both voice sources to be yngectomees with a severely hypotonic or atonic highly significant (t(13) ¼ 6.728; p < .001). In PE segment (patients F1–F5) gained on average female patients with a severely hypotonic or ato- 4.7 dB (SD, 3.3 dB) by using the SPVP, whereas nic PE segment (patients F1–F5), the mean intel- the rest of the studied group lost 0.8 dB (SD, 3.7 ligibility loss caused by the SPVP was 7.2 dB (SD, dB). This difference was significant (t(18) ¼ 2.926; 3.4), whereas it was 8.4 (SD, 5.0) in the other 15 p ¼ .009). laryngectomees. This between-group difference The intelligibility of nine lists of sentences, was not significant (t(12) ¼ 0.443; p ¼ .666). read aloud by a total of six patients, was too poor Figure 5 shows the combined results of 14 at the maximum S/N ratio (þ30 dB) to determine patients on speech level and intelligibility for the the SRT using the described adaptive up-and- regular TE shunt valve and the SPVP. Only in down procedure. For these nine lists, the percent- patient F4, a female laryngectomee with an atonic age of words repeated correctly by the 12 listeners PE segment, did the speech level improve enough was counted when played out at conversational to outweigh the intelligibility loss in noise with loudness without interfering noise. The results the SPVP. are shown in Table 3. With the SPVP, significantly fewer words (mean, 31%; SD, 17) were repeated Acoustic and Aerodynamic Assessment. For all correctly (t(5) ¼ 2.600; p ¼ .048) compared with patients, the fundamental frequency range, the regular TE shunt voice (mean, 63%; SD, 41). vocal intensity range, the intratracheal pressure For the remaining 14 patients, Table 3 shows range, and airflow range are shown in Table 4 for the SRT in noise for both voice prostheses. A lower both types of voice. For some recordings with a S/N ratio at the SRT indicates a better intelligibil- very low S/N ratio (regular TE shunt voices of

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 407 FIGURE 5. Vulnerability of speech to masking noise, expressed as the signal-to-noise ratio at threshold, versus the long-term average speech level. Each laryngectomee is presented by an arrow pointing from the result for their regular tracheoesophageal shunt voice to the result using the sound-producing voice prosthesis. The diagonal mesh represents lines of iso-intelligibility: for voices that are more vulnerable to interfering noise, each decibel increase in speech reception threshold (SRT) needs to be compensated for by a decibel in speech level. patients F1, F4, M7, M9, and M11; SPVP voice of SPL. All women classified as having a hypertonic patient F3), fundamental frequency could not be PE segment (patients F7, F8, and F9) and all male calculated despite the robust acoustic analysis subjects found it difficult not to evoke the low-fre- method.27 Patient M9 did not tolerate our experi- quency mucosal vibrations that form their regular mental setup when straining to speak with the TE shunt voice. For them the apparently unavoid- SPVP. Patient F3 could not handle the tracheos- able vibrations of their PE segment dominated, toma adapter when using the SPVP. thereby masking the SPVP voice. Exerted tracheal pressures (0.3–16.3 kPa), air- flow rates (0.01–0.9 L/s), and vocal intensities Self-assessment by Patients. Table 5 shows for (44–89 dB[A]) were comparable for both types of each patient the judgment for each of the seven voice but varied widely among patients. Using the questionnaire items. Patient F4 was not able to SPVP, all female patients attained considerably compare pitch of the SPVP voice to her regular TE higher fundamental frequencies, approximating shunt voice, because the latter is an aphonic whis- 14 the mean speaking f0 of laryngeal female voices. per. According to separate Wilcoxon matched- To a smaller extent, this also occurred in most pairs signed-rank tests, none of the questionnaire male subjects, sometimes considerably exceeding items significantly differentiated between regular the mean speaking f0 of laryngeal men, which is TE shunt voice and SPVP voice. Female laryngec- 124 Hz (SD, 2.9 semitones).14 With the SPVP, tomees with a severely hypotonic or atonic PE seg- female laryngectomees with an atonic PE segment ment (patients F1–F5) gained on average 0.6 scale (patients F1–F4) required less tracheal pressure values (SD, 1.1) of vocal intensity by using the and a lower airflow rate to attain an identical SPVP, whereas the other 15 subjects lost 0.5 scale

408 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 Table 4. Fundamental frequency range, vocal intensity range, tracheal pressure range, and airflow range associated with sustained phonation of a vowel /a:/ by 20 laryngectomized patients.

Regular TE shunt voice Sound-producing voice prosthesis Tracheal Airflow SPL Tracheal Airflow

Patient f0 (Hz) SPL (db[A]) pressure (kPa) rate (L/s) f0 (Hz) (db[A]) pressure (kPa) rate (L/s) F1 N.A. 50–76 1.3–5.8 0.01–0.4 164–278 53–71 1.2–3.6 0.04–0.2 F2 40–105 58–60 4.5–13.0 0.03–0.5 150–183 55–60 1.9–3.3 0.03–0.1 F3 151–183 53–69 1.8–10.3 0.01–0.4 N.A. 66–75 N.A. 0.09–0.3 F4 N.A. 44–64 0.6–5.6 0.01–0.3 155–195 44–56 1.0–3.9 0.04–0.05 F5 50–177 45–80 2.5–12.0 0.07–0.35 137–180 45–82 3.0–15.0 0.03–0.4 F6 39–70 45–73 1.2–5.5 0.11–0.16 145–186 57–78 1.4–5.7 0.01–0.3 F7 77–187 53–82 2.1–13.0 0.12–0.4 82–185 53–73 3.7–11.0 0.08–0.14 F8 109–185 58–74 4.0–9.0 0.09–0.1 137–187 54–76 1.2–14.5 0.07–0.5 F9 75–158 44–72 0.4–11.0 0.06–0.8 154–193 56–72 2.1–2.8 0.05–0.1 M1 72–120 64–77 4.2–13.0 0.02–0.03 72–199 60–76 2.3–15.6 0.02–0.04 M2 112–144 55–74 1.0–10.3 0.01–0.2 132–240 49–85 2.3–10.0 0.01–0.05 M3 79–130 44–77 0.8–9.6 0.03–0.8 122–130 44–78 0.3–10.0 0.02–0.2 M4 62–115 52–79 2.1–12.6 0.2–0.9 124–138 49–81 1.1–13.5 0.06–0.4 M5 57–134 49–80 1.6–16.0 0.05–0.8 98–185 55–66 4.6–12.0 0.05–0.7 M6 97–185 55–65 3.0–7.5 0.03–0.3 91–161 54–84 2.5–16.3 0.03–0.4 M7 N.A. 44–76 0.4–8.9 0.02–0.4 77–144 49–75 1.0–10.0 0.01–0.2 M8 166–169 62–69 3.2–6.0 0.05–0.3 100–150 68–70 1.3–3.9 0.02–0.2 M9 N.A. 64–78 2.3–8.7 0.02–0.3 N.A. N.A. N.A. N.A. M10 50–98 44–81 1.7–16.0 0.01–0.3 115–158 51–77 1.2–15.5 0.01–0.2 M11 N.A. 44–72 1.2–9.2 0.01–0.3 134–154 44–89 1.2–16.2 0.01–0.5

Abbreviations: TE, tracheoesophageal; fo, frequency range; SPL, sound pressure level; N.A., not available. values (SD, 0.7). This between-group difference sistent with our preceding studies,17,18 the mecha- was significant (U ¼ 15.5; p ¼ .042). The effects of nism of alaryngeal voice by means of a SPVP the SPVP on the other six scales did not differ sig- proves feasible and does not result in increased nificantly between these two subgroups of laryng- airflow resistance. In addition, the current proto- ectomees. type hardly protrudes in front of the tracheal Three women (patients F3, F7, and F9) pre- flange of a regular Groningen ULR shunt valve, ferred the higher vocal pitch attained with the fitting without difficulty in the trachea of the 20 SPVP, whereas four men (M1, M2, M3, and M6) studied laryngectomees. The objective of this rejected it for exceeding even the mean speaking study was to determine for which group of laryng- 14 f0 of laryngeal men. Five patients (patients F1, ectomees an SPVP might be beneficial, in particu- F3, M9, M10, and M11) observed no deteriorating lar with respect to the patient’s PE segment tonic- items, only some improvements with the SPVP ity and their sex. compared with their regular TE shunt. Patients F5 and M2, however, judged the SPVP voice infe- Pitch of Voice. Male TE shunt speakers are rior to their regular TE shunt voice on all items. regarded as having significantly better, more ac- Negative judgments on the availability of the ceptable, and more pleasant voices than women.13 voice, the effort required for speaking, and the flu- This may be partly because of the unnaturally low ency of the speech were partly attributable to tra- pitch of regular TE shunt voices, which for women cheal phlegm clogging the vibrating silicone lip of deviates more from the mean speaking f0 of their the SPVP momentarily or permanently in these preoperative laryngeal voices than for men.14 In patients. Seven patients thought their pitch con- this study, pitch of voice is significantly elevated trol was limited with the SPVP, resulting in poor when using the SPVP according to the perceptual intonation and monotonous voices. voice evaluation by the expert listener (Table 2). For most subjects, this is exemplified by the f0 range of several sustained phonations of the vowel DISCUSSION /a:/ with both types of voice prosthesis (Table 4). The purpose of the SPVP is to improve pitch for For female laryngectomees, the rise of vocal pitch, female laryngectomees and vocal strength for lar- approximating the pitch of laryngeal female voi- yngectomees with a hypotonic PE segment. Con- ces, is a desired effect of the SPVP. Yet, it is an

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 409 Table 5. Self-assessment of the sound-producing voice prosthesis by 20 patients compared with their regular tracheoesophageal shunt voice on a four-point scale.

Questionnaire items (number of scale values*) Vocal Pitch of Availability of Effort General Patient intensity voice voice required Fluency Intonation impression F1 2 F2 2 2 2 1 F3 1 1 2 F4 1 N.A. 2 31 F5 1 1 1 2 1 2 2 F6 1 1 2 F7 11 2 1 1 2 F8 1 12 1 F9 11 1 11 M1 11211 M2 2 2 1 2 2 2 2 M3 1 21 1 M4 M5 1 1 1 2 M6 1 1 1 1 2 M7 1 2 1 1 1 M8 1 1 M9 1 M10 2 1 1 M11 1 1 1 Mean 0.25 0.05 0.15 0.25 0.40 0.50 0.45 Significancey .260 .791 .366 .361 .112 .054 .063

Abbreviation: N.A., not applicable. *A positive number of scale values indicates improvement by the sound-producing voice prosthesis (SPVP) compared with regular tracheoesophageal shunt voice: better vocal intensity, improved pitch, better availability of the voice, less effort required for speaking, better fluency, better pitch control for intonation, and a better general impression of the voice. A negative number of scale values indicates the opposite. Items unaltered by the SPVP were left out. ySignificance of differences caused by prosthesis is based on Wilcoxon matched-pairs signed-rank tests. undesired oddity for the male subjects. Despite Intelligibility and Communicative Suitability. In the promising results of the previous in vitro contrast to clear-cut voice parameters such as benchmarking,19 the silicone lip configuration of vocal intensity and pitch of voice, intelligibility the SPVP for men still yields an unsuitable high f0 and communicative suitability are multifaceted range for them in vivo. and more comprehensive speech parameters. Both intelligibility and communicative suitability Vocal Intensity. Overall, vocal intensity seems deteriorate significantly when using the SPVP. not to be affected by the SPVP (Tables 2–5). For This holds even for female laryngectomees with a laryngectomees with a PE segment of adequate to- severely hypotonic or atonic PE segment (patients nus, vocal intensity is usually not the prime voice- F1–F5). Possible explanations for this drawback related problem. In laryngectomees with an atonic of the SPVP are listed following: or severely hypotonic PE segment, however, the TE-shunt voice is weak and breathy or merely a 1. SPVP voice has an audibly artificial timbre, coarse whisper.10,11 For this relevant subgroup although it is not significantly more monoto- (patients F1–F5), vocal strength is improved by nous than the regular TE shunt voice accord- the SPVP according to the self-assessment by the ing to the self-assessment by the patients patients, the perceptual voice evaluation by the and the perceptual voice evaluation by the expert listener, and the long-term average speech expert listener. This artificial timbre may re- levels. None of the male patients included in this duce the communicative suitability of speech study was classified as having an atonic or as judged by naive listeners. severely hypotonic PE segment, which brings 2. The new pneumatic sound source may gener- about the methodologic impossibility to draw sep- ate voice during the pronunciation of voice- arate conclusions with respect to the patient’s PE less consonants. In regular TE shunt speech, segment tonicity and sex. this results in misperception of voiced for

410 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 voiceless phonemes.30 Voiceless stops are less vides her temporarily with a tonal voice (164– 278 vulnerable to these voiced-for-voiceless confu- Hz) that is markedly stronger according to both sions than voiceless fricatives. Using the the self-assessment and the perceptual voice eval- SPVP, intelligibility may be further reduced uation by the expert listener. The 11-dB intelligi- than in regular TE shunt speech, because the bility loss in noise with the SPVP is almost com- articulatory PE-segment adjustments essen- pensated for by an 8-dB SPL increase in speech tial for the production of several fricatives in level. Using the SPVP, her communicative suit- regular TE shunt speech do not affect the ar- ability slightly increases. tificial sound source. 3. Using the SPVP, subjects with a PE segment of adequate tonus find it difficult not to evoke CONCLUSION the low-frequency mucosal vibrations that The prototype SPVP is successful in raising vocal form their regular TE shunt voice. Appa- pitch. Female laryngectomees with an atonic or rently, pulmonary airflow through the PE severely hypotonic PE segment can benefit from a segment passively evokes these vibrations. stronger voice with the current prototype SPVP, For them the apparently unavoidable vibra- generally in exchange for a deterioration of intelli- tions of their PE segment dominate, thereby gibility and communicative suitability. Future masking the SPVP voice. This phenomenon research concerning the SPVP should include is discussed and illustrated in our previous evaluation of long-term use. Furthermore, it reports17,18 and may reduce both intelligibil- should be directed at improvement of the SPVP ity and communicative suitability of speech. voice quality to enhance intelligibility in noise. 4. Contrary to regular TE shunt speech, sub- jects had minimal training or previous expe- Acknowledgments. The authors wish to rience with this novel mechanism of alaryng- thank C. Klok and D. Koops for the CAD/CAM eal voice. This probably reduces their ability injection molding of the silicone rubber lips and for prosodic accentuation and, thereby, may for preparing the stainless steel containers that reduce intelligibility and communicative suit- include these lips. J. H. M. van Beek, BSc, is grate- ability. A more extensive extramural training fully acknowledged for writing the software for period might possibly yield more favorable the adaptive up-and-down SRT procedure. M. D. results with the SPVP. This, however, was de Bruin, MA, and B. A. Boon-Kamma, MA, are not feasible in this study because of the mod- gratefully acknowledged for rating PE segment ular design of the sound source that was tonicity of the study group. We acknowledge M. N. inserted only temporarily in the patients’ reg- Koster, MA, and J. Henstra, MA, for their help in ular TE shunt valves. Prolonged use would conducting the experiments with listeners. require a more permanent fixation or inte- gration of the sound source in the shunt valve. Furthermore, the current observations REFERENCES concerning clogging of the vibrating silicone 1. Pruyn JFA, de Jong PC, Bosman LJ, et al. Psychosocial lip of the SPVP by tracheal phlegm suggest aspects of head and neck cancer—a review of the litera- that a cleansing mechanism is required ture. Clin Otolaryngol 1986;11:469–474. 2. Nalbadian M, Nikolaou A, Nikolaidis V, Petridis D, The- before a shunt valve could be permanently melis C, Daniilidis I. Factors influencing quality of life fitted with a sound source. in laryngectomized patients. Eur Arch Otorhinolaryngol 2001;258:336–340. 3. Blom ED, Hamaker RC. Tracheoesophageal voice resto- ration following total laryngectomy. In: Meyers EN, Suen Prime Example. Patient F1 is an excellent exam- J, editors. Cancer of the head and neck. Philadelphia: ple of the group of laryngectomees for whom a WB Saunders Publishers; 1996. p 839–852. 4. Brown DH, Hilgers FJM, Irish JC, Balm AJM. Postlaryn- SPVP might be beneficial. She was laryngectom- gectomy voice rehabilitation: state of the art at the millen- ized at a relatively young age (47) and still works nium. World J Surg 2003;27:824–831. full time in a cafeteria. After pectoralis major 5. Lebrun Y. The artificial larynx. Amsterdam: Zwets & Zeitlinger BV; 1973. myocutaneous flap reconstruction, primary pha- 6. Singer MI, Blom ED. An endoscopic technique for resto- ryngeal myotomy, and postoperative radiother- ration of voice after laryngectomy. Ann Otol Rhinol Laryn- apy, her PE segment is atonic. Her regular TE gol 1980;89:529–533. 7. Mahieu HF, Schutte HK, Annyas AA. Intelligibility, shunt voice consists of no more than whispering. vocal intensity, and long-term average spectra of Gro- She barely has tracheal phlegm. The SPVP pro- ningen button-oesophageal speech. In: Hermann IF, edi-

Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006 411 tor. Speech restoration via voice prostheses. Berlin, Hei- 19. Van der Torn M, Mahieu HF, Festen JM. Aero-acoustics delberg, New York: Springer-Verlag; 1986. p 139–147. of silicone rubber lip reeds for alternative voice produc- 8. Max L, Steurs W, de Bruyn W. Vocal capacities in esopha- tion in laryngectomees. J Acoust Soc Am 2001;110:2548– geal and tracheoesophageal speakers. Laryngoscope 1996; 2559. 106:93–96. 20. Fletcher NH. Autonomous vibration of simple pressure- 9. Moerman M, Pieters G, Martens JP, van der Borgt MJ, controlled valves in gas flows. J Acoust Soc Am 1993; Dejonckere P. Objective evaluation of the quality of substitu- 93:2172–2180. tion voices. Eur Arch Otorhinolaryngol 2004;261:541–547. 21. Van der Torn M, Verdonck-de Leeuw IM, Kuik DJ, 10. Mahieu HF, Annyas AA, Schutte HK, van der Jagt EJ. Mahieu HF. Communicative suitability of voice following Pharyngoesophageal myotomy for vocal rehabilitation of radiotherapy for T1 glottic carcinoma: testing the reli- laryngectomees. Laryngoscope 1987;97:451–457. ability of a rating instrument. J Voice 2002;16:398–407. 11. van As CJ, op de Coul BMR, van den Hoogen FJA, Koop- 22. Nieboer GLJ, de Graaf T, Schutte HK. Esophageal voice mans-van Beinum FJ, Hilgers FJM. Quantitative video- quality judgments by means of the semantic differential. fluoroscopy: a new evaluation tool for tracheoesophageal J Phonetics 1988;16:417–436. voice production. Arch Otolaryngol Head Neck Surg 2001; 23. Van As CJ, Hilgers FJM, Verdonck-de Leeuw IM, Koop- 127:161–169. mans-van Beinum FJ. Acoustical analysis and percep- 12. Smithwick L, Davis P, Dancer J, Hicks GS, Montague J. tual evaluation of tracheoesophageal prosthetic voice. Female laryngectomees satisfaction with communication J Voice 1998;12:239–248. methods and speech-language pathology services. Per- 24. Festen JM, Verdonck-de Leeuw IM. Perceptual rating and cept Mot Skills 2002;94:204–206. acoustic analysis of tracheo-esophageal voice. J Acoust Soc 13. Eadie TL, Doyle PC. Auditory-perceptual scaling and Am 1999;105:1247. quality of life in tracheoesophageal speakers. Laryngo- 25. Plomp R, Mimpen AM. Improving the reliability of test- scope 2004;114:753–759. ing the speech reception threshold for sentences. Audiol- 14. Krook MIP. Speaking fundamental frequency character- ogy 1979;18:43–52. istics of normal Swedish subjects obtained by glottal fre- 26. Bloothooft G, de Boer C. Measuring the intelligibility of quency analysis. Folia Phoniatr 1988;40:82–90. alaryngeal speakers. Progr Rep Inst Phonetics Utrecht 15. Trudeau MD, Qi Y. Acoustic characteristics of female 1984;9:24–40. tracheoesophageal speech. J Speech Lang Hear Res 27. Schroeder MR. Period histogram and product spectrum: 1990;55:244–250. new methods for fundamental frequency measurement. 16. De Vries MP, van der Plaats A, van der Torn M, Mahieu J Acoust Soc Am 1968;43:829–834. HF, Schutte HK, Verkerke GJ. Design and in-vitro test- 28. Jacobson BH, Johnson A, Ggrywalski C, et al. The Voice ing of a voice-producing element for laryngectomized Handicap Index (VHI): development and validation. Am patients. Int J Artif Organs 2000;23:462–472. J Speech Lang Pathol 1997;6:66–70. 17. Van der Torn M, Verdonck-de Leeuw IM, Festen JM, De 29. Bjordal K, Hammerlid E, Ahlner-Elmqvist M, et al. Vries MP, Mahieu HF. Female-pitched sound-producing Quality of life in head and neck cancer patients: valida- voice prostheses – initial experimental and clinical tion of the European Organization for Research and results. Eur Arch Otorhinolaryngol 2001;258:397–405. Treatment of Cancer Quality of Life Questionnaire- 18. Van der Torn M, Verdonck-de Leeuw IM, Festen JM, de H&N35. J Clin Oncol 1999;17:1008–1019. Vries MP, Mahieu HF. Alternative voice after laryngec- 30. Searl JP, Carpenter MA, Banta CL. Intelligibility of tomy using a sound-producing voice prosthesis. Laryngo- stops and fricatives in tracheoesophageal speech. J Com- scope 2001;111:336–346. mun Disord 2001;34:305–321.

412 Sound-producing Voice Prosthesis HEAD & NECK—DOI 10.1002/hed May 2006