Acoustic Analysis of Four Common Voice Diagnoses:

Moving Toward Disorder-Specific Assessment

Amanda I. Gillespie, Christina Dastolfo, Naomi Magid, and Jackie Gartner-Schmidt, Pittsburgh, Pennsylvania

Summary: Objectives. To assess treatment outcomes via acoustic voice laboratory measurements before and after
intervention in patients with common voice problems and Determine if outcome sensitivity of certain voice laboratory
measures varies with disorder type.
Study Design. Retrospective and single-blinded.
Methods. In this study, 40 patients with a single voice disorder diagnosis of either benign vocal fold lesions (lesions),
primary muscle tension dysphonia (MTD-1), vocal fold atrophy (atrophy) or unilateral vocal fold paralysis (UVFP) un-
derwent baseline testing, a single intervention-type (phonosurgery/voice therapy), and follow-up testing at uniform time
points. Ten patients per diagnosis group were analyzed before and after treatment. Time- and frequency-based acoustic
measures taken from vowels and sentences as well as patient-perceptual analysis (Voice Handicap Index-10) were
reviewed.
Results. Statistically significant improvements were observed for three of four groups. Patients with muscle tension
dysphonia displayed an improvement in Cepstral Spectral Index of Dysphonia speech (CSID) (P < 0.05). Patients with
lesions had improved Voice Handicap Index-10 (P < 0.05), cepstral peak prominence (CPP) vowel standard deviation
(P < 0.05), and CPP speech (P < 0.05). Patients with atrophy did not demonstrate significant improvement in any mea-
sure. Patients with unilateral vocal fold paralysis showed an improvement in CSID speech (P < 0.05) and CPP speech
(P < 0.05). In addition, strong effect sizes were observed for many of the acoustic parameters studied.
Conclusions. For all groups except atrophy, treatment was successful in improving patient perception of voice hand-
icap and/or some acoustic voice parameters. A disorder-specific response to frequency-based acoustic measures was
found.
Key Words: Cepstral–Voice–Voice laboratory.

INTRODUCTION voice laboratory measures. No correlation between acoustic

Speech-language pathologists (SLPs) and laryngologists measurements and voice handicap was found.9 Moreover, no
(specialized ear, nose, and throat physicians) struggle with a changes in acoustic measurements were observed between the
lack of standard measures with which to quantify outcomes af- time points studied, despite large VHI-10 improvements. How-
ter treatment of voice problems. Traditional assessment tools ever, that study was not designed to assess voice change out-
for analyzing vocal acoustics may be sufficient for identifica- comes after a known intervention, and the acoustic measures
tion of abnormal from normal voices1–3; however, these selected in that study were traditional time-based ones (eg,
analyses are not sensitive to detecting treatment change noise-to-harmonic ratio [NHR], fundamental frequency [F0]).
across all voice problems.4,5 In addition, most measures are The body of literature refuting the sensitivity, reliability, and
appropriate only for analysis of sustained vowels, not validity of time-based acoustic measures as outcomes to treat-
connected speech, which questions their ecological validity. ment is growing.4,5,10 In addition, because of their reliance on
Voice disorders and their treatments are heterogeneous; a periodic signal, time-based measures are inadequate for mea-
therefore, a one-size-fits-all approach to assessing treatment surement of severely dysphonic voices (eg, signal type II and
outcomes may be inappropriate, yet this approach represents III), which may represent nearly a quarter of all patients.4,5
the norm across voice centers.6 Furthermore, often when out- Frequency-based measures, specifically those that incorporate
comes are reported in the literature for a single disorder, the cepstral analyses, have been investigated as an alternative or
treatments received by the patients were not homogeneous, or complement to time-based measures.5 Frequency-based mea-
only poorly described, which impedes generalization.7 sures analyze the dominance of the F0 over additional noise
As a first step in developing disorder-specific analyses, our in the signal. These techniques may provide greater sensitivity
author group’s past work attempted to correlate change in pa- in detecting baseline dysphonia and possibly voice change after
tient perception of voice handicap using the Voice Handicap treatment.5,11–13
Index-10 (VHI-10)8 with change in acoustic and aerodynamic Although little literature exists regarding treatment
response via frequency-based measures, numerous studies
Accepted for publication February 5, 2014.
confirm the ability of cepstral-based measures to differen-
From the Department of Otolaryngology, University of Pittsburgh Voice Center, Univer- tiate normal from abnormal voices.12,14–16 Cepstral
sity of Pittsburgh Medical Center, Pittsburgh, Pennsylvania.
Address correspondence and reprint requests to Amanda I. Gillespie, Department of
measures have also shown to be accurate in tracking vocal
Otolaryngology, University of Pittsburgh Voice Center, University of Pittsburgh Medi- change over time in patients with spontaneous recovery of
cal Center Mercy, 1400 Locust Street, Suite 11-500, Building B, Pittsburgh, PA 15219.
E-mail: [email protected]
vocal fold mobility after thyroid surgery.17 Similarly, ceps-
Journal of Voice, Vol. 28, No. 5, pp. 582-588 tral peak prominence (CPP) has shown to be more consistent
0892-1997/$36.00
Ó 2014 The Voice Foundation
over time in healthy individuals compared with traditional
http://dx.doi.org/10.1016/j.jvoice.2014.02.002 time-based measures.10
Amanda I. Gillespie, et al Acoustic Analysis of Four Common Voice Diagnoses 583

The extant treatment literature has documented treatment five sessions of physiologically based voice therapy, which con-
change using algorithms of time and frequency-based measures sisted of combinations of resonant voice, flow phonation, artic-
for more than one frequency measure, such as the dysphonia ulatory precision, and intonation training, as determined by the
severity index and Cepstral Spectral Index of Dysphonia treating SLP. Patients with atrophy underwent injection
(CSID).5,18 These algorithms successfully demonstrated augmentation with calcium hydroxyapatite (n ¼ 5) or lipoinjec-
treatment success; however, therapeutic changes were not tion (n ¼ 5). Finally, patients with UVFP were treated with thy-
measured in the individual frequency-based components of roplasty medialization with Gore-Tex (Gore, Newark, DE).
the algorithms, such as CPP, low-high spectral ratio (L/H ratio), Patients with missing data during initial or follow-up visits
and their respective standard deviations (SDs).18,19 Considering were not included in the study. All patients included underwent
that all voice disorders are not acoustically similar, individual only one type of intervention (either surgery or behavioral voice
cepstral outcomes data would be valuable to determine if any therapy). After exclusion based on all the aforementioned
single cepstral-based measure can capture therapeutic change criteria, a total of 40 patients were included in the study, 10 pa-
and if some cepstral-based measures are disorder specific tients per diagnosis group.
with regard to sensitivity to change.
Given a pervasive lack of outcomes detection after treatment Procedures
by time-based measurements alone, and a growing need for The following information was gathered as part of routine clin-
voice disorder-specific outcomes, the present study aimed to ical examinations. Patients were asked to produce a sustained
build on our author group’s past work redefining the voice lab- /a/ and read the sentence ‘‘we were away a year ago’’ from
oratory assessment. Specifically, the goals of the present study the Consensus Auditory-Perceptual Evaluation of Voice
were to assess treatment success via acoustic voice laboratory (CAPE-V) protocol at their most comfortable pitch and loud-
measurements before and after a single well-defined interven- ness. Recordings of these tasks were completed using the Anal-
tion in patients with four common voice problems and to deter- ysis of Dysphonia in Speech and Voice (ADSV; KayPENTAX,
mine if outcome sensitivity of certain acoustic voice laboratory Montvale, NJ) and Multi-Dimensional Voice Profile (MDVP)
measures varies with disorder type. software from the Computerized Speech Lab (KayPENTAX,
Montvale, NJ). A Shure Beta-54 WBH54 head-mounted micro-
phone (SHURE, Chicago, IL), positioned at approximately 45
MATERIALS AND METHODS
degrees from the participant’s mouth, was used for all
All study procedures were approved by the University of Pitts-
recordings.
burgh Institutional Review Board (IRB #PRO13030372).
Data reduction
Participants Data from each task (sustained vowel and connected speech)
Data were collected retrospectively from patients presenting to were analyzed independently after being identified for inclu-
the University of Pittsburgh Voice Center. Informed consent sion in the study by an individual blinded to the experimental
was obtained from all patients before data entry into a clinical hypotheses using the appropriate protocols within the ADSV
research database. The data were retrieved from the clinical and MDVP software (ie, sustained vowel and all-voiced sen-
research database by a research coordinator blinded to experi- tence protocols). ADSV was used to provide measures of the
mental hypotheses. Patient records from January 2009 to July CPP in the sentence (CPP speech) and vowel (CPP vowel)
2013 were included if records indicated the following inclusion and respective SDs, low-high spectral ratio in the sentence
criteria: age older than 18 years, primary diagnosis of benign (L/H ratio speech) and vowel (L/H ratio vowel) and respective
midmembranous vocal fold lesion(s) (lesions), primary muscle SDs. In addition, a multifactorial estimate of dysphonia
tension dysphonia (MTD-1), vocal fold atrophy (atrophy), or severity, referred to as CSID was calculated for the all-voiced
unilateral vocal fold paralysis (UVFP). Only patients with sentence. The CSID correlates with the labeled visual analog
single-category diagnoses were included (ie, atrophy alone, scale for severity used in CAPE-V. The predicted severity of
not atrophy and UVFP). Diagnoses were determined via a CAPE-V sentences, Ss is calculated using the following
team consisting of a fellowship-trained laryngologist and a formula:
voice-specialized SLP. Data were specifically chosen as pre-
and postintervention measures. To assess the measures’ ability SS ¼ 148:68  ð5:913CPPÞ  ð11:173sCPP Þ
to capture change, and to make a statement about the interven-  ð1:313SRÞ  ð3:093sSR Þ
tion efficacy for common voice problems, the interventions and
follow-up time points for each member of each group were where CPP is the cepstral peak prominence, sCPP is the SD of
identical, and were as follows: Lesion group baseline and 12 the CPP, SR is the L/H spectral ratio, and sSR is the SD of the
months after phonomicrosurgery, all other groups (MTD-1, L/H spectral ratio.20 MDVP was used to analyze NHR of the
atrophy, UVFP), baseline and 6 months after treatment. sustained vowel.
Follow-up time points were chosen based on our standard clin- Connected speech samples were screened for extraneous
ical practice; patients with benign lesions do not typically re- background noise and consonant aspiration into the micro-
turn at a 6-month time point but do routinely follow-up at a phone; only samples with subjectively clean background noise
1-year time point. Patients with MTD-1 underwent, on average, were included. Cursors were placed at the onset of the sentence
584 Journal of Voice, Vol. 28, No. 5, 2014

‘‘we were away a year ago’’ and at the offset; this was aurally In the final group of patients, two of those with lesions were
confirmed with playback. Data were subsequently analyzed ac- considered professional voice users, that is, these individuals
cording to the ADSV program specifications. The middle one relied on their voices as part of their occupation. For patients
second portion of the vowel was selected for vowel anal- with MTD-1, two were amateur singers and two professional
ysis.21,22 Auditory-perceptual evaluations were made during voice users. In the atrophy group, two were amateur singers,
the clinical evaluation from the CAPE-V sentences using the two professional voice users, and one a semiprofessional singer.
zero to three Grade, Roughness, Breathiness, Asthenia and Finally, for the UVFP group, two patients were professional
Strain scale.23 voice users.
Audio-perceptual analyses of the participants revealed the
Statistical methods following. For patients with lesions, the most prominent charac-
Means and SDs for all analyses were calculated. A paired- teristic preintervention was hoarseness, then roughness, fol-
samples t test was used to detect statistical differences between lowed by breathiness and strain, which were rated as equally
acoustic measurements and for the VHI-10, across disorders severe. After intervention, the hoarseness rating improved the
before and after intervention. SPSS, version 20.0 (SPSS, Inc., most, followed by equal decreases in roughness and strain.
Chicago, IL) for Windows was used for all analyses. For patients with MTD-1, pretreatment roughness was rated
as the most severe, followed by hoarseness, strain, and then
breathiness. Hoarseness decreased the most as a result of inter-
RESULTS vention, followed by other categories. For patients with atrophy,
Participants roughness was rated as most severe, followed by hoarseness,
About 8300 patients received treatment at the University of then breathiness, and finally, strain. All parameters except strain
Pittsburgh Voice Center from January 2009 to July 2013. Le- improved after intervention. Finally, for patients with UVFP,
sions, MTD-1, atrophy, or UVFP were diagnosed in 2348 pa- roughness was rated as most severe, followed by breathiness,
tients. Of these, 250 were identified as having only one hoarseness, and strain. Breathiness improved the most after
diagnosis and as having undergone the predetermined interven- treatment.
tion for study inclusion. These records were reviewed chrono-
logically until 10 patients with complete pre- and Statistically significant changes
postintervention data at the determined follow-up time points For the patients with lesions, a significant difference was found
were obtained. Figure 1 depicts a flowchart of participant inclu- in CPP speech (t ¼ 2.97, P < 0.05), CPP SD vowel (t ¼ 3.20,
sion in the study. Table 1 depicts age and sex information of par- P < 0.05), and VHI-10 (t ¼ 3.18, P < 0.05). For patients with
ticipants in each diagnostic group. MTD-1, a significant difference was found in CSID (t ¼ 2.84,

FIGURE 1. Flowchart of all potential participants from identification through final data analysis.
Amanda I. Gillespie, et al Acoustic Analysis of Four Common Voice Diagnoses 585

TABLE 1. TABLE 2.
Mean, SD, and Significance of Each Measurement Before Mean, SD, and Significance of Each Measurement Before
and 12 Months After Excision of Benign Vocal Fold Lesion and 6 Months After Behavioral Voice Therapy for Patients
With MTD-1
Measure Mean SD t P
CPP speech pre 6.12 1.74 2.97 0.016* Measure Mean SD t P
CPP speech post 8.03 1.44 CPP speech pre 6.10 2.20 0.330 0.749
CPP SD speech pre 2.76 0.67 1.28 0.232 CPP speech post 6.35 1.09
CPP SD speech post 2.97 0.58 CPP SD speech pre 2.83 0.74 1.74 0.115
CPP vowel pre 9.58 3.89 1.94 0.084 CPP SD speech post 3.26 0.39
CPP vowel post 11.80 2.59 CPP vowel pre 8.58 2.11 1.53 0.161
CPP SD vowel pre 0.89 0.88 3.2 0.011 CPP vowel post 9.80 2.04
CPP SD vowel post 0.60 0.67 CPP SD vowel pre 1.38 0.93 0.119 0.908
L/H ratio speech pre 28.84 3.74 2.34 0.44 CPP SD vowel post 1.33 0.88
L/H ratio speech post 32.11 3.21 L/H ratio speech pre 31.32 3.62 0.178 0.863
L/H ratio SD speech pre 6.66 1.30 1.96 0.081 L/H ratio speech post 31.05 3.84
L/H ratio SD speech post 5.63 1.24 L/H ratio SD speech pre 6.33 1.79 2.23 0.053
CSID speech pre 22.48 21.51 1.96 0.081 L/H ratio SD speech post 7.33 1.53
CSID speech post 8.54 16.54 CSID speech pre 39.87 30.80 2.84 0.020
VHI-10 pre 23.00 7.24 3.18 0.011 CSID speech post 18.01 20.57
VHI-10 post 13.10 8.62 VHI-10 pre 20.20 8.81 1.72 0.119
NHR pre 0.25 0.19 2.161 0.059 VHI-10 post 14.5 10.42
NHR post 0.13 0.02 NHR pre 0.17 0.05 0.777 0.455
Notes: Statistically significant values are given in italics. NHR post 0.15 0.04
Notes: Statistically significant values are given in italics.

P < 0.05). For patients with atrophy, no comparisons were sta-

tistically significant. For patients with UVFP, a significant dif- with lesions, MTD-1, and UVFP demonstrated statistically sig-
ference was found in CPP SD speech (t ¼ 2.98, P < 0.05) and nificant improvement before and after intervention in at least
CSID (t ¼ 2.31, P < 0.05). No other results reached statistical one acoustic measure. All groups, except vocal fold atrophy,
significance. also showed an improvement in VHI-10 after treatment, indi-
Because of the small sample size, Cohen d effect sizes were cating that most patient groups experienced a reduction in voice
calculated for those measures with P values between 0.05 and handicap after treatment.
0.10. Effect size measures from nonsignificant results may still
indicate the potential for a subtle but important treatment effect
that should be further interpreted by the researcher(s).24–26 For
patients with lesions, a large effect was observed for L/H ratio TABLE 3.
sentence SD (d ¼ 0.81) and NHR (d ¼ 0.88), and a medium Mean, SD, and Significance of Each Measurement Before
effect was observed for CPP vowel (d ¼ 0.67). For patients and 6 Months After CaHA and Lipoinjection
Laryngoplasty for Patients With Vocal Fold Atrophy
with MTD-1, a medium effect was observed in L/H ratio sen-
tence SD (d ¼ 0.60) and VHI-10 (d ¼ 0.59). For patients Measure Mean SD t P
with atrophy, a medium effect size was calculated for L/H ratio CPP speech pre 5.37 2.99 0.112 0.914
sentence SD (d ¼ 0.72). For patients with UVFP, a large effect CPP speech post 5.48 3.22
was observed for CPP sentence (d ¼ 1.01), L/H ratio sentence CPP SD speech pre 2.58 1.22 0.012 0.990
(d ¼ 0.75), and VHI-10 (d ¼ 0.99). Means, SDs, and results of CPP SD speech post 2.59 1.25
paired-samples t test with Cohen d for each disorder are dis- CPP vowel pre 6.03 4.81 1.08 0.309
played in Tables 2–5. CPP vowel post 8.07 5.83
CPP SD vowel pre 0.75 0.54 1.00 0.341
CPP SD vowel post 0.92 0.54
DISCUSSION L/H ratio speech pre 28.50 5.77 0.401 0.698
This study was the first to analyze time- and frequency-based L/H ratio speech post 27.78 6.28
acoustic analyses, independent of an algorithm, after single L/H ratio SD speech pre 7.72 2.17 0.798 0.445
known treatments at identical follow-up time points for patients L/H ratio SD speech post 7.08 1.35
with four carefully selected and mutually exclusive voice CSID speech pre 28.62 33.79 0.002 0.999
disorders: lesions, MTD-1, atrophy, and UVFP. In general, CSID speech post 28.64 42.54
improvement in patient perception of voice handicap, VHI-10 pre 22.00 10.35 0.999 0.344
auditory-perceptual evaluation, and acoustic analyses of voice VHI-10 post 19.10 11.93
NHR pre 0.23 0.14 0.031 0.976
after intervention support the behavioral and surgical treat-
NHR post 0.23 0.13
ments for voice problems in the patients in this study. Patients
586 Journal of Voice, Vol. 28, No. 5, 2014

characteristics of breathiness and hoarseness were improved

TABLE 4.
Mean, SD, and Significance of Each Measurement Before
in these patients, which supports evidence suggesting the domi-
and 6 Months After Thyroplasty Medialization for nance of these characteristics in CPP structure.20 The improve-
Patients With Unilateral Vocal Fold Paralysis ment in L/H ratio in patients with UVFP indicates a more
consistently stable voice posttreatment than pretreatment20
Measure Mean SD t P
and possible reduction in high-frequency spectral noise. Partic-
CPP speech pre 3.72 2.20 1.99 0.078 ipants who improved in CSID also improved in VHI-10, a well-
CPP speech post 5.81 1.91 established voice treatment outcome measure.8 CSID can be
CPP SD speech pre 2.07 .93 2.98 0.016 considered a robust objective acoustic measure because it com-
CPP SD speech post 3.21 .56 prises both cepstral and spectral measures and has been shown
CPP vowel pre 6.20 3.96 1.56 0.153
to correlate with listener-perceived voice severity in response to
CPP vowel post 8.78 2.51
CPP SD vowel pre 1.08 0.93 1.28 0.232
treatment.18
CPP SD vowel post 1.611 0.99 Although it may be premature to revise voice laboratory eval-
L/H ratio speech pre 25.13 6.28 2.05 0.071 uation protocols based on this small-n investigation alone, some
L/H ratio speech post 29.06 3.92 general recommendations can be made from the current data set
L/H ratio SD speech pre 5.85 0.81 0.589 0.571 and supporting literature. First, findings in the present study
L/H ratio SD speech post 6.16 1.34 support past investigations that, for all patients (except atro-
CSID speech pre 52.54 30.30 2.31 0.047 phy), the measure that most reflected change after treatment
CSID speech post 21.29 19.91 was CSID, and therefore CSID may be a cornerstone measure-
VHI-10 pre 24.40 12.28 2.13 0.062 ment for acoustic analyses for all voice problems. This finding
VHI-10 post 13.50 9.51 is in agreement with previous studies on CSID as an outcome
NHR pre 0.30 0.25 0.856 0.417
measure.18 In addition, the measures that used connected
NHR post 0.23 0.08
speech demonstrated greater response to treatment than sus-
Notes: Statistically significant values are given in italics.
tained vowels. CPP speech may be a worthwhile measure to
regularly collect and analyze, especially on patients with pa-
thologies more likely to be treated with surgery, such as large
When results are collapsed across all groups, statistically sig- lesions not appropriate for voice therapy and UVFP. Past liter-
nificant changes were observed in CPP speech, CPP vowel, ature has confirmed that CPP, L/H ratio, their SDs, and CSID
their SDs, CSID, and VHI-10. As hypothesized, no measure re- also correlate well with auditory-perceptual ratings of voice
vealed significant change for all disorders. These findings sup- severity.15,27,28 In the present study, L/H ratio in speech or its
port the hypothesis that a one-size-fits-all approach to voice SD showed large effect sizes in all groups and may be an
outcomes may not be appropriate. When the findings were important outcome measure to track response to surgical and
analyzed by disorder, with the exception of the atrophy group, behavioral intervention.
which did not demonstrate significant change in any parameter, Regarding the second aim of the present study, disorder-
CSID and CPP speech were the most consistent indicators of specific assessment findings and recommendations, some clear
change in response to treatment. The robust and significant trends are evident. The lesion and UVFP groups demonstrated
changes in CSID provide an example whereby a multifactor for- change in more measures compared with the other two disorder
mula appears to be more sensitive to change than individual groups. Patients with lesions may present clinically with a va-
measures. In addition, both measures are taken from connected riety of perceptual voice severities, but as the current data set
speech, which indicates that ecologically valid measures—such indicate, often have substantial phonatory noise impeding the
as those taken during speech and not a single phoneme—may periodic vibratory signal. After phonomicrosurgical lesion
be most appropriate for phonatory analysis. excision, patients showed an improvement in single cepstral
Improvement in CPP speech in the patients with lesions, and measures (CPP) and conglomerate cepstral-based measures
UVFP after surgical treatment, indicates greater harmonic en- (CSID). Analysis of CPP speech, CPP vowel, L/H ratio SD
ergy, a decrease in spectral noise, and corresponds with less se- in speech, and CSID is recommended as part of the outcomes
vere vocal quality as a result of the interventions. Perceptual battery for patients with lesions. Consistent with the literature,
patients with UVFP presented as the most severe and made the
greatest gains across measurements.9 Assessment of both
TABLE 5. vowels and connected speech in frequency-based measures
Measures Demonstrating Both Clinically Meaningful and and vowels in time-based measures may be appropriate for
Statistically Significant Measures for Each Patient Group these patients. This strong response to cepstral-based analyses
is consistent with the literature showing a strong correlation of
Disorder CPP Speech CPP Vowel CSID Speech VHI-10
CPP to the perceptual correlate of breathy vocal quality. In the
Lesions X X X present sample, patients with UVFP had breathy vocal quality,
MTD-1 X and CPP, which is sensitive to this perceptual characteristic,
Atrophy
demonstrated change in response to treatment in UVFP
UVFP X
patients.
Amanda I. Gillespie, et al Acoustic Analysis of Four Common Voice Diagnoses 587

Disorder groups that responded the least to acoustic voice Conclusions

laboratory measurements in general were MTD-1 and atrophy. Results of the present study support surgical and behavioral in-
Patients with MTD-1 demonstrated change in only two acoustic terventions for patients with common voice problems. Stated
measures. This finding is not surprising given the great hetero- more simply, treatment works. The present study also demon-
geneity of presentations of MTD-1.29,30 Patients with MTD-1 strated that a one-size-fits-all approach to voice outcomes
may present with a range of perceptual manifestations testing may not be appropriate. Voice change is not uniformly
including breathy, pressed, or even normophonic. The disorder reflected by all measures for all disorders. A disorder-specific
is also one that may be completely based on the patient’s feel approach to voice outcomes analysis for both clinical and
(ie, increased phonatory effort and fatigue), and in these cases, research purposes is recommended.
acoustic analyses tools would be inappropriate for measuring
change with treatment. The composite measure CSID may be
Acknowledgments
most appropriate for assessing change in patients with MTD-
The authors thank Shaheen Awan, PhD, Christine Harrison, and
1, as it was in the present study.
Olivia Carnes for their invaluable support with the present
Patients with atrophy did not demonstrate significant change
study.
on the measures studied, including the patient-perceptual VHI-
10. Furthermore, not only was no or minimal raw change
observed on these acoustic measures but also their values re- REFERENCES
mained outside the normal range (ie, dysphonic) after treat- 1. Ma EP, Yiu EM. Suitability of acoustic perturbation measures in analysing
ment. Only L/H ratio in speech SD showed a medium effect periodic and nearly periodic voice signals. Folia Phoniatr Logop. 2005;57:
38–47.
size. These results are troubling and perhaps should be taken 2. Roy N, Barkmeier-Kraemer J, Eadie T, et al. Evidence-based clinical voice
in context with a growing body of literature demonstrating a assessment: a systematic review. Am J Speech Lang Pathol. 2013;22:
lack of—or only minimal—improvement in patients with vocal 212–226.
fold atrophy after treatment.31,32 On one hand, arguments could 3. Ma EP, Yiu EM. Multiparametric evaluation of dysphonic severity. J Voice.
be made that an appropriate evaluative metric is not yet in place 2006;20:380–390.
4. Carding PN, Steen IN, Webb A, MacKenzie K, Deary IJ, Wilson JA. The
to reflect change in this patient group. However, data indicate reliability and sensitivity to change of acoustic measures of voice quality.
that multidimensional evaluations—including patient- Clin Otolaryngol Allied Sci. 2004;29:538–544.
perceptual, auditory-perceptual, acoustic, and aerodynamic an- 5. Awan SN, Roy N. Outcomes measurement in voice disorders: application
alyses—are repeatedly unsuccessful in demonstrating a change of an acoustic index of dysphonia severity. J Speech Lang Hear Res.
after treatment in patients with atrophy. Perhaps the problem 2009;52:482–499.
6. Behrman A. Common practices of voice therapists in the evaluation of pa-
lies not in our assessment tools but rather with available inter- tients. J Voice. 2005;19:454–469.
ventions. Although it is beyond the scope of the present study 7. Ziegler A, Gillespie AI, Abbott KV. Behavioral treatment of voice disorders
to hypothesize, future research is needed in both the behavioral in teachers. Folia Phoniatr Logop. 2010;62:9–23.
and surgical interventions for patients with vocal fold atrophy. 8. Rosen CA, Lee AS, Osborne J, Zullo T, Murry T. Development and valida-
tion of the voice handicap index-10. Laryngoscope. 2004;114:1549–1556.
9. Gillespie AI, Gooding W, Rosen C, Gartner-Schmidt J. Correlation of VHI-
Limitations 10 to voice laboratory measurements across five common voice disorders. J
The present study analyzed only a small sample of voice pa- Voice. 31 March 2013; [Epub ahead of print].
tients (n ¼ 40); therefore, clinical significance may hold more 10. Leong K, Hawkshaw MJ, Dentchev D, Gupta R, Lurie D, Sataloff RT. Reli-
conclusive weight than statistical significance for reasons of ability of objective voice measures of normal speaking voices. J Voice.
2013;27:170–176.
statistical power. However, where statistical significance was
11. Heman-Ackah YD, Michael DD, Goding GS Jr. The relationship between
found, greater import can be placed on the robustness of those cepstral peak prominence and selected parameters of dysphonia. J Voice.
particular analyses in the evaluation of voice before and after 2002;16:20–27.
treatment than on measures where only clinical significance 12. Watts CR, Awan SN. Use of spectral/cepstral analyses for differentiating
was found. The small n was a result of meticulous subject normal from hypofunctional voices in sustained vowel and continuous
and treatment selection and therefore may not represent a study speech contexts. J Speech Lang Hear Res. 2011;54:1525–1537.
13. Lowell SY, Kelley RT, Awan SN, Colton RH, Chan NH. Spectral- and
limitation. cepstral-based acoustic features of dysphonic, strained voice quality. Ann
A second potential limitation lies in the measures selected. Otol Rhinol Laryngol. 2012;121:539–548.
Only one spectral-only-based measure was studied, that is, 14. Balasubramanium RK, Bhat JS, Fahim S 3rd, Raju R 3rd. Cepstral analysis
NHR. Other spectral-based measures are commonly collected of voice in unilateral adductor vocal fold palsy. J Voice. 2011;25:326–329.
during acoustic voice evaluations, specifically jitter and 15. Lowell SY, Colton RH, Kelley RT, Mizia SA. Predictive value and discrim-
inant capacity of cepstral- and spectral-based measures during continuous
shimmer. NHR was chosen because it incorporates jitter and speech. J Voice. 2013;27:393–400.
shimmer into its algorithm and is therefore an appropriate 16. Radish Kumar B, Bhat JS, Prasad N. Cepstral analysis of voice in persons
single-measure replacement for three separate measures.21 with vocal nodules. J Voice. 2010;24:651–653.
Measurements of F0 were also not assessed in the present study 17. Solomon NP, Awan SN, Helou LB, Stojadinovic A. Acoustic analyses of
thyroidectomy-related changes in vowel phonation. J Voice. 2012;26:
because the sample comprises males and females; therefore,
711–720.
analysis of averages would be inappropriate. In addition, our 18. Peterson EA, Roy N, Awan SN, Merrill RM, Banks R, Tanner K. Toward
past work, and the work of other author groups, has demon- validation of the cepstral spectral index of dysphonia (CSID) as an objective
strated no change in F0 after voice treatment.9,33 treatment outcomes measure. J Voice. 2013;27:401–410.
588 Journal of Voice, Vol. 28, No. 5, 2014

19. Awan SN, Roy N, Dromey C. Estimating dysphonia severity in continuous 26. Schuele CM, Justice LM. The importance of effect sizes in the interpreta-
speech: application of a multi-parameter spectral/cepstral model. Clin tion of research: primer on research: part 3. ASHA Leader. 2006.
Linguist Phon. 2009;23:825–841. 27. Awan SN, Roy N, Jiang JJ. Nonlinear dynamic analysis of disordered voice:
20. Awan SN, Roy N, Jette ME, Meltzner GS, Hillman RE. Quantifying the relationship between the correlation dimension (D2) and pre-/post-treat-
dysphonia severity using a spectral/cepstral-based acoustic index: compar- ment change in perceived dysphonia severity. J Voice. 2010;24:285–293.
isons with auditory-perceptual judgements from the CAPE-V. Clin Linguist 28. Awan SN, Solomon NP, Helou LB, Stojadinovic A. Spectral-cepstral esti-
Phon. 2010;24:742–758. mation of dysphonia severity: external validation. Ann Otol Rhinol Lar-
21. Baken RJ, Orlikoff RF. Clinical Measurement of Speech and Voice. 2nd ed. yngol. 2013;122:40–48.
San Diego, CA: Singular; 2000. 29. Gillespie AI, Gartner-Schmidt J, Rubinstein EN, Abbott KV. Aerodynamic
22. Awan SN, Giovinco A, Owens J. Effects of vocal intensity and vowel type profiles of women with muscle tension dysphonia/aphonia. J Speech Lang
on cepstral analysis of voice. J Voice. 2012;26:670.e15–670.e20. Hear Res. 2013;56:481–488.
23. Karnell MP, Melton SD, Childes JM, Coleman TC, Dailey SA, 30. Morrison MD, Rammage LA. Muscle misuse voice disorders: description
Hoffman HT. Reliability of clinician-based (GRBAS and CAPE-V) and and classification. Acta Otolaryngol. 1993;113:428–434.
patient-based (V-RQOL and IPVI) documentation of voice disorders. J 31. Thomas LB, Harrison AL, Stemple JC. Aging thyroarytenoid and limb
Voice. 2007;21:576–590. skeletal muscle: lessons in contrast. J Voice. 2008;22:430–450.
24. Ferguson CJ. An effect size primer: a guide for clinicians and researchers. 32. Gartner-Schmidt J, Rosen C. Treatment success for age-related vocal fold
Prof Psychol Res Pract. 2009;40:532–538. atrophy. Laryngoscope. 2011;121:585–589.
25. Kraemer HC, Morgan GA, Leech NL, Gliner JA, Vaske JJ, Harmon RJ. 33. Hufnagle J, Hufnagle K. An investigation of the relationship between
Measures of clinical significance. J Am Acad Child Adolesc Psychiatry. speaking fundamental frequency and vocal quality improvement. J Com-
2003;42:1524–1529. mun Disord. 1984;17:95–100.