Relationships Between Laryngoscopic Analysis Metrics of Supraglottic Compression and Vocal Effort in Primary Muscle Tension Dysphonia

Summary:

Purpose.

Supraglottic compression is thought to underlie vocal effort in patients with primary muscle tension dysphonia (pMTD). However, the relationship between supraglottic compression and vocal effort in this clinical population remains unclear. Gold standard laryngoscopic assessment metrics for supraglottic compression are also lacking. The goals of this study were to identify metrics proposed in the literature that could distinguish patients diagnosed with pMTD from typical voice users and determine their relationships to the vocal effort.

Methods.

Flexible laryngeal endoscopy was performed on 50 participants (25 pMTD, 25 controls). The presence of supraglottic compression was characterized using a categorical (nominal) scale and severity was quantified on ordinal and continuous scales. The three laryngoscopic metrics were correlated with self-perceived ratings of vocal effort on a 100 mm visual analog scale.

Results.

Inter-rater reliability was strongest for the continuous scale (P’s < 0.0001) compared to categorical (P’s < 0.001) and ordinal (P’s < 0.001) scales. The presence of different supraglottic compression patterns varied in both groups, and there were no significant group differences on categorical (P’s > 0.05) scales. Mediolateral (M-L) supraglottic compression was significantly greater in the pMTD group (P < 0.0001), and anteroposterior (A-P) compression was significantly greater in the control group (P = 0.001) using continuous scales. There were no significant relationships between any of the three laryngoscopic metric types and vocal effort ratings (P’s > 0.05), except for a significantly positive relationship between anterior-posterior compression on the ordinal scale and vocal effort in the control group (P = 0.047).

Conclusions.

Continuous scales are reliable and valid for distinguishing individuals with pMTD from those without voice disorders, especially occupational voice users. M-L supraglottic compression may be a better indicator of pMTD than A-P compression. However, the poor correlation between supraglottic compression and vocal effort suggests that one may not influence the other. Future studies should focus on other mechanisms underlying vocal effort in patients with pMTD.

INTRODUCTION

Supraglottic laryngeal compression on laryngoscopy is considered a key component in the diagnosis of primary muscle tension dysphonia (pMTD), a type of functional voice disorder.^1,2 Its characteristics are thought to involve anteroposterior (A-P) aryepiglottic compression from the arytenoid cartilage-to-epiglottic petiole approximation, mediolateral (M-L) approximation of the ventricular (false) folds toward the midline, or a combination of both supraglottic patterns (sphincteric configuration).^3–5 Conventional wisdom is that these supraglottic laryngeal patterns are not only key diagnostic indicators in patients with pMTD,^4–6 but also lead to pathophysiological levels of increased vocal effort in these patients.^1,5,7,8 However, these supraglottic patterns have also been observed in vocally healthy voice users,^3,6,9,10 and the relationship between supraglottic compression and vocal effort is poorly understood.

Various laryngoscopic assessment metrics have been applied to study supraglottic compression patterns in patients with pMTD. These metrics include descriptive categories based on the presence or absence of specific supraglottic features, like A-P, M-L, and sphincteric supraglottic compression (nominal scales).^4,5,11 Other metrics include incremental Likert methods to assess the severity of supraglottic compression (ordinal scales).¹² Still other methods rely on endolaryngeal outlet areas to quantify the severity of supraglottic compression (continuous scales). Many of these laryngoscopic metrics have been developed as means to an end to the study of laryngeal patterns in patients with pMTD without first conducting validation studies on these metrics for the pMTD population. Furthermore, methodological limitations across studies have led to inconsistency of findings, and as a result, supraglottic patterns that define pMTD have remained inconclusive. For example, the presence of supraglottic compression patterns using nominal scales, originally proposed to indicate a pMTD diagnosis,^4,5 has also been observed in vocally healthy voice users.^3,6,9,10 This discrepancy is likely due to the omission of control groups from the original studies proposing these nominal scale methods to define pMTD.

Other studies have proposed that the severity of supraglottic compression may better distinguish pMTD from healthy voice users than its presence, where patients with pMTD have “excessive” supraglottic compression compared to controls that could result in increased perceptions of vocal effort.^6,13,14 Several methods on ordinal and continuous scales have been used to identify compression severity levels to better define laryngeal patterns in pMTD. For example, the Voice-Vibratory Assessment with Laryngeal Imaging (VALI), a classification system based on 0–5 Likert scale clinician ratings (ordinal scale), has been used to quantify supraglottic compression severity¹² and implemented to study relationships between supraglottic compression and vocal effort.⁷ However, only 4 out of the 66 original laryngoscopic examinations used to validate the VALI ordinal scale as a metric of supraglottic compression involved patients with pMTD.¹² Quantification of supraglottic compression using endolaryngeal outlet areas has also been proposed to study supraglottic compression severity on continuous scales.⁶ However, the validation of these methods relies on the subjective identification of laryngeal landmarks for width calculations without clear endolaryngoscopic anatomical boundaries and was originally validated using rigid stroboscopy, which has known effects of anterior tongue pull on laryngeal configurations.^6,15

Finally, previous studies using these laryngoscopic metrics to investigate relationships between supraglottic compression and vocal effort have been validated in vocally healthy controls but not in patients with hyperfunctional voice disorders.^7,8 These studies often ask subjects to manipulate their level of vocal effort to simulate vocal hyperfunction, which can introduce cognitive bias to self-reported ratings of self-perceptual physiological constructs, like vocal effort. Previous studies have also demonstrated that supraglottic compression occurs in vocally healthy individuals without pathophysiological levels of vocal effort.^3,10 Whether supraglottic compression results in or contributes to symptoms of vocal effort requires elucidation.

Therefore, the first goal of this study was to identify laryngoscopic metrics previously proposed in the literature that could best distinguish patients diagnosed with pMTD from those of typical voice users (vocally healthy controls)—comparing these methods across the same dataset of 50 subjects with and without pMTD and accounting for methodological limitations previously noted. The second goal of this study was to determine relationships between these laryngoscopic metrics and self-reported measures of vocal effort.

METHODS

Participants

The protocol was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center. Laryngoscopic examinations via flexible nasoendoscopy were performed in 50 subjects (25 with a diagnosis of pMTD and 25 subjects with no voice deficits or complaints). Patients diagnosed with pMTD were recruited from the UT Southwestern Center for Voice Care (age mean: 49.08 years of age; age range: 23–84 years of age; gender: 76% women, 24% men). Diagnosis of pMTD was made on the basis of the overall impression from laryngoscopy, case history, and symptom reporting using the criteria in the Classification Manual for Voice Disorders¹⁶ by one of three fellowship-trained laryngologists and confirmed by a speech-language pathologist (SLP). Patients diagnosed with pMTD were only included in the study if they scored more than 11 out of 40 on the Voice Handicap Index-10 (VHI-10)¹⁷ based on normative values by Arffa et al¹⁸ and more than 24 on Part 1 of the Vocal Fatigue Index (herein referred to as VFI-Part1). VFI-Part1 has previously been shown to be robust in distinguishing pathological levels of vocal fatigue from typical vocal fatigue experienced with increased vocal demands.¹⁹ This inclusion criterion was chosen because vocal effort is thought to be a precursor to vocal fatigue in patients with pMTD.^5,7,20 We did not include vocal effort as an inclusion criterion because this parameter was used as the dependent (outcome) variable in this study.

Subjects in the control group were recruited from performing arts and education communities in the greater Dallas area (age mean: 29.95 years of age; age range: 20–60 years of age; gender: 80% women, 20% men). For the control group, subjects were required to score less than 5 out of 40 on the VHI-10¹⁸ and not score more than 1 (almost never) out of 4 (always) on any of the 11 questions on the VFI-Part1. The decision to include occupational voice users in the control cohort was made because occupational voice users are especially at risk for pMTD.^21–24 Comparing patients with pMTD (many of which are occupational voice users) to vocally healthy occupational voice users would provide direct comparisons to better distinguish supraglottic contributions underlying pathological levels of vocal effort in patients with pMTD from adaptive supraglottic patterns that might occur with heavy vocal demands common in occupational voice users with and without pMTD.

Exclusion criteria for both groups included history of voice therapy as well as pulmonary pathology, throat clearing, chronic cough, and laryngopharyngeal reflux—all of which could potentially increase supraglottic compression. Additional exclusion criteria included structural, neurological, or phonotraumatic laryngeal or voice pathologies that could be attributed to the clinical presentations of vocal deficits or vocal production dysfunction.

Laryngoscopy

Participants underwent laryngoscopy under xenon light using a standard flexible nasoendoscope (Olympus distal chip, model ENF-VH) with lidocaine and oxymetazoline spray as part of standard-of-care clinical procedures. As only supraglottic compression was measured in this study, no stroboscopic examinations were conducted. Laryngoscopy was also used to rule out any structural or neurological pathologies in both the pMTD and control groups. Six sustained /i/ (“eee” as in “neat”) utterances were each held for about 3–5 seconds at a comfortable pitch and loudness and recorded for each subject, per clinical standards and subsequent analysis for each of the 50 laryngoscopic examinations.

Inter-rater reliability laryngoscopy training

Two raters blinded to the condition and diagnosis completed all laryngeal analyses. Both were voice-specialized SLPs. One had 2 years of experience, while the other had 15 years of experience. Prior to applying categorical, ordinal, and continuous assessment methods to the same set of laryngeal exams (n = 50; described in the Measures of Supraglottic Comperssion section), raters underwent interrater reliability training for each laryngoscopic assessment method. The training was performed on a random set of 50 additional laryngeal examination videos and 50 digital still images from 134 laryngoscopic videos and 600 digital still images. Inter-rater reliability with training on the nominal categorical scale was Fleiss’ kappa (κ) = 0.872, P < 0.001 for identification of laryngeal exams without supraglottic compression; (κ) = 1.000, P < 0.001 for identification of M-L compression, and (κ) = 0.894, P < 0.001 for identification of A-P compression. The weighted quadratic kappa for the ordinal scale was (κ_w) = 0.593, P < 0.001 for A-P compression, and (κ_w) = 0.603, P < 0.001 for M-L compression. The intraclass correlation coefficients (ICCs) for the continuous rating scale were 0.957, P < 0.0001 for A-P compression, and ICC = 0.965, P < 0.0001 for M-L compression.

Measures of supraglottic compression using categorical, ordinal, and continuous scales

Categorical Supraglottic Laryngoscopic Scale

Categorical (nominal) scales based on previous work by Koufman and Blalock and Morrison and Rammage were used to characterize supraglottic compression patterns for each laryngeal video (Table 1). An “X” under each type was placed in an Excel spreadsheet by each rater to denote the presence of each supraglottic pattern on laryngoscopy. If multiple features were present within the same laryngeal examination, more than one X was placed in the row for each applicable subject. When there were discrepancies between the two raters, the raters came together to reach a consensus, and final consensus ratings were used to determine group differences via statistical analysis. Inter-rater reliability was assessed between the two raters’ independent ratings prior to consensus with Fleiss’ kappa (κ) using SPSS software (IBM SPSS Statistics, v29.0.0.0., Chicago, IL).

TABLE 1.

Prevalence of Mediolateral and Anteroposterior Supraglottic Compression Using Categorical (Nominal) Scale

	A. No supraglottic Compression	B. Mediolateral Compression	C. Anteroposterior Compression
pMTD	24%	44%	76%
Control	8%	8%	52%

Notes: A. Typical exam with no supraglottic compression present; B. mediolateral hyperadduction of the false vocal folds; C. anteroposterior supraglottic contraction of the aryepiglottic folds, resulting in reduced space between the epiglottis and arytenoids. Percentages were calculated as $\frac{\mathit{\text{number}}of\mathit{\text{times}}\mathit{\text{laryngoscopic}}\mathit{\text{feature}}\mathit{\text{present}}}{25\left(\mathit{\text{group}}n\right)}$ and as such, will not add up to 100%.

The percentage prevalence of each feature divided by group (pMTD, control) was calculated, and group differences were statistically analyzed using Fisher’s exact test. Correlations between categorical supraglottic metrics and vocal effort in patients with pMTD and controls were analyzed using the eta (η²) coefficient.

Ordinal Supraglottic Laryngoscopic Scale

The VALI scale (described in the INTRODUCTION) was used to rate the severity of A-P and M-L supraglottic compression on a 0–5 Likert scale for each laryngeal examination video.¹² Each examination was rated separately for the A-P and M-L supraglottic planes, according to the VALI instructions. The two sets of ratings were averaged for statistical analysis. Inter-rater reliability between the two raters was determined using weighted quadratic kappa (κ_w). Differences between groups (pMTD, control) were statistically analyzed using the Mann-Whitney U test. Correlations between VALI and vocal effort ratings in patients with pMTD and typical voice users were analyzed using Spearman’s rank-order correlation.

Continuous Supraglottic Laryngoscopic Scale

Digital still images of the steady-state vowels of each of the six /i/ productions were captured from each laryngeal exam (50 exams × 6 steady-state vowels = 300 images). Endolaryngeal outlets for each image were traced manually using the boundaries of the true vocal folds, interarytenoid mucosa, petiole of the epiglottis, and ventricular folds, using a customized ImageJ macro in Fiji (ImageJ version 1.53q, Bethesda, MD). The A-P length of the endolaryngeal outlet was also manually measured in ImageJ by placing a line within the medial space of the vocal folds, starting at the anterior commissure or petiole of the epiglottis and posteriorly into the interarytenoid mucosa to normalize laryngeal exams. The M-L width of each endolaryngeal outlet parameter was automatically obtained via pixels (light blue shaded areas in Figure 1) by identifying all widths along the length of the A-P outlet (red lines in Figure 1) and then averaging all these widths to obtain the mean width (dark blue lines in Figure 1). This method reduced the subjective nature of identifying the middle of the larynx without overt laryngeal landmarks from which to work. M-L supraglottic compression was determined by (LO/W²) × 100, where LO is the endolaryngeal outlet area and W is the average width of the outlet (in pixels). A-P supraglottic compression was determined by (LO/AP²) × 100, where LO is the endolaryngeal outlet area, and AP is the A-P outlet length (in pixels). The set of A-P and M-L compression measurements acquired from each rater’s tracings was averaged so that each subject received one area measure for A-P supraglottic compression and one for M-L supraglottic compression. Inter-rater reliability was assessed using ICC. Group differences (pMTD, control) were analyzed using mixed-model analyses of variances (ANOVAs). Correlations between M-L and A-P supraglottic compressions and vocal effort in patients with pMTD were determined using Pearson’s correlations.

FIGURE 1.

Normalization using endolarynx length and width. A. Supraglottic configuration in “typical” laryngeal exam, B. mediolateral supraglottic compression, and C. anteroposterior supraglottic compression.

Vocal effort ratings

Participants were asked to place a tick mark on a sheet of paper with a 100 mm visual analog line based on the current amount of physiological exertion it took to get their voice out prior to the lidocaine/oxymetazoline administration and laryngeal examination. Tick marks were measured with a ruler for each participant, and the assigned number representing the mark on the 100 mm scale was placed in a spreadsheet for subsequent correlation analysis for each of the three laryngoscopic methods.

RESULTS

Categorical Supraglottic Laryngoscopic Scale

The distribution of descriptive laryngeal features using the categorical scale is presented in Table 1. Fleiss’ kappa was (κ) = 0.612, P < 0.001 for inter-rater reliability of identification of exams without the presence of supraglottic compression; (κ) = 0.686, P < 0.001 for M-L compression; and (κ) = 0.728, P < 0.001 for A-P compression. There were no statistically significant differences between the group distributions across supraglottic features (P = 0.050). No one feature occurred in all participants in either group, and many features occurred variably in both groups (see percentages in Table 1). Categorical supraglottic metrics accounted for 2% of the vocal effort rating scores in patients with pMTD (η² = 0.023) and < 1% of vocal effort ratings in typical voice users (η² = 0.008).

Ordinal Supraglottic Laryngoscopic Scale

The weighted kappa for the inter-rater reliability rating level was (κ_w) = 0.698, P < 0.001 for A-P compression and (κ_w) = 0.856, P < 0.001 for M-L compression using the VALI ordinal rating scale. VALI severity scores were significantly higher in the pMTD group for the M-L (P = 0.027) but not in the A-P supraglottic planes (P = 0.067) (Figure 2, frequency of each 0–5 level). There were no significant correlations between VALI severity scores and vocal effort in patients with pMTD for either M-L (r_s = −0.146; P = 0.487) or A-P (r_s = 0.249; P = 0.231) supraglottic compression. There were also no significant correlations between VALI severity scores and vocal effort in typical voice users for M-L supraglottic compression (r_s = 0.00; P = 0.990). However, there were significant positive correlations between A-P supraglottic compression and vocal effort in the control group (r_s = 0.401; P = 0.047).

FIGURE 2.

Quantitative Measures of Supraglottic Compression using Ordinal Scale. Frequency of supraglottic compression using 0–5 ordinal VALI scale for A. mediolateral compression and B. anteroposterior compression.

Continuous Supraglottic Laryngoscopic Scale

The ICC on the continuous rating scale was 0.941 (P < 0.0001) for A-P supraglottic compression and 0.949 (P < 0.0001) for M-L supraglottic compression. Subjects with pMTD had significantly more M-L supraglottic compression (P < 0.0001), whereas the control group had significantly more A-P supraglottic compression (P < 0.001) (Figure 3). There were no significant correlations between M-L compression and vocal effort (r = −0.126, P = 0.547) or A-P compression and vocal effort (r = 0.186, P = 0.374) in the pMTD group. There were also no significant correlations between M-L compression and vocal effort (r = −0.333, P = 0.111) or A-P compression and vocal effort (r = 0.212, P = 0.321) in the control group using this continuous scale.

FIGURE 3.

Quantitative Measures of Supraglottic Compression using Continuous Scale. A. Typical occupational voice users had significantly more anteroposterior supraglottic compression when normalized to endolarynx length; B. patients with pMTD had significantly higher mediolateral supraglottic compression when normalized to endolarynx width. ***P < 0.001.

DISCUSSION

The goals of this study were to identify laryngoscopic metrics of supraglottic compression proposed in the literature that could distinguish patients diagnosed with pMTD from typical voice users and determine their relationships to self-perceived measures of vocal effort. The results of this study demonstrate that the presence of supraglottic laryngeal patterns, thought to be a sign of hyperfunction (A-P compression, M-L compression, sphincteric compression), occurs in both pMTD and typical occupational voice users. There were no laryngoscopic supraglottic features specific to, or only present in, the pMTD population using the nominal scaling system.

Although M-L supraglottic compression was prevalent in both groups, it was significantly more severe in the pMTD group when quantified using ordinal and continuous scales. M-L supraglottic compression (ie, ventricular activity) has previously been attributed to poor vocal quality and may be more active when increased intrathoracic or intra-abdominal pressures are needed, possibly in the context of poor breath support.^25,26 Ventricular movement is also involved in upper airway protection, suggesting laryngeal sensitivities may play a role in patients with pMTD.²⁷ However, using M-L supraglottic compression to define pMTD should be used with caution, as not everyone in the pMTD cohort exhibited M-L supraglottic compression patterns based on nominal scale findings. Furthermore, the poor correlation between M-L compression and vocal effort in patients with pMTD suggests that these M-L compression patterns do not underlie vocal effort. Other mechanisms may play a role in the self-perception of vocal effort (eg, vocal effort may have more to do with the lower respiratory system than the laryngeal system²⁸).

Interestingly, A-P supraglottic compression was more prevalent and severe in occupational voice users without vocal dysfunction on a continuous scale. Morrison and Rammage⁵ previously suggested that these supraglottic patterns are indicative of vocal fatigue. However, this cohort experienced little to no vocal fatigue, as determined on the VFI-Part1. These results raise questions regarding the relationship between vocal fatigue and A-P supraglottic laryngeal patterns. These data also support previous literature that found increased A-P supraglottic patterns in rock and musical theater singers,²⁹ which may better reflect laryngeal adaptation or compensation than pathophysiology. Several studies have suggested that these laryngeal patterns may also be an inertance mechanism for boosting vocal output.^30,31 The study findings in the present study are not surprising, considering participants in the control group were all occupational voice users.

In this study, a positive correlation was observed between vocal effort and A-P supraglottic compression using the VALI ordinal scale in the control group. These relationships suggest that this (possibly adaptive) supraglottic mechanism comes with sensations of increased vocal effort in healthy voice users. These patterns are supported by previous studies demonstrating a positive correlation between supraglottic compression and vocal effort in healthy controls⁷ and could have something to do with the types of vocal demands the occupational voice users in the control group encounter. Further investigations on the relationships between these supraglottic patterns, vocal effort, and vocal fatigue are needed in cohorts with and without hyperfunctional voice disorders. Studies on the effects of vocal training type on vocal effort and supraglottic configuration patterns are also needed.

Limitations

This study had several potential limitations. The first is that the introduction of the laryngoscope into the nasopharynx could influence supraglottic configurations; the second was the use of lidocaine during laryngoscopy. Although the goal was to emulate laryngeal exams performed as close to the standard-of-care as possible—as findings would be most relevant to clinicians using these methods in the voice clinic—there is always the possibility these standard-of-care laryngeal visualization procedures could influence laryngeal somatosensory feedback or laryngeal motor patterns. However, because self-perceived severity of vocal effort was conducted prior to lidocaine administration and laryngoscopy, measures of vocal effort should not have been impacted.

The third limitation is that a laryngologist was not one of the raters. However, we previously utilized both SLPs and laryngologists as blinded raters for a similar study using laryngoscopic metric methods with similar rater outcomes.¹⁰ Thus, we did not expect the profession of the raters to have a significant impact on the results of the present study.

The fourth possible limitation is the difference in ages between the pMTD and control cohorts. Although we recruited more women than men in the control group to mimic gender demographics of pMTD, we did not account for age. The mean difference between the two groups was 20 years (50 in the pMTD group and 30 in the control group). Although we do not anticipate supraglottic patterns to change drastically between 30 and 50 years of age, there is a possibility that age, even in the absence of vocal fold atrophy, could influence laryngeal patterns. Future investigations are needed to determine the impact of supraglottic compression patterns between individuals with and without pMTD, across different age groups.

The fifth limitation is that although we recruited participants in the control group who were occupational voice users to keep the two groups as homogenous as possible and to better distinguish pathological levels of vocal effort and maladaptive supraglottic compression from typical vocal effort and adaptive supraglottic compression that can occur with heavy vocal demands, we did not collect data on the level and type of formal vocal training each subject received. Future investigations on the relationships between level and type of training, supraglottic patterns, and vocal effort are needed.

The final limitation is the potential of circular logic. pMTD is not well defined, and although inclusion criteria beyond clinical diagnosis were considered, it is possible that patients in the pMTD cohort may have had another type of voice disorder and may have been erroneously included in the experimental group.

CONCLUSION

There are two main conclusions drawn from this study. First, measures of M-L supraglottic compression severity appear to be the most sensitive laryngoscopic metrics to define patients with pMTD and distinguish them from typical occupational voice users, with some stipulations. This M-L supraglottic pattern may be present in some patients with pMTD and, on average, may be more severe in these patients. However, the presence of M-L supraglottic compression is not a diagnostic sign or indication of pMTD, since these patterns were not universally present in the pMTD group and also occurred in the control group. These supraglottic laryngeal patterns also do not appear to significantly contribute to the symptoms of pathological vocal effort patients with pMTD experience, suggesting that other pathophysiological mechanisms underlying vocal effort in pMTD may be at play.

Data Availability

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.