Comparative Study of Two Semi-Occluded Vocal Tract Protocols: A Randomized Clinical Trial

Abstract

Introduction:

Semi-occluded vocal tract exercises (SOVTEs) are widely used as a therapeutic tool to create flow resistance in the upper airway. The current study was a randomized controlled clinical trial to establish the efficacy of two SOVTE protocols, flow-resistant tube (FRT) and Lessac-Madsen Resonant Voice Therapy (LMRVT). Exploratory investigations included a noninferiority analysis of FRT to the widely adopted therapy protocol (LMRVT), as well as examining the dosing required to improve acoustic measures and subjective ratings.

Method:

Sixty-seven participants with voice disorder were randomized into one of five groups: 4-week FRT (n = 14), 8-week FRT (n = 19), 4-week LMRVT (n = 15), 8-week LMRVT (n = 5), and control (n = 14). Voice Handicap Index (VHI) and Vocal Fatigue Index scores were collected pre- and posttreatment. Acoustic analysis using the Acoustic Voice Quality Index was completed. We compared VHI between controls and 8-week FRT and LMRVT, adjusting for pre-VHI using linear regression. We examined the efficacy of 4-week protocols relative to controls and conducted a noninferiority comparison of FRT (4 and 8 weeks) to LMRVT (4 and 8 weeks) using 5- and 10-point margins. Finally, we compared the 4- versus 8-week sessions for both therapies.

Results:

A statistically significant reduction of VHI in both 8-week FRT relative to controls (−10.60, 95% CI [−19.80, −1.40], p = .025) and 8-week LMRVT (−15.74, 95% CI [−29.40, −2.08], p = .025) was found. We also found an improvement in 4-week FRT relative to controls (−10.11, 95% CI [−20.03, −0.20], p = .046), but the 4-week LMRVT result was not statistically significant (p = .057). FRT was found to be noninferior to LMRVT in terms of VHI using a 10-point margin (FRT − LMRVT: 0.69, 95% CI [−5.76, 7.15], p = .01), but not using a 5-point margin (p = .054). There were no statistically significant differences in VHI scores between 4- and 8-week sessions for either therapy.

Conclusions:

Both FRT and LMRVT improved VHI scores relative to controls. FRT was noninferior to LMRVT in terms of VHI scores. There were no statistically significant differences in VHI scores between 4- and 8-week therapy sessions.

Semi-occluded vocal tract exercises (SOVTEs) have been validated as a mainstay treatment technique in the rehabilitation of voice disorders for many years (Kapsner-Smith et al., 2015; Meerschman et al., 2017, 2019; Roy, 2012; Roy et al., 2001, 2003; Stemple, 2005; Titze, 2006a, 2006b). Exercises that can be classified as SOVTEs include tongue trills, lip trills, lip–tongue trills (raspberries), flow-resistant tube (FRT) phonation, vocal tract elongation with tubes, hums, voiced fricatives, and lip-rounded vowels such as /o/ and /u/ (Cox & Titze, 2023). Structured approaches have been coined as vocal function exercises (Stemple et al., 1994), Lessac-Madsen Resonant Voice Therapy (LMRVT; Verdolini Abbott, 2008), the accent method (Kotby et al., 1991), LaxVox (Sihvo & Denizoglu, 2007), straw phonation (Kapsner-Smith et al., 2015; Laukkanen et al., 2008, 2012; Titze, 2006a, 2006b), and more. An expanded bibliography is available in Cox and Titze (2023). Although numerous targeted options for semi-occluding the vocal tract exist, the goal of creating a flow resistance (and therefore a steady oral pressure behind the semi-occlusion) remains common to all (Titze, 2009; Titze et al., 2002). Physiologically, intraoral pressure created above the glottis can dramatically unpress the vocal folds. This unpressing (reduced vocal fold collision) allows a large degree of stretching of the vocal folds without injury (Titze & Verdolini Abott, 2012). It also produces a more favorable medial surface posture of the vocal folds for self-sustained oscillation (Titze et al., 2021). Additionally, recent literature has supported using a variably occluded face mask (Awan et al., 2019) in healthy adults to improve aerodynamic and acoustic voice quality characteristics. The work of Awan et al. (2019) was translated to vocally disordered populations (i.e., functional and structural pathologies) by Gillespie et al. (2022), and similar improvements in acoustic and aerodynamics were found.

SOVTEs have been demonstrated to passively expand the airway (pharynx, larynx, and trachea), resulting in increased space for resonance options by the speaker (Vampola et al., 2011). Guzman, Jara, et al. (2017) and Guzman, Laukkanen, et al. (2013) documented a lowered laryngeal position within the airway, a 38% increase in pharyngeal area, and a 73% increase in epilaryngeal space during FRT exercises. The effects of improved pharyngeal area, lowered laryngeal position, and increased epilaryngeal space were also found following straw phonation exercises. Story et al. (2000) used computer simulation to show that acoustic airway impedances can be optimized by FRT. In current literature, it has been established that tube length is less critical if the inner diameter of the tube is 3.0 mm or less, particularly when trying to target specific impedance matching between the source and the airway (Titze et al., 2021, 2022).

Acoustically, the source–filter interaction theory (Titze, 2008) has demonstrated that SOVTEs can optimize the source for maximum power transfer and spectral variety, offering a return of energy from the vocal tract to the source. The benefits of greater choices in airway expansions and contractions—and therefore greater options in vowel and voice quality contrast—are brought about by oral semi-occlusions (Titze et al., 2021).

Historically, understanding dosage in voice therapy and the optimal dose for the rehabilitation of voice disorders has been questioned (Bane et al., 2019; Roy, 2012). Most of the evidence used to determine the dosage of frequency of exercises to be completed, the number of rehabilitation sessions, and home programming has been based chiefly on clinical impressions and anecdotal evidence. More recently, researchers have begun to explore optimal dosage following surgical procedures, pharmaceutical management, and behavioral voice therapy (Bane et al., 2019; Belsky et al., 2021; Kapsner-Smith et al., 2015; Meerschman, 2018).

Two of the above-mentioned rehabilitation therapies, FRT (known clinically as straw phonation exercises; Titze, 2006a, 2006b) and vocal function exercises (Stemple et al., 1994), have been shown to be equally efficacious for the treatment of varying voice disorders (Frisancho et al., 2020; Kapsner-Smith et al., 2015). Here, we make a second comparison between two SOVTEs, LMRVT and FRT. These two therapy protocols were chosen for this study because of their contrasting approach to daily practice. A significant difference is that LMRVT relies heavily on the immediate embedding of semi-occlusions into speech, while FRT is a nonspeech practice expected to carry over into speech more or less automatically. Despite both being considered gold-standard treatments, several questions regarding the adherence and economy of these approaches remain unanswered, primarily due to the uncontrolled, observational nature of current reports.

The current study sought to compare LMRVT and FRT in treating patients with functional voice disorders. We aimed to understand the commonalities of these two approaches and make a case for practical application in a rehabilitation setting. We hypothesized that both therapies would elicit significant improvements in clinical voice measures relative to controls. Exploratory analyses included a comparison of treatment dosages within and across treatment types.

Method

Subjects

Study procedures and recruitment processes were approved by The University of Utah Institutional Review Board (IRB 00083506). Potential participants were referred to the study following evaluation and diagnosis by a qualified otolaryngologist and speech-language pathologist (SLP) specializing in voice disorders. Laryngoscopy was performed on all participants, and all participants met the criteria of functional dysphonia based on the visualization of the larynx and clinical impressions. Functional dysphonia was defined as a voice disorder in the “absence of any structural or neurologic laryngeal pathology” (Roy, 2003). Eligibility for the study required participants to be at least 18 years old. One hundred eighty individuals were evaluated for eligibility for the study. Of that, 76 met the inclusion criteria that were based on the following: The patient was experiencing a current speaking voice problem (i.e., vocal fatigue, voice quality changes, and/or increased vocal effort), had not previously undergone voice therapy, and were determined to be good candidates for voice therapy by the evaluating SLP or otolaryngologist. Participants were excluded if (a) there was evidence of laryngeal pathology or lesions that required immediate medical attention (i.e., surgery, pharmaceutical treatment), (b) required voice rest and/or restricted voice use due to their pathology, (c) had a neurogenic etiology to their dysphonia, and/or (d) voice therapy was contraindicated by the first author (A.H.-S.). Three subjects were eliminated from the study due to a laryngeal infection that presented during the study and required medical attention and treatment. Another subject's data were eliminated due to a motor vehicle accident during the study. Lastly, two subjects withdrew from the study prior to completing their therapy.

Sixty-seven subjects completed all therapy sessions, and pre- and posttherapy data were evaluated. The average age of subjects was 43.0 years old (SD = 15.2), with a range of 19–84 years old. Demographic information is available in Table 1.

Table 1.

Summary of on-study variables stratified by semi-occluded vocal tract exercise protocol.

Vocal tract protocol		Control	LMRVT protocol		FRT protocol
Variable			4 weeks	8 weeks	4 weeks	8 weeks
		n = 14	n = 15	n = 5	n = 14	n = 19
Age, years	M ± SD	37.2 ± 15.1	51.9 ± 15.8*	41.2 ± 19.0	41.5 ± 17.9	41.7 ± 8.7
	Missing, n (%)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
Sex, n (%)	Female	10 (71.4)	11 (73.3)	4 (80.0)	10 (71.4)	11 (57.9)
	Male	4 (28.6)	4 (26.7)	1 (20.0)	4 (28.6)	8 (42.1)
VHI pre	M ± SD	46.8 ± 14.7	39.0 ± 20.2	54.0 ± 25.8	41.0 ± 22.3	45.2 ± 18.6
	Missing, n (%)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
VHI post	M ± SD	48.1 ± 16.7	33.1 ± 19.8	37.4 ± 15.4	33.9 ± 19.2	36.4 ± 20.1
	Missing, n (%)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
VFI pre	M ± SD	41.5 ± 13.8	37.4 ± 11.0	45.6 ± 20.9	44.0 ± 18.9	42.8 ± 14.8
	Missing, n (%)	0 (0.0)	1 (7.1)	0 (0.0)	0 (0.0)	0 (0.0)
VFI post	M ± SD	42.7 ± 14.4	35.5 ± 12.9	34.0 ± 11.2	30.2 ± 14.6	34.2 ± 16.5
	Missing, n (%)	0 (0.0)	2 (15.4)	0 (0.0)	1 (7.7)	0 (0.0)
AVQI pre	M ± SD	3.5 (0.57)	3.8 (0.66)	3.5 (0.65)	4.1 (1.3)	3.1 (0.76)
	Missing, n (%)	0 (0.0)	0 (0.0)	0 (0.0)	1 (7.1)	1 (5.3)
AVQI post	M ± SD	3.4 (0.71)	3.5 (0.68)	3.1 (0.52)	4.2 (1.2)	3.0 (0.72)
	Missing, n (%)	1 (7.1)	1 (6.7)	0 (0.0)	1 (7.1)	1 (5.3)

Note. LMRVT = Lessac-Madsen Resonant Voice Therapy; FRT = flow-resistant tube; VHI = Voice Handicap Index; VFI = Vocal Fatigue Index; AVQI = Acoustic Voice Quality Index.

^*Significant at p < .05 compared to controls. Statistical significance was assessed for age and sex only to assess randomization. Results for comparing controls to VHI and VFI are provided in Tables 2 and 3.

Sample Size Consideration

Sample size calculations were conducted using data from Kapsner-Smith et al. (2015) for our primary comparisons of LMRVT to control and FRT to control. However, there were slight differences in power calculations between the time the data were collected (several years ago) and more recent analyses. Inputs to our sample size calculations included LMRVT mean Voice Handicap Index (VHI) ± standard deviation at baseline of 45 ± 18, posttreatment of 30 ± 23, correlation between baseline and post of r = .66; FRT, 39 ± 15, 20 ± 15, r = .37; and control 40 ± 18, 38 ± 14, which resulted in sample sizes of n = 27 per group to achieve 80% power for comparing LMRVT to control using a two-sided alpha (.05) comparison in an analysis of covariance (ANCOVA) model. A sample size of n = 22 per group was needed to compare FRT to control.

Procedure and Randomization

Participants were randomized into one of five groups: (a) control, (b) 4-week LMRVT, (c) 8-week LMRVT, (d) 4-week FRT, or (e) 8-week FRT. Group assignment was randomized using a computerized random number generator, meaning that each group did not accrue subjects at the same rate. The COVID-19 pandemic ended this trial abruptly and prematurely, resulting in unequal allocations across groups for final analysis. Participants assigned to the control group completed the initial assessment and data collection protocol to allow participants with a voice disorder to receive treatment. An intent-to-treat statistical design and control of outcome variables were employed (Hariton & Locascio, 2018). The participants then waited 4 weeks without receiving therapy before completing a “postcontrol” assessment. Finally, members of this group were re-randomized into one of the four treatment groups. All participants underwent a 2-hr clinical voice evaluation for the treatment groups at the National Center for Voice and Speech (NCVS) main office at The University of Utah in Salt Lake City, UT. The initial case history portion of the evaluation was conducted by a licensed SLP (A.H.-S.). All further acoustic data were collected by a staff member (S.S.) who was not involved in administering voice therapy to avoid bias. All therapy sessions were conducted by A.H.-S., who had over 10 years of experience treating patients with voice disorders. A.H.-S. was blinded to group assignment (case history) until after the initial evaluation. Given the nature of the two treatment groups, it was impossible to keep the treating clinician blinded to the group. However, S.S. was blinded to all group assignments for the entirety of the study.

During the initial evaluation, imaging was obtained if the patient had not had a prior laryngoscopy or had not been evaluated by the referring ENT physician. The laryngeal images were saved and then reviewed later by an otolaryngologist. Images were reviewed from the referring SLP and/or physician for all other participants. Initial evaluations included a case history, completion of the VHI (Jacobson et al., 1997), Vocal Fatigue Index (VFI; Nanjundeswaran et al., 2015), a vocal health questionnaire, and a series of voicing tasks. Acoustic tasks were obtained in a 125-ft² acoustically treated room with an ambient noise floor of 42 dB (C weighting). Each patient was equipped with a head-mounted microphone and laryngeal electroglottograph. The evaluator wore a lapel microphone and saved the recorded signal data. All signals were obtained via ADInstruments PowerLab analog/digital converter, saved in a proprietary LabChart7 file format, and analyzed using MATLAB and PRAAT (see Figure 1).

Figure 1.

CONSORT diagram for randomization. FRT = flow-resistant tube; LMRVT = Lessac-Madsen Resonant Voice Therapy.

Questionnaires

Vocal Health Questionnaire

A comprehensive voice questionnaire was originally designed by the NCVS and has been utilized in previous studies by the NCVS (Kapsner-Smith et al., 2015). Questions from this intake form are directed toward habitual voice use, medications, medical history, current voice complaints, and strategies the person uses to assist with their voice complaints.

VHI

The VHI is a 30-item self-reported quantitative measure of the psychological, functional, and physical severity of a voice disorder. This validated questionnaire assesses the magnitude of the voice-related problem the participant experiences (Jacobson et al., 1997) based on a rating score of 0–4, wherein 4 represents the question always applies and 0 indicates the statement never applies.

VFI

The VFI is a psychometrically validated self-reported questionnaire that uses a 19-item question to quantify vocal fatigue in individuals (Nanjundeswaran et al., 2015). On this scale, 0 represents the statement always applies and 0 indicates the statement never applies.

Acoustic Voice Quality Index

The Acoustic Voice Quality Index (AVQI) is an automated acoustic measure developed to provide a more robust objective assessment of voice quality using salient acoustic features. This analysis is sensitive to mildly dysphonic voices and is clinically relevant to pre- and posttreatment (Barsties & Maryn, 2015). A 3-s segment of sustained voicing /ɑ/ is concatenated with a running speech sample (the second and third sentences from the Rainbow Passage). This acoustic recording is then analyzed for periodicity and perturbation measures. The underlying acoustic analysis is described by Maryn et al. (2010).

Treatment programs. Participants were seen once weekly, in person, for a total of either four or eight sessions, depending on group randomization. All participants were given a handout on “caring for the voice” (see the Appendix) at their initial evaluation, whether in a treatment or control group. Participants were asked to comply with all handout recommendations, including treating mucus, reducing laryngeal irritants, behavioral reflux precautions, and good vocal hygiene.

FRT

Treatment sessions ran between 55 and 60 min, depending on the instruction needed by the participant, and focused exclusively on nonspeech SOVTEs. Each participant was also given a home exercise program with a prerecorded audio demonstration of the exercises to be completed 4 times per day for approximately 4 min (1 min per exercise). Kapsner-Smith et al. (2015) provide a complete description. In brief, the program consisted of four exercises performed with a thin straw, 5 cm long and 2.5 mm in diameter. They completed 10 tokens of each exercise, more if targeted production was not achieved. In Exercise 1, the patient was asked to glissando up and down, taking a breath partway through if needed. Exercise 2 was a series of accents/hills where loudness and pitch were varied. In Exercise 3, the participant was asked to sing a song, targeting an excessive/stretched production through the straw. Lastly, Exercise 4 was a reading exercise through the straw without articulation for 1 min. Participation in home programming was tracked on a worksheet, and participants returned the worksheet weekly at their session.

LMRVT

Participants randomized to the LMRVT arm were also seen once weekly, in person, for a total of either four or eight sessions. Treatment sessions ran between 55 and 60 min and followed the clinician and patient manual for LMRVT (Verdolini Abbott, 2008). LMRVT, created and studied extensively by Katherine Verdolini Abbott, initially focuses on a nasal resonance that is transitioned to generalized forward placement in speech. Resonant voicing is a primary speech-centered voice therapy approach used by voice clinicians and has been codified in programs such as LMRVT (Verdolini Abbott, 2008). LMRVT uses semi-occluded mouth shapes such as voiced fricatives or nasals /m/ and /n/ to identify and habituate “resonant” voice productions, which have been defined clinically as “easy to produce and buzzy in the facial tissues” and scientifically as “a reinforcement of the source by the vocal tract” (Titze et al., 2011). Subjects participated in a series of exploratory exercises that utilized semi-occlusions (e.g., voiced consonants) and moved into connected speech and conversation. The training goal is to establish “resonant” voicing, characterized by barely adducted vocal folds and sensations of ease and oral vibrations.

The original LMRVT protocol was designed to be completed over 8 weeks. Participants in the 8-week LMRVT group received the protocol as initially designed. Participants in the 4-week LMRVT group received a condensed protocol where content from all eight original sessions was still provided but in half the number of sessions. Home programming for participants in both 4- and 8-week groups was to practice twice daily as indicated in the LMRVT manuals.

Statistical Analysis

Total VHI and VFI were calculated by summing the scores for each individual question to create a cumulative score for each index. Given that the last three questions of the VFI questionnaire are inversely related to the score, we reversed the scoring of these questions so that larger values indicate worse VFI. We reported means and standard deviations for our continuous variables and reported counts and column percentages for our categorical variables. We checked randomization by comparing age and sex between LMRVT (at 4 and 8 weeks separately) versus controls and FRT (at 4 and 8 weeks separately) versus controls using two-sample t tests for age and chi-square tests for sex. There were no notable deviations from normality among these five groups; however, sample sizes were small for these assessments.

We assessed efficacy of FRT and LMRVT protocols at 4 and 8 weeks considered separately, relative to controls in terms of postprotocol VHI, using univariable and multivariable linear regression models adjusting for preprotocol VHI. Adjusting for baseline VHI resulted in an ANCOVA modeling framework that improves statistical power for randomized controlled trials (Fitzmaurice et al., 2004). In this model, a five-level therapy indicator variable (control, LMRVT 4 weeks, LMRVT 8 weeks, FRT 4 weeks, FRT 8 weeks) was included, where the comparisons of LMRVT 8 weeks to controls and FRT 8 weeks to controls were our primary comparisons of interest (the 4-week protocol versions were secondary). Due to adjustment for baseline VHI, the regression coefficients for LMRVT at 8 weeks and FRT at 8 weeks represent both the mean difference in VHI at 8 weeks between the therapies and controls and also the change from baseline in VHI between the therapies and controls. These results were accompanied by 95% confidence intervals (CIs) and p values. The same analysis was repeated for our exploratory outcomes of VFI and AVQI. Regression assumptions were checked using plots of residuals versus fitted values, residuals versus leverage, and qq plots.

As an exploratory analysis, we compared the 4-week to the 8-week protocol using VHI and VFI posttreatment measures within each therapy type (LMRVT, FRT) using two-sample t tests. We report the mean differences (4-week post − 8-week post) with their 95% CIs and p values.

As a further exploratory analysis, we assessed the noninferiority of FRT (4 and 8 weeks combined) to LMRVT (4 and 8 weeks combined) using the same ANCOVA regression framework described above on the postprotocol VHI and VFI outcomes. For this analysis, we constructed a two-sided 90% CI for the mean difference in therapies (FRT – LMRVT) and compared it to a noninferiority margin (δ). If the upper bound of the 90% CI for this difference did not include the margin, then the noninferiority of FRT relative to LMRVT was demonstrated (Piaggio et al., 2012). We considered margins of both δ = 5 and δ = 10 points in our noninferiority assessments between FRT and LMRVT based on the published data by Kapsner-Smith et al. (2015).^1,2 We reported both unadjusted and adjusted noninferiority results, where the adjusted results included the baseline values of the outcomes, similar to the ANCOVA regression framework described above. SAS (Version 9.4; SAS Institute Inc.) was used to perform all analyses. Statistical significance was assessed at the .05 level.

Results

There were 15 patients in the 4-week LMRVT, five patients in the 8-week LMRVT, 14 patients in the 4-week FRT, 19 patients in the 8-week FRT, and 14 patients in the control group. There were no statistically significant sex differences between the LMRVT groups versus controls and FRT groups versus controls (see Table 1). Patients were younger in the control group (37.2 ± 15.1) relative to the LMRVT 4-week group (51.9 ± 15.8, p = .031), but there were no other age differences.

The total VHI post for participants who were assigned to the resonant voice arm at 8 weeks was significantly lower than the control group (coefficient = −15.74, 95% CI [−29.40, −2.08], p = .025), adjusting for VHI pre (see Table 2). VHI post was also significantly lower in the 4- and 8-week FRT groups, where the 4-week group was 10.11 points lower (95% CI [−20.03, −0.20], p = .046) and the 8-week group was 10.60 points lower (95% CI [−19.80, −1.40], p = .025) compared to controls after adjusting for VHI pre. Results were similar for our secondary VFI outcome, where the VFI post was significantly lower in the resonant voice arm at 8 weeks, relative to controls, and significantly lower in the 4- and 8-week FRT groups relative to controls (see Table 3).

Table 2.

Comparison of each group to the control group adjusting for VHI pre.

Model	Variables	Coefficients [95% CI]	p
Univariable	VHI pre	0.71 [0.53, 0.88]	< .001
	Therapy indicator
	LMRVT 4 weeks	−14.94 [−28.99, −0.88]	.038
	LMRVT 8 weeks	−10.67 [−30.37, 9.03]	.283
	FRT 4 weeks	−14.18 [−28.47, 0.12]	.052
	FRT 8 weeks	−11.70 [−25.02, 1.62]	.084
	Control	(reference)	(reference)
Multivariable	VHI pre	0.70 [0.53, 0.87]	< .001
	Therapy indicator
	LMRVT 4 weeks	−9.49 [−19.28, 0.30]	.057
	LMRVT 8 weeks	−15.74 [−29.40, −2.08]	.025
	FRT 4 weeks	−10.11 [−20.03, −0.20]	.046
	FRT 8 weeks	−10.60 [−19.80, −1.40]	.025
	Control	(reference)	(reference)

Note. Boldface indicates the significance level is p < .05. VHI = Voice Handicap Index; CI = confidence interval; LMRVT = Lessac-Madsen Resonant Voice Therapy; FRT = flow-resistant tube.

Table 3.

Comparison of each group to the control group adjusting for VFI pre.

Model	Variables	Coefficients [95% CI]	p
Univariable	VFI pre	0.72 [0.54, 0.89]	< .001
	Therapy indicator
	LMRVT 4 weeks	−7.14 [−18.44, 4.16]	.211
	LMRVT 8 weeks	−8.68 [−23.96, 6.60]	.26
	FRT 4 weeks	−12.45 [−23.75, −1.15]	.031
	FRT 8 weeks	−8.47 [−18.80, 1.86]	.106
	Control	(reference)	(reference)
Multivariable	VFI pre	0.74 [0.58, 0.91]	< .001
	Therapy indicator
	LMRVT 4 weeks	−3.80 [−11.19, 3.59]	.308
	LMRVT 8 weeks	−11.72 [−21.69, −1.75]	.022
	FRT 4 weeks	−12.62 [−19.97, −5.26]	.001
	FRT 8 weeks	−9.42 [−16.15, −2.69]	.007
	Control	(reference)	(reference)

Note. Boldface indicates the significance level is p < .05. VFI = Vocal Fatigue Index; CI = confidence interval; LMRVT = Lessac-Madsen Resonant Voice Therapy; FRT = flow-resistant tube.

There were no statistically significant differences between the 4- and 8-week protocols for either LMRVT or FRT (see Table 4). For the VHI post, the unadjusted mean difference between FRT and LMRVT was 1.12 (90% CI [−7.97, 10.20], δ = 5 p = .239, δ = 10 p = .054) and the mean difference after adjusting for VHI pre was 0.69 (90% CI [−5.76, 7.15], δ = 5, p = .134; δ = 10, p = .010; see Table 5). The reduction in VHI from the FRT method versus the LMRVT method was only comparable at the δ = 10 margin.

Table 4.

Mean difference between 4 and 8 weeks by therapy and outcome.

Variables	Resonant voice		FRT
	Mean difference [95% CI]	p	Mean difference [95% CI]	p
VHI post	−4.3 [−23.8, 15.2]	.632	−2.5 [−16.6, 11.6]	.723
VFI post	1.5 [−12.6, 15.7]	.809	−4.0 [−15.3, 7.4]	.478

Note. Mean difference is defined as 4 week post minus 8 week post. FRT = flow-resistant tube; CI = confidence interval; VHI = Voice Handicap Index; VFI = Vocal Fatigue Index.

Table 5.

Noninferiority tests from a multivariable regression model adjusting for VHI pre.

Protocol	Unadjusted mean (SE) or [90% CI]	p		Adjusted mean (SE) or [90% CI]	Adjusted p
		δ = 5	δ = 10		δ = 5	δ = 10
FRT (4 and 8 weeks)	35.32 (3.39)a			35.16 (2.37)
LMRVT (4 and 8 weeks)	34.20 (4.15)a			34.46 (3.04)
FRT – LMRVT	1.12 [−7.97, 10.20]b	.239	.054	0.69 [−5.76, 7.15]a	.134	.010

Note. CI = confidence interval; δ = noninferiority margin; FRT = flow-resistant tube; LMRVT = Lessac-Madsen Resonant Voice Therapy.

^aStandard error.

^b90% CI.

For the VFI post, the unadjusted mean difference of FRT and LMRVT was −1.48 (90% CI [−7.94, 4.97], δ = 5 p = .049, δ = 10 p = .002), and the adjusted mean difference was −3.26 (90% CI [−7.96, 1.44], δ = 5 p = .002, δ = 10 p < .001; see Table 6). Thus, noninferiority of FRT to LMRVT was demonstrated for the VFI outcome at both the 5- and 10-point margins. The average post-VFI score was 28.77 (SE = 1.67) for the FRT group and 32.03 (SE = 2.24) for the LMRVT group in a model adjusting for VFI pre.

Table 6.

Noninferiority tests from a multivariable regression model adjusting for VFI pre.

Protocol	Unadjusted mean (SE) or [90% CI]	p		Adjusted mean (SE) or [90% CI]	Adjusted p
		δ = 5	δ = 10		δ = 5	δ = 10
FRT (4 and 8 weeks)	29.41 (2.30)a			28.77 (1.67)
LMRVT (4 and 8 weeks)	30.89 (3.09)a			32.03 (2.24)
FRT – LMRVT	−1.48 [−7.94, 4.97]b	.049	.002	−3.26 [−7.96, 1.44]	.002	< .001

Note. Boldface indicates the significance level is p < .05. VFI = Vocal Fatigue Index; CI = confidence interval; δ = noninferiority margin; FRT = flow-resistant tube; LMRVT = Lessac-Madsen Resonant Voice Therapy.

^aStandard error.

^b90% CI.

Acoustic Findings

In the analysis that did not adjust for AVQI pre, the FRT 4-week group had a higher AVQI post than controls (0.82, 95% CI [0.19, 1.45], p = .013). In the analysis adjusting for AVQI pre, there were no differences in AVQI post for any therapy group relative to controls (see Table 7).

Table 7.

Comparison of each group to the control group adjusting for AVQI pre.

Model	Variables	Coefficients [95% CI]	p
Univariable	AVQI pre	0.80 [0.65, 0.95]	< .001
	Therapy indicator
	LMRVT 4 weeks	0.15 [−0.47, 0.76]	.638
	LMRVT 8 weeks	−0.32 [−1.16, 0.52]	.463
	FRT 4 weeks	0.82 [0.19, 1.45]	.013
	FRT 8 weeks	−0.41 [−0.99, 0.17]	.172
	Control	(reference)	(reference)
Multivariable	AVQI pre	0.74 [0.57, 0.90]	< .001
	Therapy indicator
	LMRVT 4 weeks	−0.03 [−0.44, 0.38]	.875
	LMRVT 8 weeks	−0.30 [−0.86, 0.25]	.288
	FRT 4 weeks	0.33 [−0.11, 0.76]	.146
	FRT 8 weeks	−0.08 [−0.47, 0.31]	.683
	Control	(reference)	(reference)

Note. Boldface indicates the significance level is p < .05. AVQI = Acoustic Voice Quality Index; CI = confidence interval; LMRVT = Lessac-Madsen Resonant Voice Therapy; FRT = flow-resistant tube.

Discussion

SOVTEs, a class of vocalization techniques, have been used for almost a century (Engel, 1927) as therapeutic exercises targeted at improving voicing (e.g., voicing efficiency, voice quality, voice function). SOVTEs intend to either lengthen, occlude (narrow), or both lengthen and occlude the vocal tract at one or more locations. The hypothesized purpose of the occlusion during these exercises is to reduce glottal resistance, maximize power transfer from the source to the airway, position the vocal folds for a low threshold pressure, and widen the entire airway to optimize cross-sectional area differences along the vocal tract (Cox & Titze, 2023; Story et al., 2000; Titze, 2006b; Titze et al., 2021). Within the last decade, a proliferation of research articles on the scientific findings underscoring the utility of SOVTEs has emerged (Andrade et al., 2014; Bele, 2005; Bonette et al., 2020; Calvache et al., 2020; Costa et al., 2011; Croake et al., 2017; Dargin et al., 2016; Dargin & Searl, 2015; Frisancho et al., 2020; Guzman, 2017; Guzman, Castro, et al., 2013; Guzman, Laukkanen, et al., 2013; Guzman et al., 2016; Guzman, Jara, et al., 2017; Guzman, Miranda, et al., 2017; Kapsner-Smith et al., 2015; Laukkanen et al., 2008, 2012; Meerschman et al., 2019). These findings, combined with clinical anecdotes of improved voice functioning in patients and clients, have challenged researchers to understand the complex interaction of SOVTEs and improvements in vocalization.

The current study compared the two different mainstay approaches in voice therapy, FRT and LMRVT, to controls. For both therapies, there was a significant improvement using the 8-week protocols in terms of both VHI and VFI outcomes. The FRT group also had improved outcomes under the 4-week protocol.

In an exploratory analysis, findings suggest that FRT is noninferior to LMRVT, such that FRT participants performed as well as participants in the LMRVT group based on subjective participant reports (VHI and VFI). However, for our primary VHI outcome, we only concluded noninferiority using the wider 10-point margin with a two-sided 90% CI. We also found no difference in improvement across the 4- versus 8-week groups. However, due to differences in group size and being underpowered for this comparison, the data cannot be interpreted to indicate that eight sessions of either treatment are equivalent to four. Thus, further research is needed on the appropriate dosage in voice therapy.

AVQI has become a clinically useful tool in tracking even minimal changes in voice quality (Barsties v. Latoszek et al., 2019) for the determination of dysphonia. However, the current study found no significant improvement in AVQI values.

Mechanistically, SOVTEs are beneficial through increased inertive reactance of the air column in the vocal tract. Acoustic impedance is defined as “a ratio of sinusoidal acoustic pressure applied to the resulting sinusoidal acoustic flow at the input of a wave-carrying tube” (Story et al., 2000). For vocalization, acoustic impedance is the measurement of the relationship between pressure and flow on a given cross-sectional area of the vocal tract. Several key findings on inertive reactance of the vocal tract have come from computation modeling. Overall, it has been found that all SOVTEs increase the inertive reactance of the vocal tract air column in the frequency range most commonly used in conversational speech (Story et al., 2000; Titze & Laukkanen, 2007). The increase of inertive reactance in the fundamental frequency range (100–200 Hz) alters the interaction and subsequent functioning of the vocal tract and vocal folds by reducing phonation threshold pressure (Horáček et al., 2019; Kang et al., 2019; Story et al., 2001; Tangney et al., 2021).

As part of the study's original design, adherence to home practice was an aim across sessions. Given several missing data points, an accurate representation of home practice adherence could not be investigated. On average, participants in the LMRVT groups practiced zero to once per day, whereas those in the FRT group practiced at least twice a day. Documentation from clinician notes indicated that participants reported the FRT practice sessions were easier to complete and less cognitively challenging. It could be speculated that the easier tasks resulted in higher rates of practice adherence in the FRT group.

Several limitations existed in the current study. First, the study was not adequately powered, and the group sizes were unequal. The COVID-19 pandemic severely impacted our ability to finish the course of voice therapy with several of our participants. Although underpowered, FRT and LMRVT showed improvement relative to controls. Another limitation was that all therapy was conducted by one SLP. To reduce bias, the treating clinician did not participate in postrecordings or assessments.

Conclusions

This is the largest prospective, randomized clinical trial to date evaluating two mainstay voice therapy approaches, FRT and LMRVT. FRT and LMRVT were found to improve the VHI and VFI relative to controls based on the participant's subjective reports. In an exploratory analysis, FRT was found to be noninferior to LMRVT, though mean changes in pre/post-VHI were on the small limit of clinical relevance—largely due to the mild average VHI scores at intake.

This study was the first to compare four versus eight sessions of treatment duration. The findings suggest that four sessions of FRT or LMRVT improved their subjective report and acoustic output, as well as eight sessions for participants with mild to mild-to-moderate dysphonia. More studies are needed to explore the effect of dosage on different voice disorders and treatment types.

Data Availability Statement

The data that support the findings of this study are available upon request and approval from The University of Utah Institutional Review Board. Restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available.

Appendix

Caring for Your Voice

HYDRATION

It is important to keep the vocal folds moist, both internally and externally. External dehydration can come from breathing dry air, breathing with an open mouth, smoking, and medications that can be drying. Internal dehydration comes from too much caffeine, alcohol, drying drugs, or sweating without replacing fluids.

How do I keep my vocal folds externally hydrated? Inhaling steam or possibly using a medication called a “mucolytic” that helps keep secretions thin.

How do I keep my vocal folds internally hydrated? The best way is to drink plenty of water. (**There are some instances in which a person should not drink large amounts of water, such as those with congestive heart failure.**)

Why is hydration important for the voice? Well-hydrated vocal folds vibrate with less “push” from the lungs, as compared to dry folds. Therefore, it requires less effort to voice. Also, well-hydrated vocal folds may resist injury from voice use more than dry folds, and may recover better from existing injury than dry folds

IRRITANTS

Laryngopharyngeal reflux (LPR) refers to a spill-over of acids from the esophagus onto the vocal folds. This is different from Gastroesophageal reflux (GERD) in which stomach acids back up into the esophagus only.

How does LPR affect the voice? LPR can inflame the vocal folds and make people more likely to get certain vocal fold pathologies. At the very least, it aggravates existing pathologies and slows healing.

Which types of voice conditions may benefit from LPR treatments? Any patient with a diagnosis of LPR should receive treatment for it, which is established by a physician. There are three types of treatment available:

1. Behavioral: Sometimes these precautions are enough to reduce LPR voice symptoms

   1. Limiting foods that trigger reflux. Common foods that trigger reflux include: spicy foods (including tomatoes, oranges, grapefruits, and their juices), alcohol, caffeine, and fatty foods (e.g., fried foods, chocolate).

   2. Elevate the head of the bed.

   3. Do not eat 3–4 hours before lying down.

   4. If heavy around the waist, consider losing weight.

2. Medicines: These are prescribed and monitored by a physician

   1. H2 blockers: block histamine, which is one element involved in acid formation in the stomach. Examples include: Tagamet and Zantac.

   2. Proton Pump Inhibitors (PPI): block formation of acid secretions altogether by blocking hydrogen ions, which allow stomach acids to be formed. Examples of PPIs include: Nexium (esomeprazole magnesium), Prilosec (omeprazole), and Prevacid (lansoprazole).

3. Surgery: Reserved for more persistent cases, especially those that do not get better with behavioral treatments and medicines.

   1. This approach involves a very delicate surgery called “fundoplication.” In this surgery, a tiny scope is inserted through the belly button. There are different methods, but basically a part of the stomach, called the “fundus,” is wrapped around the lower part of the esophagus, to construct a tight sphincter that keeps acids from entering the esophagus.

A few comments on smoking: Smoking is detrimental to your health and ranks as the most significant factor as the cause of head, neck, lung, and other cancers. In addition to general health issues, smoking is also bad for your voice. Smoke is a vocal fold irritant. The chemicals in the smoke and the heat of the smoke constantly irritate the vocal fold tissues and dry them out. Over time, smoke can cause long-term changes in vocal fold mucosa. One result may be a drop in the speaking pitch or singing range. Also, various kinds of vocal fold lesions can develop, including cancerous ones. Please also be aware of second-hand smoke. Avoid smoky environments whenever possible.

SCREAMING, HOLLERING, & BACKGROUND NOISE

What are screaming, hollering, and background noise, and how may they affect the voice? These terms are fairly self-explanatory. What is not so obvious is how often they can occur, sometimes without even knowing it. Sports events, or environments with background noise such as restaurants, bars, parties, choirs, and even church and professional gathering may be particularly “tricky” for the voice. In these situations, you may end up using your voice much more strongly, and for much longer, than you realize. By the time the event is finished, you may discover that you are hoarse. But by that time, some damage to the vocal fold tissue may already have been done.

What are solutions for voice in these situations? First of all, you should simply strike screaming from your list of what you expect to do in your life, period. This recommendation applies not only to people with voice problems, but to everybody. The exception is clearly when danger is involved, and screaming is needed to obtain help or for defense purposes. Another exception is if you have received training in safe screaming from a specialized theatre trainer. Otherwise, you should realize that every scream literally could be your last. Single, strong screams are sometimes sufficient to damage vocal fold tissue permanently, in a way that is difficult to fix later.

Funding Statement

This work was supported by National Institute on Deafness and Other Communication Disorders Grant 1R01DC017998 awarded to Ingo Titze. The authors thank David Palmer, Karin Titze Cox, and the team at ENT Specialist in Salt Lake City, UT, for their invaluable contributions to participant referrals. Additionally, we extend great appreciation to Zhining Ou and Angela Presson, biostatisticians at The University of Utah Division of Epidemiology, for their comprehensive statistical analyses.