Changes in Cough Airflow and Acoustics After Injection Laryngoplasty

Abstract

OBJECTIVE/HYPOTHESIS

We explored the following hypotheses in a cohort of patients undergoing injection laryngoplasty: (1) glottic insufficiency affects voluntary cough airflow dynamics and restoring glottic competence may improve parameters of cough strength, (2) cough strength can be inferred from cough acoustic signal, and (3) glottic competence changes cough sounds and correlates with spectrogram morphology.

STUDY TYPE/DESIGN

Prospective Interventional study

METHODS

Subjects with glottic insufficiency secondary to unilateral vocal fold paresis, paralysis, or atrophy, and scheduled for injection laryngoplasty completed instrumental assessment of voluntary cough airflow using a pneumotachometer and a protocolized voluntary cough sound recording. A Wilcoxon signed-rank test was used to compare the differences pre- and post- injection laryngoplasty in airflow and acoustic measures. A Spearman rank order correlation was used to evaluate the association between airflow and acoustic cough measures.

RESULTS

Twenty-five patients (13F:12M, mean age 68.8) completed voluntary cough airflow measurements, and 22 completed cough sound recordings. Following injection laryngoplasty, patients had a statistically significant decreased peak expiratory flow rise time (PEFRT) (Mean change: −0.03s, SD, 0.06, p=0.04) and increased cough volume acceleration (CVA) (Mean change: 13.1L/s², SD: 33.9, p=0.03), suggesting improved cough effectiveness. Correlation of cough acoustic measures with airflow measures showed a weak relationship between PEFRT and acoustic energy (Coefficient: −0.31, p=0.04) and peak power density (Coefficient: −0.35, p=0.02).

CONCLUSIONS

Our study thus indicates that injection laryngoplasty may help avert aspiration in patients with glottic insufficiency by improving cough effectiveness, and that improved cough airflow measures may be tracked with cough sounds.

INTRODUCTION

The larynx plays a critical role in airway protection both during swallow and through the ejection of aspirated material with cough.¹ Deterioration in this protective mechanism may severely impact health. Aspiration or the entry of secretions or foods below the level of the true vocal folds may lead to pneumonia. In particular, glottic insufficiency, or the presence of a gap between the vocal folds during phonatory adduction leading to unintentional air escape,² may affect airway protection mechanism. Glottic insufficiency may be the result of vocal fold paralysis, paresis, atrophy or scar. Presence of glottic insufficiency has been associated with increased risk of aspiration on objective swallow evaluation, such as video-fluoroscopy. For instance, aspiration has been noted on video-fluoroscopy of 23 to 38% of patients with glottic insufficiency due to unilateral vocal fold paralysis.^3,4 Though evidence is not robust and mostly based on uncontrolled case series, vocal fold medialization has been suggested as an intervention to prevent aspiration when glottic insufficiency is present.⁵ The underlying mechanistic assumption is that improved glottic competence improves airway protection during swallow, and may also improve cough strength, thereby reducing aspiration risk. Prospective investigation has however been noted to be lacking to support this intervention.⁵

The inverse relationship between cough strength and pneumonia risk is well established. Increased peak cough flow, as a measure of cough strength, has been associated with decreased pneumonia risk in patients with stroke, dysphagia and head and neck radiation.^6,7 The cough mechanism involves an aerodynamic sequence of inspiration, compression and expulsion, protecting the lungs through the ejection of secretions and foreign material from the airway.⁸ As such, dystussia, or disordered cough, can be ominous in the dysphagic population. Cough restoration and strengthening techniques have centered around rehabilitative strategies, such as expiratory muscle strength training (EMST), and pharmaceutical cough triggering agents (e.g. angiotensin-converting enzymes inhibitors).⁹ Improvement in peak expiratory pressure capacity and voluntary cough peak expiratory flow with EMST may lead to improved airway protection in part through more effective clearance subglottic and tracheal secretions.^10,11

The larynx has a central place in cough generation, aside from its sensory input in the cough reflex. Glottic closure is necessary for the compressive phase of cough, during which thoracic muscle contraction forces air to the trachea and the closed glottis, thus increasing subglottic pressure, allowing acceleration of airflow through opening glottis during the expulsive phase. Glottic closure may thus be important for the compressive phase of cough and cough airflow acceleration, the key component of cough effectiveness. Aerodynamic parameter measurements (including compression phase duration, expiratory rise time, expiratory phase peak airflow and cough volume acceleration) with pneumotachograph represents the gold standard for cough strength.¹² There is emerging data that acoustic measures of cough can provide quantitative information on cough strength. Prior studies have found that sound pressure level, power and peak energy parameters, may correlate with airflow measures of cough strength.^13–16 This area of investigation is novel, and never explored in otolaryngology.

Remarkably, little research has focused on enhancing glottic closure in improving cough strength for aspiration prevention.¹⁷ Currently, there are only three publications examining the value of injection laryngoplasty in increasing cough strength, excluding an early publication that mentions a qualitative improvement in cough strength after injection medialization.¹⁸ The first study evaluated airflow measures of voluntary cough using a pneumotachograph before and immediately after injection laryngoplasty in only three patients with glottic insufficiency, finding improved airway flow measures.¹⁹ The second study evaluated the peak airflow measurements of maximal voluntary cough using a flow meter in fourteen patients with unilateral vocal fold paralysis before, immediately after or at follow-up clinic visit after injection laryngoplasty.¹⁷ This study was limited by the low number of patients and the measurement technique of cough airflow. The third and most recent study examined the impact of injection laryngoplasty on cough airflow in six patients with stroke-related aspiration risk and laryngoscopic findings of glottic insufficiency. The authors reported significantly improved cough peak expiratory airflow after medialization, albeit with limited number of subjects and airflow measurements made with peak flow meters and not a pneumotachometer.²⁰ Thus, to date, there is no definitive clinical outcomes study establishing the value of injection laryngoplasty in improving cough strength.

Our central hypothesis was that glottic insufficiency may affect cough airflow dynamics and restoring glottic competence may improve voluntary cough strength, as measured by cough peak expiratory airflow. Improving voluntary cough strength in patients with glottic insufficiency and swallowing dysfunction could thus be beneficial in improving airway clearance of aspirated material. Our secondary hypotheses were that cough strength may be inferred from cough acoustic signal and that glottic competence changes cough spectrogram morphology. The objectives of our study were threefold: (1) to prospectively evaluate the effect of injection laryngoplasty on voluntary cough strength in patients with glottic insufficiency with gold standard pneumotachometer airflow measures, (2) to assess the correlation of acoustic measures of cough with airflow measures, and (3) to determine whether improved glottic closure after injection laryngoplasty may amplify the voiced portion of the cough sound signal.

METHODS

Study population

After Institutional Review Board approval, prospective recruitment began for patients with a diagnosis of glottic insufficiency secondary to unilateral vocal fold paresis, paralysis, or atrophy, scheduled for injection laryngoplasty, at a single institution with three participating laryngologists.

Inclusion criteria were as follows:

1. Adult patients with a diagnosis of unilateral vocal fold paralysis, paresis or bilateral vocal fold atrophy, based on strobo-video-laryngoscopy by a laryngologist, with associated glottic insufficiency and normal contralateral vocal fold mobility and absence of cricoarytenoid fixation;

2. Adult patients with diagnosis of laryngeal neuropathy presumed to be secondary to iatrogenic (e.g. cervical or thoracic surgery, endovascular cardiac instrumentation), inflammatory, cerebrovascular or idiopathic etiologies;

3. Adult patients scheduled to undergo injection laryngoplasty with calcium hydroxylapatite, hyaluronic acid or carboxymethylcellulose in the office or in the operating room.

Exclusion criteria were as follows:

1. Patients under 18 years of age;

2. Patients with Unilateral vocal fold immobility or hypomobility thought to be secondary to cricoarytenoid fixation or laryngeal neoplasm;

3. Patients without glottic insufficiency;

4. Patients with bilateral vocal fold paralysis or paresis;

5. Patients with findings of other laryngeal pathologies on office laryngoscopy;

6. Patients with current respiratory infection or a known history of laryngotracheal stenosis.

Materials and methods

Injection laryngoplasty, also called injection augmentation, has been a mainstay in the management of glottic insufficiency. It is an alternative to an Isshiki type I thyroplasty, and is preferred to the latter when there is uncertainty about the permanency of a paresis/paralysis. Initially commonly performed in the operating room, the procedure has gained popularity as an office procedure in recent years with the development of distal chip laryngoscopy, availability of new injectables and improved delivery of laryngeal anesthesia. All participants who chose to undergo injection laryngoplasty, whether in the office or in the operating room, were considered for the study.

The injectable most commonly used at our practice is calcium hydroxylapatite, known for its long-term stability, without significant inflammation and mucosal wave perturbation. Other injectables, such as autologous fat, hyaluronic acid or carboxymethylcellulose were also included, with recognition of their variable durability. Injection laryngoplasty may be performed via several routes: peroral (both in clinic and in the operating room), transcricothyroid, transthyroid cartilage, and transthyrohyoid. All of these approaches were available to included subjects, with the exception of the clinic peroral approach. Patients who meet inclusion criteria were offered participation in the study, regardless of the technique used. We excluded patients with bilateral vocal fold hypomobility/paralysis, as these patients may undergo staged bilateral injections and they are more prone to transient shortness of breath, which could affect the inspiratory phase of cough.

Patients meeting our inclusion criteria were offered to participate in the study. Demographic data were collected, including age, sex, and body mass index (BMI), along with the laryngeal diagnosis causing glottic insufficiency (vocal paralysis, vocal fold paresis or vocal fold atrophy) and a history of pneumonia or dysphagia based on the electronic medical record. Patient reported dysphagia was assessed before and after injection augmentation using the Eating Assessment Tool (EAT-10), a validated and widely used outcome instrument for dysphagia.²¹ In consenting patients, the research team collected cough strength measurements in an existing soundproof room in the clinic area using a pneumotachometer for airflow measurements and acoustic recording for sound analysis. Pneumotachograph airflow measurements is noninvasive and is considered the gold standard for cough airflow measurements. These measurements were performed at diagnosis in the clinic, and at the follow-up visit after injection laryngoplasty. Following American Thoracic Society (ATS) spirometry guidelines, subjects were placed in the standard seated position and were asked to perform a maximal voluntary cough following demonstration of the same by a research team member.²² Subjects were instructed to take a deep inspiration, then cough as forcefully as possible for two practice attempts, or more if needed. Once the subject demonstrated proficiency with voluntary cough, three consecutive voluntary coughs interspaced by deep inhalation were recorded with the pneumotachograph. The patients were instructed to hold the facemask tightly against their face to create a seal before performing the three recorded voluntary coughs.

Cough data were acquired by fitting participants with a facemask for mouth breathing only with the nasal compartment sealed off (Adult Mask 7920 and 7930, Hans Rudolph, Shawnee, KS). The facemask was coupled to a pneumotachograph (Non-heated pneunomotachometer 0–800L/min, model 4813, Hans Rudolph, Shawnee, KS) and amplifier (Model 1110B, Hans Rudolph, Shawnee, KS). After pneumotachometer calibration, cough airflow signals were digitized and recorded using LabChart Pro (ADInstruments, Dunedin, New Zealand) software to a desktop computer via PowerLab 4/35 Data Acquisition System (ADInstruments, Dunedin, New Zealand).

We analyzed maximal voluntary cough following the usual analytic method in the literature.¹² A trained rater identified cough airflow waveforms on the LabChart tracings (Figure 1), and completed the following measurements and calculations in LabChart Data Pad:

1. Peak expiratory flow rate (PEFR; Liters/second): the peak airflow rate during the expiratory phase of cough

2. Peak expiratory pressure (PEP, mmHg): the peak pressure during the expiratory phase of cough

3. Compression phase duration (CPD; seconds): the time from the end of the inspiratory phase to the beginning of the expiratory phase¹⁹

4. Peak expiratory flow rise time (PEFRT; seconds): the time from the start of the expiratory phase start to the peak expiratory flow, which represents the time from glottal opening to maximal airflow¹²

5. Cough volume acceleration (CVA; L/s/s) was calculated as the ratio of the peak expiratory flow to expiratory rise time (PEFR/PEFRT).

Figure 1 –

Airflow and pressure waveforms produced during voluntary cough. Circles represent peak measurements, and shaded boxes represent durations. A: peak expiratory flow rate (PEFR), B: peak expiratory pressure (PEP), C: compression phase duration (CPD) represents the time of minimal airflow when the glottis is closed while thoracic muscle contract, maximizing subglottal pressure necessary for the expulsive phase, D: peak expiratory flow rise time (PEFRT).

Three additional voluntary coughs were then performed for acoustic measurements. Cough sounds was recorded using a free-field microphone (Zoom H-6 Six Track Portable Handy Recorder Bundle, Zoom America, Hauppauge, NY), equipped with a shotgun microphone (Zoom SGH-6 Shotgn Microphone, Zoom America, Hauppauge, NY) positioned at one meter from and directed towards the patient’s mouth. We recorded in uncompressed audio format with a sample rate set at 48 kHz with 24 bits resolution and saved each file as a waveform audio (WAV) file. At the beginning of each recording, acoustic gain was adjusted based on a practice cough to avoid clipping. Then, a 1000 Hz calibration tone obtained from a cell phone with the application Frequency Sound Generator (iPhone App, version 18.0, Developer Alexandar Mlazev, 2016) with maximal volume in both the app and the cell phone was played next to the subject’s mouth, while a decibel reading was obtained with a calibrated sound level meter (REED Instruments R8060, Type 2, Wilmington, NC) co-located with the recorder. Gain and settings remained unchanged after recording the calibration tone. These recordings were obtained in a soundproof room, to minimize background noise in the recording. Prior researchers have described the acoustics of cough as made out of three phases: explosive, intermediate and voiced phases, respectively corresponding to glottal opening, steady state flow and interruption of flow due to glottic closure.²³ The voiced phase is not always clearly present and the termination of the intermediate phase cannot be determined without the presence of the voiced phase. We thus decided to primarily focus the acoustic analysis on the explosive phase, and secondarily analyze the voiced phase (Figure 2).

Figure 2 –

Example of a cough waveform and spectrogram tracings. Cough sounds can be divided into (1) an explosive phase, (2) an intermediate phase, and (3) a voiced phase.

Acoustic data were extracted and analyzed using Raven Pro 1.6.1 sound analysis software (K. Lisa Yang Center for Conservation Bioacoustics. (2019). Raven Pro: Interactive Sound Analysis Software (Version 1.6.1) [Computer software]. Ithaca, NY: The Cornell Lab of Ornithology. Available from http://ravensoundsoftware.com/). Using Raven Pro 1.6.1, we made acoustic waveforms and spectrograms with a 512-point (10.7 ms) Hann window (3 dB bandwidth 135 Hz), with 50% overlap and a 512-point DFT, yielding time and frequency measurement precision of 5.3 ms and 93.8 Hz. Calibration of each digital audio recordings were performed in the software settings using the 1000 Hz calibration tone and the associated decibel reading in order to obtain absolute measurements for power, energy and amplitude. The following acoustic parameters were analyzed with Raven Pro 1.6.1 after selecting the cough event with boundary measurements (Figure 3):

1. Energy flux density expressed in Joules per meter squared (J/m²), as the sound energy flux density within the selection window;

2. Acoustic rise time in seconds, as the time from the beginning of the cough sound to the maximum sound amplitude reached during the explosive phase of cough;

3. Duration 90% in seconds, as the time within the selection window containing 90% of the acoustic energy;

4. Peak power density in decibels relative to full scale (dB FS), as the maximal power spectral density within the matrix elements of the spectrogram selection window. Power density describes the distribution of acoustic power over the frequency range of the recording.

5. Inband power in dB (following calibration of the acoustic recording), as a measure of sound power within the confines of the time and frequency bounds of each selection. For selections spanning the range 0 Hz to the Nyquist frequency, Inband Power should be approximately equal to the sound pressure level.

6. Peak frequency in Hertz (Hz), as the maximal frequency reached within the spectrogram selection window;

7. 50% bandwidth in Hz, as the smallest frequency bandwidth within which 50% of the sound power is contained;

8. 90% bandwidth in Hz, as the smallest frequency bandwidth within which 90% of the sound power is contained;

9. Peak sound pressure level (SPL) in dB, which was obtained by using the maximal root mean square (RMS) sound pressure from the amplitude time waveform view and the sound pressure reading for the calibration sound in the following formula:

   $$\text{SPL}=20{\text{Log}}_{\text{10}}\left(\text{RMS}\text{sound}\text{pressure}\text{/}\text{reference}\text{sound}\text{pressure}\right)$$

   where reference sound pressure is 20 uPa.

Figure 3 –

Examples of a waveform and spectrograms of cough sounds and associated measurements obtained with Raven Pro 1.6.1. RT = acoustic rise time. Segment S1 represents the explosive phase of cough, while S2 represents the voiced phase. Window A displays a cough with distinct S1 and S2 phases, while window B demonstrates a cough with less distinct phases. In such cases, we used qualitative assessments of the acoustic recording and visual assessment of the spectrogram looking for increased energy or distinct harmonics to determine the boundaries of each phase.

Two research team members used qualitative assessments of the recorded cough sounds and the corresponding acoustic waveforms and spectrograms obtained for each patient before and after injection laryngoplasty to delineate the voiced component of each cough sound (Figure 3). The ratio of the peak amplitude of the voiced phase of the cough sound to peak amplitude of the explosive phase of the same cough sound was calculated for each cough. It is reported as the voiced to explosive phase ratio (VE Phase ratio) in the result tables.

Statistical Analysis

Study data were collected and managed using Research Electronic Data Capture (REDCap), a secure, web-based software platform designed to support data capture for research studies.²⁴ Data were then imported into a spreadsheet using Microsoft Excel (2018; Microsoft Corporation, Redmond, WA).

All analysis was performed using Stata Version 15.1 (2017, StataCorp, College Station, TX). Summary statistics and probability plots were used to describe the data. Due to the data distribution, non-parametric methods of analyses were selected. A Wilcoxon signed-rank test was used to compare the differences pre and post injection laryngoplasty in airflow and acoustic measures, as well as in EAT-10 questionnaire. A Wilcoxon rank-sum test was used to evaluate differences among groups based on gender, history of pneumonia, and history of dysphagia. A Kruskal–Wallis test was used to compare the groups of diagnoses. A Spearman rank order correlation was used to evaluate the association between airflow and acoustic cough measures. Given the nonparametric nature of this data, and linear relationship limitation of a Pearson coefficient, a Spearman’s coefficient was deemed a more reliable measure of correlation.

RESULTS

A total of 47 patients were enrolled in the study between August 2019 and March 2020. A total of 22 subjects were excluded: 6 for deviation from the research protocol and improper data collection, 15 for incomplete data due to lack of follow-up related to the COVID-19 pandemic and one due to technical issues during data collection (defective pneumotachometer filter), resulting in a final sample of 25 patients (13F:12M, mean age 68.8 years). Follow-up data were collected during the post-procedure examination date, which ranged from 11 to 49 days, with a mean of 26 days. Fourteen patients underwent injection laryngoplasty for vocal fold paralysis, eight for vocal fold paresis and three for vocal fold atrophy. Demographic characteristics are summarized in Table I.

Table I-.

Demographics (N=25)

Characteristic	Overall Group
Age (years)	68.8 ± 13.0
Sex % (#)	Female: 52 (13);
	Male: 48 (12)
BMI (kg/m2)	23.6 ± 2.90*
Diagnosis % (#)	Vocal fold paresis: 32 (8)
	Vocal fold paralysis: 56 (14)
	Vocal fold atrophy: 12 (3)
History of Pneumonia % (#)	Yes: 20 (5); No: 80 (20)
History of Dysphagia % (#)	Yes: 48 (12); No: 52 (13)

^*BMI was calculated based on N=24 due to missing information of one patient

An evaluation of the pneumotachometer airflow measures showed significant differences in patients’ cough strength in the parameters of PEFRT and CVA (Table II). Following injection laryngoplasty, patients had a statistically significant decreased PEFRT (Mean change: −0.03s, SD, 0.06, p=.04) and increased CVA (Mean change: 13.1L/s², SD: 33.9, p=.03). Subgroup analysis was performed to evaluate changes in airflow measures between patients of different genders (Table III), diagnoses (Table IV), with a history of pneumonia (Table V) and a history of dysphagia (Table VI). Patients who had a history of dysphagia had a decrease in CPD, with a mean change of −0.06 sec (SD: 0.17), while patients without dysphagia had an increase in CPD, with a mean change of 0.07 sec (SD :0.14) post procedure at p=0.02. None of the remaining airflow parameters were significantly different. Though PEFR was increased as hypothesized, this was not statistically significant (p=.86). Comparison between categories of gender, diagnosis, and history of pneumonia showed no statistically significant differences between any of the airflow parameters.

Table II–

Airflow Measures Pre vs. Post Injection Laryngoplasty (N=25)

Parameter	Pre	Post	Δ
	Mean (SD)	Mean (SD)	Mean (SD)	P-value
PEFR (L/s)	5.67 (1.42)	5.88 (2.53)	0.22 (1.23)	0.86
PEP (mmHg)	2.90 (1.42)	3.00 (1.47)	0.10 (0.94)	0.73
CPD (s)	0.27 (0.16)	0.28 (0.17)	0.01 (0.17)	0.06
PEFRT (s)*	0.12 (0.07)	0.09 (0.04)	−0.03 (0.06)	0.04
CVA (L/s2)*	72.3 (61.6)	85.4 (66.7)	13.1 (33.9)	0.03

^*designates statistically significant improvement at P ≤ 0.05.

Table III–

Comparison Between Groups of Gender

Parameter	Female (N=13)	Male (N=12)	P-value
	Mean Δ (SD)	Mean Δ (SD)
Airflow measures (N=25)
PEFR (L/s)	−0.23 (0.53)	0.71 (1.58)	0.09
PEP(mmHg)	−0.01 (0.37)	0.22 (1.33)	0.59
CPR (s)	0.05 (0.18)	−0.04 (0.15)	0.30
PEFRT (s)	−0.03 (0.06)	−0.02 (0.07)	0.91
CVA (L/s2)	4.57 (37.9)	22.4 (27.8)	0.25
Subjective Questionnaires(N=25)
EAT-10	−5.80 (7.6)	−0.25 (7.19)	0.07
Acoustic Measures (N=22)
Energy (J/m^2)	1.46 (5.39)	0.12 (5.41)	0.49
Rise Time (s)	−0.04 (0.07)	0.01 (0.03)	0.11
Duration (s)	−0.01 (0.10)	−0.02 (0.08)	0.87
Peak Power Density (dB FS)	2.02 (6.43)	0.21 (6.00)	0.53
Inband Power (dB FS)	2.02 (5.59)	0.09 (5.35)	0.45
Peak Frequency (Hz)	−42.7 (227)	34.1 (391)	0.09
50% Bandwidth (Hz)	176 (1190)	−93.7 (643)	0.67
90 % Bandwidth (Hz)	673 (2230)	−170 (1790)	0.45
Sound Pressure Level (dB) †	−4.26	1.19	0.49
VE Phase Ratio	0.11 (0.40)	−0.25 (0.72)	0.49

^†The SPL parameter is logarithmic, so a median is reported instead of the mean and standard deviation

Table IV–

Comparison Between Groups of Diagnoses

Parameter	Vocal Fold Paresis (N=8)	Vocal Fold Paralysis (N=l 4)	Vocal Fold Atrophy (N=3)	P-value
	Mean Δ (SD)	Mean Δ (SD)	Mean Δ (SD)
Airflow measures (N=25)
PEFR (L/s)	0.17 (1.00)	−0.17 (0.49)	2.18 (2.54)	0.41
PEP (mmHg)	0.27 (1.17)	−0.13 (0.49)	0.73 (1.81)	0.78
CPD (s)	0.02 (0.15)	0.01 (0.19)	−0.03 (0.10)	0.97
PEFRT (s)	−0.01 (0.04)	−0.05 (0.06)	−0.03 (0.10)	0.38
CVA (L/s2)	5.09 (50.1)	14.0 (21.2)	30.3 (37.9)	0.76
EAT-10	0 (6.35)	−3.86 (8.17)	−3.12 (7.78)	0.37
Energy (J/m^2)	3.20 (5.15)	0.84 (5.38)	−4.24 (5.15)	0.11
Rise Time (s)	−0.04 (0.09)	−0.01 (0.05)	−0.01 (0.01)	0.74
Duration (s)	0.01 (0.06)	−0.03 (0.10)	−0.01 (0.05)	0.70
Peak Power Density (dB FS)	3.72 (5.93)	0.95 (6.53)	−3.40 (2.89)	0.25
Inband Power (dB FS)	3.01 (4.97)	1.43 (5.60)	−4.50 (0.73)	0.11
Peak Frequency (Hz)	78.1 (157)	−38.5 (208)	−20.8 (831)	0.19
50% Bandwidth (Hz)	−380 (695)	248 (1070)	−10.4 (657)	0.37
90 % Bandwidth (Hz)	−224 (1900)	476 (2330)	229 (547)	0.76
Sound Pressure Level (dB) †	2.89	−6.13		0.19
VE Phase Ratio	0.18 (0.53)	−0.04 (0.26)	−0.73 (1.32)	0.70

^†The The SPL parameter is logarithmic, so a median is reported instead of the mean and standard deviation

Table V–

Comparision Between Patients With and Without history of Pneumonia

Parameter	History of Pneumonia (N=5)	No History of Pneumonia (N=20)	P-value
	Mean Δ (SD)	Mean Δ (SD)
Airflow measures (N=25)
PEFR (L/s)	0.80 (1.65)	0.08 (1.11)	0.25
PEP (mmHg)	0.10 (0.35)	0.10 (1.05)	0.45
CPD (s)	−0.11 (0.15)	0.04 (0.16)	0.09
PEFRT (s)	−0.03 (0.09)	−0.03 (0.06)	0.50
CVA (L/s2)	13.1 (17.6)	13.1 (37.3)	1.00
Subjective Questionnaires(N=25)
EAT-10	−5.40 (4.16)	−2.55 (8.43)	0.19
Acoustic Measures (N=22)
Energy (J/m^2)	−1.02 (6.36)	1.19 (5.18)	0.39
Rise Time (s)	−0.01 (0.02)	−0.01 (0.07)	0.50
Duration (s)	−0.03 (0.07)	−0.01 (0.09)	0.86
Peak Power Density (dB FS)	0.01 (7.08)	1.36 (6.11)	0.73
Inband Power (dB FS)	−0.82 (7.17)	1.47 (5.13)	0.50
Peak Frequency (Hz)	−31.2 (105)	1.70 (347)	1.00
50% Bandwidth (Hz)	−242 (966)	104 (953)	0.61
90 % Bandwidth (Hz)	1030 (1760)	78.4 (2080)	0.39
Sound Pressure Level (dB) †	−13.0	−0.32	0.06
VE Phase Ratio	−0.11 (0.43)	0.06 (0.64)	0.80

^†The The SPL parameter is logarithmic, so a median is reported instead of the mean and standard deviation

Table VI–

Comparison Between Patients With and Without History of Dysphagia

Parameter	History of Dysphagia (N=12)	No History of Dysphagia (N=13)	P-value
	Mean Δ (SD)	Mean Δ (SD)
Airflow measures (N=25)
PEFR (L/s)	0.01 (1.26)	0.42 (1.23)	0.19
PEP (mmHg)	−0.26 (0.54)	0.44 (1.12)	0.06
CPD (s)*	−0.06 (0.17)	0.07 (0.14)	0.02
PEFRT (s)	−0.03 (0.06)	−0.03 (0.06)	0.45
CVA (L/s2)	11.8 (20.1)	14.3 (50.0)	0.55
Subjective Questionnaires (N=25)
EAT-10	−6.5 (8.71)	0 (5.43)	0.08
Acoustic Measures (N=22)
Energy (J/m^2)	−0.71 (4.70)	2.29 (5.69)	0.25
Rise Time (s)	−0.01 (0.04)	−0.01 (0.08)	0.82
Duration (s)	−0.03 (0.10)	0.01 (0.08)	0.25
Peak Power Density (dB FS)	−0.18 (5.02)	2.41 (7.10)	0.41
Inband Power (dB FS)	−0.26 (5.14)	2.37 (5.63)	0.38
Peak Frequency (Hz)	−82.4 (290)	73.8 (332)	0.58
50% Bandwidth (Hz)	116 (1150)	−34 (725)	0.95
90 % Bandwidth (Hz)	231 (2200)	273 (1930)	0.84
Sound Pressure Level (dB) †	−3.97	−4.26	0.77
VE Phase Ratio	−0.01 (0.26)	−0.13 (0.82)	0.45

^*designates statistically significant improvement at P ≤ 0.05.

^†The SPL parameter is logarithmic, so a median is reported instead of the mean and standard deviation

A comparison of the EAT-10 scores pre and post procedure showed an average decrease of 3.12 points (SD: 7.78) following injection laryngoplasty. Preoperative patients had a mean of 13.68 (SD:14.3), which decreased to a mean of 10.6 (SD: 11.3) post-procedure, although this change was not statistically significant (p=.08). Subgroup analysis showed no statistically significant differences between groups of gender, diagnosis, history of pneumonia, or dysphagia, as shown in Tables III–VI.

While 25 patients completed pre- and post- procedure measures, one patient declined the follow-up acoustic recording due to mild discomfort while coughing, and two patients’ cough samples could not be analyzed, one due to lack of recording system calibration during the pre-injection session, and one due to an uninterpretable waveform, leaving 22 paired cough samples for the acoustic analysis. A Wilcoxon signed-rank test showed no statistically significant differences in the acoustic parameters following injection laryngoplasty, as shown in Table VII. The VE Phase ratio was calculated to evaluate whether the voiced portion of the cough signal would change following injection augmentation. Even though the ratio slightly decreased, it was not a statistically significant difference (p=.81). Similarly, the SPL parameter decreased from a median of 62.2 dB to 57.6 dB re 20 uPa, but did not reach significance (p=.09). However, when looking at just the cohort of patients with vocal fold paralysis and complete acoustic measures (n=13), the change in SPL was statistically significant, going from a pre-procedure median of 64.6 dB to post-procedure median of 56.7 dB at p=.01. There were no statistically significant differences when comparing the change in SPL of subjects with vocal fold paralysis to those with paresis, or atrophy (p=.49). Further group analysis showed no statistically significant differences in the remaining acoustic measures between groups of gender, diagnosis, history of pneumonia, or dysphagia, as shown in Tables III–VI.

Table VII–

Acoustic Cough Measures Pre vs. Post INejction Laryngoplasty (N=22)

Parameter	Pre	Post	Δ
	Mean (SD)	Mean (SD)	Mean (SD)	P-value
Energy (J/m^2)	11.4 (5.08)	12.2 (5.4)	0.80 (5.31)	0.61
Rise Time (s)	0.05 (0.05)	0.04 (0.03)	−0.01 (0.06)	0.54
Duration (s)	0.31 (0.13)	0.30 (0.13)	−0.02 (0.09)	0.41
Peak Power Density (dB FS)	−28.6 (6.39)	−27.5 (6.34)	1.11 (6.14)	0.52
Inband Power (dB FS)	−33.1 (5.15)	−31.9 (5.25)	1.10 (5.43)	0.45
Peak Frequency (Hz)	724 (290)	720 (331)	−4.28 (315)	0.38
50% Bandwidth (Hz)	1260 (731)	1301 (907)	41.2 (940)	0.99
90 % Bandwidth (Hz)	4570 (1970)	4820 (2030)	250 (2020)	0.55
Sound Pressure Level (dB) †	62.2	57.6	−4.11	0.09
VE Phase Ratio	0.54 (0.58)	0.47 (0.36)	−0.07 (0.60)	0.81

^†The SPL parameter is logarithmic, so a median is reported instead of the mean and standard deviation

A correlation of acoustic measures of cough with airflow measures showed some significant relationships. For the correlation analysis of the airflow and acoustic data, four acoustic samples were excluded, resulting in a total of 46 encounters. A Spearman rank order correlation showed a weak negative relationship between PEFRT and acoustic parameters of energy (Coefficient: −0.31, p= .04) and peak power density (Coefficient: −0.35, p=.02). A weak positive correlation was noted between PEFRT and peak frequency (Coefficient: 0.35, p=.02). PEFRT also had a moderate statistically significant relationship with acoustic rise time (Coefficient: 0.40, p=.01). CVA was also weakly correlated with peak frequency (Coefficient: −0.35, p= .02) and moderately negatively correlated with acoustic rise time (Coefficient: −0.43, p<0.01). The remaining combinations did not show any significant findings, as shown in the correlation matrix (Table VIII).

Table VIII–

–Spearman Correlation Between Airflow and Acoustic Measures (N=46)

		Airflow Measures of Cough
		Peak expiratory flow rate (L/s)		Peak expiratory pressure(mmHg)		Compression phase duration (s)		Peak expiratory flow rise time (s)		Cough volume acceleration (L/s2)
		Coeff.	P-value	Coeff.	P-value	Coeff.	P-value	Coeff.	P-value	Coeff.	P-value
Acoustic Measures of Cough Sounds	Energy (J/m^2)	0.09	0.53	0.11	0.46	−0.09	0.56	−0.31*	0.04	0.25	0.09
	Rise Time (S)	−0.22	0.14	−0.25	0.10	−0.13	0.40	0.40**	0.01	−0.43**	0.00
	Duration (S)	0.15	0.32	0.09	0.57	−0.18	0.24	−0.15	0.32	0.16	0.29
	Peak Power Density (dB FS)	0.03	0.82	0.06	0.68	0.05	0.74	−0.35*	0.02	0.27	0.07
	Inband Power (dB FS)	.08	0.60	0.10	0.51	−0.05	0.76	−0.33	0.03	0.26	0.09
	Peak Frequency (Hz)	−0.10	0.50	−0.12	0.41	0.10	0.54	0.35*	0.02	−0.35*	0.02
	50% Bandwid th (Hz)	0.28	0.06	0.05	0.75	−0.05	0.72	0.13	0.41	0.09	0.57
	90% Bandwidth (Hz)	0.17	0.26	−0.04	0.76	−0.03	0.83	−0.01	0.93	0.14	0.34
	Cough Ratio	0.02	0.89	−0.12	0.43	−0.21	0.15	−0.12	0.43	0.03	0.83
	Sound Pressure Level (dB)	0.19	0.21	0.12	0.43	0.04	0.81	0.04	0.79	−0.04	0.79

^*designates statistically significant improvement at P ≤ 0.05.

^**designates statistically significant improvement at P ≤ 0.01.

DISCUSSION

Effective cough and airway protection mechanisms are essential in preventing aspiration. In both, glottic competence appears to be an important factor, though evidence is overall lacking.^5,17,19,20 The primary objective of our study was to test the hypothesis that restoring glottic competence with injection laryngoplasty may improve voluntary cough strength, as measured by cough PEFR. While our data did not support this hypothesis, we found a statistically significant decrease in PEFRT and increase in CVA, corresponding to an increase in cough effectiveness after injection laryngoplasty. In other words, voluntary cough strength was not affected by glottic function, while cough effectiveness improved. These results are consistent with prior studies that looked at the impact of glottic insufficiency on cough airflow dynamics. Ruddy et al. noted improved CVA in their case series of three patients undergoing injection laryngoplasty.¹⁹ Enver et al. found a significant decrease in cough effectiveness in patients with vocal fold bowing and progressive supranuclear palsy, with both decreased PEFR and CVA as measured with pneumotachometry.²⁵ Also using a pneumotachometer, Murty et al. compared cough airflow in subjects with unilateral vocal fold paralysis to that of age and sex matched controls, and noted a statistically significant increase in PEFRT and no change in PEFR in subjects with paralysis.²⁶ Patients with laryngectomy have been noted to have reduced CVA compared to aged-matched controls, indicating a possible role for the functional larynx in ensuring cough effectiveness.²⁷ Finally, while Yiu et al. did not find an association between the vocal fold bowing index and cough airflow measures in subjects with Parkinson’s Disease, they recommended prospective trials evaluating the impact of injection laryngoplasty on cough strength in this specific patient population, given promising results of medialization on cough airflow in prior case series.²⁸ Our study is, to the best of our knowledge, the largest prospective interventional trial investigating the effect of injection laryngoplasty on voluntary cough airflow dynamics, and confirms prior findings that glottic competence may be a key factor in the production of an effective cough. Of note, the use of pneumotachography is not routine in laryngology clinical practice, and there is no established minimal clinical important difference in cough airflow measurements that is currently agreed upon among laryngologists.

Prior retrospective studies have described an improvement in dysphagia following injection laryngoplasty.^29,30 In this study, outcome measure of dysphagia was obtained for all patients before and after injection laryngoplasty with the EAT-10 questionnaire²¹ - irrespective of the treatment goal for injection laryngoplasty. There was an overall improvement in dysphagia symptoms, with an average decrease in EAT-10 score of 3.12 points, but this fell short of statistical significance (p=.08). Subjects with a history of dysphagia documented in the electronic medical record had a statistically significant decrease in CPD, in contrast to those without dysphagia who had an increase in CPD. Other parameters of cough aerodynamics were not significantly different between the two groups. This finding suggests that the benefits of injection laryngoplasty in improving cough production may not be as pronounced in patients with swallowing dysfunction. In fact, the literature has conflicting evidence on the benefits of medialization on swallowing function, though symptoms of dysphagia may be improved. For instance, Kammer et al. performed high resolution manometry on patients with swallowing dysfunction and vocal fold paralysis/paresis before and after injection laryngoplasty, and found that these patients had abnormal pharyngeal pressure measures compared to normative data that did not improve with treatment.³¹ In contrast, Bhattarachyya et al. reported a significantly decreased incidence of pneumonia in an early vs. late injection laryngoplasty in patients with unilateral vocal fold paralysis following thoracic surgery, concluding that early injection medialization decreases aspiration and thereby pneumonia incidence.³² In light of our finding of increased voluntary cough effectiveness after injection laryngoplasty, we posit that injection laryngoplasty may improve outcomes of patients with swallowing dysfunction by ameliorating airway clearance through effective cough, though their swallowing function may be unchanged.

Our secondary objective was to harness acoustic analysis of cough sounds to infer cough airflow measures and glottic competence. Acoustic screening, if meaningful, would offer an easy, safe and cheap solution for bedside evaluation. The correlation of cough acoustic measures with cough airflow measures, irrespective of pre- or post-medialization status, demonstrated a few significant relationships. This includes a weak negative relationship between PEFRT and acoustic energy and peak power density, indicating that more effective coughs had increased sound energy and peak power. Another notable significant relationship was a moderate positive correlation between PEFRT and acoustic rise time, which means that PEFRT may be inferred from acoustic rise time of cough sounds. Finally, a weak positive correlation was noted between PEFRT and peak frequency, indicating higher frequencies overall in sounds of less effective coughs - a finding previously noted in voluntary cough of patients with vocal fold paralysis.³³ No statistically significant differences were noted when comparing pre- and post-medialization cough acoustic data, except for a significant decrease in SPL in patients with vocal fold paralysis. It could be speculated that the observed decrease of SPL in vocal fold paralysis may reflect the fact that paralyzed vocal fold not only cannot adduct correctly, but also has limited abduction capacity. A medialized paralyzed vocal fold may impact inhaled volumes, thus altering cough airflow dynamics comparatively to atrophy or paresis. No difference in VE phase ratio was noted either, signifying that no significant change in the amplitude of the voice portion of cough sound occurred after injection laryngoplasty. Our findings of significant correlations between cough sounds and airflow measures are in line with prior research efforts. Laciuga et al.³⁴ and Miles and Huckabee³⁵ examined clinicians’ perceptual judgement of cough strength based on sound in voluntary and involuntary cough respectively, finding moderate agreement. Smith-Hammond et al.³⁶ found that expiratory airflow parameters of cough (PEFRT, PEFR and CVA) as well as cough SPL were sensitive predictors of aspiration in patients with ischemic strokes. Lee et al.¹³ found a strong correlation between sound power and energy and cough expiratory airflow, and Umayahara et al.^14,15 built a non-linear model to accurately predict peak cough flow from cough SPL. Overall, our acoustic analysis confirms and adds to the existing literature examining the interplay between cough sounds and airflow dynamics.

Our investigation on the role of laryngeal competence in voluntary cough generation highlights deficiencies in the English medical literature. Described by Avicenna as “the movement by which the body’s Nature throws the torment away,”³⁷ cough is vital in homo sapiens. While the descent of the hyoid and larynx has allowed for the development of the faculty of speech in humans, the verticality of the supra-laryngeal vocal tract and the proximity of the larynx to the esophageal inlet have also led to a perilous configuration placing humans at a greater risk of choking and aspiration – necessitating a sensitive and effective cough defense mechanism. ³⁸ Despite the larynx active role in cough production, the literature evaluating its contribution in physiology and pathophysiology is limited. Much of the published work on the laryngeal involvement in cough focuses on its sensory input in the cough reflex and dystussia, either as a therapeutic target in chronic cough or, when deficient, as a risk factor in silent aspiration and dysfunctional deglutition.³⁹ Yet, much remains to be uncovered on the impact of laryngeal mechanics on cough generation.

One of the most comprehensive evaluation of laryngeal physiology during cough dates back to 1965 when von Lenden and Isshiki used ultra-high-speed photography to elucidate vibrations and other aerodynamic phenomena in the upper airway.⁸ They described three phases of laryngeal activity during cough: the opening - inspiratory phase, the closing phase reinforced by supraglottic structures, and the opening - expiratory phase which leads to expulsion of air, mucous and debris. In addition, they noted specific vibratory patterns during the expiratory phase of cough, with four progressive phases: initial vocal folds vibrations, a medial excursion of the aryepiglottic folds followed by steady vibration of the vocal folds, which then subsides. With this work, von Lenden and Isshiki not only confirmed the aerodynamic phases of cough at the level of the larynx, but also hinted at the acoustic phases of cough resulting from laryngeal contributions: a first vibrational event involving vocal folds during the initial explosive airflow, followed by an intermediate stage with supraglottic involvement and finally another vibrational event involving vocal folds – three acoustic phases, which are now commonly referred to as the explosive, intermediated and voiced phases of cough sounds. Other important scientific discoveries on laryngeal action during cough in the mid-twentieth century include Faaborg-Andersen’s investigation of intrinsic laryngeal muscles electromyographic activity during cough⁴⁰ and Yanagihara, von Lenden and Werner-Kukuk’s airflow studies determining the role of the larynx in cough power generation and resistance modulation to prevent re-aspiration of ejected material.⁴¹

Following these early investigations, interest in laryngeal activity during cough waned in the otolaryngology literature. In addition to the notable publications previously mentioned,^{17–20,25,26,28} Sant’Ambrogio et al. studied the pattern of discharge of the posterior cricoarytenoid (PCA), cricothyroid (CT), thyroarytenoid (TA), and arytenoideus transversus (AR) muscles during cough using electromyography in anesthetized dogs, finding that the PCA and CT are activated during the inspiratory phase while the laryngeal adductors (TA and AR) are recruited during the compressive phase.⁴² More recently, Britton et al. used rigid and flexible laryngoscopy to examine the true vocal folds (TVF) during voluntary cough and establish normative data related to TVF movement during cough in young healthy adults. TVF abduction angles during cough were reliably measured from laryngeal video-endoscopy and TVF movements were noted to be faster for expulsion abduction than for pre-compression adduction.⁴³ Studies on cough airflow dynamics following laryngectomy have also indirectly shed light on laryngeal function during cough. Fontana et al. determined a reduction in PEFR and CVA in subjects with laryngectomy compared to aged-matched healthy subjects.²⁷ The decrease in PEFR with laryngectomy was further corroborated by Fullerton et al.,⁴⁴ thus indicating a possible function of the larynx in catalyzing effective and maximal expiratory airflow during cough to achieve airway clearance.

The contribution of the larynx in cough sound generation received little attention in otolaryngology, and most of the published literature on cough acoustics is in pulmonary medicine and engineering. Juraj Korpás of Slovakia has undoubtedly produced the most comprehensive exploration of cough sounds as it relates to physiology starting in the 1980s, and his investigations have offered important insights on the laryngeal component of cough sounds. His work reflects and integrates key technical developments in the field of bioacoustics in the second half of the 20^th century.⁴⁵ This includes the use of (1) tussiphography, i.e., oscillography of cough sounds, to determine cough sounds pattern specific to airway pathologies for diagnostic purposes, severity of disease and response to treatment, (2) histographic analysis or the study of sound amplitude according to their frequency of occurrence, which can differentiate healthy from non-healthy cough sound, (3) spectral analysis or assessment of intensity bands at different frequencies for determination of cough sound timbre through identification of fundamental frequency, harmonics and overtones, and (4) Fast Fourier Transform spectrum analysis (otherwise known as spectrograms), i.e., the algorithmic representation of acoustic signal in the frequency-intensity scale over time for pathologic classification of cough sounds.²³ Korpás’ studies have highlighted the laryngeal origin of the third phase of cough sound, noting its changes with various laryngeal pathologies, such as its prolongation in laryngitis, its absence in laryngectomy, chordotomy and severe vocal fold paralysis and its presence in milder cases of paralysis.⁴⁵ In recent years, new capabilities in data analytics, such as machine learning, have brought back an interest in cough sounds and its spectrogram, with the goal of creating noninvasive and inexpensive screening tools for respiratory assessment.⁴⁶ Despite impressive reported diagnostic accuracy, these efforts have been restricted by the limited availability of labelled cough sound datasets, the lack of a standard for cough sound collection, and the computing power required for continuous real-time signal processing.⁴⁷ Collaboration between clinicians and engineers is required to advance bioacoustics in human diagnostics, and otolaryngologists have a key role to play, given their expertise in laryngeal physiology and cough management.

Our study contributes uniquely to existing knowledge on the role of glottic competence in cough aerodynamics and the interplay between cough acoustics and airflow measures. However, we acknowledge a number of significant limitations. First, our subject population was heterogeneous with different etiology for glottic insufficiency. Degree of glottic insufficiency could have been estimated by acoustic and aerodynamic measures of breathiness, the volume of injectable used or quantification of vocal fold bowing on laryngoscopy frames.²⁵ Also, we did not analyze our data according to the different injectable materials used, though we recognize that injectable varying viscoelastic properties may affect the physiology of cough generation in different ways. Second, respiratory muscle strength and pulmonary function, which affect cough strength and may vary in the individual subject over time, were not measured and may confound some of the cough airflow measurements. Respiratory muscle strength can be measured non-invasively with a MicroRPM manometer, a validated tool for measurement of maximum inspiratory and expiratory pressure. Pulmonary function can be measured with pulmonary function testing, and may be variable in the same subject over a short time period in the presence of underlying lung disease, a variable we did not inquire about. Third, in addition to voluntary cough, reflexive cough could have been investigated, as it is physiologically different and more clinically relevant for airway clearance in patients with dysphagia.⁴⁸ Reflexive cough can be induced with the delivery of nebulized capsaicin in the pneumotachometer circuit.²⁸ However, prior research has also established that voluntary cough strength does inversely correlate with aspiration risk,³⁶ hence the relevance of determining the impact of injection augmentation on voluntary cough strength. Future studies should evaluate the impact injection augmentation on reflexive cough strength. Fourth, though subjects were coached to give their maximal voluntary cough, variability in effort cannot be avoided. In particular, some patients may have been hesitant to produce a maximal cough after injection laryngoplasty, perhaps explaining the decrease in SPL in patients with vocal fold paralysis after injection laryngoplasty. There may be other confounding factors affecting cough sounds, such as subject smoking history or pre-existing pulmonary disease. Fifth, airflow measurements and acoustic recordings were performed asynchronously, introducing more variability in the experimental designs. This limitation could be mitigated by recording cough sounds during airflow measures with a contact microphone⁴⁹ or a microphone introduced in the pneumotachometer circuit,⁵⁰ although both of these approaches may be prone to clipping and acoustic signal distortion and do not mimic real-life scenarios. Sixth, both Lee et al.¹³ and Umayahara et al.¹⁵ have noted that microphone position may affect cough acoustic measurements such as power and SPL, and therefore cough acoustic recordings must be protocolized. Our study followed a recording protocol under the guidance of expert bio-acousticians to avoid such variability in measurements, but meaningful small variations in microphone position cannot be excluded. Seventh, we did not include any subjective assessment of cough strength, as there is no such validated patient reported outcome measure. Currently, validated cough questionnaires focus on cough severity and cough-related quality of life measures.⁵¹ Finally, our study, though the largest of its kind, is limited by a small sample size. After starting the study in August 2019, our goal was to prospectively recruit 50 patients. This goal was cut short by the COVID-19 pandemic, which interrupted all institutional clinical research in March 2020. Further delays were encountered in re-opening the protocol due to infection control limitations related to aerosols generated with voluntary cough production in the research protocol.

CONCLUSION

Our study is the largest prospective interventional study examining the impact of injection laryngoplasty on voluntary cough airflow dynamics, with significant findings of decreased PEFRT and increased CVA, indicating improved cough effectiveness. We also collected voluntary cough sounds, and though no significant change in acoustic parameters occurred pre- and post-procedure, significant correlations were noted between acoustic recordings and airflow measurements. Future studies should focus on examining the impact of injection laryngoplasty on reflexive cough airflow dynamics, as involuntary cough is more impactful in aspiration prevention. It is also essential that otolaryngologists and other physicians involved in the care of patients with dystussia collaborate alongside engineers to create large, protocolized and labelled cough sound databases with the aim of developing validated, inexpensive, non-invasive automated acoustic diagnostic algorithms.