Introduction
Inhaled asthma medication has the potential to directly affect the larynx. Therefore, asthma, or its treatment, may have a direct effect on patients' voices. Up to 50 per cent of patients using inhaled corticosteroids may suffer from dysphonia, which is usually reversible.Reference Barnes, O'Byrne and Thomson1 Such dysphonia is usually attributed to fungal infection or to steroid-induced adductor myasthenia of the larynx,Reference Williamson, Matusiewicz, Brown, Greening and Crompton2 although laryngoscopy or voice laboratory assessment may reveal more complicated problems such as apposition abnormalities and cycle-to-cycle irregularity.Reference Crompton, Sanderson, Dewar, Matusiewicz, Ning and Jamieson3, Reference Lavy, Wood, Rubin and Harries4 Recent work has demonstrated the impact of dysphonia without significant structural laryngeal disease on patients' quality of life, compared with normal subjects,Reference Wilson, Deary, Millar and MacKenzie5 as measured by the short form 36 (SF-36) questionnaire. Much of the voice literature is focused on those patients who attend otolaryngology clinics; however, Baker et al. found that 50 per cent of 80 young adults with asthma or allergy had vocal quality abnormalities (as assessed by speech and language therapists).Reference Baker, Baker and Le6 Voice morbidity in patients with asthma has not been extensively studied.
The voice assessment performed by voice scientists, speech and language therapists, and others can include sophisticated perceptual, acoustic, aerodynamic and endoscopic methods which enable diagnosis and treatment planning for voice disorders.Reference Carding7 There are a number of protocols available for perceptual analysis of the voice, with the grade–roughness–breathiness–asthenicity–strain (GRBAS) score scale being the recognised ‘gold standard’ tool in the UK.Reference Carding, Carlson, Epstein, Mathieson and Shewell8 Speech and language therapists use this tool to assess a number of aspects of vocal impairment, using a four-point scale.Reference Hirano9 The tool assesses the voice's overall grade (i.e. overall degree of voice deviance), roughness (i.e. impression of irregular pulses or of low frequency noise), breathiness (i.e. audible turbulent air leakage), asthenicity and strain. Several studies have shown reasonable inter-rater reliability in the use of this scale.Reference De Bodt, Wuyts, Van de Heyning and Croux10–Reference Webb, Carding, Deary, MacKenzie, Steen and Wilson13
It is unclear how the degree of voice impairment determined by speech and language therapists relates to patients' own perceptions. The voice symptom scale is a 30-item questionnaire which has been thoroughly evaluated as a tool for the self-assessment of voice quality.Reference Deary, Wilson, Carding and MacKenzie14, Reference Wilson, Webb, Carding, Steen, MacKenzie and Deary15 This scale comprises a total score plus three robust subscales assessing voice impairment, emotional reaction and laryngopharyngeal symptoms (i.e. physical component).
The aim of this study was therefore to characterise the vocal quality and laryngeal appearance in patients attending a problem asthma clinic, and to relate this to patients' and speech and language therapists' perceptions of vocal morbidity. This evaluation was conducted in parallel with assessment of upper airway physiology and nasal disease, the results of which have been previously reported.Reference Stanton, Johnson, MacKenzie, Carter and Bucknall16–Reference Stanton, McGarry, Carter and Bucknall18 A secondary aim was to assess the local inter-rater reliability of the GRBAS scale among speech and language therapists; these results will be reported separately.
Methods
Patients were recruited to the study from a problem asthma clinic based in a central teaching hospital. All patients attending the problem asthma clinic were eligible for inclusion in the study. Letters of invitation to take part in the study were sent to patients attending the clinic, and this invitation was reinforced by telephone or in person when patients attended the clinic. The study protocol involved attendance on a single afternoon. The study was conducted in accordance with the recommendations of the Helsinki Declaration of 1975, and was approved by the North Glasgow University Hospitals NHS Trust local research and ethics committee (REC reference number 03RE002). All patients gave written, informed consent for their participation in the study.
Patients completed the 30-item, self-administered VoiSS questionnaire. Voice recordings were then performed in a soundproof booth housed within the otolaryngology department. Recordings were made using digital tape recorder and digital audio tape. Patients were instructed to speak approximately 10 to 15 cm away from the microphone. They were asked to speak spontaneously for a few seconds (giving their name, how they had travelled to hospital that day, and what they had eaten for dinner and had watched on television the previous evening) before reading the standard ‘Rainbow passage’.Reference Fairbanks19 These recordings were made in the presence of one of two independent observers who were not involved in any further data analysis.
Following the topical application of co-phenylcaine to the nose, each patient had their larynx and laryngopharynx visualised by a single observer (KMacK) using standard naso-laryngoscopic technique. The larynx structure and function were assessed. Laryngeal appearance was noted, as was the mobility of the vocal folds on phonation, inspiration and expiration.
The patients' digital audio tape voice recordings were transferred onto two compact discs (CDs), such that both CDs contained every patients' voice recording plus anonymised personal details, but in a different, randomised order. Individual patient's recordings therefore corresponded to different individual tracks on each CD. A master list was kept in which track numbers were linked to patient names; this list was not seen by the voice raters.
The voice raters were three experienced speech and language therapists familiar with the GRBAS scale. Raters graded patients' voices according to this scale, with a further assessment of voice quality fluctuation (i.e. instability).Reference De Bodt, Van de Heyning, Wuyts and Lambrechts20 Each subscale was assessed on a four-point scale (i.e. zero to three) in order to determine the degree of vocal impairment. Each CD was listened to and independently rated by the three speech and language therapists during listening sessions at least seven days apart. Every patient therefore had their voice recording scored by three speech and language therapists on two occasions.
The voice symptom score questionnaires were scored according to the total score and the three subscales of voice impairment (15 items; score range zero to 60), emotional reaction (eight items; score range zero to 32) and physical symptoms (seven items; score range zero to 28). A higher score indicated greater vocal morbidity.
Data analysis
Mean values for each GRBAS(I) scale score were calculated, from the six scores given by the three raters, for each patient. The total GRBAS scores were calculated using the means for each of these subscales; the mean for the instability subscale was not used in this calculation as this is not in widespread use. Spearman rank correlations were then calculated between the GRBAS score subscales and each VoiSS subscale, as well as the total VoiSS. Mann–Whitney U tests were used to compare the VoiSS of subgroups of patients with abnormalities identified laryngoscopically. MinitabTM software (version 14; Minitab Inc., Coventry, UK) was used to perform these calculations. We used conventional methods to calculate the sensitivity and specificity and the positive and negative predictive values of the VoiSS and the GRBAS scores for detecting an abnormality at laryngoscopy.Reference Altman21
Results
Sixty patients initially agreed to take part in the study but 17 subsequently withdrew; therefore, 43 patients were ultimately included. Thirty-four patients had documented, objective evidence of asthma (i.e. significant diurnal variability in peak expiratory flow rate, significant bronchodilator reversibility or response of airflow obstruction to oral corticosteroid trial, or histamine challenge confirming bronchial hyper-reactivity). Four patients gave a very good clinical history of asthma from an early age but had normal lung function whenever this was tested. Five patients had no definite evidence of asthma. Patients' median dose of inhaled corticosteroids was 1000 µg beclomethasone dipropionate or equivalent (interquartile range 800–2000), and all but three patients were taking inhaled corticosteroids at the time of assessment.
The VoiSS are shown in Table I. There was no relationship between VoiSS and inhaled corticosteroid dose (r = 0.23; p = 0.117).
Table I Voice symptom scores
SD = standard deviation; IQR = interquartile range
The mean of the six GRBAS(I) score assessments (two each from the three speech and language therapists) was calculated for each patient, to give their final GRBAS(I) scores (Table II). Thirteen of our 43 (30.2 per cent) patients had an overall grade score, for the GRBAS scale, of one or more, which is recognised as denoting definite abnormality. A statistically significant correlation was observed between patients' overall grade score, for the GRBAS scale, and their inhaled corticosteroid dose (Spearman r = 0.56; p < 0.001; Figure 1).
Fig. 1 Relationship between inhaled corticosteroid dose and overall grade score for grade–roughness–breathiness–asthenicity–strain (GRBAS) scale. Spearman r = 0.56; p < 0.001. BDP = beclomethasone dipropionate
Table II GRBAS(I) scores
GRBAS(I) = grade–roughness–breathiness–asthenicity–strain–instability scale; SD = standard deviation; IQR = interquartile range
Patients' laryngoscopy results were divided into structural and functional findings, with each patient being assessed separately for each category (Figure 2). Laryngitis was defined as diffuse reddening and swelling of the glottis consistent with an inflammatory process, and was graded subjectively as mild, moderate or severe. Only one patient had laryngeal thrush in addition to mild laryngitis.
Fig. 2 Patients' laryngoscopy results. Each patient was assessed on both structural and functional appearance. Other abnormalities comprised arytenoid abnormality, vocal fold polyp, and benign thickening of tongue base causing hypopharyngeal narrowing. Mod = moderate
Patients with a normal laryngoscopic appearance had lower median VoiSS than those with abnormalities, as shown in Table III. The mean overall grade score, for the GRBAS scale, did not differ between these two groups. However, there were significant differences in GRBAS subscale scores in the group with functional abnormalities, compared with those with no functional abnormalities, as shown in Appendix A (which relates VoiSS and GRBAS(I) subscale scores to laryngoscopic findings). Spearman Rank correlations between patients' median GRBAS(I) rating and their VoiSS (totals and subscale scores) are shown in Table IV. Nonsignificant p values are not shown.
Table III VoiSS and overall GRBAS scores, by laryngoscopic appearance
* n = 17; †n = 26. ‡p = 0.044; **p = 0.15. GRBAS = grade–roughness–breathiness–asthenicity–strain scale; VoiSS = voice symptom score; CI = confidence intervals
Table IV Spearman rank correlations between VoiSS and GRBASI score
GRBAS(I) = grade–roughness–breathiness–asthenicity–strain–instability scale; VoiSS = voice symptom score
Relationship between voice assessment scores and laryngoscopic appearance
We found that the VoiSS and the GRBAS score predicted laryngoscopic abnormality equally (Table V). A total VoiSS of 30 was chosen as a cut-off because of its relationship with the median total VoiSS for the study group. An overall grade of one or more on the GRBAS scale is recognised as identifying significant vocal morbidityReference Hirano9 and was therefore used as the cut-off for this analysis.
Table V Predictive value of overall GRBAS score and VoiSS for laryngoscopic abnormality*
* Any abnormality. †Six scorings; ‡single, self-completed scoring. GRBAS = grade–roughness–breathiness–asthenicity–strain scale; VoiSS = voice symptom score; NPV = negative predictive value; PPV = positive predictive value
Discussion
The patients we studied were representative of those attending our problem asthma clinic, with a majority having objective evidence of asthma and receiving moderately high doses of inhaled corticosteroids. We identified voice morbidity as a problem in this population. While other authors have investigated the frequency of voice problems in patients with asthma,Reference Williamson, Matusiewicz, Brown, Greening and Crompton2, Reference Ihre, Zetterstrom, Ihre and Hammarberg22 these studies used only self-administered questionnaires to identify voice problems. Our study is novel in that it included a comprehensive vocal assessment undertaken variously by patients (using the VoiSS), by speech and language therapists (using the GRBAS scale) and by an ENT specialist (using direct visualisation of the larynx). Although selection bias may have operated, the impact of this would have been diluted by the fact that patients were invited to take part in a broad-based assessment of their upper airway (including the nose). Additional data from a control group would have strengthened our findings. However, the present research was a pilot study designed to generate hypotheses. Further evaluation of the VoiSS and the GRBAS score in an asthmatic cohort should take into account the need for a control group. Our analyses of the inter-relations between VoiSS, GRBAS score and laryngeal findings were not affected by these considerations.
A significant proportion of patients had abnormal laryngeal findings, with abnormalities of function in 28 per cent (12/43) and laryngitis in 35 per cent (15/43). Five patients had glottic chinks, a further five were phonating with the false vocal folds and two had reduced vocal fold mobility; all these are likely to affect voice quality. All laryngoscopies were performed by a single observer; thus, although consistency was achieved, the impact of recognised inter-observer variability in the reporting of laryngoscopic findings was not assessed.
The VoiSS has been extensively investigated and refined in over 800 subjects, and its subscales have shown good internal consistency, in contrast to the self-administered questionnaires used in earlier studies of patients with asthma.Reference Williamson, Matusiewicz, Brown, Greening and Crompton2, Reference Ihre, Zetterstrom, Ihre and Hammarberg22 Although there are other instruments available for the self-assessment of voice quality, such as the vocal handicap index and the voice-related quality of life questionnaire, we chose the VoiSS because of its rigorous development process in a UK population, which included laryngopharyngeal symptom assessment.Reference Deary, Wilson, Carding and MacKenzie14, Reference Wilson, Webb, Carding, Steen, MacKenzie and Deary15 Hitherto, there have been no published VoiSS data for patients with asthma. Our patients' scores were less abnormal than those reported for 144 functional dysphonics and 145 patients with structural laryngeal pathology (mean total scores of 43.3 and 46.5, respectively).Reference Wilson, Webb, Carding, Steen, MacKenzie and Deary15 The GRBAS scores from our study are not readily comparable with those reported elsewhere, as the latter have been assessed differently (e.g. using visual analogue scales)Reference Millet and Dejonckere23 or in such a manner as to determine inter-rater reliability.Reference De Bodt, Wuyts, Van de Heyning and Croux10
Higher voice symptom scores were associated with laryngeal pathology (Table III). We did not evaluate quality of life, but the emotional domain of the VoiSS may reflect this parameter. Dysphonia has been shown to adversely affect patients' quality of life;Reference Wilson, Deary, Millar and MacKenzie5 in another study, self-rated voice quality was significantly related to a range of personality, psychological distress and quality of life measures.Reference Deary, Wilson, Carding and MacKenzie24 The contribution of dysphonia to asthma patients' impaired quality of life merits further exploration.
• Patients with asthma often complain of voice symptoms, which are usually attributed to treatment with inhaled corticosteroids
• Vocal morbidity in patients with asthma is a complex issue; it can be characterised using research instruments (e.g. the VoiSS and the GRBAS score) more routinely used in non-asthma populations
• In only a small proportion of cases can vocal morbidity be directly attributed to laryngeal candidiasis
• The VoiSS is a potential screening tool for vocal morbidity in patients with asthma
As well as investigating the relationship between VoiSS responses and laryngoscopic findings, we also observed a relationship between the specialist, objective GRBAS scoring and the self-reported, subjective VoiSS. Since the overall grade score for the GRBAS scale is taken as a summary measure of voice deviance, we expected to observe a relationship between this component and the impairment domain of the VoiSS. We found weakly positive correlations between the related GRBAS score subscales and the total GRBAS score, and the VoiSS. The lack of a relationship between the physical subscale of the VoiSS and the GRBAS score was expected, as this subscale assesses non-vocal laryngopharyngeal symptoms. Murry et al. found a moderate correlation between total GRBAS score and voice-related quality of life scores (derived from a 10-item, self-administered questionnaire).Reference Murry, Medrado, Hogikyan and Aviv25 This study observed no breakdown of the relationship with individual GRBAS score subscales, and it is these, rather than the total score, which are pertinent to clinical practice. Our study therefore adds to the evidence that patients' perception of vocal morbidity correlates with that of an experienced observer. In addition, the overall grade score for the specialist, labour-intensive GRBAS scale was no better than the VoiSS in predicting laryngoscopic abnormality (Table V). For these reasons, we believe that further validation of the voice symptom score as a screening test for asthma patients with vocal morbidity is warranted.
We found a low incidence of laryngeal thrush (i.e. only one patient), suggesting that dysphonia should not immediately be attributed to this; we also found little evidence of steroid-induced myopathy (two patients). Lavy et al. also found a low incidence of candidiasis in a group of asthmatics complaining of dysphonia (four out of 22 patients), and they identified other laryngoscopic abnormalities (i.e. mucosal changes, apposition problems and supraglottic hyperfunction) which could explain patients' symptoms.Reference Lavy, Wood, Rubin and Harries4
Conclusion
We found that laryngeal structural and functional abnormalities occurred in a significant proportion of patients attending a problem asthma clinic. We also found that such abnormalities were associated with significant differences in patients' (self-reported) VoiSS but not in their (more labour-intensive, objective) GRBAS scores. Very few patients were found to have fungal infections or myopathy as a result of using inhaled corticosteroids, and we suggest that vocal morbidity should not be attributed to these without positive evidence. Our findings confirm relevant, positive correlations between VoiSS and GRBAS scores (the latter being our gold standard measurement), suggesting that these two instruments measure similar attributes. The results of this pilot study suggest that the VoiSS is a useful screening tool in asthma patients; however, further work is required, as is comparative data for normal subjects.
Acknowledgements
The authors would like to thank Catherine Dunnet and Linda Chapman for their assistance with the GRBAS(I) score data collection. This work was supported by the Ritchie Trust Research Fellowship from the Royal College of Physicians and Surgeons of Glasgow.
APPENDIX A Relation of VoiSS and Grbas(I) Scores to Laryngoscopic Findings
GRBAS(I) = grade–roughness–breathiness–asthenicity–strain–instability scale; SD = standard deviation; IQR = interquartile range; VoiSS = voice symptom scale; NS = not significant
