Real-world evidence and optimization of vocal dysfunction in end-stage renal disease patients with secondary hyperparathyroidism

Patients with end-stage renal disease (ESRD) may demonstrate secondary hyperparathyroidism (SHPT), characterized by parathyroid hormone oversecretion in response to electrolyte imbalance (e.g., hypocalcemia and hyperphosphatemia). Moreover, this electrolyte imbalance may affect vocal cord muscle contraction and lead to voice change. Here, we explored the effects of SHPT on the voices of patients with ESRD. We used data of 147,026 patients with ESRD from the registry for catastrophic illness patients, a sub-database of Taiwan National Health Insurance Research Database. We divided these patients into 2 groups based on whether they had hyperparathyroidism (HPT) and compared vocal dysfunction (VD) incidence among them. We also prospectively included 60 ESRD patients with SHPT; 45 of them underwent parathyroidectomy. Preoperatively and postoperatively, voice analysis was used to investigate changes in vocal parameters. In the real-world database analysis, the presence of HPT significantly increased VD incidence in patients with ESRD (p = 0.003): Cox regression analysis results indicated that patients with ESRD had an approximately 1.6-fold increased VD risk (p = 0.003). In the clinical analysis, the “jitter” and “shimmer” factors improved significantly after operation, whereas the aerodynamic factors remained unchanged. In conclusion, SHPT was an independent risk factor for VD in patients with ESRD, mainly affecting their acoustic factors.


Scientific Reports
| (2021) 11:653 | https://doi.org/10.1038/s41598-020-79810-0 www.nature.com/scientificreports/ diagnosis of vocal dysfunction (VD). The possible causes of VD in patients with ESRD include dehydration of the vocal fold mucosa or weakness of the vocal fold muscle. In patients with ESRD, dehydration is the predominant cause of VD. In this study, video strobe laryngoscopy (VSL) and voice analysis were performed on days between dialysis to minimize dehydration. Electrolyte imbalance, which was reported in patients with ESRD, may affect voice production 14 . Therefore, we performed a real-word data analysis using a nationwide database of Taiwan to investigate whether ESRD patients with SHPT have VD risks. Then, we prospectively recruited ESRD patients with SHPT and analyzed their vocal characteristics through VSL and voice analysis before and after parathyroidectomy and investigated vocal motion abnormalities due to SHPT. The aim of this study is to investigate voice changes in patients with SHPT and voice recovery after the surgical procedure. Our findings can help the clinician to evaluate vocal symptoms when diagnosing SHPT.

Methods
Part I-real-world data study. We used data from Taiwan's National Health Insurance (NHI) Research Database (NHIRD). The NHI program, implemented by Taiwan's government in 1995, had covered almost all Taiwan residents by 2017 [15][16][17][18] . Subsequently, Taiwan National Health Research Institutes had constructed NHIRD for research use. NHIRD contains NHI beneficiaries' medical records, including disease diagnosis at outpatient visits and hospitalizations, drug types and dosages, examination items, operation contents, medical expenditure, area of residence, and income level 19 . In the NHIRD, diagnosis coding follows the International Classification of Diseases, Ninth Revised Edition, Clinical Modifications (ICD-9-CM) 20,21 . The NHI system defines chronic kidney failure with long-term renal dialysis as a "catastrophic illness. " The concerned patients are registered in the Registry for Catastrophic Illness Patients (RFCIP) 20 after audit and are given catastrophic illness-related treatment at subsidized rates 22 . However, the audit process is rigorous, with regular dialysis treatment for ≥ 3 months as an eligibility criterion. Therefore, RFCIP data were used for identifying and recruiting chronic renal failure patients requiring kidney dialysis.
Data of patients with a diagnosis of ESRD between January 1997 and December 2013 were retrieved from the RFCIP by using ESRD-associated ICD-9-CM codes:  20 . Consequently, data of 147 026 patients with ESRD were extracted from RFCIP after the exclusion of patients with comorbid cancer, cerebrovascular accident, dementia, parkinsonism ESRD, and any voice-related diseases before ESRD and HPT. The index date for the study cohort was the HPT diagnosis date.
ICD-9-CM codes of HPT were employed to divide the included patients into the ESRD-HPT (study) and ESRD-non-HPT (comparison) cohorts. Then, all ESRD patients with HPT were matched 1:2 by sex, age, urbanization level, and income level with ESRD patients without HPT, who then formed the comparison cohort. An index date matching that of the study cohort was allocated to the comparison cohort.
The main outcome here was VD occurrence (ICD-9-CM 478.5, 478.7, and 784.5). Both cohorts were followed from the index date to VD diagnosis, death, or December, 2013, whichever occurred first.
Part II-clinical study. For the clinical study, we prospectively recruited patients with SHPT undergoing hemodialysis. The exclusion criteria were past history of vocal mucosal lesion (e.g., nodules or polyps) and vocal fold paralysis as well as current diagnoses of other vocal etiologies (e.g., severe asthma) or other systemic diseases. Of the enrolled hemodialysis patients with SHPT, patients with had refractory symptomatic SHPT received total parathyroidectomy and autotransplantation, which were successful.
The vocal functions within 2 weeks before surgery and 3 months after surgery were documented through VSL on a stroboscopy system (Model 9400) from KayPENTAX (Lincoln Park, NJ, USA). VSL and voice analysis were performed on a day between dialysis sessions. For analysis of the laryngeal image, the following data were considered: absence/presence of vocal fold lesion; wave of vocal fold mucosa; pattern of glottal closure and any involvement of supraglottic structures. The reports from laryngeal examination before and after surgery of each patient were taken from two laryngologists. If there was any difference between two reports from two laryngologists, reevaluation was necessary.
We measured acoustic parameters including jitter, shimmer, mean fundamental frequency (F0), and noiseto-harmonics ratio (NHR) by using a Computerized Speech Laboratory (CSL) system (Model 4500; KayPentax, Lincoln Park, NJ, USA) 27,28 . The participants were asked to phonate the sustained vowel /a/ at their habitual pitch and comfortable loudness, after inhaling deeply. The acoustic parameters of the voice, including fundamental frequency (Hz), jitter (%), shimmer (dB), and noise-harmonic ratio (NHR; dB), were evaluated. The phonation time was 3 s and a segment from the middle of the vowel phonation was analyzed, and bad voice quality data was discarded. Here, F0 is the average of all extracted fundamental frequency values. Shimmer (%) was defined as the average absolute difference between peak amplitudes of consecutive periods divided by the average peak amplitudes during the phonatory segment. Jitter (%) was defined as the average absolute difference between successive periods divided by the average period duration. Moreover, here, the NHR reflected the relative spectral energy contributions of the noise and harmonic components of the acoustic voice signal 29  www.nature.com/scientificreports/ Moreover, aerodynamic parameters including maximum phonation time (MPT) and durations of the sounds "s" and "z" were analyzed by a speech pathologist using a stopwatch 27 . Here, the sound "s" is voiceless (i.e., produced without vibration of the vocal folds), whereas the sound "z" is a voiced counterpart of the sound "s" 30 . The voice of patients who underwent parathyroidectomy was analyzed 1 or 2 weeks before surgery and then 3 months after surgery.
Preoperative and postoperative perceptual evaluation was performed using the GRBAS scoring system (G = grade, R = roughness, B = breathiness, A = asthenia, and S = strain; 0 = normal, 1 = mild, 2 = moderate, and 3 = severe). The GRBAS-sum is the summation of each GRBAS score, and reevaluation was required if a difference of more than 2 points was recorded between the two GRBAS scores. A speech pathologist and an otolaryngologist analyzed the above-mentioned voice parameters in a double-blinded manner. All participants in this study gave written informed consent prior to study begin. The experiment was conducted in line with the relevant guidelines and regulations.
Both parts of this study were approved by the Institutional Review Board of Chang Gung Memorial Hospital (IRB No. 201800401B0).

Statistical analysis.
For Part I of this study, we compared the comorbidities and demographic characteristics of the study and comparison cohorts; here, we used Pearson's chi-square test and an unpaired Student t test to compare categorical and continuous variables, respectively 31 . Next, we included control variables such as sex, age, income level, urbanization level, and comorbidities as covariates in our univariate model to perform a univariate analysis 32 . The variables with p < 0.1 in the univariate analysis were included in the multivariate analysis. The cumulative incidence of VD in the 2 cohorts was estimated using a Kaplan-Meier analysis, and then, a twotailed log-rank test was used determine the relevant differences between the cohorts. We also used multivariable Cox proportional hazard regression models to measure the hazard ratios (HRs) and corresponding 95% CIs for VD. The stability of HRs was also examined using subgroup and sensitivity analyses so as to evaluate whether the comorbidity-SHPT interaction effects on VD were significant. www.nature.com/scientificreports/ For Part II of this study, paired t and Wilcoxon signed-rank tests were used to analyze the statistical significance of our parametric and nonparametric measurement data, respectively 28 . Pearson correlation coefficient analysis was also applied 28 .

Results
Part I-real-world data study. After the 1:2 matching of sex, age, urbanization level, and income level, the study and comparison cohorts contained 7988 patients with HPT (mean ± standard deviation [SD] follow-up period: 4.6 ± 3.3 year) and 15,976 patients without HPT (mean ± SD observation period: 4.5 ± 3.3 year), respectively ( Table 1). The study cohort had a significantly higher DM, COPD, CRS, and GERD prevalence than did the comparison cohort. VD incidence rate was also significantly higher in the study cohort (p = 0.003; 2.2 vs 1.4 per 1000 person-years). The incidence rate ratio of VD thus was 1.56 (95% CI 1.17-2.09; p = 0.003), and the mean duration from HPT diagnosis to VD development was 2.9 ± 2.3 years.
Of the 45 patients with SHPT who underwent total parathyroidectomy, 32 patients completed the preoperative and postoperative vocal evaluations through VSL and voice analysis.
There was no significant difference in laryngeal analysis between pre and post operation. No vocal fold lesion and no glottal gap was found among 32 patients before and after operation. The wave of vocal fold mucosa was normal and symmetric among these 32 patients. No supraglottic constriction was found in the study. Table 4 lists the levels of calcium, phosphorus, and iPTH in these patients, all of which significantly decreased after surgery. In the objective vocal measurements, F0, jitter, and shimmer values demonstrated significant improvements postoperatively (F0: 178.0 ± 42.9 vs 236.5 ± 83.7 Hz, p = 0.01; jitter: 1.41% ± 0.84% vs 0.84% ± 0.35%, p = 0.047; shimmer: 0.38 ± 0.18 vs 0.24 ± 0.09 dB, p = 0.033). However, the preoperative and postoperative differences in perceptual evaluation and aerodynamic parameters, including MPT and duration of "s" and "z" sounds, were nonsignificant.

Discussion
ESRD-complicated with SHPT can influence the calcium and phosphorus balance, potentially leading to fine vocal muscle weakness and consequent functional vocal disturbances. Here, by using real-world and clinical data, we proved the aforementioned hypothesis and for the first time investigated the influence of HPT in voice change. Our results confirmed that SHPT in patients with ESRD is a predisposing factor for VD, such that ESRD patients with SHPT have an approximately 1.6 times higher VD risk than do ESRD patients without SHPT.
Insufficient water in the body may lead to decrease in the water content of the vocal cords and decrease in the secretion of the respiratory tract. The drying of the vocal cord surface may lead to changes in voice quality. Therefore, patients on hemodialysis often experience transient hoarseness at the end of dialysis [33][34][35] .
To clarify whether the VD was due to dialysis-associated dehydration or SHPT, we used a nationwide database and enrolled hemodialysis patients and divided them to cohorts of patients with and without HPT to investigate VD occurrence. Consequently, we could eliminate dialysis-associated VD as the causative mechanism, confirming HPT as a risk factor for VD.
Calcium blocks sodium ion flux through voltage-gated sodium channels. Thus, hypercalcemia results in inhibition of depolarization and impairment of action potential generation in neurons and muscle cells 36 . Moreover, hypercalcemia can damage the recurrent laryngeal nerve and intrinsic laryngeal muscle, and this can result in dysphonia. These findings support our hypothesis that iPTH influences vocal production. www.nature.com/scientificreports/ On the basis of Taiwan Renal Registry Data System, a large-scale study reported that a serum phosphorus level of ≥ 6.5 mg/dL was associated with increased mortality 37 . However, a significant imbalance of such electrolytes can be life-threatening.
Thus, when hemodialysis patients report changes in their voices, clinicians must include SHPT in the differential diagnosis. If in doubt, plasma phosphorus, calcium and iPTH levels could be estimated. Taken together, these steps could facilitate early detection of SHPT.
In our current clinical study, SHPT was found to significantly increase plasma calcium, phosphate, and iPTH levels before surgery. In the voice analysis after surgery, only the acoustic indices F0, jitter, and shimmer demonstrated significant differences. Studies have reported that abnormal vocal cord lesions typically increase jitter and shimmer. If Jitter exceeds the threshold of 1% and shimmer exceeds the threshold of 0.35 dB, abnormal performance is regarded to have occurred 38 . In the current study, the mean preoperative jitter was 1.41% (i.e., > 1%), but normalized to 0.84 postoperatively; similarly, the mean preoperative shimmer was 0.38 dB (i.e., > 0.35), but normalized to 0.24 dB postoperatively. Therefore, we speculated that SHPT mainly causes changes in the acoustic indices jitter and shimmer in the voice analysis, with the thresholds identical to those reported previously; this is a clinically significant observation. Jitter and shimmer represent variations in vocal fold vibrations. The change in vibration may result from changes in vocal muscle strength. An improvement (i.e., decrease) in the acoustic parameters produces a more harmonic voice.
This study had several strengths. First, we analyzed a large amount of data from numerous dialysis patients. Second, we verified the real-world evidence regarding HPT as an independent risk factor for VD. We also analyzed the clinical data of ESRD patients with and without SHPT and even evaluated their data before and www.nature.com/scientificreports/ after parathroidectomy to further verify the relationship between SHPT and voice disorders. Consequently, we provided reference indexes for clinicians when performing vocal examination through VSL and voice analysis.
Our study however has limitations related to real-world data investigations. First, the NHIRD does not provide patient data for clinical symptoms, laboratory data, and VSL results; therefore, we could not determine the VD severity in the included patients. Nevertheless, in our clinical study, we performed clinical examinations to obtain the data unavailable in the NHIRD. Although only ESRD patients with SHPT were enrolled in this study, our research could still provide clinical evidence supporting our hypothesis. In future studies, laryngeal electromyography (LEMG) should be applied to evaluate laryngeal muscle activity. Table 3. Characteristics of Enrolled Patients with ESRD (N = 60). iPTH, intact parathyroid hormone; F0, fundamental frequency; MPT, maximum phonation time; s, voiceless "s" produced without vibration of vocal folds; z, voiced "z" produced with vibration of vocal folds; NHR, noise-to-harmonics ratio. † Aerodynamic parameters with stop watch. ‡ Acoustic parameters in a Computerized Speech Laboratory.  www.nature.com/scientificreports/ Based on our current findings, the use of VSL examinations and voice analysis is ideal for analyzing ESRD patients with voice disturbances. Moreover, the abnormal acoustic indices could be used for speculating whether SHPT may be an underlying cause of VD. Nevertheless, future studies should confirm the accuracy and sensitivity of these factors. Moreover, voice changes after nonsurgical treatment of SHPT warrant further research.
In conclusion, by using real-world and clinical patient data, we confirmed that SHPT is a predisposing factor for VD. In addition, the comparative results before and after surgery indicated that the acoustic indices jitter and shimmer in voice analysis might indicate the influence of SHPT on the voice of patients with ESRD. Thus, when a clinician encounters a patient with ESRD complaining of voice disturbances, they must consider including SHPT in the differential diagnosis to facilitate early detection and treatment of the underlying problem.