The objective of this study was to identify the prognostic factors for hearing preservation that would allow the more accurate stratification of patients who undergo stereotactic radiosurgery (SRS) for unilateral, sporadic vestibular schwannoma (VS).
In total, 119 patients with VS who had serviceable hearing underwent SRS as primary treatment. The mean (±standard deviation) patient age was 48 ± 11 years, and the mean (±standard deviation) follow-up duration was 55.2 ± 35.7 months. The median marginal radiotherapy dose was 12.0 grays (Gy), and the mean (±standard deviation) tumor volume was 1.95 ± 2.24 cm3. The mean (±standard deviation) pure tone average (PTA) score was 26 ± 12 decibels (dB) (range, 4-50 dB), and the mean (±standard deviation) maximum speech discrimination score was 91 ± 12% (range, 52-100%). The mean (±standard deviation) baseline values for the interlatency (IL) of waves I and III (IL I-III) and the IL of waves I through V (IL I-V) on auditory brainstem response were 2.58 ± 0.60 milliseconds (mS) (range, 1.92-4.30 mS) and 4.80 ± 0.61 mS (range, 3.80-6.40 mS), respectively.
In multivariate analysis, the PTA score and IL I-V were significant and independent prognostic factors (hazard ratio, 1.072; 95% confidence interval, 1.046-1.098; P < .001; and hazard ratio, 1.534; 95% confidence interval, 1.008-2.336; P = .046, respectively). By using the PTA score and IL I-V, the patients were classified into 4 groups. The ratios of patients with serviceable hearing after SRS were 89.6%, 64.0%, 25.8%, and 6.7%, respectively, in Groups A through D (P < .001).
In 1988, Gardner and Robertson described a classification system for predicting hearing preservation in patients who underwent surgery for unilateral vestibular schwannoma (VS).1 This system, called the Gardner-Robertson classification system (G-R classification), is simple, straightforward, and can be performed from memory. Thus, the G-R classification also has been useful for classifying hearing levels both before and after stereotactic radiosurgery (SRS) for VS,2-5 although this system was based on the results from a surgical series of patients with VS.
SRS was first designed in 1951 by Lars Leksell,6 and the use of SRS was expanded to include the treatment of VS in 1969.7 Since the publication of Leksell's 1971 report,7 numerous efforts have been made to increase the tumor control rate and reduce the risk of cranial nerve injury. Consequently, many reports on long-term outcomes after SRS for VS indicated not only excellent progression-free survival rates of 92% to 100% but also outstanding preservation of the trigeminal nerve and facial nerve in 92% to 100% and 94% to 100% of patients, respectively, even with the lower marginal tumor doses of 12 to 13 grays (Gy).4, 8-18 However, hearing preservation still falls short of our expectations; the range after SRS with long-term follow-up has been reported as 32% to 71%.3, 4, 8, 10-13, 15-19
Although the pathophysiology of hearing loss after SRS in patients with VS is still under debate,5, 20, 21 a coherent factor throughout the previous studies is the hearing level before SRS.2, 11, 22 The representative hearing level classification systems are the G-R classification and the American Academy of Otolaryngology-Head and Neck Surgery guidelines for hearing preservation in VS.1, 23 These 2 classification systems rely on the results of audiometry, pure tone average (PTA) and maximum speech discrimination score (SDS).
However, to efficiently stratify patients according to the outcomes of treatment modalities used, a thorough statistical analysis is required. In that sense, the aforementioned classification systems seem to be deficient in their ability to predict the exact rates of hearing preservation among patients who undergo SRS for VS, although a trend toward differences in hearing preservation between the classes is obvious. In fact, from 10.9% to 39.2% of patients with a hearing level of G-R Class 1 lose their hearing. Conversely, hearing can be preserved in approximately 50% of patients who have a hearing level of G-R Class 2 after SRS for VS.2, 22, 24
Therefore, there appears to be a need for better stratification of patients with VS to refine the ability to predict hearing preservation among patients who undergo SRS. Thus, we performed this study based on our experience over 10 years in treating patients with VS using SRS to identify the prognostic factors of hearing preservation that would allow us to stratify patients more appropriately.
MATERIALS AND METHODS
Between 1998 and 2009, 728 patients with VS underwent SRS at Seoul National University Hospital, Seoul, Korea. The inclusion criteria were as follows: 1) SRS as the primary treatment for VS; 2) sporadic unilateral VS; 3) patients with serviceable hearing, defined as a hearing level of G-R Class I or II,1 at the time of SRS; 4) regular audiometry follow-up; 5) patients who had auditory brainstem response (ABR) testing before SRS as a routine examination of their auditory function; and 6) no surgical resection of VS within 12 months after SRS. Therefore, 119 patients (66.5%) among the 179 patients who had sporadic unilateral VS and serviceable hearing underwent radiosurgery as a primary treatment and were enrolled in the current study (Fig. 1). All patient data were collected from hospital charts and radiologic studies in accordance with the case record form approved by the institutional review board. The clinical characteristics are summarized in Table 1.
Table 1. Patient Characteristics and Stereotactic Radiosurgery Parameters
Abbreviations: dB, decibels; Gy, grays; IL I-III, the interlatency between waves I and III on the auditory brainstem response; IL I-V, the interlatency between waves I and III or between waves I and V on the auditory brainstem response; mS, milliseconds; SD, standard deviation.
No. of men (%)
No. of women (%)
Follow-up duration, mo
Tumor volume, cm3
Marginal dose prescribed, Gy
Mean cochlear dose, Gy
Maximum cochlear dose, Gy
No. of shots
Pure tone average, dB
Speech discrimination score, %
IL I-III, mS
IL I-V, mS
SRS was performed using the Leksell gamma knife (model B, model C, or Perfexion; Elekta Instrument AB, Stockholm, Sweden). A treatment plan was generated using the Leksell GammaPlan system (Elekta Instrument AB) based on thin-sliced magnetic resonance (MR) images. T1-weighted, 3-dimensional, multiplanar, rapid-acquisition, gradient echo MR scan images were obtained before and after gadolinium enhancement to determine the target volume. The median marginal dose was 12.0 Gy (mean ± standard deviation, 12.2 ± 0.9 Gy; range, 11-20 Gy), and the median prescription isodose line was 50% (mean ± standard deviation, 50.3 ± 2.3%; range, 45-70%). The final prescription dose, expressed as a marginal dose, and the associated treatment parameters are summarized in Table 1.
Baseline evaluations, including audiometry and ABR, were performed within 1 to 3 months before SRS. After SRS, patients usually were followed at 1 month, 3 months, 6 months, 12 months, 18 months, and 24 months and annually thereafter. A clinical evaluation, which included a physical and neurologic examination, was performed at each visit. Audiometry and ABR usually were performed every 6 months for the first 2 years after SRS. Additional audiometry and/or ABR assessments were based on patients' perceptions of hearing deterioration.
PTA was calculated by averaging audiometric masked bone conduction responses at 500 Hz, 1000 Hz, and 2000 Hz. SDS was recorded to establish a pretreatment audiometric grade, as previously described.1
Auditory Brainstem Response Measurements
ABR testing was performed with a 4-channel evoked potentials system (Spirit model; Nicolet Electronics, Madison, Wis). Insert receivers were used in all patients. The stimuli were 95-decibel (dB), rarefaction polarity clicks with a stimulus duration of 100 milliseconds (mS). The contralateral masking was set at 50 dB. The surface electrode montage was the frontal zone to A1 or A2 (earlobe) with the common ground at the frontal pole zero. The stimulus rate was 7.7 Hz. The average sample size was 1000 repetitions performed twice. Parameters were changed based on selection of the ABR with the best morphologic characteristics and replicability. Subdermal needle electrodes were used.
When ABR testing produced no response, regardless of the level of hearing, the maximum values of the baseline ABR testing obtained in this cohort were coded for each interlatency (IL) of wave I and wave III (IL I-III) and IL I through V (IL I-V), which was regarded as the maximum value for ABR of the auditory pathway with serviceable hearing in the presence of VS. However, the real ABR values in patients who had no response on ABR testing may be greater than those estimated in this study. The values used were 4.3 mS for IL-III and 6.4 mS for IL I-V.
Definition of Hearing Response and Other Covariates
PTA and SDS were recorded to establish a post-treatment audiometric grade in the same way as in the pretreatment evaluation. After assessing post-SRS hearing status during follow-up, audiometries were treated as censored observations if serviceable hearing was preserved within G-R Class I or II. Radiation dose to the cochlea was determined from retrieved treatment backup files, including T2-weighted, volume-acquisition MR imaging divided into axial images at 1-mm to 1.5-mm intervals.
Kaplan-Meier survival plots were used to estimate the overall survival with serviceable hearing. The log-rank test (level of significance, α = .05) was used to assess differences in overall survival with serviceable hearing between the groups. The chi-square test was used to compare nominal variables between the groups. A Pearson correlation analysis was performed to identify bivariate correlations between 2 continuous variables. The Cox proportional hazards model (level of significance, α = .05) was used to adjust for covariates. A backward, stepwise method was used for the multivariate analysis to identify the possible new prognostic factors of hearing preservation and to reduce the chance of type II error; variables were considered for multivariate analysis only if they were associated with a dependent variable in each analysis at the P < .10 level. These statistical analyses were performed using PASW Statistics (version 17.0.2; SPSS Inc., Chicago, Ill). To identify the statistical cutoff value for each covariate and classification and regression tree (CART) of hearing preservation, the free software R (version 2.12.2; rpart package; R Foundation for Statistical Computing, Vienna, Austria; available from: http://www.r-project.org/ accessed February 24, 2012) was used.
Forty-five patients (37.8%) were men, and 74 (62.2%) were women. The mean patient age (±standard deviation) was 48 ± 11 years (range, 21-71 years), and the mean follow-up duration (±standard deviation) was 55.2 ± 35.7 months (range, 12.3-158 months). The mean tumor volume (±standard deviation) was 1.95 ± 2.24 cm3 (range, 0.03-9.10 cm3). The mean number of shots was 8 (range, 1-20 shots). The mean PTA score (±standard deviation) was 26 ± 12 decibels (dB) (range, 4-50 dB). The mean SDS (±standard deviation) was 91% ± 12% (range, 52%-100%). Thus, according to the G-R classification, 79 patients (66.4%) had a Class 1 hearing level, and 40 patients (33.6%) had a Class 2 hearing level.
The mean (±standard deviation) baseline values of IL I-III and IL I-V were 2.58 ± 0.60 mS (range, 1.92-4.30 mS) and 4.80 ± 0.61 mS (range, 3.80-6.40 mS), respectively. The mean (±standard deviation) interaural differences in IL I-III (ID-IL I-III) and ID-IL I-V were 0.45 ± 0.59 mS (range, −0.48 to 2.30 mS) and 0.70 ± 0.60 mS (range, −0.33 to 2.40 mS), respectively.
The mean (±standard deviation) of the maximum cochlear dose was 8.13 ± 3.02 Gy (range, 2.40-16.6 Gy). The mean (±standard deviation) of the mean cochlear dose was 4.30 ± 1.51 Gy (range, 1.40-8.30 Gy). The clinical and radiologic characteristics are summarized in Table 1.
Hearing-Preservation Rate and Its Prognostic Factors
Until the last clinical follow-up, in total, 51 patients (42.9%) lost their serviceable hearing. The median survival with serviceable hearing was 67 months after SRS. The actuarial rates of hearing preservation were 79.7%, 68.5%, 62.5%, 59.9%, and 56.2% at 6 months, 12 months, 24 months, 36 months, and 60 months after SRS, respectively (Fig. 2).
In the univariate analyses, G-R class, PTA score, SDS, IL I-V, and ID-IL I-V were associated significantly with hearing outcome. In addition, the tumor volume and mean cochlear dose were associated with the preservation of serviceable hearing; however, they did not reach statistical significance. On multivariate analysis, only PTA score and IL I-V remained as significant and independent prognostic factors (hazard ratio, 1.072 [95% confidence interval, 1.046-1.098; P < .001] and 1.534 [95% confidence interval, 1.008-2.336; P = .046], respectively). The remaining covariates, including the G-R class, mean cochlear dose, and SDS, were removed during the process of the multivariate analysis that involved use of the backward-stepwise method. Results of the statistical analyses are summarized in Table 2.
Table 2. Prognostic Factors Related to Hearing Preservation After Stereotactic Radiosurgery for Vestibular Schwannoma
Multivariate Analysis (Backward Stepwise)
Abbreviations: CI, confidence interval; IL I-III, the interlatency between wave I and II on the auditory brainstem response; IL I-V, the interlatency between wave I and III or between wave I and V on the auditory brainstem response; HR, hazard ratio.
These covariates were analyzed in the multivariate analysis using the backward stepwise method with a Cox proportional hazards model.
The difference in these covariates was statistically significant.
Classification and Regression Tree Analysis and New Classification for Hearing Preservation
Using PTA score and IL I-V, which we identified as significant and independent prognostic variables of hearing preservation, a CART comprising 119 patients was created. We identified 3 nodes (ie, branching points). The tree was split by a PTA score of 21 mS and then by an IL I-V of 5.225 mS. Finally, a PTA of 31 mS split the tree (Fig. 3). On the basis of the terminal nodes, we categorized the patients into 4 groups (Table 3, Fig. 4).
Table 3. Risk Group Splits Based on the Results From Classification and Regression Tree Analysis
No. of Patients
No. With Hearing Preservation
Hearing Preservation, %
Abbreviations: IL I-V, the interlatency between wave I and V on the auditory brainstem response; mS, milliseconds; PTA, pure tone average score.
Group A: PTA≤20
Group B: IL I-V<5.225 mS and 21≤PTA≤30
Group C: IL I-V<5.225 mS and 31≤PTA≤50
Group D: IL I-V≥5.225 mS and PTA≤50
The ratios of patients who preserved their serviceable hearing during follow-up differed significantly (P < .001) between groups, and the ratios of patients with serviceable hearing at the last clinical follow-up were 89.6%, 64%, 25.8%, and 6.7% for groups A through D, respectively.
Although numerous studies have suggested possible prognostic factors for hearing preservation, such as radiation dose to the cochlea or its structures, the length of irradiated cochlear nerve, marginal dose to the tumor itself, and age,2, 21, 24-32 these factors still are debated, and we do not clearly understand why patients lose their serviceable hearing after SRS for VS. However, a consistent prognostic factor of hearing preservation in the management of patients with VS is the level of hearing before treatment, even in surgical series.1, 33 Therefore, the previous classification systems have been based on results from audiometry using the PTA score and SDS.1, 23 In the current study, the PTA score was identified as 1 of the significant and independent covariates related to hearing preservation, which appears to be the same result obtained in previous studies.2, 11, 22, 34
However, SDS and G-R class itself were removed in the multivariate analysis. One possible reason is that the PTA score had a significant correlation with SDS (correlation coefficient, −0.688; P < .001), although the PTA score does not usually match up exactly with SDS, as described by Gardner and Robertson Gardner in their report.1 Thus, the classification systems based on 2 covariates with significant correlation to each other may fall short in predicting the exact possibilities of hearing preservation in patients who undergo SRS for VS. Consequently, there is huge variation in the level of hearing preservation even within each subgroup,2, 22, 24 which may create the need for a new classification system.
One of the noteworthy findings of the current study is that 1 value of ABR—IL I-V—was identified as a significant prognostic factor for hearing preservation, even in the multivariate analysis. Similar findings about the possible relation between ABR results and hearing preservation have been recently reported, although the results were not confirmed in multivariate analysis.22 Badie et al35 directly measured the intracanalicular pressure of the internal auditory canal in patients with VS and demonstrated that the intracanalicular pressure correlates well with absolute wave V latency and IL I-V of ABR. Although ABR is not a direct measure of hearing function, studies have indicated an association between hearing loss and delays in wave V and IL I-V.35, 36 In addition, patients who had G-R Class 1 hearing tended to have lower intracanalicular pressure compared with patients who had G-R Class 2 hearing, but the difference between the groups did not reach statistical significance.35 This finding suggests that intracanalicular pressure may not correlated exactly with hearing function in the setting of VS but may have an independent effect on hearing outcome. Therefore, the huge intraclass variation of hearing preservation in the same G-R class may be caused by the differences in intracanalicular pressure among patients who undergo SRS for VS, which could explain the finding that patients with G-R Class 1 hearing lose their serviceable hearing. Conversely, patients with G-R Class 2 hearing retain their serviceable hearing after SRS.
By using the CART analysis, we were able to categorize our patients into 4 groups and refine the relative risk of losing serviceable hearing in each group after SRS for VS. This finding seems to be important in the proper selection of patients whose serviceable hearing should be preserved and in predicting the hearing loss of each individual.
In summary, the level of hearing, expressed as the PTA score, and the intracanalicular pressure, measured as the IL I-V on the ABR examination, appear to be significant and independent prognostic factors for hearing preservation in patients who undergo SRS for VS. The classification system using these 2 covariates may be more useful and specific for predicting the hearing-preservation rate in an individual than the systems that use only the correlations among audiometry results. To preserve serviceable hearing in each patient with VS, an individualized treatment strategy may be beneficial according to the classification system. The findings of this study and the accuracy of the classification system should be scrutinized in the near future in a larger group of patients.
This study was supported by grant 02-2011-006 from the Seoul National University Bundang Hospital Research Fund.