The retrospective cohort consisted of 523 primary sporadic vestibular schwannomas from 261 female and 262 male patients, representing a balanced gender ratio. The mean age of the included patients was 48.4 years, ranging from 18.0 to 79.1 years. Preoperative hearing was serviceable in the majority of cases (74.8%). Preoperative pure tone audiograms met the criteria for Gardner Robertson (GR) class 1 in 241 cases (46.1%) and for GR class 2 in 150 cases (28.7%). Non-serviceable hearing was present preoperatively as GR class 3 in 112 cases (21.4%), while GR class 4 and 5 were observed in 5 and 15 patients, respectively (1.0 and 2.9%). Regarding tumor extension according to the Koos classification, most tumors were T3 (n = 214, 40.9%) or T4 (n = 167, 31.9%), followed by T2 (n = 116, 22.2%) and a limited number of purely intrameatal tumors (T1: n = 26, 5.0%). The mean preoperative tumor volume was 4.26 cm3, with a range from 0.04 to 52.14 cm3. Widening of the internal acoustic meatus compared to the contralateral healthy side stretched from non-widened IACs to a difference of 15.1 mm. The mean immunohistochemical expression of MIB1 was 1.34% with a minimum of 0.25% and a maximum of 4.2%. The distribution of these parameters is delineated in Fig. 4; Table 1.

Distribution of clinical factors in the study cohort: age (A), preoperative hearing according to the Gardner Robertson classification (B), IAC widening in mm compared to the healthy side (C), immunohistochemical MIB1 expression in % (D), tumor extension according to the Koos classification (E) and preoperative tumor volume in cm3 (F)

CART-specific cut offs based on the most pronounced differentiation between serviceable (GR 1 or 2) and non-serviceable hearing (GR 3, 4 or 5) were calculated. A higher rate for non-serviceable hearing was observed for age >/=55.47 years, preoperative tumor volume >/= 6.57 cm3, IAC widening >/= 1.2 mm and MIB1 immunopositivity >/= 1.2%.

There was a significantly more advanced age of patients with worse preoperative hearing. The mean age increased with each GR class of preoperative hearing (p < 0.0001). When dividing the cohort based on the CART-specified cut off at 55.47 years, patients in the younger group had significantly more often serviceable preoperative hearing (82.3%, 312/379) compared to older patients (54.9%, 79/144, p < 0.0001).

Regarding tumor extension in the cerebellopontine angle (Koos classification), serviceable preoperative hearing was significantly more common in smaller schwannomas (p = 0.0055). T1 tumors showed preoperative serviceable hearing in 84.6% (22/26), T2 in 75% (87/116), T3 in 78.0% (167/214) and T4 in 68.9% (115/167). A more precise look at tumor size revealed that the mean preoperative tumor volume within each GR class was higher with worse preoperative hearing class, ranging from 3.38 cm3 for GR 1 to 7.89 cm3 for GR 5 (p = 0.0013). The CART-specific cut off at 6.57 cm3 divided the study cohort into 409 smaller (78.2%) and 114 larger tumors (21.8%), with a significantly lower rate of cases with preoperative serviceable hearing for patients with larger tumors (59.6% vs. 79.0%, p = 0.0005).

Widening of the internal acoustic canal also showed a significant association with poor preoperative hearing. The mean IAC widening increased with each higher GR class, except GR 5 (p = 0.0353). When dividing the cohort at the CART-specific cut off at 1.2 mm IAC widening, the most pronounced difference in preoperative hearing was described. Cases with less or no IAC widening (n = 134) had a serviceable hearing in 85.8% (115/134) compared to tumors with IAC widening of 1.2 mm or more (68.6%, 276 /389).

The mean immunohistochemical expression of the proliferation marker MIB1 was 1.34% throughout the complete cohort, without a significant difference between different GR classes (p = 0.9.49) or when using the CART-specific cut off of 1.2% (0.7456).

No difference in preoperative hearing was observed between female and male gender (p = 0.2006). The results of the univariate analysis are listed in Table 1.

Potential factors that were associated with IAC widening were included into a multiple linear regression model. Overall, the regression model was statistically significant (p < 0.0001). Younger age (p = 0.0049), larger tumor volume (p < 0.0001), non-serviceable hearing (p = 0.0060) and male gender (p = 0.0458) were all independently associated with widening of the IAC. MIB1 expression did not have an independent impact on IAC widening (see Table 2).

When including all relevant parameters that influence preoperative hearing in the univariate analysis into a multivariate model, several factors were identified as independently influencing preoperative hearing. In the multiple nominal logistic regression focusing on serviceable vs. non-serviceable hearing, older age and larger preoperative tumor volume were the parameters with the most pronounced effect, with an odds ratio of 27.60 (CI 9.17–87.18, p < 0.0001) and 20.20 (CI 3.43–128.58, p = 0.0011), respectively (per change in regressor over the entire range). Additionally, widening of the internal acoustic canal was also an independent significant factor (odds ratio 7.86 (CI 1.77–35.46, p = 0.0068). Gender or differences in MIB1 immunopositivity were without statistical significance, but both parameters showed a mentionable statistical trend, with female patients showing a potentially lower odds ratio for poor preoperative hearing (0.69 (CI 0.45–1.06), p = 0.0907) and higher MIB1 expression a rather negative trend on preoperative hearing (4.29 (CI 0.87–21.07), p = 0.0715).

An additional multiple nominal logistic regression was done with all continuous parameters after CART-specific cut offs based on the best dichotomization regarding serviceable vs. non-serviceable hearing. The analysis rendered similar results. Older age, larger preoperative tumor volume and IAC widening remained independent factors for non-serviceable preoperative hearing (p < 0.0001, p = 0.0006 and p = 0.0044, respectively). A statistical trend for a potentially negative impact of higher MIB1 expression and male gender was also present (p = 0.0799 and 0.0795, respectively). Details of the multiple nominal logistic regression analyses are displayed in Tables 3 and 4.