Hearing Loss Risk in Pediatric Patients Treated with Cranial Irradiation and Cisplatin-Based Chemotherapy

Abstract

Purpose:

Cranial radiation therapy (RT) and cisplatin-based chemotherapy are essential to treating many pediatric cancers but cause significant ototoxicity. The objective of this study is to determine the relationship of RT dose with risk for subsequent hearing loss in pediatric patients treated with cisplatin.

Methods and Materials:

This retrospective study of cisplatin-treated pediatric patients examined ototoxicity from cranial RT. Ototoxicity was graded for each ear according to International Society of Pediatric Oncology (SIOP) consensus ototoxicity scale. RT dose to cochlea was calculated using mean, median, maximum, and minimum dose received to determine most predictive parameter for hearing loss. Multivariable logistic regression models then examined risk factors for hearing loss.

Results:

In 96 children (161 ears) treated with RT+cisplatin, minimum cochlear RT dose was most predictive of hearing loss. Higher cochlear RT dose was associated with increased hearing loss (Odds Ratio [OR] per 10 Gy dose increase = 1.64, p = 0.043), with added risk in those receiving an autologous bone marrow transplantation (aBMT) (HR = 10.47, p < 0.001).

Conclusions:

This research supports further testing of minimum cochlear RT dose as a more predictive dose parameter for risk of ototoxicity. Cochlear RT dose was additive to risk of hearing loss from underlying cisplatin-based chemotherapy. Exposure to aBMT was the strongest predictor of developing hearing loss, placing these children at particularly high risk for hearing loss across all cochlear doses. Future prospective studies are crucial to further inform RT dose thresholds and minimize risk of hearing loss in childhood cancer survivors.

Introduction

Radiation therapy (RT) is often combined with cisplatin-based chemotherapy to treat a wide variety of childhood cancers.¹ Although incorporation of cisplatin is associated with survival improvements in many malignancies, it is also ototoxic, damaging the cochlea and causing permanent hearing loss in most affected patients.^1,2 Cumulative cisplatin dose has been shown to be consistently associated with hearing loss, with younger children at a higher risk than older individuals.³ RT similarly damages the cochlea resulting in hearing loss. These effects are potentially additive and particularly damaging to patients receiving multimodal therapy with cisplatin and RT in combination.^3–9 The implications of this hearing loss in pediatric patients is profound, as hearing loss impacts neurocognitive, academic, and social development, and has been shown to decrease quality of life for survivors.^10,11

Although ototoxicity resulting from RT is well-documented, detailed characterization of its ototoxicity is lacking and conflicting.¹² Pediatric data validating the routine ototoxic cochlear limit of 30 to 35 Gy is limited and the magnitude of the interaction between RT and cisplatin on hearing loss not fully elucidated.^3–9,13,14 While several studies indicate the dose of cisplatin supersedes the importance of cochlear RT dose, others note that prior cranial irradiation drastically reduces the cisplatin dose threshold for hearing loss.^15–20 Data from these studies were limited by the size of population in each study, different treatment eras, variable inclusion of ototoxic medications in combination with RT, and outmoded hearing assessments. Moreover, when fluctuations in RT doses to an organ may be high, as one would expect in non-tumor, off-target locations such as the cochlea, it is unclear which of several RT dose parameters (mean, median, minimum, maximum) is most predictive of hearing loss. Thus, while it is evident that this population is at additive risk for hearing loss, significant knowledge gaps remain for optimal RT dosing for children receiving cisplatin-containing regimens.

In addition to cisplatin exposure, RT dose thresholds for hearing loss following RT are also impacted by multiple treatment factors including chemotherapy, other ototoxic medications, presence of central nervous system tumor, age, and baseline hearing.^3,5,8,16,21 Yet, despite increasing use of autologous bone marrow transplant (aBMT), and the high prevalence of hearing loss in this population, its additive impact on risk of hearing loss from cochlear RT exposure is not well-characterized.^8,11,22–24 Given current trends toward aBMT in younger patient populations with neural tumors to reduce or delay RT and its associated late effects, this has emerged as a critical knowledge gap.^25–27

To minimize risk for off-target toxicity, intensity-modulated radiation therapy (IMRT) utilizes computerized tomography (CT) planning to reliably calculate and limit RT dose exposure to key structures.^28,29 Understanding the additive ototoxic impact from chemotherapy will allow for radiation oncologists to determine safer radiation exposures to the cochlea.³⁰ We therefore examined a cohort of pediatric patients treated with both cisplatin and RT to determine the relationship of RT dose and risk for subsequent hearing loss in the context of contemporary, multimodal pediatric treatment regimens.

Methods

Patient population and variables

We conducted a retrospective cohort study of cisplatin-treated patients between 1–21 years of age at diagnosis of previously untreated malignancy who were also treated with RT to the cranium at Children’s Hospital Los Angeles. This study was approved by the Institutional Review Board (IRB). Inclusion criteria required radiation exposure to the cochlea and audiometric testing after any administration of cisplatin. Data was extracted for demographic and disease information (age at diagnosis, sex, ethnicity, malignancy type including location [i.e., central nervous system or extracranial]), treatment as well as therapy provided (treatment regimen, cumulative cisplatin dose delivered [in mg/ m²]), and use of autologous bone marrow transplant (aBMT). Where applicable, cisplatin dosage per patient was converted from mg/kg to mg/ m² (mg/ m²= mg/kg x 30). All patients treated with cisplatin at our institutionare routinely referred for audiometric evaluations during and after therapy, and thus all patients in this cohort received formal audiology assessment regardless of symptoms of hearing loss. Ototoxicity was assessed via central review of audiometric evaluations (V.M., E.O.).

After an assessment of tympanic function, hearing thresholds were assessed at frequencies between 250 and 12,500 Hz. Ototoxicity was graded according to the International Society of Pediatric Oncology (SIOP) consensus ototoxicity scale.³¹ A grade was assigned to each ear and for each patient (where discrepancy between ears, the “better” ear was used to determine the patient-specific grade). Moderate or severe hearing loss was defined as SIOP Grade ≥2. ³²

Description of radiation therapy

All patients underwent CT simulation for either 3D or IMRT treatment. Mean, median, maximum, and minimum RT dose delivered to cochlear was collected from the treatment planning system (Eclipse, Varian Medical Systems, Palo Alto, CA). Cochlear contouring was performed as part of routine CT planning for treatment by two radiation oncologists throughout the study period (Supplemental Figure 1), and centrally reviewed by the study’s radiation oncologist who was blinded to the hearing outcome (K.W.).

Statistical Analyses

Each ear was analyzed individually.^4,33 Distribution of radiation exposure parameters by laterality (right versus left) were compared using the t-test. The study primary endpoint was moderate/severe hearing loss (SIOP Grade ≥2) at most recent audiometric evaluation. To determine which RT cochlear dose parameter was most predictive of hearing loss (i.e., minimum/maximum/mean/median), a multivariable logistic regression model was constructed for the primary endpoint of SIOP Grade ≥2 hearing loss beginning with a base model of a priori specified predictors (cisplatin dose, age, time to audiology assessment, and aBMT) and stepwise forward selection with additional parameters (sex, ethnicity, CNS tumors, timing of radiation) and α <0.15 to retain covariables. To the final model, each radiation dosing parameter was added and the resulting four non-nested models were compared using Akaike information criterion (AIC) and associated weighted AIC (AICw).³⁴ RT dose thresholds were then estimated to maintain moderate/severe hearing loss in ≤20% (as an a priori clinically relevant threshold). A second multivariable logistic regression model was constructed using the same approach and the same endpoint of SIOP Grade ≥2 hearing loss for patient-level data. To provide a real-world evidence correlation to SIOP grade, chi-square analysis compared patients requiring a hearing aid versus no hearing aid and SIOP <2 versus SIOP ≥2. As RT-induced hearing loss increases with time from exposure,³⁵ and as per below, prevalence of SIOP Grade≥2 increased with time in our data, all multivariable models included time to audiology assessment (in years). This also serves to account for length bias. Additionally, a sensitivity analysis was performed for the subset of patients with evaluable audiology follow-up more than 1 year from radiation exposure. All tests were two-sided with significance set at p<0.05. All calculations were conducted using STATA, version 15 (Statacorp LLC, College Station, TX).

Results

Patient Population

Patient characteristics are described in Table 1 (n=96). Median time from RT to follow-up was 2.14 (range 0.04 to 12.37 years), and 2.2 (range 0.1 to 12.1 years) from cisplatin to follow-up. From the 96 patients, 161 ears were evaluable for SIOP grading (16 patients did not have sufficient ear-specific testing to determine SIOP grade). No cochlear contours required adjustment after review by the radiation oncologist. Notably, the majority of patients had a CNS primary tumor, and approximately one-third of patients (29%) had aBMT as part of their treatment. The lowest cochlear RT dose received by patient was a minimum dose of 3.9 Gray (Gy) and a mean dose of 4.9 Gy There was no difference in cochlear RT exposure by ear noted by any dose parameter (Supplemental Table 1). Approximately 40% of the cohort (n=38 of 96) had cisplatin dose-reductions for toxicity during therapy; mean cumulative cisplatin exposure was therefore higher in those without toxicity-induced dose reductions (mean cisplatin dose 348 mg/ m² versus 292 mg/ m², p=0.019). Of the 161 ears, 62 (39%) were graded with moderate/severe hearing loss (SIOP Grade ≥2). Similarly, 34% (32 of 95) of patients were graded with moderate or severe hearing loss. In the cohort, 24% (23 of 96) of all patients were prescribed hearing aids, with a significant difference in the recommendation for a hearing aid present in those with SIOP grade <2 versus ≥2 (5% vs 63%, p<0.001, Supplemental Figure 2).

Table 1:

Description of Cohort (n=96)

Variable	n (%)*
Age, years
median (range)	8.6 (0.8–19.9)
<5 years	22 (23)
5–14.9 years	55 (57)
>15 years	19 (20)
Sex
Male	37 (38)
Female	59 (62)
Ethnicity
Hispanic/Latino	49 (51)
Not Hispanic/Latino	46 (48)
Unknown	1 (1)
Diagnosis
CNS tumor**	85 (88)
Non-CNS tumor†	11 (12)
ABMT
No	68 (71)
Yes	28 (29)
Single	19 (20)
Tandem	5 (5)
Unknown	4 (4)
Cumulative cisplatin delivered (mg/m2)††
Mean ±SE	326±11.8
<200	14 (14.6)
200–399	55 (57.3)
>400	27 (28.1)
SIOP Grade hearing loss§
0	37 (41)
1	21 (23)
2	15 (17)
3	14 (16)
4	3 (3)

^*Unless otherwise specified for a covariable;

^**CNS tumors include medulloblastoma (n=58), ependymoma (n= 11), ATRT (n=6), CNS GCT (n=3), and other CNS (n=7).

^†Non-CNS tumors include metastatic neuroblastoma (n=5) and nasopharyngeal carcinoma (n=6).

^††Where applicable converted from mg/kg (mg/m²= mg/kg x30).

^§SIOP Grade ≥2 considered moderate/severe hearing loss warranting intervention. Unevaluable for specific grade (n=6); unevaluable for ±SIOP2 (n=1)

Analysis of RT dosing parameter

While minimum, maximum, mean, and median RT dose to the cochlea were correlated, variability in actual RT dose delivered to the cochlea remained present across the dose-range (Figure 1). Final multivariable models for evaluation of the impact of the RT dose parameter included cumulative cisplatin dose, age, aBMT, and time from RT to audiology assessment. Comparison of the models supported minimum dose as the “best” RT dose predictor of moderate/severe hearing loss, as all models compared to minimum dose had a ΔAIC >2. Comparison of evidence ratios resulting from AICw indicated a 15–20% greater likelihood that minimum dose was the best fit model versus mean or median dose, with maximum dose demonstrating a less than 33% chance of being the best fit (Table 2). Subsequent analyses were therefore performed using minimum dose as the predictive RT dose parameter.

Figure 1:

Correlation of RT dose parameters (doses in Gy)

Table 2:

Model selection of radiotherapy dosing parameter

Model*	n	Dosing Parameter	Log likelihood	Corrected AIC (AICc)	∆AICc	AIC weight (AICw)	Cumulative AICw	Evidence Ratio
1	161	Min	−79.806	169.9991	0.000	0.337	0.337	n/a
2	161	Median	−79.975	170.3371	0.338	0.285	0.622	0.84
3	161	Mean	−80.013	170.4131	0.414	0.274	0.896	0.81
4	161	Max	−80.977	172.3411	2.342	0.104	1.000	0.31

^*All models include the covariables: cumulative cisplatin dose, age, autologous bone marrow transplant (aBMT), time from audiology assessment.

Association of RT dose with hearing loss

In this cohort receiving chemoradiation, hearing loss was present even at the lowest cochlear exposures and increased with minimum cochlear RT dose received by the ear (Figure 2). Multivariable analyses showed that after accounting for cisplatin dose and other covariables, risk for developing moderate/severe hearing loss (SIOP ≥2) increased with every 10 Gy of exposure to the cochlea (Odds Ratio [OR] = 1.64, 95% Confidence Interval [95%CI] 1.04–2.75, p=0.043). In the model, aBMT was also a significant predictor of hearing loss after adjusting for dose of RT (OR =10.47, 95%CI 3.30–33.36, p < 0.001) as was time from RT to audiology assessment (OR = 1.42, 95% CI 1.15–1.75, p = 0.001). Similar findings were seen within the sensitivity analysis of the subset of patients with audiology assessment ≥ 1 year from RT (Supplemental Table 2). Sex, ethnicity, diagnosis of CNS tumor, and timing of RT relative to cisplatin did not significantly influence the model (all p>0.15 during model selection). Analysis of the patient-level data using the same approach, albeit with fewer numbers, identified similar trends predictive of hearing loss, including the association of aBMT and hearing loss (Supplemental Table 3).

Figure 2: Predicted probability for developing moderate/severe hearing loss (SIOP Grade ≥2).

Probability of SIOP Grade ≥2 hearing loss predicted according to minimal dose of rdaiotherapy to cochlea (in Gy) following adjustment for covariables in the multivariable model. 95% confidence intervals surrounding each point estimate are depicted.

For patients without aBMT, the minimal cochlear RT dose associated with 20% or greater estimated probability of hearing loss at most recent follow-up was 20 Gy (predicted probability SIOP Grade ≥2 = 20.4%, 95% CI 11.8–29.1, p < 0.001). Patients undergoing aBMT in addition to chemoradiation, however, had a ≥20% expected probability of hearing loss at all cochlear RT doses and at 31 Gy exceeded 75% (95%CI 58.7–91.3, p < 0.001) (Figure 3). Examination of audiology assessments as a function of time showed generally increasing prevalence of moderate/severe SIOP Grade ≥2 hearing loss with further time from RT exposure (Figure 4).

Figure 3: Predicted probability for developing moderate/severe hearing loss (SIOP Grade ≥2) in those receiving/not receiving aBMT.

Probability of SIOP Grade ≥2 hearing loss predicted according to minimal dose of rdaiotherapy to cochlea (in Gy) in patients being treated with chemoradiatoin with or without autologous bone marrow transplant (aBMT) following adjustment for covariables in the multivariable model. 95% confidence intervals bands as depicted.

Figure 4: Prevalence of moderate/severe hearing loss (SIOP Grade ≥2) by time to audiology.

Percentage of ears with SIOP ≥2 hearing loss for assessments within two-year increments.

Discussion

In our cohort, RT dose to the cochlea was found to be clearly additive to the risk from underlying cisplatin-based chemotherapy in causing permanent hearing loss.^3–9,14,36 In contrast to previous reports,¹⁵ RT exposure remained a significant predictor of hearing loss even after accounting for cisplatin dose. The greatest predictor of hearing loss from underlying therapy was exposure to aBMT. Our analysis also found that while minimum, median, and mean RT dose were closely related, minimum cochlear RT dose was the most relevant parameter to assess the probability of developing ototoxicity from RT exposure to the cochlea. This latter finding is important for organ toxicity thresholds in radiotherapy planning and indicates the need for consistent and precise contouring of the cochlea as routine practice in order to ensure the calculation of the most predictive cochlear dose for risk of ototoxicity. RT remains integral to cure for many childhood cancers and this data for its off-target impact will aid in treatment planning to spare long-term toxicity.

Specifically, as advanced technology radiotherapy now provides increased precision to limit RT dosing exposure to off-target organs,³⁰ it is important to note that our findings for most relevant cochlear RT dose parameter lie in contrast to the more commonly referenced “mean dose.” However, there is a limited body of evidence to support reliance on any specific dose parameter. In a cohort of pediatric medulloblastoma patients treated with cisplatin and IMRT,³⁶ similar analysis found that only median cochlear RT dose correlated with hearing loss. Of note, that study used an older and less sensitive grading system to define hearing loss (Pediatric Oncology Group (POG) scale),³⁷ as well as analyzed a more homogeneous cohort with a narrower RT dose-range and relatively few patients with severe hearing loss. A study of nasopharyngeal cancer patients³⁸ did find that the RT minimum dose most strongly correlated with hearing loss, however, this study was of adult patients. Other reports analyze several RT dose parameters but do not directly compare with mean dose which was used for reporting in those studies.^13,39 In our data, though the RT dose parameters were correlated as would be expected, minimum dose was most predictive of likelihood of hearing loss, a finding that warrants further investigation. We propose that minimum and not mean dose may be the best RT dose predictor of ototoxicity.

In examining which aspects of underlying therapy were similarly ototoxic, we found that cumulative dose of cisplatin and timing of RT (i.e., prior to or after cisplatin exposure) were not independently predictive of hearing loss. However, this is not to state that cumulative cisplatin dose does not confer risk for ototoxicity, but rather, the lack of association may be due to cisplatin dose-reductions for ototoxicity during therapy in this contemporary patient cohort. Such ototoxicity-related cisplatin dose adjustments were similarly found to confound the association of cumulative cisplatin dose in a cohort of pediatric medulloblastoma patients.¹³ It is interesting to note that timing of RT exposure prior to cisplatin did not increase cochlear sensitivity to cisplatin or aBMT as it was not associated with increased risk for hearing loss in the cohort. Rather, these findings emphasize that increasing RT dose significantly worsens the already substantial underlying risk of hearing loss from chemotherapy.

A novel and notable finding from our research was the profound impact of autologous BMT on the development of hearing loss. This was a statistically significant predictor of hearing loss at both patient and ear-level, and the probability of developing hearing loss across the RT doses range was significantly higher in patients who received aBMT. Several etiologies may explain the additive risk. The majority of patients receiving aBMT were either medulloblastoma or neuroblastoma patients. Although no interaction with age was identified, many of these patients were younger and therefore may be at a baseline greater risk for ototoxicity. In addition, undergoing aBMT itself confers risk for ototoxicity from exposure to ototoxic conditioning chemotherapies and co-treatment with aminoglycosides, diuretics, or other ototoxic agents for aBMT-associated toxicities.²⁴ This risk would be further compounded in those requiring tandem (multiple) aBMTs. Should treatment considerations allow, adjustment of cranial RT doses and increased cochlear sparing in those patients planned for aBMT might be considered, particularly as these patients have substantially increased risk of hearing loss even at the lowest RT doses. This finding indicates the need for close audiology surveillance at potentially lower RT thresholds than previously considered in patients who receive both chemoradiation and aBMT as more than half of this population would be predicted to develop hearing loss at cochlear doses <20 Gy. In those receiving higher doses of RT to the cochlea additive to chemotherapy and aBMT, close surveillance is clearly warranted, as the large majority of patients are expected to develop hearing loss.

While ototoxicity surveillance is recommended for childhood cancer survivors exposed to cochlear radiation, aBMT, or cisplatin, scarce data exists to guide the timing or frequency.⁴⁰ Of note, median time to follow-up in this study is a relatively short duration of 2.14 years. Nonetheless, as seen in our cohort, many patients treated with combined modality cisplatin and RT develop hearing loss even during therapy³⁰ as the combination of these treatments likely results in earlier ototoxicity. Time from RT to audiology assessment was significant throughout our analyses, demonstrating the risk for developing hearing loss continues and may even increase long after therapy concludes. These analyses may therefore underestimate the significance of contributory factors to hearing loss with even greater magnitude of effects seen with longer duration of follow-up. As such, our data supports ototoxicity surveillance should begin soon after completion of therapy for those receiving combined modality therapy, and as seen with RT-only pediatric populations,⁴¹ confirms that surveillance for ototoxicity is also required long after completion of therapy.

As a retrospective, hypothesis-generating study, there were several limitations that may affect our findings, including innate biases of a retrospective investigation. Notably, hearing loss was analyzed within individual ears as per convention in radiation-induced hearing loss research^4,12,33 and due to asymmetrical risk in hearing loss from chemoradiation. However, as hearing loss by ears alone may not fully capture differences in host sensitivity nor the clinical impact to the patient, future studies are necessary to investigate the relative impact of cochlear dose and aBMT in larger cohorts. As a retrospective study, it is not possible to balance patient factors that may introduce confounding of the results, though the use of a multivariable analysis mitigated the extent of this impact on the findings. This study did not include data to detect differences in effect from the number of aBMTs per patient, limiting further analysis of the impact of single versus tandem aBMTs on hearing loss. The disease itself was also characterized by category, but there may be subtle differences in tumor location that could possibly contribute to hearing loss through direct involvement of hearing-related intracranial pathways.¹⁶ Similarly, despite patients likely receiving other ototoxic medications, only cisplatin dosing was captured, with reliance on disease and regimen serving as surrogate measures for the underlying chemotherapy platforms. Additionally, cochleae are small organs, and contouring is inherently subjective, though the limited number of radiation oncologists performing the contouring minimized variability, as did blinded review prior to RT dose calculation. Finally, only the most recent audiology assessments were included, and while this was controlled for in the analysis, the study could not assess for the rare occurrence of preexisting congenital or disease-related hearing loss nor whether RT dose impacted the rapidity of developing hearing loss following therapy.

Conclusions

Our study clearly demonstrates the additive risk of minimum cochlear RT dose to risk for ototoxicity and provides data that may aid in treatment planning to spare ototoxicity. Moreover, the profound impact of aBMT on risk for ototoxicity in RT-treated patients is substantial and identifies a population that warrants both greater audiology surveillance but also further research into limiting this toxicity. Hearing loss has a substantial impact on the lives of pediatric cancer survivors, affecting their development, school performance, and quality of life.¹⁰ Prospective investigation is needed to validate these findings to further inform RT dose thresholds and to minimize the risk of hearing loss in childhood cancer survivors.

Table 3:

Multivariable model of hearing loss SIOP Grade ≥2 (n=161 ears^*)

Covariables	Univariable Analysis			Multivariable analysis**
	OR	95% CI	p-value	OR	95% CI	p-value
Age, per 5 years	0.43	0.30–0.61	<0.001	0.95	0.43–1.27	0.322
Sex
Female		reference
Male	0.96	0.50–1.82	0.891
Ethnicity
Not Hispanic or Latino1		reference
Hispanic or Latino	1.36	0.62–3.00	0.441
Diagnosis
Non-CNS tumor		reference
CNS tumor	0.65	0.24–1.79	0.406
Cisplatin dose, per 100mg/m2	1.02	0.77–1.35	0.907	0.85	0.59–1.21	0.365
RT prior to cisplatin
No		reference
Yes	0.39	0.21–0.76	0.005
Treatment with aBMT
No		reference
Yes	7.25	3.32–15.85	<0.001	10.47	3.30–33.36	<0.001
Time to audiology, years	1.34	1.15–1.56	<0.001	1.42	1.15–1.75	0.001
Minimum RT dose,per +10 Gy	1.12	0.80–1.57	0.499	1.64	1.04–2.75	0.043

^*16 patients did not have sufficient ear-specific testing to determine SIOP grade.

^**Stepwise forward selection, retention α<0.15, significance p<0.05 (see methods).

¹excludes 1 unknown patient/2 ears; CNS = central nervous system, aBMT = autologous bone marrow transplant, RT = radiotherapy.