- Email: [email protected]
Prospective study of inner ear radiation dose and hearing loss in head-and-neck cancer patients
Int. J. Radiation Oncology Biol. Phys., Vol. 61, No. 5, pp. 1393–1402, 2005 Copyright © 2005 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/05/$–see front matter
doi:10.1016/j.ijrobp.2004.08.019
CLINICAL INVESTIGATION
Head and Neck
PROSPECTIVE STUDY OF INNER EAR RADIATION DOSE AND HEARING LOSS IN HEAD-AND-NECK CANCER PATIENTS
RHONDA
CHARLIE C. PAN, M.D.,* AVRAHAM EISBRUCH, M.D.,* JULIA S. LEE, M.S.,† M. SNORRASON, PH.D.,‡ RANDALL K. TEN HAKEN, PH.D.,* AND PAUL R. KILENY, PH.D.‡
Departments of *Radiation Oncology, †Biostatistics, and ‡Otolaryngology, University of Michigan, Ann Arbor, MI Purpose: To determine the relationship between the radiation dose to the inner ear and long-term hearing loss. Methods and Materials: Eligible patients included those receiving curative radiotherapy (RT) for head-and-neck cancer. After enrollment, patients underwent three-dimensional conformal RT planning and delivery (180 –200 cGy/fraction) appropriate for their disease site and stage. The inner ear was contoured on axial CT planning images. Dose–volume histograms, as well as the mean and maximal dose for each structure, were calculated. Patients underwent pure tone audiometry at baseline (before treatment) and 1, 6, 12, 24, and 36 months after RT. The threshold level (the greater the value, the more hearing loss) in decibels was recorded for 250, 500, 1000, 2000, 4000, and 8000 Hz. For patients receiving predominantly unilateral RT, the contralateral ear served as the de facto control. The differences in threshold level between the ipsilateral and contralateral ears were calculated, and the temporal pattern and dose–response relation of hearing loss were analyzed using statistical methods that take into account the correlation between two ears in the same subject and repeated, sequential measurements of each subject. Results: Of the 40 patients enrolled in this study, 35 qualified for analysis. Four patients who received concurrent chemotherapy and RT were analyzed separately. The 31 unilaterally treated patients received a median dose of 47.4 Gy (range, 14.1– 68.8 Gy) to the ipsilateral inner ear and 4.2 Gy (range, 0.5–31.3 Gy) to the contralateral inner ear. Hearing loss was associated with the radiation dose received by the inner ear (loss of 210dB was observed in ears receiving >45 Gy) and was most appreciable in the higher frequencies (>2000 Hz). For a 60-year-old patient with no previous hearing loss in either ear, after receiving 45 Gy, the ipsilateral ear, according to our clinical model, would have a 19.3-dB (95% confidence interval [CI], 15.5–23.0) and 5.4-dB (95% CI, 3.5–7.5) hearing decrement compared with the contralateral ear for 8000 Hz and 1000 Hz, respectively. Age and an initial hearing difference within an ear pair also affected hearing loss. The baseline hearing threshold was inversely related to radiation-induced hearing loss. The degree of hearing loss was dependent on the frequency tested, age, baseline hearing, and baseline difference in hearing between a patient’s two ears. Conclusion: High-frequency (>2000 Hz) hearing acuity worsens significantly after RT in a dose-dependent fashion. A larger number of patients needs to be studied to validate these results. This knowledge can be applied to create guidelines regarding future dose limits to the auditory apparatus for patients undergoing head-andneck RT. © 2005 Elsevier Inc. Hearing, Inner ear, Radiotherapy, Head-and-neck cancer.
Intensity-modulated radiotherapy (RT) has allowed for better dose conformation, improving dose escalation to the target and sparing of more normal tissue. To best use these capabilities, one needs a thorough understanding of the relationship between the normal tissue damage and the radiation dose to make rational decisions regarding the tradeoffs between the dose to the target and the sparing of normal tissue. For head-and-neck tumors, many normal tissue structures coexist in this region, but our understand-
ing of their response to RT varies by structure. Sensorineural hearing loss (SNHL) as a result of RT to the inner ear and cochlea is a radiation dose-limiting toxicity that needs additional investigation. Approximately 24,000 patients each year present with malignancy of the nasopharynx, parotid gland, paranasal sinuses, or brain (1). The inner ear is often included in the radiation field for treatment of these patients, and a substantial number of these patients will develop transient serous otitis media during or immediately after RT. More importantly, permanent SNHL due to radiation effects on the
Reprint requests to: Avraham Eisbruch, M.D., Department of Radiation Oncology, University of Michigan, 1500 E. Medical Center Dr., UH B2C490, Box 0010, Ann Arbor, MI 48109. Tel: (734) 936-4300; Fax: (734) 763-7370; E-mail: [email protected] Supported in part by National Cancer Institute Grant P01
CA59827. Acknowledgments—We thank Mr. Steve Kronenberg for assistance with figure preparation. Received May 4, 2004, and in revised form Aug 6, 2004. Accepted for publication Aug 16, 2004.
INTRODUCTION
1393
1394
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
cochlea will also develop in many of these patients. SNHL has been shown to result in significant cognitive impairment, depression, and reduction in functional status (2). The incidence of post-RT sensorineural deficit has been reported to range from 0% to 50% (3–5). Retrospective reports have also reported radiation-induced hearing loss to begin anywhere from 30 Gy to ⬎65 Gy (6, 7). The influence of confounding variables such as concurrent chemotherapy and other comorbidities that predispose to hearing loss has been unclear in many of these studies. Also, with these historical studies, a reliable dose estimation for the cochlea was often difficult because CT planning might not have been used. Therefore, because hearing loss is an especially morbid sequelae of head-and-neck RT, we sought to determine the relationship between the radiation dose to the inner ear and the development of hearing loss in head-and-neck patients treated at our institution with three-dimensional conformal RT.
METHODS AND MATERIALS Population This study was an institutional review board-approved prospective study involving 40 patients with head-and-neck tumors undergoing curative RT that was anticipated to involve one or both inner ears in the Department of Radiation Oncology at the University of Michigan. Patients agreed to undergo hearing tests, and all patients provided written informed consent for participation in this prospective study.
Treatment Each patient underwent three-dimensional planning and treatment as practiced at the Department of Radiation Oncology. Most patients received unilateral neck RT, and the inner ear contralateral to the primary tumor primarily received scatter radiation. All RT was planned using an in-house three-dimensional treatment planning system (UMPlan) (8 –10). Each patient underwent CT simulation, using a standard 3-mm slice thickness and interval, except for 5 patients who underwent simulation with a 5-mm slice thickness and interval, all with no gap. One radiation oncologist (C.C.P.) contoured both cochleae for each patient (Fig. 1). Dose– volume data for each cochlea were obtained. Because the cochleae are small (average volume, 0.56 cm3; range, 0.15– 0.91), the mean cochlear dose was used in our analysis. The median prescription dose was 64.0 Gy (range, 40.0 –70.0 Gy). No modification in the plan or prescribed radiation dose was made due to the patient participation in this study. Most patients were treated primarily with surgery, followed by postoperative RT. Four patients, all with nasopharyngeal cancer, were treated primarily with concurrent, cisplatin-based chemoradiotherapy. The patient characteristics are presented in Table 1.
Hearing tests Audiologic testing was conducted at baseline (before treatment) and at 1, 6, 12, 24, and 36 months after RT completion. Each test battery consisted of otoscopy, acoustic immittance testing (tympanometry and acoustic reflex), pure-tone audiometry, including air and bone conduction testing, speech reception threshold, and
Fig. 1. Computed tomography anatomy of middle and inner ear.
speech recognition determination. A hearing decrement of ⱖ10 dB was considered clinically significant (11). To avoid contamination of the hearing threshold data by the presence of conductive hearing loss caused by middle ear effusion, bone conduction thresholds were used to establish the changes in hearing for frequencies ⱕ4000 Hz. Bone conduction testing is a routine component of an audiologic assessment and involves the delivery of pure tone stimuli via a bone oscillator placed on the mastoid of the ear to be tested, thus bypassing the conductive mechanism. These measurements, therefore, reflect only sensory hearing loss, when present. Because bone conduction hearing testing is limited to 4000 Hz, measurements ⬎4000 Hz were performed using air conduction testing alone. Sensorineural hearing at high frequencies (8000 Hz) tested by air conduction is unaffected by, and independent of, middle ear effusion.
Audiologic variables The baseline threshold level (bTL) of an irradiated ear is defined as the hearing threshold level (measured in decibels) obtained at baseline audiologic testing. The difference in the threshold levels (dTL) is defined as the difference between the threshold levels of the irradiated ear and contralateral ear for each patient. The four concurrent chemoradiotherapy patients received significant bilateral inner ear radiation doses, and these patients were analyzed separately.
Statistical analysis The outcome variable was the dTL between ears for each patient, reflecting the sensory hearing loss associated with an increased radiation dose to the ipsilateral ear. To examine the
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1395
Table 2. Average threshold difference between each ear pair for each frequency and time point
Table 1. Patient characteristics Characteristic
Value
Total (n) Average age (y) Gender (n) Male Female Average inner ear volume (cm3) Concurrent cisplatin (n) Medical comorbidities (n) Hypertension Diabetes mellitus Inner ear dose for patients receiving unilateral RT (n ⫽ 31) Ipsilateral, treated side (Gy) Median Range Contralateral, untreated side (Gy) Median Range Tumor sites (n) Nasopharynx Oropharynx Oral cavity Salivary gland Paranasal sinus (mostly maxillary sinus) Skin Stage (n) Nonmalignant Stage I/II Stage III Stage IV
35 57.9 18 17 0.56 4 (11) 11 (31) 5 (15)
Time after RT (mo)
Frequency (Hz)
Patients (n)
Average dTL (dB)
SD
Baseline (before RT)
250
29
⫺0.9
7.6
500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000
30 30 30 30 25 22 22 22 22 22 19 20 20 20 20 20 15 17 17 17 16 16 13 13 13 13 13 12 10
1.3 0.3 4.2 3.3 5.6 ⫺0.2 4.1 3.6 8.2 9.8 12.9 2.3 1.0 0.0 7.0 7.3 15.0 1.5 7.4 2.9 6.9 10.3 15.4 2.7 1.9 0.0 8.1 7.5 17.5
9.0 11.1 10.8 16.2 13.2 6.3 8.8 13.7 11.9 15.4 21.8 10.8 7.4 11.5 12.3 13.4 18.6 8.6 7.5 10.6 9.8 18.8 17.3 9.5 8.3 8.2 11.8 11.4 20.8
1
47.4 14.1–68.8 6 4.2 0.5–31.3 4 1 11 8 9 3 3 7 8 17
12
24
Abbreviation: RT ⫽ radiotherapy. patterns of hearing loss across time and across frequencies, line graphs of hearing thresholds vs. time and hearing thresholds vs. frequency were produced for individual subjects using Splus (Mathsoft, version 6.0, release 1, Sun OS 5.3, Cambridge, MA). Additional line graphs were produced for the overall sample to establish whether the variance in hearing loss changed either over time or across frequencies. The repeated hearing threshold measurements among the individual subjects were modeled as a function of time and frequency by fitting linear mixed models using PROC MIXED (Statistical Analysis Systems proprietary software, release 8.2, SAS Institute, Cary, NC). This accounts for the correlation among repeated measures from individual subjects. Curvilinear relations of threshold with frequencies were examined by including quadratic and cubic terms for frequencies. With the above analysis, a statistical model was created that would allow for the estimation of radiation-induced hearing loss. In addition to radiation dose, other covariates of interest included the difference in hearing threshold at baseline (dTL0), baseline threshold, difference in radiation dose between irradiated and nonirradiated ears, natural logarithm of the test frequency, and time. Adjustments for potential confounding by gender, age, and comorbidities associated with hearing loss (hypertension and diabetes mellitus) were also included in the model. The final results of the model are presented in nomogram form. The model used in this study is presented in the Appendix. A two-tailed p value of ⱕ0.05 was considered statistically significant.
Abbreviations: RT ⫽ radiotherapy; dTL ⫽ difference in threshold level. Positive dTL denotes increased hearing loss in ipsilateral ear compared with contralateral ear.
RESULTS We enrolled 40 patients into this prospective study, and 35 patients participated in the required hearing tests. Of these 35 patients, 4 patients received bilateral RT and con-
Table 3. Incidence of clinically significant hearing loss (ⱖ10 dB) for 8000 Hz and 4000 Hz Time after RT (mo)
Frequency (Hz)
1 6 12 24 36 1 6 12 24 36
8000
4000
Patients (n)
Patients with ⱖ 10 dB loss (n)
19 16 13 10 2 23 20 14 12 3
9 7 6 4 2 5 5 7 6 1
1396
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
Fig. 2. Difference in hearing between ipsilateral and contralateral ears (dTL) vs. differences in mean cochlea dose, over time.
current cisplatin chemotherapy for their primary nasopharynx cancer. We present their data in brief later in this paper. The 31 unilaterally treated, RT-alone patients were used in the primary analysis. Of the 31 patients, 14 were women and 17 were men, with the average age of 58 years (SD, 15.7). The median radiation dose was 47.4 Gy and 4.2 Gy for the ipsilateral and contralateral ear, respectively. Table 2 lists the average threshold difference between the ipsilateral and contralateral ears at each frequency and time point. At each point after RT, a greater difference in the dTL was observed at the higher frequencies (ⱖ2000 Hz). The incidence of clinically significant hearing loss is seen in Table 3. Overall, 6, 6, and 4 patients had a ⱖ10-dB hearing loss at two or more follow-up visits at 8000, 4000, and 2000 Hz, respectively. Longitudinally, for those patients with a
ⱖ10-dB hearing loss noted during the first 6 months, recovery to baseline hearing after 6 months was seen in 3 of 11, 2 of 7, and 0 of 6 patients at 8000, 4000, and 2000 Hz, respectively. The dTL for each test frequency was plotted against the differences in the mean cochlea dose for each time point (Fig. 2). A positive dTL indicated increased hearing loss in the ipsilateral, heavily irradiated ear compared with the contralateral ear. Hearing loss was most predominant and clinically significant (ⱖ10 dB) at the higher frequencies (ⱖ2000 Hz). Throughout this study, 20 of the 31 tested patients had some degree of conductive hearing loss secondary to middle ear effusion at one or more visits, confirmed by an air-bone gap, tympanometry, and otoscopy. This condition was for
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1397
Fig. 2. (Continued).
the most part self-limiting, but when indicated, treatment was provided. The association of the dTL between the ipsilateral and contralateral ears with the radiation dose, age, test frequency, bTL, and threshold difference between two ears at baseline (dTL0) was statistically significant (p ⬍ 0.05), as described below and in the nomograms in Figs. 3– 6. The pattern of hearing loss was not significantly related to time; however, this may have been because of the lack of enough long-term data. In this study, only 3 patients had undergone
Fig. 3. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), baseline threshold level (bTL) (0 dB), and differences in threshold levels (dTL)0 (0 dB), by test frequency.
their 36-month audiologic evaluation, because of poor patient compliance and loss to follow-up. Finally, the potential confounding variables of gender and the presence of hypertension and/or diabetes mellitus did not have any statistically significant effects on hearing loss. To illustrate the effects of each covariate on hearing loss, nomograms based on the final model are presented. These demonstrate the predicted hearing loss for varying radiation dose levels for a constant age, bTL, and dTL0 by test frequency (Fig. 3), for a constant bTL, dTL0, and frequency by age (Fig. 4), for a constant age, dTL0, and frequency by bTL (Fig. 5), and for a constant age, bTL, and frequency by dTL0 (Fig. 6). Radiation dose effect An increase in the mean radiation dose to the inner ear was associated with increased hearing loss (Fig. 3). This effect varied across frequencies, with identical radiation doses causing more hearing loss in the higher than in the lower frequencies. The damage was clinically significant (ⱖ10 dB) at frequencies of ⱖ2000 Hz. For example, using our final model, in a hypothetical 60-year-old patient with normal hearing in both ears at baseline across frequencies, after receiving 45 Gy, the ipsilateral ear would later present with 19.3-dB (95% confidence interval [CI], 15.5–23.0) and 5.4-dB (95% CI 3.5–7.5) more hearing loss compared with the contralateral ear at 8000 Hz and 1000 Hz, respectively. With 60 Gy, the ipsilateral ear would be expected to have 25.1-dB (95% CI, 20.6 –29.7) and 6.6 dB (95% CI, 4.4 – 8.9) more hearing loss than the contralateral ear at 8000 Hz and 1000 Hz, respectively. Within this study, we observed that for almost all cases in which significant hearing loss occurred in the ipsilateral ear receiving a high dose compared with the contralateral ear, the dose received was ⱖ45 Gy.
1398
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
Fig. 4. Nomogram of hearing loss after radiation dose delivered for constant baseline threshold levels (bTL) (0 dB), differences in threshold levels (dTL0) (0 dB), and frequency (2000, 4000, and 8000 Hz), by age.
Fig. 5. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), differences in threshold levels (dTL)0 (0 dB), and frequency (2000, 4000, and 8000 Hz), by baseline hearing threshold level (bTL).
Age effect The age effect on hearing loss is most noticeable at the higher frequencies, especially 4000 Hz and 8000 Hz (Fig.
4). Older patients are more susceptible to hearing loss. For example, for a 70-year-old hypothetical patient with normal hearing in both ears at baseline, after receiving 45 Gy, the
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1399
tralateral ear. For 50-year-old patient, 20 years younger, the ipsilateral ear would only be about 13.6 dB (95% CI, 10.3–16.9) and 10.8 dB (95% CI, 8.6 –13.0) poorer than the contralateral ear at 8000 Hz and 4000 Hz, respectively. Baseline hearing effect Ears with lower (better) thresholds at pre-RT baseline developed more hearing loss (greater thresholds) after RT (Fig. 5). For a 60-year-old hypothetical patient with normal or 0-dB baseline hearing in both ears, after 45-Gy RT, the ipsilateral ear threshold would be 20 dB (8000 Hz) or 5 dB (1000 Hz) poorer than the contralateral ear. In contrast, if this patient had a worse, baseline hearing of 20 dB in both ears, the ipsilateral ear threshold would only be 15 dB (8000 Hz) or ⬍5-dB (1000 Hz) poorer than the contralateral ear. Effect of pre-RT hearing threshold differences between ears The larger the differences between the ipsilateral and contralateral ears at baseline at which the ipsilateral ear presented with elevated (poorer) thresholds, the larger the difference after RT (Fig. 6). For example, if a 60-year-old hypothetical patient had the same hearing in both ears at baseline, after 45-Gy RT, the ipsilateral ear would be 19.3 dB (95% CI, 15.6 –23.0) or 8.6 dB (95% CI, 6.4 –10.7) worse than the contralateral ear for 8000 Hz and 2000 Hz, respectively. However, if the ipsilateral ear was 15 dB worse than the contralateral ear at baseline, after RT, it would be 30.7 dB (95% CI, 26.9 –34.4) and 18.2 dB (95% CI, 15.8 –20.5) worse than the contralateral ear for 8000 Hz and 2000 Hz, respectively. The effect of RT on the ipsilateral ear would be smaller if the contralateral ear had poorer baseline hearing (smaller difference) compared with the ipsilateral ear. Cisplatin effect on hearing The difference in hearing acuity for the 4 patients (eight inner ears receiving ⬎50 Gy) who received cisplatin was not statistically significant from that of the seven inner ears receiving similar doses in patients who had not received cisplatin. However, given the small number of patients, no strong conclusion about the effect of cisplatin and its effects on radiation-induced hearing loss could be drawn from this comparison. DISCUSSION
Fig. 6. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), baseline threshold level (bTL) (0 dB), and frequency (2000, 4000, and 8000 Hz), by differences in threshold level (dTL)0.
ipsilateral ear threshold would be elevated by 25.0 dB (95% CI, 20.3–29.8) or 15.5 dB (95% CI, 12.3–18.8), respectively, at 8000 Hz and 4000 Hz compared with the con-
Effect of RT on one inner ear—a controlled study Previous studies examining the effect of RT on SNHL mostly focused on patients with nasopharyngeal carcinoma. Although selection of this population ensured a group of patients whose inner ears received a clinically significant dose, the drawback was that typically both inner ears were uniformly treated. Without a “control” ear, the true relationship between the radiation dose and its effect is difficult to tease out from the confounding effects of other risk factors, time, and interventions such as chemotherapy. In our study,
1400
I. J. Radiation Oncology
● Biology ● Physics
we chose primarily to examine patients who received unilateral RT, and thus, each irradiated inner ear was compared with its built-in, contralateral control inner ear. This is unique to our study and adds a measure of confidence in our results lacking from all other published series. Radiation dose–response relation for hearing loss Dose–response data have been scant to date. Grau et al. (12) reported on 22 patients and found significantly greater SNHL for cochlear doses ⬎50 Gy. Kwong et al. (4) prospectively examined nasopharyngeal carcinoma patients treated with RT and cisplatin, but the radiation dose had no statistically significant effect on SNHL. Honoré et al. (13) demonstrated a dose–response curve for SNHL and dose for hearing loss at 4000 Hz. Most recently, Merchant et al. (14) studied 72 children treated with conformal RT and suggested that the average cochlear dose should be kept to ⬍32 Gy to prevent hearing loss. Our observation that clinically apparent hearing loss (ⱖ10 dB) after ⱖ45 Gy is consistent with these series. Although our model suggests a nearly linear relationship, the degree of hearing loss within the model did not reach clinical significance (ⱖ10 dB) until around 40 – 45 Gy, consistent with our observations. Additional data are needed to validate this observation. Our data also suggest that radiation-induced SNHL occurs throughout the entire frequency range of hearing but is only clinically significant (ⱖ10 dB) at higher frequencies, ⱖ2000 Hz. The audiometric frequency range of 2000 – 8000 Hz overlaps with the frequency range of human speech (500 – 4000 Hz). The upper end of the speech frequency range (2000 – 4000 Hz) is especially important for consonant sound recognition. Hearing loss in this frequency range can result in reduced speech recognition abilities. Confounding risk factors and interventions Confounding factors such as preexisting hearing loss, age, hypertension, diabetes mellitus, and treatment with cisplatin were also examined in this study. Because of the design of our trial, the effect of RT alone was isolated, as well as the effects of these confounders. Age and preexisting hearing loss had an effect on radiation-induced SNHL, consistent with the pathophysiology of this deficit (15). Cisplatin ototoxicity is well documented, and its relationship to radiation-induced SNHL has been described by Walker et al. (16). They studied a cohort of children with brain tumors who had received cisplatin chemotherapy and whose radiation portals encompassed the inner ear. The resulting hearing loss was compared with children who had received comparable doses of cisplatin alone, and synergized ototoxicity was found. In our study, we had only 4 patients who received cisplatin chemotherapy, and when analyzed, found no statistically significant hearing loss difference between the inner ears treated to similar doses. We could not, however, draw any strong conclusions from our data, because we examined only 8 inner ears. Middle ear effusion causes conductive hearing loss by interrupting the mechanical transmission of sound to the
Volume 61, Number 5, 2005
sensory apparatus. This confounder was avoided by using bone conduction measurements only, which is a direct measurement of sensory hearing. Another tool used to measure sensory hearing includes the auditory brainstem response, a noninvasive electrophysiologic test that may be used for hearing threshold estimation (17) or for neurodiagnostic purposes (18). It is an excellent tool for the determination of hearing thresholds in infants and young children or in cognitively impaired adults. However, thresholds determined using auditory brainstem responses may differ from actual hearing thresholds measured behaviorally by as much as 5–15 dB (17). Because the auditory brainstem response is generally not recommended as a substitute for behavioral audiometry unless absolutely necessary or for neurodiagnostic purposes, its use was not warranted for our study.
Time course The persistence of SNHL is difficult to assess from the literature, because the data can be biased as some of the patients will develop otitis media in the first few months after RT that will resolve in the first year (19). Another study (5) suggested that SNHL is found between 1.5 and 2 years after RT. In our study, we did not find any time association between hearing loss and RT. Also, on the basis of our data, we could not yet determine whether SNHL improved after 2 years, because our follow-up was limited after 2 years.
Radiobiologic correlations Animal studies have demonstrated SNHL with histopathologic correlation to damage to the organ of Corti and endolymph beginning at 40 Gy (20, 21). After conventionally fractionated RT, a relationship was established among the radiation dose received, cellular loss in the organ of Corti, and hearing loss. Additionally, the vast amount of “radiosurgery” literature has demonstrated the vulnerability of additional hearing apparatus structures, such as the acoustic nerve, to radiation. Although this was not specifically examined in our study, radiation effects on other portions of the auditory pathway will be examined in the future.
Future hearing-loss models We have presented a fitted model based on the population in our study. Although this model accurately represents our data, additional data will allow us to refine our model to predict hearing loss more accurately. Current work is in progress to apply normal tissue complication probability modeling techniques to the dose–volume data obtained in this study and to apply this model to patients receiving intensity-modulated RT in which low total doses received by the inner ear are also associated with a lower fraction dose compared with the target fraction dose.
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1401
Table 4. Fixed effect parameter estimates on final model Fixed effect
Estimate
Standard error
df
p
95% CI
Fixed effect intercept Baseline contralateral hearing threshold Radiation dose* Baseline difference between ear pair ⫻ Frequency† Radiation dose ⫻ age Radiation dose ⫻ frequency Radiation dose ⫻ (frequency) (23) Radiation dose ⫻ age ⫻ frequency Radiation dose ⫻ age ⫻ (frequency) (23)
2.4 ⫺0.2 ⫺5.8 0.09 0.11 1.7 ⫺0.13 ⫺0.04 0.003
2.7 0.04 1.9 0.006 0.03 0.5 0.04 0.009 0.0006
153 226 143 137 138 144 145 139 141
0.4 ⬍0.0001 0.004 ⬍0.0001 0.0007 0.002 0.001 0.0003 ⬍0.0001
⫺2.9, 7.6 ⫺0.3, ⫺0.1 ⫺9.6, ⫺1.9 0.07, 0.10 0.05, 0.18 0.6, 2.8 ⫺0.2, ⫺0.05 ⫺0.05, ⫺0.02 0.001, 0.004
Abbreviation: CI ⫽ confidence interval. * Radiation dose represents radiation dose received in ipsilateral ear. † Frequency was logarithm transformed.
CONCLUSION These results, including the dose effect and the effect of other factors, should allow clinicians to better weight any
necessary tradeoffs among sparing the inner ears, sparing other neighboring critical structures, and RT of the adjoining targets.
REFERENCES 1. Jemal A, Murray T, Samuels A, et al. Cancer statistics, 2003. CA Cancer J Clin 2003;53:5–26. 2. Cacciatore F, Napoli C, Abete P, et al. for the Osservatorio Geriatrico Campano Study Group. Quality of life determinants and hearing function in an elderly population. Gerontology 1999;45:323–328. 3. Anteunis LJ, Wanders SL, Hendriks JJ, et al. A prospective longitudinal study on radiation-induced hearing loss. Am J Surg 1994;168:408 – 411. 4. Kwong DL, Wei WI, Sham JS, et al. Sensorineural hearing loss in patients treated for nasopharyngeal carcinoma: A prospective study of the effect of radiation and cisplatin treatment. Int J Radiat Oncol Biol Phys 1996;36:281–289. 5. Ho WK, Wei WI, Kwong DL, et al. Long-term sensorineural hearing deficit following radiotherapy in patients suffering from nasopharyngeal carcinoma: A prospective study. Head Neck 1999;21:547–553. 6. Schot LJ, Hilgers FJ, Keus RB, et al. Late effects of radiotherapy on hearing. Eur Arch Otorhinolaryngol 1992;249: 305–308. 7. Chen WC, Liao CT, Tsai HC, et al. Radiation-induced hearing impairment in patients treated for malignant parotid tumor. Ann Otol Rhinol Laryngol 1999;108:1159 –1164. 8. Fraass BA, McShan DL. 3-D treatment planning: III. Complete Beam’s-Eye-View planning capabilities. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 193–196 9. Fraass BA, McShan DL, Ten Haken RK, et al. 3-D treatment planning: V. A Fast 3-D photon calculation model. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 521–525 10. McShan DL, Fraass BA. 3-D treatment planning: II. Integration of gray scale images and solid surface graphics. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 41– 44 11. Raaijmakers E, Engelen AM. Is sensorineural hearing loss a possible side effect of nasopharyngeal and parotid irradiation? A systematic review of the literature. Radiother Oncol 2002; 65:1–7. 12. Grau C, Moller K, Overgaard M, et al. Sensori-neural hearing
13.
14.
15. 16.
17.
18.
19.
20.
21.
22.
23.
24.
loss in patients treated with irradiation for nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1991;21:723–728. Honoré HB, Bentzen SM, Moller K, et al. Sensori-neural hearing loss after radiotherapy for nasopharyngeal carcinoma: individualized risk estimation. Radiother Oncol 2002;65:9 –16. Merchant TE, Gould CJ, Xiong X, et al. Early neuro-otologic effects of three-dimensional irradiation in children with primary brain tumors. Int J Radiat Oncol Biol Phys 2004;58: 1194 –1207. Pickles JO. An introduction to the physiology of hearing. (2nd ed). London: Academic Press; 1999. Walker DA, Pillow J, Waters KD, et al. Enhanced cis-platinum ototoxicity in children with brain tumours who have received simultaneous or prior cranial irradiation. Med Pediatr Oncol 1989;17:48 –52. Kileny PR, Magathan MG. Predictive value of ABR in infants and children with moderate to profound hearing impairment. Ear Hear 1987;8:217–221. El-Kashlan HK, Eisenmann D, Kileny PR. Auditory brain stem response in small acoustic neuromas. Ear Hear 2000;21: 257–262. Tang NL, Choy AT, John DG, et al. The otological status of patients with nasopharyngeal carcinoma after megavoltage radiotherapy. J Laryngol Otol 1992;106:1055–1058. Johannesen TB, Rasmussen K, Winther FO, et al. Late radiation effects on hearing, vestibular function, and taste in brain tumor patients. Int J Radiat Oncol Biol Phys 2002;53:86 –90. Winther FO. X-ray irradiation of the inner ear of the guinea pig: Early degenerative changes in the vestibular sensory epithelia. Acta Otolaryngol 1969;68:514 –525. Verbeke G, Molenberghs G. Linear mixed models in longitudinal data. In: Verbeke G, Molenberghs G, editors. Inference for the random effects. New York: Springer-Verlag; 2000. Brant LJ, Fozard JL. Age changes in pure-tone hearing thresholds in a longitudinal study of normal human aging. J Acoust Soc Am 1990;88:813– 820. Brant LJ, Pearson JD. Modeling the variability in longitudinal patterns of aging. In: Crews DE, Garruto RM, editors. Biological anthropology and aging: Perspectives on human vari-
1402
I. J. Radiation Oncology
● Biology ● Physics
ation over the life span. New York: Oxford University Press; 1994. p. 373–393
Volume 61, Number 5, 2005
25. Morrell CH, Brant LJ. Modelling hearing thresholds in the elderly. Stat Med 1991;10:1453–1464.
APPENDIX A linear mixed model with random intercept (22, 23) was used to investigate the relationship between hearing threshold and clinical variables of interest. Specifically, the model is as follows: yijK ⫽ ⫹ XijK ⫹ aij ⫹ ijK where i is subject 1 through 31; j is the frequency 250, 500, 1000, 2000, 4000, 8000 Hz; k is the time point 1 through 5 (1, 6, 12, 24, and 36 months); aij is N(0, 12); ⑀ijk ⬃ N(0, 22); yijk is the column vector of hearing threshold difference between the ipsilateral and contralateral ears at time k, frequency j, for subject i; is the population average; and Xijk is the row vector that contains the subject level variables, which included the following main effects: age, gender, hypertension, diabetes mellitus, time point, radiation dose difference between ear pair, radiation dose of ipsilateral/contralateral ear, baseline hearing difference between ear pair, baseline hearing of ipsilateral/contralateral ear, frequency (logarithm transformed), and quadratic term of logarithm transformed frequency (23). Two-way interaction terms included age by baseline hearing difference between ear pair, age by time, age by radiation dose difference between ear pair, age by frequency, age by quadratic frequency (23–25), time by baseline hearing difference between ear pair, time by radiation dose difference between ear pair, time by frequency, time by quadratic frequency, baseline hearing difference between ear pair by frequency, baseline hearing difference between ear pair by radiation dose difference between ear pair, frequency by radiation dose between ear pair, baseline hearing difference between ear pair by quadratic frequency,
radiation dose difference between ear pair by quadratic frequency, time by ipsilateral ear radiation dose, age by ipsilateral ear radiation dose, baseline hearing difference between ear pair by ipsilateral ear radiation dose, frequency by ipsilateral ear radiation dose, and quadratic frequency by ipsilateral ear radiation dose. Three-way interaction terms included age by ipsilateral ear radiation dose by frequency, age by ipsilateral ear radiation dose by quadratic frequency, ipsilateral ear radiation dose by time by frequency, ipsilateral ear radiation dose by time by quadratic frequency, ipsilateral ear radiation dose by baseline hearing difference between ear pair by frequency, and ipsilateral ear radiation dose by baseline hearing difference between ear pair by quadratic frequency. aij are independent (subject by frequency) level random effects that account for variations between frequencies due to external, unmeasured factors. ⑀ijk are independent, normally distributed, between-frequency and between-time point differences. The final model was chosen through backward deletion, Akaike Information Criterion, and Schwarz’s Bayesian Information Criterion model selection criteria. The preliminary analysis included the main effects, and the time point was modeled as either a linear variable or five binary factor variables. Two-way and three-way interaction terms were added based on data from Brant and Fozard (23), Brant and Pearson (24), and Morrell and Brant (25) and additional interactions of interest. Among the three equally fitted models, the final one presented here (Table 4) was chosen on the basis of the literature and the relevance of the clinical variables remaining in the model.
doi:10.1016/j.ijrobp.2004.08.019
CLINICAL INVESTIGATION
Head and Neck
PROSPECTIVE STUDY OF INNER EAR RADIATION DOSE AND HEARING LOSS IN HEAD-AND-NECK CANCER PATIENTS
RHONDA
CHARLIE C. PAN, M.D.,* AVRAHAM EISBRUCH, M.D.,* JULIA S. LEE, M.S.,† M. SNORRASON, PH.D.,‡ RANDALL K. TEN HAKEN, PH.D.,* AND PAUL R. KILENY, PH.D.‡
Departments of *Radiation Oncology, †Biostatistics, and ‡Otolaryngology, University of Michigan, Ann Arbor, MI Purpose: To determine the relationship between the radiation dose to the inner ear and long-term hearing loss. Methods and Materials: Eligible patients included those receiving curative radiotherapy (RT) for head-and-neck cancer. After enrollment, patients underwent three-dimensional conformal RT planning and delivery (180 –200 cGy/fraction) appropriate for their disease site and stage. The inner ear was contoured on axial CT planning images. Dose–volume histograms, as well as the mean and maximal dose for each structure, were calculated. Patients underwent pure tone audiometry at baseline (before treatment) and 1, 6, 12, 24, and 36 months after RT. The threshold level (the greater the value, the more hearing loss) in decibels was recorded for 250, 500, 1000, 2000, 4000, and 8000 Hz. For patients receiving predominantly unilateral RT, the contralateral ear served as the de facto control. The differences in threshold level between the ipsilateral and contralateral ears were calculated, and the temporal pattern and dose–response relation of hearing loss were analyzed using statistical methods that take into account the correlation between two ears in the same subject and repeated, sequential measurements of each subject. Results: Of the 40 patients enrolled in this study, 35 qualified for analysis. Four patients who received concurrent chemotherapy and RT were analyzed separately. The 31 unilaterally treated patients received a median dose of 47.4 Gy (range, 14.1– 68.8 Gy) to the ipsilateral inner ear and 4.2 Gy (range, 0.5–31.3 Gy) to the contralateral inner ear. Hearing loss was associated with the radiation dose received by the inner ear (loss of 210dB was observed in ears receiving >45 Gy) and was most appreciable in the higher frequencies (>2000 Hz). For a 60-year-old patient with no previous hearing loss in either ear, after receiving 45 Gy, the ipsilateral ear, according to our clinical model, would have a 19.3-dB (95% confidence interval [CI], 15.5–23.0) and 5.4-dB (95% CI, 3.5–7.5) hearing decrement compared with the contralateral ear for 8000 Hz and 1000 Hz, respectively. Age and an initial hearing difference within an ear pair also affected hearing loss. The baseline hearing threshold was inversely related to radiation-induced hearing loss. The degree of hearing loss was dependent on the frequency tested, age, baseline hearing, and baseline difference in hearing between a patient’s two ears. Conclusion: High-frequency (>2000 Hz) hearing acuity worsens significantly after RT in a dose-dependent fashion. A larger number of patients needs to be studied to validate these results. This knowledge can be applied to create guidelines regarding future dose limits to the auditory apparatus for patients undergoing head-andneck RT. © 2005 Elsevier Inc. Hearing, Inner ear, Radiotherapy, Head-and-neck cancer.
Intensity-modulated radiotherapy (RT) has allowed for better dose conformation, improving dose escalation to the target and sparing of more normal tissue. To best use these capabilities, one needs a thorough understanding of the relationship between the normal tissue damage and the radiation dose to make rational decisions regarding the tradeoffs between the dose to the target and the sparing of normal tissue. For head-and-neck tumors, many normal tissue structures coexist in this region, but our understand-
ing of their response to RT varies by structure. Sensorineural hearing loss (SNHL) as a result of RT to the inner ear and cochlea is a radiation dose-limiting toxicity that needs additional investigation. Approximately 24,000 patients each year present with malignancy of the nasopharynx, parotid gland, paranasal sinuses, or brain (1). The inner ear is often included in the radiation field for treatment of these patients, and a substantial number of these patients will develop transient serous otitis media during or immediately after RT. More importantly, permanent SNHL due to radiation effects on the
Reprint requests to: Avraham Eisbruch, M.D., Department of Radiation Oncology, University of Michigan, 1500 E. Medical Center Dr., UH B2C490, Box 0010, Ann Arbor, MI 48109. Tel: (734) 936-4300; Fax: (734) 763-7370; E-mail: [email protected] Supported in part by National Cancer Institute Grant P01
CA59827. Acknowledgments—We thank Mr. Steve Kronenberg for assistance with figure preparation. Received May 4, 2004, and in revised form Aug 6, 2004. Accepted for publication Aug 16, 2004.
INTRODUCTION
1393
1394
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
cochlea will also develop in many of these patients. SNHL has been shown to result in significant cognitive impairment, depression, and reduction in functional status (2). The incidence of post-RT sensorineural deficit has been reported to range from 0% to 50% (3–5). Retrospective reports have also reported radiation-induced hearing loss to begin anywhere from 30 Gy to ⬎65 Gy (6, 7). The influence of confounding variables such as concurrent chemotherapy and other comorbidities that predispose to hearing loss has been unclear in many of these studies. Also, with these historical studies, a reliable dose estimation for the cochlea was often difficult because CT planning might not have been used. Therefore, because hearing loss is an especially morbid sequelae of head-and-neck RT, we sought to determine the relationship between the radiation dose to the inner ear and the development of hearing loss in head-and-neck patients treated at our institution with three-dimensional conformal RT.
METHODS AND MATERIALS Population This study was an institutional review board-approved prospective study involving 40 patients with head-and-neck tumors undergoing curative RT that was anticipated to involve one or both inner ears in the Department of Radiation Oncology at the University of Michigan. Patients agreed to undergo hearing tests, and all patients provided written informed consent for participation in this prospective study.
Treatment Each patient underwent three-dimensional planning and treatment as practiced at the Department of Radiation Oncology. Most patients received unilateral neck RT, and the inner ear contralateral to the primary tumor primarily received scatter radiation. All RT was planned using an in-house three-dimensional treatment planning system (UMPlan) (8 –10). Each patient underwent CT simulation, using a standard 3-mm slice thickness and interval, except for 5 patients who underwent simulation with a 5-mm slice thickness and interval, all with no gap. One radiation oncologist (C.C.P.) contoured both cochleae for each patient (Fig. 1). Dose– volume data for each cochlea were obtained. Because the cochleae are small (average volume, 0.56 cm3; range, 0.15– 0.91), the mean cochlear dose was used in our analysis. The median prescription dose was 64.0 Gy (range, 40.0 –70.0 Gy). No modification in the plan or prescribed radiation dose was made due to the patient participation in this study. Most patients were treated primarily with surgery, followed by postoperative RT. Four patients, all with nasopharyngeal cancer, were treated primarily with concurrent, cisplatin-based chemoradiotherapy. The patient characteristics are presented in Table 1.
Hearing tests Audiologic testing was conducted at baseline (before treatment) and at 1, 6, 12, 24, and 36 months after RT completion. Each test battery consisted of otoscopy, acoustic immittance testing (tympanometry and acoustic reflex), pure-tone audiometry, including air and bone conduction testing, speech reception threshold, and
Fig. 1. Computed tomography anatomy of middle and inner ear.
speech recognition determination. A hearing decrement of ⱖ10 dB was considered clinically significant (11). To avoid contamination of the hearing threshold data by the presence of conductive hearing loss caused by middle ear effusion, bone conduction thresholds were used to establish the changes in hearing for frequencies ⱕ4000 Hz. Bone conduction testing is a routine component of an audiologic assessment and involves the delivery of pure tone stimuli via a bone oscillator placed on the mastoid of the ear to be tested, thus bypassing the conductive mechanism. These measurements, therefore, reflect only sensory hearing loss, when present. Because bone conduction hearing testing is limited to 4000 Hz, measurements ⬎4000 Hz were performed using air conduction testing alone. Sensorineural hearing at high frequencies (8000 Hz) tested by air conduction is unaffected by, and independent of, middle ear effusion.
Audiologic variables The baseline threshold level (bTL) of an irradiated ear is defined as the hearing threshold level (measured in decibels) obtained at baseline audiologic testing. The difference in the threshold levels (dTL) is defined as the difference between the threshold levels of the irradiated ear and contralateral ear for each patient. The four concurrent chemoradiotherapy patients received significant bilateral inner ear radiation doses, and these patients were analyzed separately.
Statistical analysis The outcome variable was the dTL between ears for each patient, reflecting the sensory hearing loss associated with an increased radiation dose to the ipsilateral ear. To examine the
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1395
Table 2. Average threshold difference between each ear pair for each frequency and time point
Table 1. Patient characteristics Characteristic
Value
Total (n) Average age (y) Gender (n) Male Female Average inner ear volume (cm3) Concurrent cisplatin (n) Medical comorbidities (n) Hypertension Diabetes mellitus Inner ear dose for patients receiving unilateral RT (n ⫽ 31) Ipsilateral, treated side (Gy) Median Range Contralateral, untreated side (Gy) Median Range Tumor sites (n) Nasopharynx Oropharynx Oral cavity Salivary gland Paranasal sinus (mostly maxillary sinus) Skin Stage (n) Nonmalignant Stage I/II Stage III Stage IV
35 57.9 18 17 0.56 4 (11) 11 (31) 5 (15)
Time after RT (mo)
Frequency (Hz)
Patients (n)
Average dTL (dB)
SD
Baseline (before RT)
250
29
⫺0.9
7.6
500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000 250 500 1000 2000 4000 8000
30 30 30 30 25 22 22 22 22 22 19 20 20 20 20 20 15 17 17 17 16 16 13 13 13 13 13 12 10
1.3 0.3 4.2 3.3 5.6 ⫺0.2 4.1 3.6 8.2 9.8 12.9 2.3 1.0 0.0 7.0 7.3 15.0 1.5 7.4 2.9 6.9 10.3 15.4 2.7 1.9 0.0 8.1 7.5 17.5
9.0 11.1 10.8 16.2 13.2 6.3 8.8 13.7 11.9 15.4 21.8 10.8 7.4 11.5 12.3 13.4 18.6 8.6 7.5 10.6 9.8 18.8 17.3 9.5 8.3 8.2 11.8 11.4 20.8
1
47.4 14.1–68.8 6 4.2 0.5–31.3 4 1 11 8 9 3 3 7 8 17
12
24
Abbreviation: RT ⫽ radiotherapy. patterns of hearing loss across time and across frequencies, line graphs of hearing thresholds vs. time and hearing thresholds vs. frequency were produced for individual subjects using Splus (Mathsoft, version 6.0, release 1, Sun OS 5.3, Cambridge, MA). Additional line graphs were produced for the overall sample to establish whether the variance in hearing loss changed either over time or across frequencies. The repeated hearing threshold measurements among the individual subjects were modeled as a function of time and frequency by fitting linear mixed models using PROC MIXED (Statistical Analysis Systems proprietary software, release 8.2, SAS Institute, Cary, NC). This accounts for the correlation among repeated measures from individual subjects. Curvilinear relations of threshold with frequencies were examined by including quadratic and cubic terms for frequencies. With the above analysis, a statistical model was created that would allow for the estimation of radiation-induced hearing loss. In addition to radiation dose, other covariates of interest included the difference in hearing threshold at baseline (dTL0), baseline threshold, difference in radiation dose between irradiated and nonirradiated ears, natural logarithm of the test frequency, and time. Adjustments for potential confounding by gender, age, and comorbidities associated with hearing loss (hypertension and diabetes mellitus) were also included in the model. The final results of the model are presented in nomogram form. The model used in this study is presented in the Appendix. A two-tailed p value of ⱕ0.05 was considered statistically significant.
Abbreviations: RT ⫽ radiotherapy; dTL ⫽ difference in threshold level. Positive dTL denotes increased hearing loss in ipsilateral ear compared with contralateral ear.
RESULTS We enrolled 40 patients into this prospective study, and 35 patients participated in the required hearing tests. Of these 35 patients, 4 patients received bilateral RT and con-
Table 3. Incidence of clinically significant hearing loss (ⱖ10 dB) for 8000 Hz and 4000 Hz Time after RT (mo)
Frequency (Hz)
1 6 12 24 36 1 6 12 24 36
8000
4000
Patients (n)
Patients with ⱖ 10 dB loss (n)
19 16 13 10 2 23 20 14 12 3
9 7 6 4 2 5 5 7 6 1
1396
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
Fig. 2. Difference in hearing between ipsilateral and contralateral ears (dTL) vs. differences in mean cochlea dose, over time.
current cisplatin chemotherapy for their primary nasopharynx cancer. We present their data in brief later in this paper. The 31 unilaterally treated, RT-alone patients were used in the primary analysis. Of the 31 patients, 14 were women and 17 were men, with the average age of 58 years (SD, 15.7). The median radiation dose was 47.4 Gy and 4.2 Gy for the ipsilateral and contralateral ear, respectively. Table 2 lists the average threshold difference between the ipsilateral and contralateral ears at each frequency and time point. At each point after RT, a greater difference in the dTL was observed at the higher frequencies (ⱖ2000 Hz). The incidence of clinically significant hearing loss is seen in Table 3. Overall, 6, 6, and 4 patients had a ⱖ10-dB hearing loss at two or more follow-up visits at 8000, 4000, and 2000 Hz, respectively. Longitudinally, for those patients with a
ⱖ10-dB hearing loss noted during the first 6 months, recovery to baseline hearing after 6 months was seen in 3 of 11, 2 of 7, and 0 of 6 patients at 8000, 4000, and 2000 Hz, respectively. The dTL for each test frequency was plotted against the differences in the mean cochlea dose for each time point (Fig. 2). A positive dTL indicated increased hearing loss in the ipsilateral, heavily irradiated ear compared with the contralateral ear. Hearing loss was most predominant and clinically significant (ⱖ10 dB) at the higher frequencies (ⱖ2000 Hz). Throughout this study, 20 of the 31 tested patients had some degree of conductive hearing loss secondary to middle ear effusion at one or more visits, confirmed by an air-bone gap, tympanometry, and otoscopy. This condition was for
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1397
Fig. 2. (Continued).
the most part self-limiting, but when indicated, treatment was provided. The association of the dTL between the ipsilateral and contralateral ears with the radiation dose, age, test frequency, bTL, and threshold difference between two ears at baseline (dTL0) was statistically significant (p ⬍ 0.05), as described below and in the nomograms in Figs. 3– 6. The pattern of hearing loss was not significantly related to time; however, this may have been because of the lack of enough long-term data. In this study, only 3 patients had undergone
Fig. 3. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), baseline threshold level (bTL) (0 dB), and differences in threshold levels (dTL)0 (0 dB), by test frequency.
their 36-month audiologic evaluation, because of poor patient compliance and loss to follow-up. Finally, the potential confounding variables of gender and the presence of hypertension and/or diabetes mellitus did not have any statistically significant effects on hearing loss. To illustrate the effects of each covariate on hearing loss, nomograms based on the final model are presented. These demonstrate the predicted hearing loss for varying radiation dose levels for a constant age, bTL, and dTL0 by test frequency (Fig. 3), for a constant bTL, dTL0, and frequency by age (Fig. 4), for a constant age, dTL0, and frequency by bTL (Fig. 5), and for a constant age, bTL, and frequency by dTL0 (Fig. 6). Radiation dose effect An increase in the mean radiation dose to the inner ear was associated with increased hearing loss (Fig. 3). This effect varied across frequencies, with identical radiation doses causing more hearing loss in the higher than in the lower frequencies. The damage was clinically significant (ⱖ10 dB) at frequencies of ⱖ2000 Hz. For example, using our final model, in a hypothetical 60-year-old patient with normal hearing in both ears at baseline across frequencies, after receiving 45 Gy, the ipsilateral ear would later present with 19.3-dB (95% confidence interval [CI], 15.5–23.0) and 5.4-dB (95% CI 3.5–7.5) more hearing loss compared with the contralateral ear at 8000 Hz and 1000 Hz, respectively. With 60 Gy, the ipsilateral ear would be expected to have 25.1-dB (95% CI, 20.6 –29.7) and 6.6 dB (95% CI, 4.4 – 8.9) more hearing loss than the contralateral ear at 8000 Hz and 1000 Hz, respectively. Within this study, we observed that for almost all cases in which significant hearing loss occurred in the ipsilateral ear receiving a high dose compared with the contralateral ear, the dose received was ⱖ45 Gy.
1398
I. J. Radiation Oncology
● Biology ● Physics
Volume 61, Number 5, 2005
Fig. 4. Nomogram of hearing loss after radiation dose delivered for constant baseline threshold levels (bTL) (0 dB), differences in threshold levels (dTL0) (0 dB), and frequency (2000, 4000, and 8000 Hz), by age.
Fig. 5. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), differences in threshold levels (dTL)0 (0 dB), and frequency (2000, 4000, and 8000 Hz), by baseline hearing threshold level (bTL).
Age effect The age effect on hearing loss is most noticeable at the higher frequencies, especially 4000 Hz and 8000 Hz (Fig.
4). Older patients are more susceptible to hearing loss. For example, for a 70-year-old hypothetical patient with normal hearing in both ears at baseline, after receiving 45 Gy, the
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1399
tralateral ear. For 50-year-old patient, 20 years younger, the ipsilateral ear would only be about 13.6 dB (95% CI, 10.3–16.9) and 10.8 dB (95% CI, 8.6 –13.0) poorer than the contralateral ear at 8000 Hz and 4000 Hz, respectively. Baseline hearing effect Ears with lower (better) thresholds at pre-RT baseline developed more hearing loss (greater thresholds) after RT (Fig. 5). For a 60-year-old hypothetical patient with normal or 0-dB baseline hearing in both ears, after 45-Gy RT, the ipsilateral ear threshold would be 20 dB (8000 Hz) or 5 dB (1000 Hz) poorer than the contralateral ear. In contrast, if this patient had a worse, baseline hearing of 20 dB in both ears, the ipsilateral ear threshold would only be 15 dB (8000 Hz) or ⬍5-dB (1000 Hz) poorer than the contralateral ear. Effect of pre-RT hearing threshold differences between ears The larger the differences between the ipsilateral and contralateral ears at baseline at which the ipsilateral ear presented with elevated (poorer) thresholds, the larger the difference after RT (Fig. 6). For example, if a 60-year-old hypothetical patient had the same hearing in both ears at baseline, after 45-Gy RT, the ipsilateral ear would be 19.3 dB (95% CI, 15.6 –23.0) or 8.6 dB (95% CI, 6.4 –10.7) worse than the contralateral ear for 8000 Hz and 2000 Hz, respectively. However, if the ipsilateral ear was 15 dB worse than the contralateral ear at baseline, after RT, it would be 30.7 dB (95% CI, 26.9 –34.4) and 18.2 dB (95% CI, 15.8 –20.5) worse than the contralateral ear for 8000 Hz and 2000 Hz, respectively. The effect of RT on the ipsilateral ear would be smaller if the contralateral ear had poorer baseline hearing (smaller difference) compared with the ipsilateral ear. Cisplatin effect on hearing The difference in hearing acuity for the 4 patients (eight inner ears receiving ⬎50 Gy) who received cisplatin was not statistically significant from that of the seven inner ears receiving similar doses in patients who had not received cisplatin. However, given the small number of patients, no strong conclusion about the effect of cisplatin and its effects on radiation-induced hearing loss could be drawn from this comparison. DISCUSSION
Fig. 6. Nomogram of hearing loss after radiation dose delivered for constant age (60 years old), baseline threshold level (bTL) (0 dB), and frequency (2000, 4000, and 8000 Hz), by differences in threshold level (dTL)0.
ipsilateral ear threshold would be elevated by 25.0 dB (95% CI, 20.3–29.8) or 15.5 dB (95% CI, 12.3–18.8), respectively, at 8000 Hz and 4000 Hz compared with the con-
Effect of RT on one inner ear—a controlled study Previous studies examining the effect of RT on SNHL mostly focused on patients with nasopharyngeal carcinoma. Although selection of this population ensured a group of patients whose inner ears received a clinically significant dose, the drawback was that typically both inner ears were uniformly treated. Without a “control” ear, the true relationship between the radiation dose and its effect is difficult to tease out from the confounding effects of other risk factors, time, and interventions such as chemotherapy. In our study,
1400
I. J. Radiation Oncology
● Biology ● Physics
we chose primarily to examine patients who received unilateral RT, and thus, each irradiated inner ear was compared with its built-in, contralateral control inner ear. This is unique to our study and adds a measure of confidence in our results lacking from all other published series. Radiation dose–response relation for hearing loss Dose–response data have been scant to date. Grau et al. (12) reported on 22 patients and found significantly greater SNHL for cochlear doses ⬎50 Gy. Kwong et al. (4) prospectively examined nasopharyngeal carcinoma patients treated with RT and cisplatin, but the radiation dose had no statistically significant effect on SNHL. Honoré et al. (13) demonstrated a dose–response curve for SNHL and dose for hearing loss at 4000 Hz. Most recently, Merchant et al. (14) studied 72 children treated with conformal RT and suggested that the average cochlear dose should be kept to ⬍32 Gy to prevent hearing loss. Our observation that clinically apparent hearing loss (ⱖ10 dB) after ⱖ45 Gy is consistent with these series. Although our model suggests a nearly linear relationship, the degree of hearing loss within the model did not reach clinical significance (ⱖ10 dB) until around 40 – 45 Gy, consistent with our observations. Additional data are needed to validate this observation. Our data also suggest that radiation-induced SNHL occurs throughout the entire frequency range of hearing but is only clinically significant (ⱖ10 dB) at higher frequencies, ⱖ2000 Hz. The audiometric frequency range of 2000 – 8000 Hz overlaps with the frequency range of human speech (500 – 4000 Hz). The upper end of the speech frequency range (2000 – 4000 Hz) is especially important for consonant sound recognition. Hearing loss in this frequency range can result in reduced speech recognition abilities. Confounding risk factors and interventions Confounding factors such as preexisting hearing loss, age, hypertension, diabetes mellitus, and treatment with cisplatin were also examined in this study. Because of the design of our trial, the effect of RT alone was isolated, as well as the effects of these confounders. Age and preexisting hearing loss had an effect on radiation-induced SNHL, consistent with the pathophysiology of this deficit (15). Cisplatin ototoxicity is well documented, and its relationship to radiation-induced SNHL has been described by Walker et al. (16). They studied a cohort of children with brain tumors who had received cisplatin chemotherapy and whose radiation portals encompassed the inner ear. The resulting hearing loss was compared with children who had received comparable doses of cisplatin alone, and synergized ototoxicity was found. In our study, we had only 4 patients who received cisplatin chemotherapy, and when analyzed, found no statistically significant hearing loss difference between the inner ears treated to similar doses. We could not, however, draw any strong conclusions from our data, because we examined only 8 inner ears. Middle ear effusion causes conductive hearing loss by interrupting the mechanical transmission of sound to the
Volume 61, Number 5, 2005
sensory apparatus. This confounder was avoided by using bone conduction measurements only, which is a direct measurement of sensory hearing. Another tool used to measure sensory hearing includes the auditory brainstem response, a noninvasive electrophysiologic test that may be used for hearing threshold estimation (17) or for neurodiagnostic purposes (18). It is an excellent tool for the determination of hearing thresholds in infants and young children or in cognitively impaired adults. However, thresholds determined using auditory brainstem responses may differ from actual hearing thresholds measured behaviorally by as much as 5–15 dB (17). Because the auditory brainstem response is generally not recommended as a substitute for behavioral audiometry unless absolutely necessary or for neurodiagnostic purposes, its use was not warranted for our study.
Time course The persistence of SNHL is difficult to assess from the literature, because the data can be biased as some of the patients will develop otitis media in the first few months after RT that will resolve in the first year (19). Another study (5) suggested that SNHL is found between 1.5 and 2 years after RT. In our study, we did not find any time association between hearing loss and RT. Also, on the basis of our data, we could not yet determine whether SNHL improved after 2 years, because our follow-up was limited after 2 years.
Radiobiologic correlations Animal studies have demonstrated SNHL with histopathologic correlation to damage to the organ of Corti and endolymph beginning at 40 Gy (20, 21). After conventionally fractionated RT, a relationship was established among the radiation dose received, cellular loss in the organ of Corti, and hearing loss. Additionally, the vast amount of “radiosurgery” literature has demonstrated the vulnerability of additional hearing apparatus structures, such as the acoustic nerve, to radiation. Although this was not specifically examined in our study, radiation effects on other portions of the auditory pathway will be examined in the future.
Future hearing-loss models We have presented a fitted model based on the population in our study. Although this model accurately represents our data, additional data will allow us to refine our model to predict hearing loss more accurately. Current work is in progress to apply normal tissue complication probability modeling techniques to the dose–volume data obtained in this study and to apply this model to patients receiving intensity-modulated RT in which low total doses received by the inner ear are also associated with a lower fraction dose compared with the target fraction dose.
Sensorineural hearing loss in irradiated patients
● C. C. PAN et al.
1401
Table 4. Fixed effect parameter estimates on final model Fixed effect
Estimate
Standard error
df
p
95% CI
Fixed effect intercept Baseline contralateral hearing threshold Radiation dose* Baseline difference between ear pair ⫻ Frequency† Radiation dose ⫻ age Radiation dose ⫻ frequency Radiation dose ⫻ (frequency) (23) Radiation dose ⫻ age ⫻ frequency Radiation dose ⫻ age ⫻ (frequency) (23)
2.4 ⫺0.2 ⫺5.8 0.09 0.11 1.7 ⫺0.13 ⫺0.04 0.003
2.7 0.04 1.9 0.006 0.03 0.5 0.04 0.009 0.0006
153 226 143 137 138 144 145 139 141
0.4 ⬍0.0001 0.004 ⬍0.0001 0.0007 0.002 0.001 0.0003 ⬍0.0001
⫺2.9, 7.6 ⫺0.3, ⫺0.1 ⫺9.6, ⫺1.9 0.07, 0.10 0.05, 0.18 0.6, 2.8 ⫺0.2, ⫺0.05 ⫺0.05, ⫺0.02 0.001, 0.004
Abbreviation: CI ⫽ confidence interval. * Radiation dose represents radiation dose received in ipsilateral ear. † Frequency was logarithm transformed.
CONCLUSION These results, including the dose effect and the effect of other factors, should allow clinicians to better weight any
necessary tradeoffs among sparing the inner ears, sparing other neighboring critical structures, and RT of the adjoining targets.
REFERENCES 1. Jemal A, Murray T, Samuels A, et al. Cancer statistics, 2003. CA Cancer J Clin 2003;53:5–26. 2. Cacciatore F, Napoli C, Abete P, et al. for the Osservatorio Geriatrico Campano Study Group. Quality of life determinants and hearing function in an elderly population. Gerontology 1999;45:323–328. 3. Anteunis LJ, Wanders SL, Hendriks JJ, et al. A prospective longitudinal study on radiation-induced hearing loss. Am J Surg 1994;168:408 – 411. 4. Kwong DL, Wei WI, Sham JS, et al. Sensorineural hearing loss in patients treated for nasopharyngeal carcinoma: A prospective study of the effect of radiation and cisplatin treatment. Int J Radiat Oncol Biol Phys 1996;36:281–289. 5. Ho WK, Wei WI, Kwong DL, et al. Long-term sensorineural hearing deficit following radiotherapy in patients suffering from nasopharyngeal carcinoma: A prospective study. Head Neck 1999;21:547–553. 6. Schot LJ, Hilgers FJ, Keus RB, et al. Late effects of radiotherapy on hearing. Eur Arch Otorhinolaryngol 1992;249: 305–308. 7. Chen WC, Liao CT, Tsai HC, et al. Radiation-induced hearing impairment in patients treated for malignant parotid tumor. Ann Otol Rhinol Laryngol 1999;108:1159 –1164. 8. Fraass BA, McShan DL. 3-D treatment planning: III. Complete Beam’s-Eye-View planning capabilities. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 193–196 9. Fraass BA, McShan DL, Ten Haken RK, et al. 3-D treatment planning: V. A Fast 3-D photon calculation model. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 521–525 10. McShan DL, Fraass BA. 3-D treatment planning: II. Integration of gray scale images and solid surface graphics. In: Bruinvis IAD, editor. The use of computers in radiation therapy. Holland: Elsevier Science; 1987. p. 41– 44 11. Raaijmakers E, Engelen AM. Is sensorineural hearing loss a possible side effect of nasopharyngeal and parotid irradiation? A systematic review of the literature. Radiother Oncol 2002; 65:1–7. 12. Grau C, Moller K, Overgaard M, et al. Sensori-neural hearing
13.
14.
15. 16.
17.
18.
19.
20.
21.
22.
23.
24.
loss in patients treated with irradiation for nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 1991;21:723–728. Honoré HB, Bentzen SM, Moller K, et al. Sensori-neural hearing loss after radiotherapy for nasopharyngeal carcinoma: individualized risk estimation. Radiother Oncol 2002;65:9 –16. Merchant TE, Gould CJ, Xiong X, et al. Early neuro-otologic effects of three-dimensional irradiation in children with primary brain tumors. Int J Radiat Oncol Biol Phys 2004;58: 1194 –1207. Pickles JO. An introduction to the physiology of hearing. (2nd ed). London: Academic Press; 1999. Walker DA, Pillow J, Waters KD, et al. Enhanced cis-platinum ototoxicity in children with brain tumours who have received simultaneous or prior cranial irradiation. Med Pediatr Oncol 1989;17:48 –52. Kileny PR, Magathan MG. Predictive value of ABR in infants and children with moderate to profound hearing impairment. Ear Hear 1987;8:217–221. El-Kashlan HK, Eisenmann D, Kileny PR. Auditory brain stem response in small acoustic neuromas. Ear Hear 2000;21: 257–262. Tang NL, Choy AT, John DG, et al. The otological status of patients with nasopharyngeal carcinoma after megavoltage radiotherapy. J Laryngol Otol 1992;106:1055–1058. Johannesen TB, Rasmussen K, Winther FO, et al. Late radiation effects on hearing, vestibular function, and taste in brain tumor patients. Int J Radiat Oncol Biol Phys 2002;53:86 –90. Winther FO. X-ray irradiation of the inner ear of the guinea pig: Early degenerative changes in the vestibular sensory epithelia. Acta Otolaryngol 1969;68:514 –525. Verbeke G, Molenberghs G. Linear mixed models in longitudinal data. In: Verbeke G, Molenberghs G, editors. Inference for the random effects. New York: Springer-Verlag; 2000. Brant LJ, Fozard JL. Age changes in pure-tone hearing thresholds in a longitudinal study of normal human aging. J Acoust Soc Am 1990;88:813– 820. Brant LJ, Pearson JD. Modeling the variability in longitudinal patterns of aging. In: Crews DE, Garruto RM, editors. Biological anthropology and aging: Perspectives on human vari-
1402
I. J. Radiation Oncology
● Biology ● Physics
ation over the life span. New York: Oxford University Press; 1994. p. 373–393
Volume 61, Number 5, 2005
25. Morrell CH, Brant LJ. Modelling hearing thresholds in the elderly. Stat Med 1991;10:1453–1464.
APPENDIX A linear mixed model with random intercept (22, 23) was used to investigate the relationship between hearing threshold and clinical variables of interest. Specifically, the model is as follows: yijK ⫽ ⫹ XijK ⫹ aij ⫹ ijK where i is subject 1 through 31; j is the frequency 250, 500, 1000, 2000, 4000, 8000 Hz; k is the time point 1 through 5 (1, 6, 12, 24, and 36 months); aij is N(0, 12); ⑀ijk ⬃ N(0, 22); yijk is the column vector of hearing threshold difference between the ipsilateral and contralateral ears at time k, frequency j, for subject i; is the population average; and Xijk is the row vector that contains the subject level variables, which included the following main effects: age, gender, hypertension, diabetes mellitus, time point, radiation dose difference between ear pair, radiation dose of ipsilateral/contralateral ear, baseline hearing difference between ear pair, baseline hearing of ipsilateral/contralateral ear, frequency (logarithm transformed), and quadratic term of logarithm transformed frequency (23). Two-way interaction terms included age by baseline hearing difference between ear pair, age by time, age by radiation dose difference between ear pair, age by frequency, age by quadratic frequency (23–25), time by baseline hearing difference between ear pair, time by radiation dose difference between ear pair, time by frequency, time by quadratic frequency, baseline hearing difference between ear pair by frequency, baseline hearing difference between ear pair by radiation dose difference between ear pair, frequency by radiation dose between ear pair, baseline hearing difference between ear pair by quadratic frequency,
radiation dose difference between ear pair by quadratic frequency, time by ipsilateral ear radiation dose, age by ipsilateral ear radiation dose, baseline hearing difference between ear pair by ipsilateral ear radiation dose, frequency by ipsilateral ear radiation dose, and quadratic frequency by ipsilateral ear radiation dose. Three-way interaction terms included age by ipsilateral ear radiation dose by frequency, age by ipsilateral ear radiation dose by quadratic frequency, ipsilateral ear radiation dose by time by frequency, ipsilateral ear radiation dose by time by quadratic frequency, ipsilateral ear radiation dose by baseline hearing difference between ear pair by frequency, and ipsilateral ear radiation dose by baseline hearing difference between ear pair by quadratic frequency. aij are independent (subject by frequency) level random effects that account for variations between frequencies due to external, unmeasured factors. ⑀ijk are independent, normally distributed, between-frequency and between-time point differences. The final model was chosen through backward deletion, Akaike Information Criterion, and Schwarz’s Bayesian Information Criterion model selection criteria. The preliminary analysis included the main effects, and the time point was modeled as either a linear variable or five binary factor variables. Two-way and three-way interaction terms were added based on data from Brant and Fozard (23), Brant and Pearson (24), and Morrell and Brant (25) and additional interactions of interest. Among the three equally fitted models, the final one presented here (Table 4) was chosen on the basis of the literature and the relevance of the clinical variables remaining in the model.