Presented in part at the 47th Annual Meeting of the American Society for Therapeutic Radiology and Oncology (ASTRO), October 16, 2005, Denver, Colorado.
Combined modality therapy has become the standard of care for nasopharyngeal carcinoma, yet the combined ototoxic effects of radiation and cisplatin are poorly understood. The incidence and severity of sensorineural hearing loss (SNHL) with combined modality therapy was evaluated and the dose–response relation between radiation and hearing loss was investigated.
Patients with newly diagnosed AJCC Stage II–IVB nasopharynx carcinoma treated from 1994–2003 were identified. The records of 44 ears in 22 patients who received a preirradiation pure tone audiogram and followup audiograms 12+ months postirradiation were included in the analysis. All patients were treated with conformal radiotherapy to 70 Gy and received platinum-based chemotherapy similar to the Intergroup 0099 trial. Composite cochlear dose distributions were calculated. Ototoxicity was measured using intrasubject audiogram comparisons and SNHL was defined as per the American Speech and Hearing Association guidelines, with standard range of speech between 2000–4000 Hz. SNHL was analyzed using Fisher exact test and linear and logistic regression models.
Patient characteristics: median age, 45; 27% Asian; 68% male; 64% WHO III. Median audiologic followup was 29 months (range, 12–76 mos). Mean cochlear dose (Dmean) ranged from 28.4–70.0 Gy (median, 48.5 Gy). SNHL was detected in 25 of the 44 ears (57%) studied. There was an increased risk of SNHL for ears receiving Dmean > 48 Gy compared with those receiving ≤ 48 Gy at all frequencies within the range of speech (P = 0.04). Using univariate logistic regression analysis, Dmean to the cochlea, cycles of cisplatin, and time postradiotherapy were independently significant factors in determining the incidence of SNHL (P = 0.02, P = 0.03, and P = 0.04, respectively). In univariate and multivariate linear regression analysis, Dmean was statistically significant at all frequencies in affecting degree of SNHL, whereas the significance of cisplatin and time was variable.
Hearing loss is one of the major complications of therapy for patients with nasopharynx carcinoma. Recently, the addition of cisplatin-based chemotherapy to radiotherapy has shown improved overall survival and, as a result, combined modality treatment has become the standard of care.1 However, both cisplatin and high-dose radiation are known to be ototoxic. Historically, at least one-third of nasopharyngeal carcinoma patients developed significant sensorineural hearing loss after treatment.2–9 Those rates are expected to increase with more widespread use of combined modality therapy.
With the introduction and popularity of intensity-modulated radiotherapy (IMRT) for head and neck cancer, greater effort has been placed on reducing long-term side effects. We are now in a situation where our knowledge lags behind our technical capabilities. We have the power to control the dose to the cochlea, but we have no idea what radiation dose is ‘safe’ for the cochlea, especially when patients are also receiving cisplatin.
In this study, we aimed to evaluate the incidence and severity of hearing loss in the setting of combined modality therapy for nasopharynx carcinoma and determine if a threshold dose exists for the cochlea. Determining the relation between radiation dose plus cisplatin exposure and cochlear function will provide radiation oncologists with guidelines to help plan the increasingly sophisticated treatment for nasopharynx and other head and neck cancers.
MATERIALS AND METHODS
After obtaining permission from our institutional review board (IRB), the Memorial Sloan-Kettering Cancer Center tumor registry and Department of Radiation Oncology database were used to retrospectively identify all patients presenting with a diagnosis of nasopharyngeal carcinoma between 1994 and November 2004. Retrospective review of patient records was performed within IRB guidelines. Patient and tumor characteristics were recorded, along with treatment variables for each patient. Patients who had no baseline preirradiation audiogram and/or no 12+ month postirradiation audiogram were excluded. Patients with locally recurrent disease and/or subsequent reirradiation were also excluded. Each ear was treated as an individual subject. The records of 44 ears in 22 patients serve as the subjects of this analysis.
Pretreatment evaluation included a complete history and physical examination with fiberoptic nasopharyngoscopy, chest radiograph, and computed tomographic (CT) scan of the nasopharynx, skull base, and neck. Twelve (55%) patients also had a pretreatment magnetic resonance imaging (MRI) scan of the nasopharynx. Laboratory studies included complete blood cell counts, blood chemistries, and urinalysis. Patients receiving chemotherapy also had a baseline calculated or collected creatinine clearance test, electrocardiogram, and audiogram. CT scans of the chest and abdomen, bone scans, and positron emission tomography (PET) scans were performed as clinically indicated on the basis of abnormal screening test results or symptoms. All patients had a biopsy to confirm the diagnosis with review at Memorial Sloan-Kettering Cancer Center. All patients also had a dental evaluation before radiotherapy.
Patients were immobilized in the supine position with custom Aquaplast masks and target localization was accomplished using CT simulation. CT images indexed every 3 mm were obtained, extending from the vertex to 5 cm inferior to the clavicular heads. The target volumes and normal tissue structures were defined using CT images, supplemented with fused diagnostic MRI and/or PET scans. Sixteen (73%) patients had volumes defined on fused plans.
The volumes of interest were identified on each axial CT slice. The gross target volume (GTV) consisted of the gross primary tumor and involved lymph nodes as defined by MRI (CT when MRI not available). CTVg (‘g’ for gross) was defined by adding a 5-mm margin to GTV to include biological uncertainties. This margin was applied in all dimensions except posterior to the primary tumor, as there is less biological uncertainty at the bones of the skull base and a smaller margin allowed lower dose delivery to critical structures, including the brainstem and cochlea. PTVg was defined by adding a 5-mm margin to CTVg in all dimensions to include technical uncertainties. CTVm (‘m’ for microscopic) consisted of CTVg plus the area encompassing the entire nasopharynx and all cervical lymph nodes bilaterally, whereas PTVm was defined by adding a 5-mm margin to CTVm for technical uncertainties. Normal structures of interest outlined in three dimensions included the globes, optic nerves, optic chiasm, brainstem, spinal cord, cochleae, parotid glands, submandibular glands, oral cavity, and larynx. During the course of this study, several different radiotherapy techniques, doses, and fractionation schedules were employed, as there was an evolution in radiotherapy techniques for nasopharyngeal carcinoma at our institution. All radiotherapy was delivered with 6 MV photons.
Seven patients were treated with opposed lateral fields with a 3D conformal or IMRT boost with, if necessary, electron boosts to the neck. Fifteen patients were treated with IMRT for the entire course of treatment beginning in 1999. The total prescribed dose to the primary tumor was 70 Gy for all patients.
Of the 15 patients treated with IMRT for the entire course of treatment, 11 were treated with a hyperfractionated concomitant boost. This involved using in the initial 20 fractions a 1.8 Gy/fraction IMRT plan that encompassed all sites of gross and presumed microscopic disease, followed by 10 days of twice-daily treatment, using each day the initial IMRT plan and a second, separate 1.6 Gy/fraction IMRT plan to boost sites of gross tumor involvement.
More recently, four patients were treated on an institutional protocol using the dose-painting capability of IMRT that delivered a simultaneous, field-within-a-field boost to treat regions of gross tumor involvement in 30 once-daily fractions. With each fraction, the PTVm (primary tumor and bilateral retropharyngeal and cervical lymph node chains) received 1.8 Gy for a total of 54 Gy, and the PTVg (including gross tumor involvement in the nasopharynx and lymph nodes) at the same time received 2.34 Gy for a total of 70.2 Gy.
As our radiotherapy techniques evolved over time, the dosimetric criteria for target coverage evolved as well, leading to nonuniform criteria for the patients in this study. In an early evaluation of our IMRT technique for nasopharynx carcinoma, Hunt et al.10 performed a retrospective comparison of treatment techniques. For patients treated with opposed lateral fields and supplementary electrons, the mean, minimum, and maximum PTVg doses were 67.9, 54.6, and 74.2 Gy, respectively. Comparable mean, minimum, and maximum PTVg doses for the 3D and early IMRT patients treated with hyperfractionated concomitant boost were 74.6, 65.7, and 80.2 Gy (3D) and 77.3, 69.4, and 81.8 Gy (IMRT). We believe these results are representative of the target doses received by patients in this study treated with these techniques. Doses for patients treated on our institutional protocol with IMRT dose-painting conformed to the protocol requirements that specify that the PTVg D95 (dose received by 95% of the volume) must be at least 70 Gy and the maximum dose must not exceed 84 Gy.
Dose Calculation of the Inner Ear
The cochlea contours were defined for every patient by a single physician (S.L.W.) to ensure accurate and consistent radiographic localization (Fig. 1). The cochlear volumes were entered before retrieval of any dosimetry information and without knowledge of audiogram data in order to prevent bias. Where necessary, each patient's individual treatment was reconstructed from the archive using patient charts and the 3D treatment planning system. At our institution, graphical plans were not routinely created when treating with standard lateral fields. Therefore, in cases where these fields were used for treatment the blocking pattern of the lateral, low anterior neck, and electron fields were reconstructed by scanning the original simulation films for these beams into the beam's-eye view module of the system. Planning data for 3D and IMRT fields were retrieved from archive and reviewed for consistency with the treatment record. Composite dose distributions were then calculated including all phases of treatment using the Memorial Sloan-Kettering Cancer Center treatment planning system11–13 with a calculation resolution of 5000 quasirandom points within each cochlea. Doses were calculated using a pencil beam algorithm, with ray tracing to account for tissue inhomogeneity. The minimum dose (Dmin), maximum dose (Dmax), mean dose (Dmean), minimum dose to the hottest 5% of the cochlea (D05), and dose to 95% of the volume (D95) were examined.
Before 1998, patients with AJCC Stage II–IV disease received two planned concurrent cisplatin chemotherapy 100 mg/m2 (single dose or split over 2 days) intravenously on weeks 1 and 4 of radiotherapy.
After publication of the Phase III randomized Intergroup study 0099,1 patients were planned for five total cycles of chemotherapy consisting of concurrent cisplatin chemotherapy 100 mg/m2 (single dose or split over 2 days) intravenously on weeks 1 and 4 of radiotherapy, plus up to three planned cycles of cisplatin 100 mg/m2 intravenously and 5-fluorouracil 1000 mg/m2/day by continuous intravenous infusion on Days 1–4 after radiotherapy.1, 2 For patients treated with the dose-painting technique, a third cycle of concurrent cisplatin was considered during Week 7, depending on patient toxicity/tolerance. All patients included for analysis received at least one cycle of cisplatin. In patients for whom continued cisplatin-based chemotherapy was contraindicated because of hearing loss or kidney dysfunction, a carboplatin-based program was substituted. Twelve (55%) patients received at least one cycle of carboplatin.
Pure-tone audiograms (PTA) were obtained before treatment to assess individual baseline hearing thresholds. The hearing assessment included audiologic history, air and bone conduction threshold, speech audiometry, and impedance audiometry. Testing was performed in a double-walled audiometric testing suite (IAC 1400 series). Masking was used, if indicated. Bone and air conduction thresholds (250–4000 Hz and 250–8000 Hz, respectively) were obtained using a Grason-Stadler 1710, Beltone 2000, or Grason-Stadler 1761 clinical audiometer with TDH-50 earphones and MX-51 earcushions.
High-frequency air conduction thresholds are generally used to assess early signs of ototoxicity. However, because of serous otitis caused by tumor obstructing the Eustachian orifice in many nasopharynx carcinoma patients, a considerable percentage of audiograms in this study demonstrated a mixed or conductive hearing loss, defined by an air–bone gap of ≥ 15 dB and/or flat tympanograms. With mixed or conductive hearing loss at baseline and/or at followup, air conduction thresholds could not be used to study early cochlear sensorineural damage and we thus relied on bone conduction thresholds less affected by outer or middle ear abnormalities, measured at 500–4000 Hz. Frequencies in the standard range of speech are between 2000–4000 Hz. Incidence of serous otitis and tympanic membrane perforation were documented at baseline and at followup.
Posttreatment audiograms were obtained at various intervals after completion of radiotherapy. The most recently performed audiograms were used for analysis. Patients with audiograms less than 12 months after completion of radiotherapy were excluded.
Ototoxicity was measured using intrasubject audiogram comparisons. Each patient served as his/her control with the pretreatment audiogram serving as the baseline. Hearing threshold change was determined relative to each patient's baseline. As per the American Speech and Hearing Association guidelines, significant SNHL was defined as a ≥ 20 dB increase in bone conduction threshold at one frequency or a ≥ 10 dB increase at two consecutive frequencies.
Statistical Analysis of Ototoxicity
Doses were correlated with incidence of hearing loss using logistic regression to determine whether a cochlear tolerance dose exists. Factors thought to influence incidence and severity of hearing loss—including cochlear radiation dose, number of cisplatin cycles, time of audiogram postradiotherapy, pretreatment hearing level, and patient age—were assessed for significance using univariate and multivariate linear and logistic regression analyses with both continuous and binary endpoints. Continuous endpoints for the study included both absolute dB change and dB change as a percentage of greatest potential individual change given baseline hearing levels. The presence or absence of SNHL as defined by the American Speech and Hearing Association was evaluated as a binary endpoint. Models were assessed in both step-up regression, where a factor subset selection method adds factors one at a time starting with no factors, and step-down regression, which starts with all factors and removes them one at a time. Fisher exact test was used to assess risk of hearing loss and cochlear radiation thresholds. A P-value of ≤ 0.05 was considered significant.
Characteristics of the 22 patients including gender, ethnicity, tumor histology according to the World Health Organization (WHO) classification,14 and 1997 AJCC stage distribution15 are outlined in Table 1. Median age at treatment was 45 years and ranged from 14–71 years. All patients received the prescription dose of 70 Gy to macroscopic disease. A summary of treatment characteristics is shown in Table 2.
The Dmean delivered to the cochlea ranged from 28.4–70.0 Gy (median, 48.5 Gy). In general, patients treated with IMRT received a lower dose to the cochlea than those treated with opposed lateral fields with standard 3D boost (median, 47.7 Gy vs. 51.1 Gy, respectively). Dosimetric data is shown in Table 3. The median time for audiologic followup after radiotherapy was 29 months (mean, 30 mos; range 12–76 mos). Hearing threshold change ranged from –10 dB to +80 dB. The mean change in speech reception threshold was 7 dB (standard deviation [SD] 17 dB). The mean change in word recognition was –5.8% (SD 11.7%). Incidence of serous otitis and tympanic membrane perforation at baseline and at followup is shown in Table 4. Serous otitis and tympanic membrane perforation were not independently predictive of changes in bone conduction. Composite audiograms of baseline and posttreatment audiometric data are provided in Figure 2. SNHL was detected in 25 of the 44 (57%) ears studied (Fig. 3). There were no cases of posttreatment osteoradionecrosis of the tympanic bone.
Table 3. Radiation Doses to Cochlea
D05: minimum radiation dose to the hottest 5% of the cochlea; Dmean: mean radiation dose to the cochlea; D95: radiation dose to 95% of the volume.
Table 4. Incidence of Serous Otitis and Tympanic Membrane Perforation
No. of patients
6 of 7 pretreatment tympanic membrane perforations were for ventilation tube.
Cochlear dosimetric data (Dmax, Dmin, Dmean, D95, and D05) was tested for correlation with intrasubject bone conduction threshold changes using univariate linear regression analysis with positive results. All subsequent analyses focused on Dmean, given the small anatomic size of the cochlea.
Using the Fisher exact test, a Dmean threshold of 48 Gy was found to be statistically significant in predicting SNHL at all frequencies in the range of speech. At the upper range of speech where SNHL is first perceptible (4000 Hz), the probability of developing SNHL was 61% for patients receiving > 48 Gy compared with 24% for patients receiving ≤ 48 Gy (P = 0.02). Similarly, the risk of SNHL was 57% versus 24% (P = 0.04) at 3000 Hz and 43% versus 10% (P = 0.02) at 2000 Hz (Fig. 4).
Bone conduction threshold shifts were larger at the higher frequencies, reflecting more severe SNHL. Speech reception thresholds (SRTs) at baseline were predictive of baseline hearing thresholds at 2000–4000 Hz (P = 0.02). Pure tone averages were predictive of SRT (P < 0.001) but were not predictive of word recognition.
Using univariate logistic regression analysis of absolute dB change, Dmean to the cochlea, cisplatin dose, and time of audiogram postradiotherapy were independently significant factors in determining the incidence of SNHL (P = 0.02, P = 0.03, and P = 0.04, respectively). In univariate and multivariate linear regression analysis, Dmean was statistically significant at all frequencies in affecting degree of SNHL, whereas the significance of cisplatin and time was variable. At 4000 Hz, a multivariate linear model to explain degree of SNHL could be made using Dmean, cisplatin, and time (P = 0.01, P = 0.02, and P = 0.001, respectively). At 3000 and 2000 Hz, Dmean and time were significant in multivariate and univariate analyses, whereas cisplatin was not shown to have a statistically significant contributing effect on the change in hearing thresholds. At 1000 Hz, Dmean was significant in univariate linear analysis (P = 0.02), and no multivariate analysis could significantly improve the Dmean correlation to SHNL.
Expressing dB change as a percentage of potential change for univariate and multivariate linear regression analysis, Dmean was again statistically significant at all examined frequencies in affecting degree of SNHL (univariate P = 0.03, multivariate P = 0.03), whereas the significance of cisplatin and time was again variable.
In predicting SNHL, there was also an extremely powerful association between the two ears (P < 0.001). In other words, hearing loss in one ear was strongly predicted by that in the other ear, even when taking into account additional factors such as radiation dose, cisplatin dose, time postradiotherapy, and patient age.
Other suspected contributing factors of SNHL such as age at treatment and baseline hearing threshold were not statistically significant at any of the speech frequencies.
As combined modality therapy becomes the standard of care in the treatment of many head and neck cancers, there are increasing concerns of ototoxicity, as cisplatin and radiation are known to independently contribute to otologic sequelae.
Irreversible hearing loss as a consequence of cisplatin administration has been closely investigated, with numerous causative factors identified, including: the cumulative dose of cisplatin received, the mode of drug administration, the concurrent use of aminoglycosides and diuretics, and renal dysfunction.
Whereas the precise mechanism of radiation-induced hearing loss is obscure, studies of irradiated animal cochleae have also shown ultrastructural changes in the outer hair cells (OHCs) and stria vascularis of the basal turn.16
Whereas many have focused on the individual toxicities of radiation and cisplatin on the auditory apparatus, few clinical studies have evaluated the synergistic ototoxic effects of radiation and cisplatin chemotherapy. Some have demonstrated that combined modality protocols with cochleotoxic compounds such as cisplatin may increase the detrimental effects of radiation on the inner ear,7, 17–22 with SNHL both immediate and delayed.23, 24 In examining the results, it is thus important to differentiate between radiation-induced and cisplatin-induced ototoxicity.
Typically, cisplatin ototoxicity occurs acutely, with effects seen as early as 3–4 days after administration. Audiologic threshold shifts are greatest initially in the high frequencies, with threshold changes extending to lower frequencies as the duration of cisplatin administration lengthens. The audiometric abnormalities are typically bilateral, irreversible, and progressive. As the cumulative dose of cisplatin reaches 200 mg/m2, significant hearing loss begins to occur in the 6000–8000 Hz range and the 2000–4000 Hz range also appears susceptible to changes.25–27 In our series, all but one patient received a cumulative cisplatin dose of at least 200 mg/m2.
Radiation-induced ototoxicity, meanwhile, is typically evident 6–12 months after completion of radiotherapy.3, 4, 28 Predisposing factors include older age and coexisting otitis media.4, 29, 30 Bohne et al.31 showed nerve fiber degeneration as well as sensory and supporting cell loss 2 years after radiation exposure to the inner ear of chinchillas. Grau et al.29 and Chen et al.5 suggested cochlea tolerance doses in human patients of 50 and 60 Gy, respectively, showing significant increase of severe SNHL in patients receiving greater than the reported thresholds. Recently, Honore et al.32 and Pan et al.33 showed increased risk of SNHL with increased patient age and decreased pretherapeutic hearing level as well. Conversely, Liberman et al.34 concluded that SNHL was not associated with total radiation dose or with dose to the ear region; the study was, however, limited by size (11 patients) and followup (audiograms range, 5–10 mos). It is important to note that these observations were made in patients receiving radiation only and no chemotherapy.
In Pan et al.,33 31 patients with various head-and-neck tumors were treated with radiotherapy alone, with ipsilateral mean cochlear doses ranging from 14.1–68.8 Gy and contralateral doses ranging from 0.4–31.3 Gy. An increase in the mean radiation dose to the inner ear was associated with increased hearing loss.
The findings that radiation dose may have significant long-term audiologic impact is particularly relevant in nasopharyngeal carcinoma; Ondrey et al.35 demonstrated that patients with cancers arising in or involving the nasopharynx were at greatest risk for receiving high radiation doses to otologic structures, and the cochleae always received nearly full tumor doses of radiation. Treatment plans based on lateral opposed fields or 3D conformal therapy provided no cochlear sparing. IMRT not only allows superior dose distribution, but also enables the delivery of high fractional doses to the tumor, while delivering a more conformal radiation dose to reduce the dose exposure to surrounding normal structures.10, 36, 37
Our findings that SNHL is more severe at higher frequencies corroborates past reports of pattern of loss. The most pronounced threshold changes occurred at the highest bone-conduction frequency evaluated, 4000 Hz, and once changes are noticed at this higher frequency, a patient is at risk for subsequent loss in the remaining speech frequencies.
Analysis suggests that limiting cochlear Dmean to 48 Gy will provide significant protection from SNHL, including the higher 4000 Hz frequency. Among the subset of patients treated using IMRT, the median cochlear Dmean was 47.7 Gy and as low as 28.4 Gy, suggesting that such constraints are feasible for most patients.
Based on multivariate linear correlation models of SNHL, we see a statistically significant contribution from Dmean but not cisplatin, suggesting that at 12+ months posttreatment the radiation dose may supersede cisplatin in affecting long-term sensorineural sequelae. It is important to recognize, however, that chemotherapy regimens were altered based on the results of routine followup audiograms, substituting carboplatin for cisplatin when SNHL was detected. This alone may explain the lack of statistical significance noted in our analyses involving cisplatin dose, particularly in the lower frequencies. That 4000 Hz was statistically significant may reflect the early SNHL that prompted the chemotherapy regimen change.
The strong association of SNHL between opposite ears of individual patients suggests an intrinsic patient sensitivity that would result in a propensity for bilateral loss consistent with earlier reports. Additionally, whereas radiation dose to bilateral cochlea may not be uniform, some treatment variables may affect both ears equally, including time and chemotherapy regimen.
Given the retrospective nature of this study, we recognize the limitations of its findings. There are several continuous variables that could differentially impact a patient's risk for developing hearing loss. One such variable is varying time, as the duration of time posttreatment that an audiogram was obtained proved to be a significant factor in predicting SNHL. However, Dmean was itself significant, even in analyses factoring for time.
Patient characteristics may also be a confounding factor. It has been shown by Kwong et al.4 that advanced age at radiotherapy treatment increases the risk of SNHL. From this, one may conclude that a series of low mean age may yield a lower incidence of SNHL than a series with a high mean age. Our cohort of patients ranged from 14–72, but age was not shown to be a significant factor in our analyses.
Another potential confounding factor may be individual baseline hearing thresholds. If a patient presents initially with severe baseline audiologic deficit, there may be little room for change after completion of treatment, or conversely the patient may be more susceptible to hearing loss. To assess its effect, we evaluated baseline threshold levels as a component of our multivariate analysis and found no statistically significant contribution. Additionally, hearing loss was assessed based on a percentage of greatest potential change and Dmean was again found to be statistically significant in all univariate and multivariate linear regression analyses.
There was a significant increase in risk of SNHL among patients receiving > 48 Gy, suggesting a threshold in cochlear radiation dose–response in the setting of combined modality therapy. This dose should serve as a Dmean constraint maximum for IMRT treatment of nasopharynx carcinoma.
With a significant percentage of patients developing hearing loss secondary to definitive therapy, it is important to recognize that modulating the radiation dose exposure can potentially mitigate the morbidity associated with therapy.