This cohort study uses data from a cancer treatment facility and clinical audiometric database to assess the validity of a previously created dynamic nomogram and model for hearing loss to be used after treatment with chemoradiation for head and neck cancer.
Key Points
Question
Is a previously created model and dynamic nomogram for hearing loss after treatment with chemoradiation in head and neck cancer externally valid?
Findings
This cohort study of 105 patients (208 ears) found that a previously created model and nomogram for treatment-induced hearing loss in patients with head and neck cancer showed a high level of accuracy. Predicted and observed 1-, 2-, and 4-kHz pure-tone averages were not significantly different, and a predicted pure-tone average greater than 35 dB with an area under the curve of 0.71 suggested good calibration and discrimination, respectively.
Meaning
These findings suggest that this model can give an estimate of hearing outcomes after radiation and chemotherapy, providing a useful counseling tool for clinicians to inform patients’ expectations and plan for necessary follow-up.
Abstract
Importance
Hearing loss affects up to 88% of patients undergoing head and neck cancer treatment; however, there are few validated models to predict this outcome. A predictive posttreatment model for hearing loss will allow clinicians and patients to make well-informed decisions about treatment with cisplatin-based chemotherapies and radiotherapy.
Objective
To validate a previously created predictive model for objective hearing outcomes and to assess barriers to using the prediction nomogram in general practice for patients newly diagnosed with head and neck cancer.
Design, Setting, and Participants
This cohort study includes an evaluation of 105 patients (208 ears) and interviews with 6 clinicians. The patients were treated at a high-volume tertiary care hospital. Patient participants were newly diagnosed with head and neck cancer and treated at Siteman Cancer Center from July 1, 2018, to December 31, 2019, with radiotherapy both with and without cisplatin-based chemotherapy. Additionally, the clinicians involved in the care of patients with head and neck cancer were interviewed to assess implementation strategies.
Exposures
Radiotherapy with and without cisplatin-based chemotherapy.
Main Outcomes and Measures
Hearing defined by the audiometric pure-tone average of 1, 2, and 4 kHz.
Results
A total of 105 patients (208 ears; mean [SD] age, 61 [11] years; 82 men [78%]) were compared with the development cohort to assess the similarities and differences in case mix. All patients underwent radiation therapy, 50 (48%) received cisplatin-based chemotherapy, and 67 (64%) had a surgical resection. The mean (SD) cochlear dose of radiation was 13 (12) Gy, and the mean (SD) total cisplatin dose was 238 (83) mg/m2 for those undergoing cisplatin therapy. A calibration curve demonstrated that predicted and observed posttreatment pure-tone average were not significantly different. The model predicted a posttreatment pure-tone average greater than 35 dB (a common threshold for hearing aid consideration) with a sensitivity of 73% and specificity of 67% with an area under the curve of 0.71, showing good discrimination. Clinician interviews suggest the nomogram requires careful integration into patient counseling to clarify risks and benefits for treatment.
Conclusions and Relevance
The findings of this cohort study confirm this model’s ability to predict posttreatment hearing outcomes in a unique population of patients. This model has the potential to inform pretreatment counseling and posttreatment hearing evaluations for this patient population.
Introduction
In 2020, more than 65 000 Americans will develop new cancers of the oral cavity, pharynx, or larynx with almost 15 000 deaths.1 The treatment options, supported by guidelines from national organizations,2 stress multimodality therapy when indicated, including surgery, radiation therapy, and chemotherapy.3 The most commonly used chemotherapies are the platinum-based agents, such as cisplatin and carboplatin. Both radiation and cisplatin are routinely used despite strong evidence of ototoxic effects.4
The percentage of patients who experience hearing loss after these treatments is variable, with estimates of between 17% and 88% associated with cisplatin5,6 and as high as 40% with radiation.7 This variability, coupled with a lack of understanding of underlying molecular pathways, leaves “few means of identifying patients at risk for developing important platinum toxicities before treatment.”2 Many studies have assessed the ototoxic effects of cisplatin and other chemoradiation therapies4,7,8; however, there are, to our knowledge, few that are externally validated and even fewer that use large-cohort, prospective data. Statistical models currently exist to predict ototoxicity from cisplatin and radiation,4,8,9,10,11 but the dosage (input) data needed for these models are typically not available until treatment is complete. Patients presenting with hearing loss must undergo multiple evaluations to receive treatment, and the process can be exhausting for someone managing posttreatment sequelae and follow-up care.12 Complex medical management, in concert with the stigma and trouble of a hearing evaluation, can generate barriers to follow-up that can affect quality of life.13
A recently completed retrospective cohort study of 242 patients newly diagnosed with head and neck cancer examined hearing loss after radiation therapy and/or cisplatin.14 The researchers were able to identify 4 pretreatment variables that were associated with hearing loss and developed a dynamic nomogram. Using this tool, oncologists, audiologists, and other health care professionals can predict hearing loss attributable to ototoxic therapies in patients with head and neck cancer using data available before treatment. The proposed benefit of this tool is its ability to help with treatment planning as well as audiologic counseling and follow-up after completion of therapy. The current study validates this nomogram using prospectively collected data from the same institution as the development cohort15 and identifies implementation strategies for its clinical use.
Methods
Study Design
This prospective cohort study included patients undergoing treatment for head and neck cancer at a single institution. Patient and treatment data were prospectively collected from patients undergoing treatment at the Siteman Cancer Center at Barnes-Jewish Hospital at Washington University in St Louis and the clinical audiometric database, AudBase (AudSoft Inc). Further treatment data were retrospectively requested and abstracted by the Washington University Oncology Data Services Cancer Registry and the radiation oncology treatment planning software, Eclipse (Varian Medical Systems). The registry contains standard data elements16 as well as pertinent comorbidity information abstracted by trained staff.17 This study was approved by the Washington University School of Medicine Institutional Review Board, which waived the need for informed consent because the data were collected as a part of the patients’ clinical care, posed no more than minimal risk to the patients, and did not adversely affect the welfare of the patients involved. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Study Population and Demographics
All patients in the included cohort were treated with radiotherapy and/or cisplatin therapy from July 1, 2018, to December 31, 2019. Patients with a history of radiation to the head or neck or those who previously received platinum-based chemotherapies were excluded. Patients with an unknown primary site, missing treatment data, or missing audiometric data were also excluded. Patients underwent an audiometric evaluation before the initiation of treatment and approximately 2 months after completion of treatment. Data were collected on patient demographics and disease characteristics, including age, sex, tobacco use, tumor stage, and tumor site. Primary tumor site was grouped into 4 categories based on distance to the cochlea as described in the initial creation of the nomogram to serve as a proxy for the amount of cochlear radiation.14 Comorbidity data in the form of the overall Adult Comorbidity Evaluation-27 score17 were abstracted from the Oncology Data Services Cancer Registry as well as anesthesia preoperative assessment notes when available.
Patient Treatments
Data were collected on the use of surgery, chemotherapy treatment regimen, total cisplatin dose, and mean cochlear radiation dose as primary disease treatments. The mean cochlear radiation dose was calculated using the radiation oncology treatment planning software. Specifically, the expected radiation per volume element was calculated within an axial computed tomography scan. The mean for all of the elements within the cochlea was then calculated. Chemotherapy treatment regimens were separated based on the original study into (1) no cisplatin, (2) induction cisplatin with high-dose cisplatin, or (3) all other cisplatin regimens (induction, low dose, induction and low dose, and high dose).
Audiometric Evaluation
Pure-tone audiometry was conducted in a double-walled IAC sound booth (IAC Acoustics) in compliance with the American National Standards Institute on Acoustics (specification S3.6-1996). Air conduction thresholds were measured at the standard frequencies of 250 to 8000 Hz via a GSI-61 (Grason-Stadler), GSI AudioStar Pro (Grason-Stadler), or Madsen Astera (GN Otometrics) audiometer. Bone conduction thresholds were measured at 250, 500, 1000, 2000, and 4000 Hz. High-frequency audiometry was performed on patients who reported cisplatin as a known agent in their treatment regimen and for those whose treatment regimen was not yet clear at the time of the baseline audiometric evaluation. High-frequency pure-tone thresholds were measured at 9 to 20 kHz. The pure-tone average (PTA) of 1000, 2000, and 4000 Hz was calculated and defined, consistent with the original study. This average was chosen because most speech information is contained within these frequencies.18 Audiometric evaluation was performed before treatment to establish a baseline PTA and approximately 2 months after completion of treatment.
Clinician Interviews
Select clinicians from different specialties who interact with these patients were interviewed using a standardized questionnaire to generate themes about the nomogram’s clinical utility and to identify barriers to implementation (eAppendix in the Supplement). The interviewees included 2 radiation oncologists, 2 audiologists working in the head and neck clinic, a medical oncologist, and a head and neck surgeon who follows up with patients in a survivorship clinic.
Statistical Analysis
Descriptive statistics were used to explore distribution of patient demographics, treatments, and audiometric evaluations, and bivariate analysis was used to compare the validation cohort with the original development cohort. Comparison was reported using mean and proportion differences with 95% CIs. A multivariable model using the established preclinical variables in the original nomogram was generated using data from each ear and the SAS/STAT MIXED procedure, version 9.4 (SAS Institute, Inc) to account for the nesting of ears in each patient and was reported as unstandardized beta (B) values with 95% CIs.
Model discrimination19 was evaluated using a receiver operating characteristic (ROC) curve to assess the model’s ability to predict a common threshold for hearing aid intervention: posttreatment PTA at 1, 2, and 4 kHz greater than 35 dB. Those with a pretreatment PTA of 35 dB or greater were not included in the analysis. This outcome was compared with the original model and to a model developed by Theunissen et al.4 The ROC curve was used to identify an optimal sensitivity and specificity for this population. Observed and predicted values for audiometric hearing loss were calculated for each ear and then compared using calibration curves by decile to assess how well the tool predicted ototoxic risk in the validation cohort. Adjusted B values and 95% CIs were reported from regression models. All analyses were performed from September 1, 2019, to March 31, 2020, using SAS software, version 9.4 (SAS Institute, Inc), and dynamic nomograms were updated using DynNom library in R and Shiny, version 1.4.0 (R Foundation).
Results
A total of 105 patients (208 ears; mean [SD] age, 61 [11] years; 82 men [78%]) being treated for head and neck cancer using radiation with or without chemotherapy were identified from July 1, 2018, to December 31, 2019. One patient was excluded for previous carboplatin therapy, and 2 additional ears were excluded for lack of audiometric data (both with unilateral anacusis, or complete deafness). This cohort was compared with the development cohort14 to assess the similarities and differences in case mix.
Demographically, the ratio of women (23 of 105 [22%] vs 56 of 242 [23%] women; proportion difference [PD], –1.2 [95% CI, –10.2 to 8.8]) and average age (mean [SD], 61 [11] years vs 60 [10] years; mean difference [MD], 1.0 [–1.4 to 3.4]) were similar in the validation and original cohorts. The original and validation cohorts also had similar comorbidity scores for no comorbidities (30% vs 34%; PD, 3.6 [95% CI, –7.0 to 14.9]), mild comorbidities (42% vs 35%; PD, –6.5 [95% CI, –17.4 to 5.1]), moderate comorbidities (18% vs 23%; PD, 5.1 [95% CI, –3.9 to 15.3]), and severe comorbidities (10% vs 8%; PD, –2.3 [95% CI, –8.3 to 5.5]) (Table 1). Notably, there was a smaller proportion of current smokers in the validation group (25 of 105 [24%] vs 83 of 242 [36%]; PD, –11.8 [95% CI, –1.1 to –21.4]). In terms of disease characteristics, there were significantly fewer stage 4 cancers in the validation cohort compared with the development cohort (54 of 105 [52%] vs 173 of 242 [71%]; PD, –19.1 [95% CI, –30.0 to –7.9]). Although tumor site was generally similar, there was a relative increase in the proportion of cancers labeled close to the cochlea in the validation cohort (20 of 105 [9%] vs 24 of 242 [5%]; PD, 4.5 [95% CI, 0.5-9.5]).
Table 1. Comparison of Patient Demographic Characteristics and Treatment Between Original and Validation Cohorts.
| Characteristic | No. (%)a | Mean, median, or proportion difference (95% CI)b | |
|---|---|---|---|
| Original | Validation | ||
| Ears | 242 (484) | 105 (208) | NA |
| Sex | |||
| Men | 186 (77) | 82 (78) | 1.2 (−8.8 to 10.2) |
| Women | 56 (23) | 23 (22) | NA |
| Age, mean (SD), y | 60 (10) | 61 (11) | −1.0 (−3.4 to 1.4) |
| Tobacco | |||
| Never | 77 (33) | 47 (45) | 11.7 (0.6 to 22.8) |
| Prior | 73 (31) | 33 (31) | 0.1 (−10.1 to 11.1) |
| Current | 83 (36) | 25 (24) | −11.8 (−1.1 to −21.4) |
| Missing | 9 | 0 | NA |
| Comorbidity status (ACE-27) | |||
| None | 68 (30) | 34 (34) | 3.6 (−7.0 to 14.9) |
| Mild | 93 (42) | 35 (35) | −6.5 (−17.4 to 5.1) |
| Moderate | 40 (18) | 23 (23) | 5.1 (−3.9 to 15.3) |
| Severe | 23 (10) | 8 (8) | −2.3 (−8.3 to 5.5) |
| Missing | 18 | 5 | NA |
| Stage | |||
| 1, 2, or 3 | 69 (29) | 49 (48) | 19.1 (7.9 to 30.0) |
| 4 | 173 (71) | 54 (52) | NA |
| Missing | 0 | 2 | NA |
| Primary tumor site by approximate distance to cochlea | |||
| Far | 144 (30) | 58 (28) | −2.3 (−9.3 to 5.3) |
| Mid far | 145 (30) | 55 (26) | −3.9 (−10.8 to 3.6) |
| Mid near | 169 (35) | 77 (37) | 1.6 (−6.0 to 9.5) |
| Close | 24 (5) | 20 (9) | 4.5 (0.5 to 9.5) |
| Treatment received | |||
| Surgery | |||
| No | 90 (37) | 38 (36) | −1.0 (−10.2 to 11.6) |
| Yes | 152 (63) | 67 (64) | NA |
| Cisplatin chemotherapy | |||
| No | 137 (57) | 55 (52) | −4.2 (−15.5 to 7.0) |
| Yes | 105 (43) | 50 (48) | NA |
| Planned cisplatin treatment | |||
| None | 137 (57) | 55 (52) | −4.2 (−15.5 to 7.0) |
| Induction and high dose | 31 (13) | 1 (1) | −11.9 (−16.7 to −6.3) |
| Other (induction, low dose, induction and low dose, high dose) | 74 (31) | 49 (47) | 16.1 (5.0 to 27.0) |
| Total cumulative cisplatin dose, No. | 105 | 50 | NA |
| Mean (SD), mg/m2 | 298 (109) | 238 (83) | −60.0 (−94.4 to −25.6) |
| MCD by ear, No. | 210 | 182 | NA |
| Mean (SD), Gy | 15 (13) | 13 (12) | −2.0 (−4.5 to 0.5) |
| Hearing tests | |||
| PTA, median (IQR), dB | |||
| Pretreatment | 20.0 (11.7 to 33.3) | 21.7 (11.7 to 31.7) | 1.7 (−2.0 to 5.7) |
| Posttreatment | 26.7 (15 to 41.7) | 26.7 (16.7 to 39.2) | 0.0 (−3.5 to 3.5) |
Abbreviations: ACE-27, Adult Comorbidity Evaluation-27; IQR, interquartile range; MCD, mean cochlear dose; NA, not applicable; PTA, pure-tone average.
Values are expressed as No. (%) unless otherwise specified.
Mean, median, and proportion differences presented as the validation cohort minus the development cohort numbers.
With regard to treatment, the validation group received comparable surgical intervention (67 of 105 [64%] vs 152 of 242 [63%]; PD, 1.0 [95% CI, –11.6 to 10.2]) and similar proportion of patients receiving some dosage of cisplatin therapy (50 of 105 [48%] vs 105 of 242 [43%]; PD, 4.2 [95% CI, –7.0 to 15.5]). Within the cisplatin-treatment therapy designation, many fewer patients in the validation cohort received induction and high-dose treatment (1 of 105 [1%] vs 31 of 242 [13%]; PD, –11.9 [95% CI, –16.7 to –6.3]); however, many more received some other cisplatin-based treatment in the validation cohort (49 of 105 [47%] vs 74 of 242 [31%]; PD, 16.1 [95% CI, 5.0-27.0]). Of those receiving cisplatin, the mean [SD] dose in the validation cohort was less (238 [83] mg/m2 vs 298 [109] mg/m2; MD, – 60.0 [95% CI, –94.4 to –25.6 mg/m2]). Of those receiving radiation to the head and neck region, the mean [SD] dose was lower in the validation cohort (13 [12] Gy vs 15 [13] Gy; MD, –2.0 [95% CI, –4.5 to 0.5 Gy]). Pretreatment and posttreatment median PTA was not different between the validation and development cohorts (21.7 dB [interquartile range (IQR), 11.7-31.7 dB] vs 20.0 dB [IQR, 11.7-33.3 dB] and 26.7 dB [IQR, 16.7-39.2 dB] vs 26.7 dB [IQR, 15-41.7 dB], respectively).
The developed predictive model was recreated to test for posttreatment average at 1, 2, and 4 kHz. Compared with no cisplatin treatment, induction and high-dose therapy were associated with a mean hearing threshold PTA increase of 19.38 dB (95% CI, 3.40-35.35 dB), whereas all other cisplatin regimens were associated with a mean increase of 4.56 dB (95% CI, 1.44-7.68 dB). Baseline average 1-, 2-, and 4-kHz PTA (95% CI, 0.64-0.86 dB) and age divided by 10 (95% CI, 0.87-3.92) were both predictors for increased posttreatment PTA. Tumor location showed a position-dependent increase in posttreatment effect, with lesions closer to the cochlea showing a more pronounced effect on posttreatment PTA (Table 2).
Table 2. Model of Posttreatment 1-, 2-, and 4- kHz PTA Using the New Validation Cohort Compared With the Development, Predictive Model.
| Variable | B (95% CI) | |
|---|---|---|
| Development model | Validation model | |
| Ears, No. | 482 | 208 |
| Intercept | −4.89 (−10.94 to 1.17) | −7.47 (−16.60 to 1.66) |
| Age, y | ||
| Age divided by 10 y | 1.08 (0.08 to 2.08) | 2.39 (0.87 to 3.92) |
| Baseline 1-2-4 kHz PTA | 0.95 (0.90 to 1.01) | 0.75 (0.64 to 0.86) |
| Planned cisplatin regimen | ||
| None | 1 [Reference] | 1 [Reference] |
| Induction and high dose | 13.57 (10.53 to 16.62) | 19.38 (3.40 to 35.35) |
| Othera | 5.35 (3.14 to 7.56) | 4.56 (1.44 to 7.68) |
| Primary tumor site by approximate distance to cochlea | ||
| Far | 1 [Reference] | 1 [Reference] |
| Mid far | 1.57 (−0.48 to 3.62) | −0.25 (−3.44 to 2.94) |
| Mid near | 1.97 (0.15 to 3.79) | 1.40 (−1.39 to 4.18) |
| Close | 2.46 (−1.02 to 5.93) | 5.37 (0.88 to 9.86) |
Abbreviation: PTA, pure-tone average.
Other includes induction, low-dose, induction and low-dose, and high-dose regimens.
The original prediction model was used to predict posttreatment hearing loss for the validation cohort as previously described by Theunissen et al4 (defined as PTA at 1, 2, and 4 kHz >35 dB). The ROC curve is shown in Figure 1 with an area under the curve of 0.71 and a predicted posttreatment PTA with an optimal sensitivity and specificity of 73% and 67%, respectively, compared with the original model with a sensitivity of 80% and specificity of 75%. Predicted and observed posttreatment PTA values using the original model and validation cohort were compared using a calibration curve (Figure 2).
Figure 1. Receiver Operating Characteristic Curve (ROC) Predicting 1-, 2-, and 4-kHz PTA Greater Than 35 dB After Treatment for Validation Cohort.
An ROC curve for prediction of posttreatment pure-tone average (PTA) > 35 dB using the developed model has an area under the ROC curve of 0.71. The optimal sensitivity is 73% and specificity is 67% for the validation cohort. AUC indicates area under the curve.
Figure 2. Calibration Curve for Validation Cohort Using Original Predictive Model for Posttreatment PTA by Decile With 95% CIs Compared With a Perfectly Predictive Model .
Calibration curve for the predictive model of posttreatment pure-tone average (PTA), plotting the observed and predicted PTA values by decile. 95% CIs are shown by the whiskers, and the diagonal, dashed line indicates a perfect model for which there is no difference in observed and predicted values.
Six clinicians (2 audiologists, 2 radiation oncologists, a head and neck surgeon, and a medical oncologist) were asked a series of questions (eAppendix in the Supplement). Their responses were recorded and are paraphrased in Table 3. All clinicians said the nomogram could be useful for counseling patients on what to expect with their hearing posttreatment. When asked whether it would affect treatment considerations, all respondents stated it would not; however, this was caveated that it could be useful for in-depth treatment counseling. Similarly, clinicians acknowledged the utility of the tool when asked about decision-making but did not suggest it would change treatment considerations. All respondents thought the information offered by the nomogram would be most useful in the initial visit with treatment counseling. Suggestions for where to record this information varied, including documenting in the chart, sending an email to the medical oncologist, or curating a database. All respondents expressed possible limitations ranging from difficulty of implementing this tool to questioning its utility when posttreatment audiology visits are already standard of care. Table 3 includes a full list of questions and responses.
Table 3. Questions and Responses to Clinician Interviews.
| Question | Common and relevant responses |
|---|---|
| Would this nomogram be useful in your practice? | Yes, for counseling (A,A) |
| Directly, no, but it could be useful for my patients (HNS, MO, RO, RO) | |
| Is the information the nomogram provides valuable to your treatment considerations? | It can be helpful for survivorship clinic and planning for potential toxicity (A, A, HNS, MO) |
| It would work well as a counseling tool (A, A) | |
| It could be, but it could also raise more questions about what range qualifies as a functional loss (MO) | |
| Not for radiation therapy (RO, RO) | |
| Do you think this nomogram would be valuable in decision-making with your patients? Why or why not? | Possibly within the context of future directions (new treatments or radiation modalities) (MO, RO, RO) |
| This information wouldn’t affect counseling of gold-standard treatments (HNS, MO) | |
| Not for treatment, but for pretreatment counseling (A) | |
| Not in direct decisions, but in posttreatment surveillance to make future decisions about hearing health (A) | |
| When would be the best time to know this information (pretreatment screening, at the initial visit with the oncologist, during treatment counseling, etc)? | Ideally at the initial visit to the head and neck clinic (A, HNS, MO, RO, RO) |
| Specifically when the patient is being counseled by the oncologist (A) | |
| Where would it be useful to have the information provided by the nomogram (in the chart, emailed to the oncologist, on a web application, etc)? | The chart (ie, audiology note or flag) (HNS,MO, RO) |
| Brief communication (ie, email) (A, HNS, MO, RO) | |
| Since it’s not 100% accurate, maybe not the chart; maybe email or a database (A) | |
| What potential problems do you foresee in using the nomogram? | Being clinically meaningful for patients (A, HNS) |
| Difficult to implement (A, MO) | |
| Posttreatment audiometry is already routinely conducted (RO) | |
| Refusing lifesaving treatment to preserve hearing (A) | |
| Adding worries to an already worried patient (A, MO) | |
| Common themes | The nomogram is more useful as a counseling tool than as a treatment–decision-making tool |
| The nomogram can be useful for future hearing health | |
| It can provide another piece of information to help inform patients of what to expect | |
| There is no clear direct benefit to decision-making | |
| There is a possibility of fewer benefits than consequences (ie, increased worry, refusing treatment) |
Abbreviations: A, audiologist; HNS, head and neck surgeon; MO, medical oncologist; RO, radiation oncologist.
The original dynamic nomograms developed using the DynNom library were updated using the most recent version as depicted by Schuette and colleagues.14
Discussion
This cohort study confirms and validates the previous findings of a preclinical nomogram capable of predicting posttreatment hearing in patients being treated for head and neck cancer.14 When recreating the original model using the validation cohort, all but 1 of the 95% CIs of the new β coefficients contained the original estimate. The 1 value that did not was baseline PTA; however, the absolute effect of this estimate is very small compared with the other variables in the model. The area under the curve of the validation cohort using the derived model was 0.71, which is adequate discrimination and falls short of the excellent discrimination reached by the development cohort. This drop in the area under the curve is to be expected, as the original cohort was used to develop the model and cannot account for the differences in other patient or clinical characteristics. The validation model is slightly overpredictive in 2 out of 10 deciles in the calibration curve (Figure 2), suggesting that some patients with higher predicted hearing PTAs may suffer less severe losses at follow-up. This finding explains the decreased specificity seen for the ability of the model to predict a loss greater than 35 dB. The clinical tool uses baseline PTA, age, type of chemotherapy treatment, and the location of the tumor and can be used clinically to counsel patients on the likely degree of hearing loss after treatment.
It is well known that hearing loss has a dramatic effect on quality of life.13,20 Both the literature and our interviews with clinicians show that patients undergoing treatment for head and neck cancer are often dealing with many sequelae.21,22 The symptom burden is often accepted as a natural progression of the treatment, a trade-off in hope of remission. Although some adverse effects have simple treatments, hearing loss is more insidious and complicated. Several of the clinicians interviewed for this study agreed that identifying those most at risk can potentially mitigate barriers to evaluation and treatment.
A patient presenting with hearing loss requires an office visit with an audiologist followed by hearing aid fittings and, if severe, possibly a cochlear implant evaluation. This process can seem daunting for a patient managing complex or competing medical problems.12 Combined with the stigma and inconvenience of undergoing a hearing loss evaluation, these competing medical problems represent some of the biggest barriers to accessing hearing health care.12 The nomogram presented here has the potential to mitigate these barriers by providing counseling to reduce stigma and to identify high-risk patients in need of more extensive follow-up care. All clinicians interviewed for this study agreed that this can be an exceptionally useful tool for identifying patients at highest risk for hearing loss. Others expanded on this idea, suggesting the creation of a “high-risk for hearing loss” list that can inform the need for closer and longer follow-up.
Furthermore, there is potential to use the nomogram as a counseling tool to empower and inform patients. Often, treatment of head and neck cancer involves many consent forms and a litany of adverse effects that the patient may experience. Not only is this overwhelming, but hearing loss is particularly stigmatized and carries many psychological, social, and socioeconomic barriers.12 Counseling patients on their likely severity of hearing loss may empower them to seek help and mitigate the stigma of hearing devices, particularly because hearing loss still remains a strong concern for patients undergoing chemoradiation for head and neck cancer.6 According to our interviews, patients undergo a single posttreatment audiogram but are not followed further unless they are referred by their physician. If the patient understands there is a way to help them with their hearing loss and that this is an expected, manageable problem, then they may be more likely to pursue that option.
There are previously created models that predict ototoxicity from cisplatin and radiation.4,8,9,10,11 The benefit of the model validated in this article is in its ability to inform individualized counseling on the risk of hearing loss before treatment is initiated. Previous models use variables such as the cumulative dose of cisplatin and mean cochlear radiation dose, which are rarely available or accurate before treatment. Cisplatin is subject to dose increases or decreases during treatment based on patient response and side effects.22 Mean cochlear dose is calculated right before the start of radiation therapy, not during treatment planning. The model validated here does not require these variables to predict treatment outcomes.
Beyond clinical utility, this tool can engage patients in their own care. All clinicians interviewed in this study agreed this nomogram would not affect their recommendations regarding treatment with radiation or platinum agents. However, they did agree that this tool can be used to inform the patient of the potential need for posttreatment care. This involvement of patients in their own care is a form of shared decision-making, which has been shown to be associated with better health outcomes.23,24
Currently, this tool is not meant to be used for treatment planning, and the focus of this work was not to suggest as such, but we believe that it has the potential to enhance posttreatment care. That being said, this nomogram can serve as a prototype for an eventual treatment–decision-making tool in 2 ways. The first will be to serve as a comparative standard for future models that may be able to inform treatments. If and when there are noninferior or superior treatments to platinum-based chemotherapies, these treatments can easily be added to this model in order to weigh the costs and benefits of disease and hearing outcomes. The second is in understanding that, although we as health care professionals may know the best treatments for disease, we may not always realize the best treatment for a patient. Patients whose quality of life would be severely worsened by hearing loss should have the option to decline treatment or pursue other, less ototoxic options.6 Although this approach requires a more complicated discussion on beneficence and respect for persons, we believe it is important to acknowledge this in patient discussions.
Limitations
This study is not without limitations. There is a potential selection bias, as we omitted patients with incomplete audiometric data and those with previous chemotherapy or radiation to the head and neck. Although the nomogram appears to be valid in an independent cohort, there is a need to further validate beyond a single geographic catchment area. Furthermore, our audiometric follow-up is limited to approximately 2 months posttreatment owing to current clinical standards at our institution. Although routine follow-up (6 months posttreatment, 1 year, and annually thereafter) is recommended by the audiologist and medical team, many patients are lost to follow-up. This short follow-up limits our ability to assess the insidious onset of hearing loss in these patients; however, this limitation suggests that we may be underreporting those eventually experiencing hearing loss in this population. We also acknowledge that the limited number of clinicians interviewed (6) decreases the potential generalizability of the qualitative findings. Future studies should focus on dissemination and implementation of this tool clinically, as well as associating these findings with patient-reported outcomes.
Conclusions
Here we present the validation of a nomogram that predicts posttreatment PTA for patients undergoing chemoradiation for head and neck cancer using data that are available before initiation of treatment. This validation confirms this model’s ability to predict posttreatment hearing outcomes in a unique population of patients, providing validity. This model has the potential to transform pretreatment counseling and posttreatment hearing evaluations for this patient population. The nomogram can identify patients at high risk for hearing loss and may allow audiologists to closely monitor them for possible intervention of this silent sequela.
eAppendix. Ototoxicity Nomogram Oncologist/Audiologist Interview
References
- 1.Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7-30. doi: 10.3322/caac.21590 [DOI] [PubMed] [Google Scholar]
- 2.National Comprehensive Cancer Network The National Comprehensive Cancer Network. Accessed August 18, 2020. https://www.nccn.org/
- 3.Prestwich RJ, Sen M, Scarsbrook AF. Management of advanced head and neck cancer. New Engl J Med. 2016;375(5):491-492. doi: 10.1056/NEJMc1606211 [DOI] [PubMed] [Google Scholar]
- 4.Theunissen EA, Zuur CL, Józwiak K, et al. Prediction of hearing loss due to cisplatin chemoradiotherapy. JAMA Otolaryngol Head Neck Surg. 2015;141(9):810-815. doi: 10.1001/jamaoto.2015.1515 [DOI] [PubMed] [Google Scholar]
- 5.Travis LB, Fossa SD, Sesso HD, et al. ; Platinum Study Group. Chemotherapy-induced peripheral neurotoxicity and ototoxicity: new paradigms for translational genomics. J Natl Cancer Inst . 2014;106(5):dju044. doi: 10.1093/jnci/dju044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Schmitt NC, Page BR. Chemoradiation-induced hearing loss remains a major concern for head and neck cancer patients. Int J Audiol. 2018;57(sup4):S48-S53. doi: 10.1080/14992027.2017.1353710 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jereczek-Fossa BA, Zarowski A, Milani F, Orecchia R. Radiotherapy-induced ear toxicity. Cancer Treat Rev. 2003;29(5):417-430. doi: 10.1016/S0305-7372(03)00066-5 [DOI] [PubMed] [Google Scholar]
- 8.Zuur CL, Simis YJ, Lansdaal PE, et al. Risk factors of ototoxicity after cisplatin-based chemo-irradiation in patients with locally advanced head-and-neck cancer: a multivariate analysis. Int J Radiat Oncol Biol Phys . 2007;68(5):1320-1325. doi: 10.1016/j.ijrobp.2007.01.042 [DOI] [PubMed] [Google Scholar]
- 9.Johnson A, Tarima S,Wong S, Friedland DR, Runge CL. Statistical model for prediction of hearing loss in patients receiving cisplatin chemotherapy. JAMA Otolaryngol Head Neck Surg . 2013;139(3):256-264. doi: 10.1001/jamaoto.2013.1713 [DOI] [PubMed] [Google Scholar]
- 10.Kennedy IC, Fitzharris BM, Colls BM, Atkinson CH. Carboplatin is ototoxic. Cancer Chemother Pharmacol . 1990;26(3):232-234. doi: 10.1007/BF02897206 [DOI] [PubMed] [Google Scholar]
- 11.Dille MF, Wilmington D,McMillan GP, Helt W, Fausti SA, Konrad-Martin D. Development and validation of a cisplatin dose-ototoxicity model. J Am Acad Audiol . 2012;23(7):510-521. doi: 10.3766/jaaa.23.7.3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Barnett M, Hixon B, Okwiri N, et al. Factors involved in access and utilization of adult hearing healthcare: a systematic review. Laryngoscope. 2017;127(5):1187-1194. doi: 10.1002/lary.26234 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hyams AV, Hay-McCutcheon M, Scogin F. Hearing and quality of life in older adults. J Clin Psychol. 2018;74(10):1874-1883. doi: 10.1002/jclp.22648 [DOI] [PubMed] [Google Scholar]
- 14.Schuette A, Lander DP, Kallogjeri D, et al. Predicting hearing loss after radiotherapy and cisplatin chemotherapy in patients with head and neck cancer. JAMA Otolaryngol Head Neck Surg. 2020;146(2):106-112. doi: 10.1001/jamaoto.2019.3550 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Justice AC, Covinsky KE, Berlin JA. Assessing the generalizability of prognostic information. Ann Intern Med. 1999;130(6):515-524. doi: 10.7326/0003-4819-130-6-199903160-00016 [DOI] [PubMed] [Google Scholar]
- 16.Commission on Cancer, Facility Operations and Data Standards (FORDS). Section One: Case Eligibility, Cancer Identification, and Overview of Coding Principles. American College of Surgeons; 2003:1-28. [Google Scholar]
- 17.Piccirillo JF, Tierney RM, Costas I, Grove L, Spitznagel EL Jr. Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA. 2004;291(20):2441-2447. doi: 10.1001/jama.291.20.2441 [DOI] [PubMed] [Google Scholar]
- 18.Mueller HG KM An easy method for calculating the articulation index. Hearing Journal. 1990;43(9):14-17. [Google Scholar]
- 19.Alba AC, Agoritsas T, Walsh M, et al. Discrimination and calibration of clinical prediction models: users' guides to the medical literature. JAMA . 2017;318(14):1377-1384. doi: 10.1001/jama.2017.12126 [DOI] [PubMed] [Google Scholar]
- 20.Brodie A, Smith B, Ray J. The impact of rehabilitation on quality of life after hearing loss: a systematic review. Eur Arch Otorhinolaryngol. 2018;275(10):2435-2440. doi: 10.1007/s00405-018-5100-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heijnen BJ, Speyer R, Kertscher B, et al. Dysphagia, speech, voice, and trismus following radiotherapy and/or chemotherapy in patients with head and neck carcinoma: review of the literature. Biomed Res Int. 2016;2016:6086894. doi: 10.1155/2016/6086894 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Iqbal MS, Chaw C, Kovarik J, et al. Primary concurrent chemoradiation in head and neck cancers with weekly cisplatin chemotherapy: analysis of compliance, toxicity and survival. Int Arch Otorhinolaryngol. 2017;21(2):171-177. doi: 10.1055/s-0036-1594020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kaplan SH, Greenfield S, Ware JE Jr. Assessing the effects of physician-patient interactions on the outcomes of chronic disease. Med Care. 1989;27(3)(suppl):S110-S127. doi: 10.1097/00005650-198903001-00010 [DOI] [PubMed] [Google Scholar]
- 24.Makoul G, Clayman ML. An integrative model of shared decision making in medical encounters. Patient Educ Couns. 2006;60(3):301-312. doi: 10.1016/j.pec.2005.06.010 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eAppendix. Ototoxicity Nomogram Oncologist/Audiologist Interview
