Global burden of ototoxic hearing loss associated with platinum-based cancer treatment: A systematic review and meta-analysis.

Platinum-based chemotherapeutic agents cisplatin and carboplatin are widely used in cancer treatment worldwide and may result in ototoxic hearing loss. The high incidence of cancer and salient ototoxic effects of platinum-based compounds pose a global public health threat. The purpose of this study was twofold. First, to estimate the prevalence of ototoxic hearing loss associated with treatment with cisplatin and/or carboplatin via a systematic review and meta-analysis. Second, to estimate the annual global burden of ototoxic hearing loss associated with exposure to cisplatin and/or carboplatin. For the systematic review, three databases were searched (Ovid Medline, Ovid Embase, and Web of Science Core Collection) and studies that reported prevalence of objectively measured ototoxic hearing loss in cancer patients were included. A random effects meta-analysis determined pooled prevalence (95% confidence intervals [CI]) of ototoxic hearing loss overall, and estimates were stratified by treatment and patient attributes. Estimates of ototoxic hearing loss burden were created with published global estimates of incident cancers often treated with platinum-based compounds and cancer-specific treatment rates. Eighty-seven records (n = 5077 individuals) were included in the meta-analysis. Pooled prevalence of ototoxic hearing loss associated with cisplatin and/or carboplatin exposure was 43.17% [CI 37.93-48.56%]. Prevalence estimates were higher for regimens involving cisplatin (cisplatin only: 49.21% [CI 42.62-55.82%]; cisplatin & carboplatin: 56.05% [CI 45.12-66.43%]) versus carboplatin only (13.47% [CI 8.68-20.32%]). Our crude estimates of burden indicated approximately one million individuals worldwide are likely exposed to cisplatin and/or carboplatin, which would result in almost half a million cases of hearing loss per year, globally. There is an urgent need to reduce impacts of ototoxicity in cancer patients. This can be partially achieved by implementing existing strategies focused on primary, secondary, and tertiary hearing loss prevention. Primary ototoxicity prevention via otoprotectants should be a research and policy priority.

Introduction

The burden of cancer presents a threat to every country. In 2020, there were an estimated 19.3 million incident cancer cases worldwide [1], [2]. Hematologic and solid tumor malignancies are often treated with platinum-based chemotherapeutic agents cisplatin and/or carboplatin, which are used as either single agents or in combination for induction and neoadjuvant therapies. Both cisplatin and carboplatin are widely used and included on the World Health Organization’s (WHO) list of essential medicines [3]. Cisplatin, in particular, is very commonly utilized globally given its high effectiveness, relatively low cost and accessibility. Although use of cisplatin and/or carboplatin is necessary for cancer treatment, widespread use of these agents puts many individuals at risk for experiencing associated toxicities and adverse events, including ototoxicity (i.e., drug-induced hearing loss). The global burden of ototoxic hearing loss associated with exposure to cisplatin and/or carboplatin has not been quantified. However, these data are necessary to prioritize the areas of research and clinical care needed to promote prevention and management of ototoxic hearing loss.

Cisplatin is a well-known ototoxic agent, often resulting in permanent, bilateral and sensorineural hearing loss [4]. Carboplatin is less ototoxic than cisplatin, although hearing loss and other ototoxic effects, such as vestibular dysfunction or tinnitus, have been reported after carboplatin exposure [5]. Studies have reported a wide range of prevalent ototoxic hearing loss estimates in patients treated with cisplatin and/or carboplatin, depending on certain risk factors including age and treatment details [6]. Prevalence estimates are also influenced by methods of measuring and defining hearing (e.g., using ototoxicity grading scales/systems, times of hearing measurement).

Given the high global incidence of cancer and the likely high incidence of ototoxic hearing loss from platinum-based compounds, efforts should focus on ototoxicity prevention and reducing the burden of ototoxic hearing loss. Impacts of hearing loss have been well-documented. In children, hearing loss often leads to delayed speech and language development, which may impact literacy and educational attainment [7], [8]. In adults, hearing loss has been associated with impaired communication, lower income, poorer quality of life and mental health, and cognitive impairment [9], [10]. Negative consequences of hearing loss are often exacerbated for individuals in low- and middle-income countries which is important given the burden of cancer is projected to rise most rapidly in these areas [2].

Undoubtedly, the priority of cancer treatment is to reduce mortality. However, researchers and clinicians recognize the importance of preserving quality of life during and after cancer treatment [11], which may be partially achieved via primary, secondary or tertiary prevention of ototoxicity. Although hearing loss prevention has historically received inadequate attention despite evidence of its cost effectiveness and benefit [12], [13], disability prevention and preservation of quality of life continues to gain traction on the global agenda [12], [14], [15].

To promote ototoxicity prevention, it is necessary to understand the burden of ototoxicity. Therefore, the purpose of article was twofold. First, to estimate the prevalence of ototoxic hearing loss associated with treatment with cisplatin and/or carboplatin, and evaluate differences in estimates by treatment details (i.e., drug type, dose, radiotherapy), age, cancer/malignancy type, and audiometric methods. Second, to estimate the annual global number of individuals at risk of exposure to cisplatin/carboplatin (‘at-risk’ population) and exposed to cisplatin/carboplatin (‘exposed’ population), and the annual global number of hearing loss cases likely resulting from exposure (‘cases’).

Material and methods

This systematic review was conducted under the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16].

Inclusion criteria

Peer-reviewed manuscripts published in English between years 2005–2018 that were case-control, cohort, or cross-sectional studies were eligible for screening. Manuscripts must have contained information on human subjects diagnosed with cancer, treated with cisplatin and/or carboplatin and measured hearing loss using standard objective hearing evaluation tests, including pure tone audiometry (250–8000+ Hz), otoacoustic emissions (OAE), and/or auditory brainstem response (ABR) waveforms. Studies reporting hearing loss based on subjective self-report measures were not included. Hearing must have been measured objectively in the entire study sample to avoid risk of underestimation. Eligible manuscripts were required to report incident or prevalent hearing loss after treatment with cisplatin and/or carboplatin.

Search strategy

The databases Ovid Medline, Ovid Embase, and Web of Science Core Collection were searched. Search strings are available in supplementary material 1. Search terms included the words “hearing loss,” “hearing impair,* ” “deafness,” “auditory” or “audiolog* ” and were used along with antineoplastic agent/ or *chemotherapy.

Study selection

A single reviewer extracted all eligible papers across the three databases and removed duplicates. This reviewer also completed the title and abstract screen, based on the eligibility criteria stated above. References were exported to Mendeley and shared with study team members for full paper screening then review. Two separate reviewers then completed a full paper screen then a full paper review based on the eligibility criteria. Any differences between the two reviewers were reconciled via discussion of manuscripts.

Data collection & analysis

Two WHO biostatisticians developed data extraction tables. Extracted data included (a) meta-study information (name of authors, year of publication), (b) sample characteristics including demographics (e.g., age, sex), (c) intervention (treatment) details (e.g., dose, frequency, duration of treatment), (d) outcome details (e.g., method of hearing loss measurement, definition of hearing loss), and (e) prevalence or incidence of ototoxic hearing loss. A single reviewer extracted study information.

The treatment drug was categorized as follows, carboplatin only, cisplatin only, cisplatin and carboplatin, and cisplatin or carboplatin (i.e., drug not specified for individuals). Age was categorized as < 5 yrs, children (< 20 yrs), children & young adults (≤ 35 yrs), adults (≥ 20 yrs), and all age categories. Age categories were chosen based on categorizations used in the studies included. The cancer/malignancy type was specified during data extraction/analyses when reported in the study, but when it was not specified, the type was categorized as ‘mixed.’ Radiotherapy was defined as present if > 30% of the study sample received radiotherapy.

The primary method of hearing measurement (audiometry, OAE, ABR, or mixed) and the ototoxicity grading scale or system (when available) were reported. If hearing loss was quantified on more than 1 ototoxicity grading scale, the American Speech-Language Hearing Association (ASHA) scale was used to report prevalence of audiometric threshold shift [17]. Given heterogeneity in outcome measures and definitions, hearing loss was classified binarily. Hearing loss was defined as any threshold shift classified by ototoxicity grading scales or systems, or hearing loss (threshold shifts or absolute hearing levels) defined by authors of the articles.

The I2 statistic was used to evaluate heterogeneity across studies. Random effects models were used to meta-analyze study results and to determine pooled prevalence estimates of ototoxic hearing loss and corresponding 95% confidence intervals (95% CI). Logistic meta-regression models evaluated the impact of mean or median cumulative dose (used continuously) on the effect estimate for cisplatin only, carboplatin only, and cisplatin and carboplatin. A modified version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) scale was used to assess risk of bias for each study, based on the following categories: selection, randomization, performance, completeness, and reporting bias [18]. Sensitivity analyses evaluated pooled prevalence estimates after a) excluding each study, individually, from analyses, and b) excluding studies with a high or unclear risk of bias.

Global annual estimates of at-risk and exposed populations and hearing loss cases

The global at-risk population was defined as the number of individuals diagnosed with cancers commonly treated with cisplatin and/or carboplatin. These cancers were bladder, breast, cervical, esophageal, lung (including small lung), mesothelioma, head and neck (nasopharynx, oropharynx, hypopharynx, larynx, salivary glands), ovarian, testicular, colon, prostate and rectal [19]. Estimates focused on cancers primarily affecting adults given the quality of available incidence and treatment data. Primarily pediatric cancers (e.g., blastomas) were not included in global estimates given paucity of global incidence and treatment data and their relatively low incidence [20], [91]. It is unlikely that including pediatric cancers would substantially change estimates given their low global incidence. Incidence estimates by cancer type were obtained from GLOBOCAN 2020 produced by the International Agency for Research on Cancer [1].

The global exposed population was the estimated number of individuals likely treated with cisplatin and/or carboplatin for the cancers stated above. Treatment data were obtained from the American Cancer Society (ACS) [21] or when ACS data were unavailable, Cancer Research UK [22], both of which estimate the proportion of prevalent cancer cases by stage and the proportion of chemotherapeutic treatment within each stage. Treatment data for head and neck cancers were unavailable from these sources and were obtained from a relevant publication [23]. To account for differences in global utilization and availability of cisplatin or carboplatin, we corrected for the global population residing in countries with chemotherapy being ‘generally available’ (≥ 50% of patients in need) in the public sector (obtained via WHO survey) [24]. The proportion of the global population with generally availability to chemotherapeutic drugs was estimated to be 71.9%. To create this estimate, we evaluated a) country-specific responses to the previously mentioned survey question on chemotherapy availability in the public sector (country-specific data are unpublished and were obtained from WHO partners), and b) country-specific population estimates from the World Bank [25]. The proportion of the global population with chemotherapy being ‘generally available’ was multiplied by 75%, which is the median value (range 50–100%) of the population estimated to have chemotherapy generally available. Finally, we accounted for mortality by subtracting estimates of proportion of deaths (deaths/cases) from GLOBOCAN data [1].

The global annual number of hearing loss cases from cisplatin and/or carboplatin exposure was estimated by multiplying the estimated number of exposed individuals (after accounting for global availability and mortality) by the pooled prevalence estimate of ototoxic hearing loss reported in this paper. A similar approach has been used to estimate at-risk and exposed populations and cases of hearing loss for multi-drug resistant tuberculosis [26].

Results

Screening results

A total of 994 non-duplicate citations were identified by using the selected keywords. After the final review, 66 articles were eligible for inclusion in this study [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90]. The study selection process is shown in Fig. 1.

Fig. 1

Study selection process.

Study characteristics

A total of 87 records from 66 studies, corresponding to data from 5077 individuals, were included in the meta-analysis. Studies were from 22 countries (country not stated in 2 studies) corresponding to representation from African (n = 1), American (n = 30), European (n = 21), South-East Asian (n = 7) and Western Pacific (n = 5) regions. All studies included males and females. The number of records including data for each drug is as follows: cisplatin only (n = 49), carboplatin only (n = 17), both cisplatin and carboplatin (n = 15), cisplatin or carboplatin (drug unspecified; n = 6). Pertinent details for records used in the meta-analyses are in supplementary materials 2.

Prevalence of ototoxic hearing loss

The pooled prevalence estimate of ototoxic hearing loss in all records was 43.17% (CI 37.93–48.56%). Table 1 shows prevalence estimates overall and stratified by drug, age group, radiotherapy, and malignancy/cancer type. Prevalence estimates were highest in individuals treated with cisplatin-based regimens (cisplatin & carboplatin: 56.05% [CI 45.12–66.43%]; cisplatin only: 49.21% [CI 42.62–55.82%]), as compared to those treated with carboplatin only (13.47% [CI 8.68–20.30%]). Prevalence estimates across both drug types were highest in records including adults (adults only: 47.39% [CI 36.72–58.30%]; all ages: 61.69%, [CI 44.83–76.14%]) and lowest for records including young children (<5 yrs; 21.00% [CI 6.30–51.23%]).

Table 1

Pooled prevalence estimates of ototoxic hearing loss, overall and stratified by drug type, age group, radiotherapy, and malignancy/cancer type.

	Number of records	Pooled n of participants	Point Estimate (%)	95% Confidence Interval (%)
Total Pooled Prevalence	87	5077	43.17%	37.93 – 48.56%
Drug
Carboplatin only	17	742	13.47%	8.68 – 20.30%
Cisplatin only	49	3224	49.21%	42.62 – 55.82%
Cisplatin & carboplatin	15	697	56.05%	45.12 – 66.43%
Cisplatin or carboplatin	6	414	47.61%	37.65 – 57.77%
Age Group
< 5 years	9	357	21.00%	6.30 – 51.23%
Children (<20 yr)	47	2086	42.92%	36.96 – 49.10%
Children/Young Adults (≤ 35 yr)	7	854	40.06%	22.85 – 60.13%
Adults (≥ 20 yr)	19	1091	47.39%	36.72 – 58.30%
All ages	5	689	61.69%	44.83 – 76.14%
Radiotherapy (>30% received)
No	50	3086	42.71%	36.05 – 49.76%
Yes	37	1991	43.79%	35.27 – 52.69%
Malignancy / Cancer Type
Germ cell tumors	2	529	67.85%	36.41 – 88.61%
Head & neck	19	1018	49.17%	38.52 – 59.90%
Hepatoblastoma	2	107	24.05%	2.43 – 80.03%
Medulloblastoma	6	154	32.48%	11.77 – 63.43%
Neuroblastoma	6	555	54.32%	33.74 – 73.52%
Retinoblastoma	6	363	13.34%	5.50 – 28.95%
Mixed	46	2351	44.84%	38.99 – 50.84%

Estimates were similar regardless of radiotherapy use (no: 42.71% [CI 36.05–49.76%]; yes: 43.79% [CI 35.27–52.69%]). Radiotherapy was used to treat head and neck cancer (16 records), retinoblastoma (1 record), medulloblastoma (6 records), hepatoblastoma (1 record) or mixed (13 records) cancer/malignancy types. Pooled prevalence estimates were highest in studies focusing on germ cell tumors (67.85% [CI 36.41–88.61]), head and neck cancers (49.17% [CI 38.52–59.90%]), and neuroblastoma (54.32% [CI 33.74–73.52%]).

Table 2 displays pooled prevalence estimates for age group and malignancy/cancer type stratified by drug (carboplatin, cisplatin only, cisplatin and carboplatin). Carboplatin was commonly used to treat children under 5 years (6 of 9 records). In children under 5 years treated with carboplatin, the prevalence estimate was lower (7.42% [CI 2.17–22.45]) than treatment regimens involving cisplatin (cisplatin only: 66.67% [CI 26.81–91.61]; cisplatin and carboplatin: 73.20% [CI 27.46–95.17]). Similarly, for other age categories, prevalence estimates for the carboplatin only group were lower than groups exposed to cisplatin.

Table 2

Pooled prevalence estimates of ototoxic hearing loss for age group and malignancy/cancer type, stratified by drug type.

	Number of records	Pooled n of participants	Point Estimate (%)	95% Confidence Interval (%)
Carboplatin only
Age Group
< 5 years	6	277	7.42%	2.17 – 22.45%
Children (< 20 yr)	10	377	19.50%	13.85 – 26.76%
Children/Young Adults (≤ 35 yr)	1	88	6.82%	3.10 – 14.36%
Malignancy / Cancer Type
Medulloblastoma	2	34	9.02%	2.93 – 24.54%
Neuroblastoma	1	30	3.33%	0.47 – 20.20%
Retinoblastoma	6	363	13.34%	5.50 – 28.95%
Mixed	8	227	15.95%	2.93 – 24.54%
Cisplatin only
Age Group
< 5 years	1	6	66.67%	26.81 – 91.61%
Children (<20 yr)	22	1079	45.05%	37.16 – 53.19%
Children/Young Adults (≤ 35 yr)	4	441	53.43%	24.14 – 80.53%
Adults (≥ 20 yr)	17	1009	48.67%	37.04 – 60.44%
All ages	5	689	61.69%	44.83 – 76.14%
Malignancy / Cancer Type
Germ cell tumors	2	529	67.85%	36.41 – 88.61%
Head & Neck	17	936	50.63%	39.02 – 62.17%
Hepatoblastoma	1	52	7.69%	2.92 – 18.77%
Medulloblastoma	4	120	49.04%	16.47 – 82.45%
Neuroblastoma	2	181	80.46%	7.69 – 99.51%
Mixed	23	1406	47.42%	40.61 – 54.32%
Cisplatin and carboplatin
Age Group
< 5 years	2	74	73.20%	27.46 – 95.17%
Children (< 20 yr)	10	284	60.42%	49.32 – 70.54%
Children/Young Adults (≤3 5 yr)	2	325	41.06%	24.25 – 60.25%
Adults (≥ 20 yr)	1	14	14.29%	3.60 – 42.68%
Malignancy / Cancer Type
Head & Neck	1	14	14.29%	3.60 – 42.68%
Hepatoblastoma	1	55	52.73%	39.65 – 65.44%
Neuroblastoma	3	344	55.50%	30.47 – 78.02%
Mixed	10	284	60.42%	49.32 – 65.44%

Hepatoblastoma, medulloblastoma, neuroblastoma and retinoblastoma are pediatric cancers [83] and prevalence estimates varied by the primary drug used (Table 2). Hepatoblastoma was treated only with cisplatin-based regimens (1 cisplatin only; 1 cisplatin and carboplatin) and retinoblastoma was treated only with carboplatin regimens. For medulloblastoma and neuroblastoma, prevalence estimates were higher in regimens involving cisplatin versus those using carboplatin only.

Diagnostic method and ototoxicity grading scales

Details on diagnostic method and ototoxicity grading scales are presented in Table 3. Most records measured hearing via audiometry or via multiple methods (i.e., audiometry and/or ABR and/or OAE). Prevalence estimates were similar for records using audiometry (41.24% [CI 35.36–47.37%]) or other/mixed (45.08% [CI 34.53–56.10%]) methods and highest in the 3 records that used OAE (54.59% [CI 24.42–81.72%]).

Table 3

Pooled prevalence estimates of ototoxic hearing loss, stratified by diagnostic method and ototoxicity grading scale.

	Number of records	Pooled n of participants	Point Estimate (%)	95% Confidence Interval (%)
Diagnostic Method
Audiometry	51	3264	41.24%	35.36 – 47.37%
OAE	3	76	54.59%	24.42 – 81.72%
Other/mixed	33	1737	45.08%	34.53 – 56.10%
Ototoxicity Grading Scale
ASHA	18	1621	58.99%	45.83 – 69.25%
CTCAE	7	447	58.36%	43.63 – 71.73%
Brock	22	903	33.00%	23.44 – 44.20%
Chang	3	296	47.91%	34.61 – 61.52%
Muenster	6	601	52.87%	34.95 – 70.08%
SIOP	4	145	23.29%	16.83 – 31.28%
Non-specified	3	101	41.23%	7.73 – 85.45%
Other/mixed	24	963	36.29%	27.46 – 46.15%

Abbreviations: OAE: Otoacoustic emissions; ASHA: American Speech-Language Hearing Association; CTCAE: Common Terminology Criteria for Adverse Events; SIOP: International Society of Pediatric Oncology

There were 6 primary ototoxicity grading scales used, although 27 records used either a non-specified scale or a different, often non-standardized, definition of ototoxic hearing loss. Pooled prevalence estimates were highest when using ASHA (58.99% [CI 45.83–69.25%]) or Common Terminology Criteria for Adverse Events (CTCAE) scales (58.36% [CI 43.63–71.73%]), which are indicated for use in adult or pediatric populations. The Brock, Chang, Muenster and International Society of Pediatric Oncology (SIOP) scales are pediatric ototoxicity grading scales [84]. Prevalence estimates were higher in records using Chang (47.91% [CI 34.61–61.52%]) or Muenster (52.87% [CI 34.95–70.08%]) scales, as compared to Brock (33.00% [CI 23.44–44.20%]) or SIOP (23.29% [CI 16.83–31.28%]) scales.

Cumulative dose

A meta-regression model did not support a relationship between mean/median cumulative dose of cisplatin or carboplatin (separately or together) and prevalent hearing loss (slope and R2 close to 0).

Heterogeneity & sensitivity analyses

The I2 statistic was 90.1, indicating a high amount of heterogeneity among studies. A sensitivity analysis evaluated change in the pooled prevalence estimate after removing each study, individually, from analyses. The prevalence estimates and confidence intervals were stable (<2% change) after removal of each study, indicating the pooled prevalence estimate was insensitive to individual study inclusion. A sensitivity analysis removing studies with a high or unclear risk of bias did not substantially change prevalence estimates (results not shown). A funnel plot of standard error by logit event rate (not shown) was symmetric and thus did not suggest publication bias.

Table 4 shows crude annual global estimates of at-risk and exposed populations and the annual number of cases of hearing loss associated with exposure to cisplatin and/or carboplatin. The global population of individuals at risk was approximately 10.5 million. The number of individuals likely exposed to chemotherapy annually was approximately one million. It was estimated that exposure to cisplatin and/or carboplatin likely results in almost half a million cases of hearing loss per year worldwide. Detailed calculations by cancer type and stage are in Supplement 3.

Table 4

Estimates of at-risk and exposed populations and associated ototoxic hearing loss cases.

Incident cancer cases per year (at-risk) (n)a	Exposed to chemotherapy per year (n)b	Prevalence ototoxicity (%)	Hearing loss cases per year (n)
10,496,286	1021,700	43.17%	441,068

Cancers included were: bladder, breast, cervical, esophageal, lung (including small lung), mesothelioma, head and neck (nasopharynx, oropharynx, hypopharynx, larynx, salivary glands), ovarian, testicular, colon, prostate and rectal.

Estimate of exposed population accounts for treatment patterns by stage, deaths, and population with chemotherapy available (see methods and supplementary materials 3).

Discussion

The high pooled prevalence of ototoxic hearing loss in cancer patients treated with cisplatin and/or carboplatin (43.17% [CI 37.93 – 48.56%]) reiterates the importance of prioritizing hearing as an adverse event of cancer treatment with platinum-based compounds. Our crude global estimate of annual ototoxic hearing loss cases associated with exposure to cisplatin and/or carboplatin was almost half a million, which demonstrates the high global burden of ototoxicity and highlights the need to prioritize primary, secondary and tertiary hearing loss prevention. Ototoxic hearing loss associated with platinum-based compounds is an important global health issue as demonstrated by high global estimates of ototoxic hearing loss cases, rising cancer incidence (particularly in low-income countries), high global use of platinum-based compounds and difficulties accessing hearing health care in many areas with high cancer burden [1], [12].

Consistent with previous research, this meta-analysis showed regimens using cisplatin resulted in the highest prevalence of ototoxic hearing loss whereas studies that reported only carboplatin use resulted in a lower, but still substantial, prevalence of ototoxic hearing loss [5]. It was not possible to conduct in-depth analyses by malignancy/cancer type as most studies included participants with multiple cancer types and did not report prevalent hearing loss stratified by cancer type. We did not observe an effect of radiotherapy despite being indicated by other studies [93]. In this study, radiotherapy occurred most often in cases of head and neck cancer or mixed malignancy types. Previous research has indicated effects of radiotherapy may be most salient in cases of head and neck cancers given the proximity of the radiotherapy to the cochlea [93]. It is possible that our definition of radiotherapy lacked sensitivity to detect effects. This meta-analysis included observational studies and clinical trials with heterogenous treatment regimens. Therefore, prevalence estimates stratified by certain risk factors are likely reflective of clinical decision-making rather than only mechanistic effects.

Previous research has indicated that both young and old age are risk factors for ototoxicity [6], [94]. In analyses evaluating prevalent hearing loss associated with exposure to cisplatin and/or carboplatin, children under 5 years had the lowest prevalence of hearing loss, likely because most of these records used carboplatin (i.e., the less ototoxic agent), which resulted in a lower pooled prevalence. However, after stratification by drug, children under 5 years treated with cisplatin-based regimens showed the highest prevalence of hearing loss, which is consistent with research indicating young age is a risk factor for ototoxicity. Notably, there was substantial variability (i.e., wide confidence intervals) in estimates focused on pediatric age groups, including the estimates focused on pediatric cancers and using pediatric ototoxicity grading scales. These estimates likely reflect differences in age distributions and treatment regimens of the records included in the meta-analysis. The age trends presented in this study are nonetheless important because they likely reflect standard clinical care for treatment of young children. For example, the age of study participants may influence choices regarding treatment, such as use of carboplatin instead of cisplatin and/or dosing.

The prevalence estimates reported in this study may also reflect systematic or non-systematic differences in measurement and classification of ototoxic hearing loss. Only 3 studies used OAE to detect ototoxic hearing loss and thus there is limited confidence in the high prevalence estimate given the small sample and high amount of variability. However, despite this substantial limitation, this estimate is consistent with evidence indicating that OAE are useful to detect early changes in hearing after exposure to ototoxic drugs [95]. In addition to OAE, other recommended tools for ototoxicity monitoring include high frequency audiometry (> 8000 Hz) in addition to the standard clinical audiogram (250–8000 Hz). High frequency audiometry and OAE are recommended because ototoxic damage develops first at the basal end of the cochlea, corresponding to the high frequencies, before progressing to frequencies most important for speech understanding [49]. Therefore, studies that included measurement of high frequency hearing loss likely reported a higher prevalence of ototoxic hearing loss. Measurement of high frequency hearing loss in children is less common as it more time consuming and less reliable, which may have contributed to a slight underestimate in children as compared to estimates for other age groups.

Differences in definitions and applications of ototoxicity grading scales may have resulted in systematic differences in prevalence estimates. Several studies have evaluated concordance among ototoxicity grading scales and have demonstrated differences in prevalence estimates and sensitivity to detect ototoxic hearing loss [92], [96], [97]. In this study, ASHA and CTCAE scales were applied to adult and pediatric populations, whereas Brock, Chang, SIOP and Muenster scales were designed exclusively for pediatric use. The differences in prevalence estimates across pediatric grading scales may, in part, reflect differences in study sample and treatment characteristics such as age distributions and treatment drug(s). However, some observed trends across pediatric grading scales are consistent with previous research. For example, studies have suggested the Brock scale may have reduced sensitivity to detect pediatric ototoxic hearing loss, which prompted development of the more clinically sensitive Chang scale [97]. In this study, we observed lower prevalence of ototoxic hearing loss in the Brock (vs Chang) scale. Because there is no true ‘gold standard’ definition of ototoxic hearing loss, it was not possible to determine whether prevalence measured with different scales resulted in a potential underestimate or overestimate of pooled prevalence of ototoxic hearing loss. While discussion of preferential use of scales is outside the scope of this study (interested readers should refer to previously mentioned reviews), ototoxicity research and clinical practice would benefit from international agreement on grading scales and hearing measurement protocols. This would promote uniformity in prevalence and incidence estimates and allow for comparisons across studies and clinical settings.

The global estimates of at-risk and exposed populations and hearing loss cases are crude yet were created with the best available global data on cancer incidence and treatment. These estimates accounted for global availability of chemotherapeutics but did not account for potential regional differences in patients’ treatment seeking, access or adherence. We attempted to capture exposure to cisplatin/carboplatin (rather than other chemotherapy drugs) by focusing on cancers primarily treated with platinum-based compounds. However, treatment regimens may vary for several reasons, including patient-level characteristics, treatment recommendations and/or regional differences in drug availability. These estimates do not capture off-label use of platinum-based compounds.

Hearing loss prevention

Promoting and prioritizing ototoxic hearing loss prevention involves use of primary, secondary and tertiary prevention strategies [26]. The goal of primary prevention is to prevent hearing loss onset. In some situations, primary prevention can be achieved via treatment modification (e.g., dosage modification or alternative pharmaceutical treatments). However, treatment modification is often not feasible given the high effectiveness of cisplatin and current lack of alternative drugs. Future research should focus on development or identification of less toxic treatment alternatives that do not compromise effectiveness. Another option is pharmaceutical prevention via otoprotectants (e.g., sodium thiosulfate, amifostine), which may be co-administered with cisplatin to inhibit cell death pathways or augment protective pathways [98]. Despite a large body of evidence on otoprotectants [4], [99], [100], some of which show promising effects in clinical trials [101], [102], most otoprotectants are not approved by international drug regulation agencies (e.g., food and drug administration) nor are they globally accessible. There is an urgent need for continued research to improve understanding on biological mechanisms of ototoxicity, which will progress the ability to identify and/or maximize effectiveness of pharmaceutical otoprotectants and to prioritize bringing these otoprotectants to market [103].

Until primary ototoxicity prevention is widespread, secondary and tertiary prevention should be prioritized. Secondary prevention occurs via serial ototoxicity monitoring, which focuses on early detection of hearing loss and when possible, prevention of its progression by monitoring hearing acuity from baseline to drug cessation. Detection of hearing loss may indicate treatment modification when possible. Otherwise, the burden of hearing loss can be reduced via counseling or further consultation with hearing health professionals, which often includes the promotion of tertiary prevention. Tertiary prevention can reduce negative impacts of hearing loss when ototoxicity is unavoidable. Aural rehabilitation aims to mitigate negative consequences of hearing loss and can occur via amplification (hearing aids, assistive listening devices, or cochlear implants). When amplification is unavailable, use of speech reading or other compensatory strategies, and/or sign language may be viable options for reducing hearing loss impacts. There is a need for context-specific and patient-centered guidance for ototoxicity management which should be embedded into existing systems of care. Priorities include development and assessment of rapidly implementable recommendations and toolkits for ototoxicity detection and management that are comprehensive yet adaptable to different settings. For example, evidence-based consensus is needed to identify tools and methods to detect ototoxicity that can be easily adapted for use in limited-resource settings, as well as tools for use in efficacy/effectiveness research that measure functional impacts of ototoxicity. Furthermore, guidance is needed on appropriate referral pathways for detection and management of ototoxicity as well as strategies for ototoxicity management for settings with limited hearing health care professionals, such as task sharing or shifting and/or telecommunications to expand workforce capacity. Some strategies for ototoxicity management in low- and middle-income countries have been previously presented [26], [104] and may be valuable in development of comprehensive guidelines and toolkits that can be adapted to specific contexts.

Cancer treatment may have lasting effects on quality of life and well-being for individuals across the age range and these effects may be particularly salient in individuals experiencing ototoxicity [105]. There are several scales that measure quality of life in cancer patients [106], [107]. In clinical practice, evaluating changes in quality of life over time may inform intervention focused on intensive ototoxicity management, as treating hearing loss may improve health-related quality of life in these individuals.

Strengths & limitations

To our knowledge, this review is the first to report pooled prevalence estimates of ototoxic hearing loss in cancer patients treated with cisplatin and/or carboplatin, and to estimate the global burden of ototoxic hearing loss associated with exposure to cisplatin and/or carboplatin. However, this meta-analysis is limited by the heterogeneity and lack of standardized research methodology of the studies included. Given the heterogeneity in measurement and reporting of ototoxic hearing loss or threshold shift (including differences in ototoxicity grading scales), it was not possible to evaluate hearing loss by the magnitude of audiometric shift or by degree or grade of hearing loss. Use of a binary definition of hearing loss may have reduced our ability to detect a dose-response relationship. We were not able to consider treatment duration or number of chemotherapy cycles as a predictor given lack of treatment details in most studies. Most studies were from American and European regions and thus may not be reflective of chemotherapeutic treatment in other regions. Age categories were not mutually exclusive because they were chosen based on the data available in individual studies. In general, studies evaluating cisplatin- or carboplatin-induced ototoxicity would benefit from provision of thorough descriptions of cancer treatment and the use of standardized methods of audiologic data collection and reporting to reduce study heterogeneity and maximize study comparability. Our global estimates of at-risk and exposed populations and hearing loss cases did not account for several potentially influential factors discussed above. Regional differences in timeliness to cancer treatment exist, and individuals residing in low- and middle-income countries may receive cancer treatment later than recommended [108]. However, delayed treatment in low- and middle-income countries (compared to US and UK data used in this study) could result in an underestimate of drug exposure and hearing loss cases.

Conclusions

Cisplatin use was associated with a high prevalence and carboplatin use was associated with a lower, but still substantial prevalence of ototoxic hearing loss. Global estimates suggested that approximately one million individuals are likely exposed to cisplatin and/or carboplatin chemotherapy annually, which would result in almost half a million cases of hearing loss cases per year. There is an urgent need to reduce impacts of ototoxicity in cancer patients. Currently, this can be partially achieved by implementing existing strategies focused on primary, secondary, and tertiary hearing loss prevention. Primary ototoxicity prevention via otoprotectants should be a research and policy priority.

CRediT authorship contribution statement

Lauren K. Dillard: Conceptualization, Methodology, Writing – original draft, Visualization. Lucero Lopez-Perez: Methodology, Formal analysis, Writing – review & editing. Ricardo X. Martinez: Methodology, Formal analysis, Writing – review & editing. Amanda M. Fullerton: Methodology, Writing – review & editing. Shelly Chadha: Conceptualization, Resources, Writing – review & editing. Catherine M. McMahon: Conceptualization, Methodology, Writing – review & editing.

Declaration of Competing Interest

None.