Effects of chemotherapy on the auditory system of children with cancer: a systematic literature review

Purpose: to identify and analyze the effects of chemotherapy on the auditory system of children and/or adolescents with cancer treated with cisplatin and carboplatin, assessed through standardized audiological procedures. Methods: studies in Brazilian Portuguese and in English were searched for, as avai-lable in the databases Science Direct, PubMed, LILACS, BIREME, Embase, SciELO, Web of Science and Cochrane. The descriptors were: Hearing Loss, Audiology, Child Cancer, Chemotherapy, and Child. Articles with levels 1 and 2 of scientific evidence, published in the last 20 years (1997 to 2017), were considered, of which the audiological results were analyzed, as well as the prevalence of hearing loss in children with cancer undergoing chemotherapy. Results: 3,625 articles were found, of which only 23 were selected for analysis in the present review. Studies have shown a high incidence of sensorineural hearing loss and decrease or even loss of otoacoustic emissions in children and adolescents with cancer, even after the first dose of chemotherapy drugs, with high frequencies being the most affected. Conclusion: there is evidence that both carboplatin and especially cisplatin from the first doses may impair the hearing of children and adolescents, mainly affecting the cochlear function, thus, the importance of long-term audiological monitoring.


INTRODUCTION
Child and adolescent cancer (involving children and adolescents between zero and 19 years old) is a disease with characteristics of its own, especially concerning histopathology and clinical behavior. In general, child and adolescent cancer presents short latency periods, grows faster, and is more aggressive. However, it responds better to treatment and has good prognosis 1 .
Chemotherapy is a mode of systemically treating the oncologic patient, more recent than surgery and radiotherapy, and it consists in administrating chemical substances, either alone or in combination, with the objective of destroying malignant neoplastic cells, while preserving the normal ones 2 .
The antineoplastic therapy includes the use of various drugs, among which are the alkylating agents, which are chemical substances capable of replacing a hydrogen atom by an alkyl radical. Platinum compounds, including cisplatin, carboplatin and oxaliplatin, can penetrate the cells by diffusion and modifying the DNA molecules, which are indispensable for the cell replication process 3 .
Among the many collateral effects resulting from chemotherapy, ototoxicity is one that may cause lesions to the inner ear structures, resulting in hearing loss, normally, sensorineural, bilateral and symmetric, affecting first the higher frequencies 4,5 . Producing free radicals with the use of antineoplastic therapeutic agents may change the cell wall and the cells' genetic material, including the cochlea's hair cells. Thus, basic mechanisms of these collateral effects involve the production of reactive oxygen species in the cochlea that induce these cells to death, making it impossible for the electrical signals to be adequately transmitted to the auditory nerve 6 .
It is known that platinum compound-induced hearing loss may vary from zero to 90.1% 7 . Furthermore, some studies have described that hearing loss may be triggered during the platinum-based therapy, or even develop after the treatment has finished 5,8,9 . Moreover, in the literature there are studies relating possible factors that contribute to the seriousness of the hearing loss. For instance, higher cisplatin dosages, as well as younger patients, can maximize the ototoxic effects of the platinum compound 10,11 . On the other hand, it is shown that even individual dosages can impair the auditory system 12 .
Although cisplatin ototoxicity has been carefully investigated, much less is known about ototoxicity through the screening of titles and summaries; after this step, the eligibility was conducted with the full reading of the texts, excluding the articles not meeting the previously established criteria; lastly, the eligible articles were included in the systematic review.

Data analysis
After finishing the collection from the databases, a table was filled out with the data from each study in order to exclude repeated articles. Two independent researchers conducted the analysis of the texts found, selecting the texts that met the inclusion criteria. For an initial filtering, the title of each article was read, and then, the summary of the remaining articles. After that, the articles that remained were read in full.
From the articles that were selected after full reading, the following data were identified for posterior analysis: number, gender and age of the participants; type of tumor; audiological procedures employed; drug(s) used, cycle and dose; hearing loss criteria; follow-up; main audiological findings; evidence of ototoxicity. Conflicts in the analysis of the studies were solved by discussion between the researchers.
For the assessment of risk of bias in each study, the Cochrane 19 tool was used, which encompasses the following criteria: randomization; allocation concealment; blinding of participants; blinding of outcome assessor; incomplete outcomes; selective outcome reports; and other sources of bias. Each individual criterion was considered as having low risk of bias, high risk of bias, or uncertain risk of bias (lack of information, or uncertainties referring to potential biases). The discrepancies between the authors were solved by consensus.

Results in the electronic databases
A total of 3,625 studies were found throughout the databases researched. Among these, 2,415 studies were repeated and so they were excluded.
The titles of the remaining 1,210 articles were read, leading to the exclusion of 622 due to their digressing from the topic approached in this review, so that 588 remained, whose summaries were read.
After the reading of the summaries, 186 articles of interest were left, which were fully read. Of these, 163 did not meet all the criteria for this research.

Selection criteria
The period selected for inclusion of studies in Portuguese or English was the last 20 years (1997 to 2017). The ones included, in general, were original articles dealing with the hearing of children and adolescents with cancer submitted to cisplatin and/or carboplatin chemotherapy treatment, with levels 1 and 2 of scientific evidence according to the Oxford Centre for Evidence-based Medicine 18 . These articles included: systematic review of randomized controlled clinical trials or of cohort studies, randomized controlled clinical trial with narrow confidence interval, "all or nothing" therapeutic results, cohort studies and therapeutic results observation. Articles with expert opinion, case reports, case-control studies and systematic literature reviews were excluded, as well as studies including the use of otoprotective agents.
The selection of the articles was carried out by two researchers. At first, the articles obtained through the search on the databases were identified; then, the articles were selected, excluding the duplicated ones and those not related to the defined descriptors, with the stages in the selection of the articles can be visualized in Figure 1. The main aspects of the selected studies are found in Figures 2 and 3     was noted in the criteria of randomization, allocation concealment, blinding of participants and blinding of outcome assessors, which may be justified by the type of population assessed and treatment conducted.
As for the criterion of incomplete outcomes, all the studies presented low risk of bias, since all data loss was justified, and the outcomes lost did not influence on the effect size observed. For the criterion selective outcome reports, five studies presented uncertain risk of bias, for the information was insufficient to enable it to be judged. Nine studies presented other sources of bias, the main one being the absence of information regarding the use of concomitant medications.

Analysis of the selected studies
There is in the literature a concern regarding the ototoxic effects in children with cancer submitted to platinum-based chemotherapy, since the administration of these substances can impair the functioning of the structures of the auditory system. Even though only 23 studies met all the inclusion criteria, it was observed in the first stages of the research a great number of articles verifying the chemotherapy effects on hearing.
Regarding the risk of bias (Figure 4), all the studies included presented methodological flaws in at least one criterion evaluated. In all the studies, a high risk of bias  assessment method had to be adapted to the child's response possibilities; therefore, for some of the studies, objective and/or subjective procedures had to be used 8,12,20,28,30,33,39 .
The acoustic immittance, though used in many of the studies 12,20,24,27,28,31,35 , was carried out mainly with the purpose of discarding possible middle ear alterations that could compromise the results of the other procedures, especially those of the otoacoustic emissions (OAE).
The OAE, besides being an objective method, an alternative to behavioral auditory assessment, had its inclusion justified by some authors for its potential in assessing outer hair cells function in patients submitted to chemotherapy, capable of providing early evidences of cochlear damage, even before the occurrence of alterations on auditory thresholds obtained through pure-tone audiometry 22,32,36,39 .
Some studies also included the brainstem auditory evoked potentials (BAEP) 8,12,20,28,30,33,39 ; however, the usefulness of this procedure was mostly in estimating the psychoacoustic threshold of babies and very small children, who are not able to respond to behavioral assessments. Of the seven studies that used the BAEP 8,12,20,21,28,33,39 , only one described the findings from such assessment 20 .
In addition to the diversity of procedures, there was also a diversity of criteria used for classifying hearing loss and/or detecting ototoxicity. Two studies considered as hearing loss auditory thresholds superior to 20 dB HL, without mentioning any other specific criterion 24,25 . Other studies also used 20 dB HL as hearing loss criterion, besides including specific criteria to identify the presence of ototoxicity, such as: Common Terminology Criteria for Adverse Events 26 ; Brock grading system 8,35 ; ASHA's criteria referring to ototoxicity 19,34,37 ; Muenster criteria 38 ; SIOP Boston ototoxicity scale 39 ; and a comparison between diverse criteria (ASHA, Brock, Muenster, POGT, BHL) 4,12,33 . On

Methodological characteristics
The sample number varied greatly, as there were studies that assessed only 10 individuals 31,32 , whereas another one assessed 451 individuals 38 (80.72 ± 109.99).
Regarding gender, only two studies 4,30 did not mention the distribution of the participants. Most of them included children and adolescents of both genders, similarly distributed (mean number of female individuals per study: 39.8 ± 57.6; and male individuals: 35.95 ± 55.4).
The mean age at diagnosis also varied considerably between the various studies, from seven months 28 to 12.3 years 22 .
Considering that the aim of this research was to verify the effects of cisplatin and carboplatin chemotherapy on the auditory system of children and/or adolescents with cancer, the selected studies tended to be longitudinal, assessing the individuals before, during and/or after chemotherapy; the follow-up after chemotherapy ranged from one month to 127 months. Only five studies performed the assessment exclusively post-chemotherapy treatment 4,22,27,29,36 .
Among the studies, different methods were employed to compose the audiological assessment. Since a wide age range was involved in the assessment, sometimes including very small children, the auditory the other hand, ten studies did not specify the criterion used to classify hearing loss and/or presence of ototoxicity 20,22,23,[27][28][29][30][31][32]36 .
The absence of this criterion for the classification of hearing loss due to ototoxicity may be justified by the variety of methods used in audiological assessment, since not carrying out the pure-tone audiometry makes it unfeasible to use many of these standardized criteria, as most of them are based on auditory thresholds and/ or on evolution throughout time.
A recent literature review on the subject 40 assessed the most used criteria for ototoxicity, discussing their benefits and limitations. The authors observed that diverse criteria are used in the various studies, including some verified in this study: ASHA, Brock, Muenster, SIOP, among others. The authors concluded that the different ototoxicity criteria result in different prevalence and degrees of hearing loss, which may result in different medical practice. Moreover, attention was called to the need of further research on the subject, emphasizing the importance of adopting a consensual criterion to follow up chemotherapy regimens, as well as the inclusion of high-frequency audiometry (>8 kHz) and/or auditory assessment through OAE.

Results of the audiological assessment and occurrence of ototoxicity
The methodological differences between the studies (different audiological procedures, time of assessment, as well as the use of different classification criteria) resulted in a great variability between the 23 studies. Despite this, in regard to prevalence or incidence of hearing loss, it can be concluded that most of the studies (82.6%) verified the occurrence of ototoxicity in children and adolescents submitted to platinum-based medication treatment.
Of the 19 studies that verified ototoxic effects during or after the treatment with the antineoplastic agents used, nine included only cisplatin, whereas the others studied cisplatin and/or carboplatin.
Regarding cisplatin, it was noted that auditory impairment could already be verified right after the first chemotherapy dose, with absence or decrease of the OAE in 7.6% of the individuals 24 , as well as significant changes of high-frequency auditory thresholds, and decrease in OAE, in approximately 50% of the individuals 21 .
Similarly, another study identified that 25% of the individuals already presented sensorineural hearing loss right after the first cycle of cisplatin (cumulative doses of 120 to 160 mg/m 2 ). Furthermore, 16.6% of the individuals developed hearing loss after the second cycle of cisplatin (cumulative doses of 200 to 320 mg/ m 2 ) and 8.3% of the individuals presented hearing loss after the third cycle of cisplatin (cumulative doses of 360 mg/m 2 ) 12 .
In addition to the immediate effect, other studies highlighted that the appearance of hearing loss may happen even after the cisplatin treatment has finished, justifying the importance of audiological monitoring even after the end of the treatment. One of the studies analyzed in this review observed that 26% of the individuals presented high-frequency sensorineural hearing loss, whose appearance began from one to 60 months after the end of the chemotherapy, regardless of individual or cumulative doses of cisplatin 20 . Another three studies also verified percentages of alterations ranging from 24 to 100%, from 6 to 44 months after, the main hearing loss being the sensorineural, bilateral and symmetric 22,23,[25][26][27] .
As for the studies that assessed carboplatin and/ or cisplatin, all of them verified ototoxic effects of the drug(s), there being found prevalence of auditory alterations ranging from 17 to 94% 4,8,[33][34][35][36][37][38][39] , including decreased amplitude of OAE and/or hearing loss of various degrees, involving mainly the high frequencies.
Two of these studies also specified the prevalence of alterations related to each drug 38,39 . The first study verified 45% of alterations in individuals treated only with cisplatin, 17% of those treated only with carboplatin, and 75% of the ones treated with both agents 38 . As for the second study, there was a prevalence of 25% those treated only with cisplatin; 19% of those treated only with carboplatin, and 35% of the individuals that underwent combined treatment (cisplatin and carboplatin) 39 . Despite the differences in prevalence, it was observed that carboplatin presented a less ototoxic effect, followed by cisplatin, whereas the combined use of the two substances presented greater ototoxic power due to its synergistic effect. Not all the studies described systematically the hearing loss characteristics in children with cancer, due to the different methodological procedures/ classification criteria employed, as previously mentioned. Nevertheless, in the studies that described these variables, it was noted that the audiological profiles were very similar after the chemotherapy treatment. There was a prevalence of sensorineural hearing loss 12,20,21,27,39 , of mild to moderate degree, bilateral and symmetric in most of the cases 21,27 , affecting mainly the high frequencies 12,[20][21][22][23]25,36 , although some studies also observed the presence of severe degree of hearing loss in some patients 8,26,36,39 .
The high-frequency audiometry analysis (between 9 and 16 kHz), which was used in only one study, showed 94.1% of bilateral ototoxicity in this frequency range 33 . Likewise, the studies revealed a decrease in the amplitude of responses to the OAE, both by transient stimuli and distortion product, the high frequencies being more sensitive to ototoxic action 21,22,24,31,33 . The analysis of the responses to the OAE was compatible with the results presented in the pure-tone audiometry 20,24 , which highlights that the basal regions of the cochlea are the most affected 33 .
It is described in the literature that the platinumbased compounds affect first the outer hair cells at the basal region of the cochlea, which is the reason for it initially affecting frequencies over 4,000 Hz 24,31 .
Regarding the brainstem auditory pathway analysis, the only study that described this finding (registered in seven patients) observed an increase of the wave V latency at the intensity of 20 dBnHL in an individual, suggesting a cochlear dysfunction, and an increase of the interpeak in two individuals, suggesting a neuralconduction slowdown along the auditory nerve. This is the only study of those selected for this review that highlighted the possibility of retrocochlear alteration as a result of the cisplatin 20 .
Hence, future studies are necessary to consider the auditory pathway integrity analysis through the BAEP, and to assess the possibility of neurotoxicity resulting from the use of chemotherapy medications.
Five studies assessed only individuals with retinoblastoma, treated with carboplatin alone [28][29][30][31][32] . Of these, only one study observed decrease in DPOAE at the frequency of 7,996 Hz in 40% of the individuals, which indicates ototoxicity 31 ; in the other four studies, no ototoxic effects resulting from chemotherapy were observed [28][29][30]32 . Some other studies have reported that carboplatin alone may not cause alterations in the auditory pathways 8,36 ; nonetheless, cumulative doses are potentially ototoxic 31 .
When observing the cumulative doses in these five studies [28][29][30][31][32] , administration of high doses was not observed in the study that observed ototoxicity 31 in comparison to the other ones [28][29][30]32 . Therefore, though countless other factors may have interfered in these results, it cannot be discarded the influence that even small doses may have on impairing the functioning of the cochlear hair cells.
In the case of children with retinoblastoma, one of the studies highlighted the priority of early detection of ototoxicity, since many children are diagnosed after presenting visual impairment. If the ototoxic potential is identified in due time, the dosage of carboplatin and/or other drugs can be changed to avoid further cochlear deterioration, considering that the negative impacts of hearing loss can be even more aggressive in a child with visual impairment 28 .
Among the most highlighted factors of great risk of developing platinum-induced hearing loss, the diagnosis of osteosarcoma 26 and the cumulative cisplatin dosage can be cited 8,20,22,27 , being this greater than 400 mg/m² 26 , though dosages superior to 200 mg/ m² have already shown to be ototoxic 12 . The influence of age has also been mentioned, as well as the concomitant use of furosemide 38 .
Considering this, it is fundamental that audiological monitoring be carried out in this population, before, during and after the treatment, emphasizing the need of long-term follow-up, as well.
Auditory monitoring can aid in decision-making related to the treatment itself and/or in determining the appropriate clinical interventions for each case, seeking to minimize possible negative impacts of hearing loss for the language, schooling and quality of life of these individuals.
Regarding the evaluative procedures, the instrument should be adapted to child's response possibilities. Nevertheless, whenever possible, conducting pure-tone threshold audiometry is desirable, as most of the time the auditory thresholds are gold standard for ototoxicity monitoring in the many criteria available.
Furthermore, the inclusion of the OAE and/or the high-frequency audiometry (>8 kHz) is fundamental, since these are clinical resources capable of showing possible auditory impairments even before any alteration is observed on the conventional pure-tone audiometry, due to the damage beginning at the basal region of the cochlea. In this study, the possible role of genetics on determining ototoxicity was not discussed, neither was the use of otoprotective agents for hearing loss in this population. Future studies on the subject should be conducted, deepening knowledge about these specific issues.