Amifostine does not prevent platinum-induced hearing loss associated with the treatment of children with hepatoblastoma

A report of the Intergroup Hepatoblastoma Study P9645 as a part of the Children's Oncology Group


  • See editorial on pages 5623–6, this issue.

  • We are extremely grateful for the hard work and diligence of the clinical research associates of our member institutions for data collection and to the Children' Oncology Group Publications Office for editorial assistance.



The current study was conducted to determine whether amifostine is effective in reducing the toxicities associated with the administration of platinum-containing regimens in children with hepatoblastoma (HB).


Patients were enrolled on P9645 beginning in March of 1999. Patients who had stage I/II disease received treatment with 4 cycles of combined cisplatin, 5-fluorouracil, and vincristine (C5V) with or without amifostine. Patients who had stage III/IV disease were randomized to receive treatment with 6 cycles of either C5V with or without amifostine or carboplatin alternating with cisplatin (CC) with or without amifostine. Patients who were randomized to receive amifostine were given a dose of 740 mg/m2 intravenously over 15 minutes before each administration of a platinum agent.


Eighty-two patients were considered in a special interim analysis of the incidence of toxicity. The disease outcome for patients who received amifostine was similar to the outcome for patients who did not receive amifostine (P = .22). The incidence of significant hearing loss (>40 dB) was similar for patients who did or did not receive amifostine (38% [14 of 37 patients] vs 38% [17 of 45 patients], respectively; P = .68). There were no differences in the incidence of renal or bone marrow toxicities evaluated. Patients who received amifostine had a higher incidence of hypocalcemia (5% vs 0.5%; P = .00006).


Amifostine in the doses and schedule used in this study failed to significantly reduce the incidence of platinum-induced toxicities in patients with HB. Cancer 2009. © 2009 American Cancer Society.

Cisplatin is 1 of the most effective agents against hepatoblastoma (HB), and the addition of this agent to HB treatment protocols in the mid-1980s by both the Pediatric Oncology Group (POG) and the Children's Cancer Group (CCG) resulted in dramatically improved survival rates for these children.1, 2 Cisplatin, however, is associated with significant toxicity, including hearing loss, myelosuppression, and nephrotoxicity.3 Carboplatin also has been used in an effort to provide antitumor efficacy with another platinum agent that has less potential toxicity.4, 5

Amifostine is an agent that was developed to selectively protect normal tissues from the cytotoxic effects of radiation and chemotherapy. Adult studies have demonstrated the ability of amifostine to reduce toxicity without altering outcome or therapeutic efficacy.6, 7 The Children's Oncology Group (COG) Study P9645 was designed to determine whether the intensification of platinum administered therapy could improve survival in newly diagnosed patients with HB and to determine whether amifostine could prevent platinum-associated toxicities, including hearing loss, renal dysfunction, and myelosuppression. This report describes the results of amifostine use and its effects in reducing ototoxicity associated with the administration of platinum-containing regimens in children with HB.



The Pediatric Intergroup Hepatoblastoma Study P9645 opened in March 1999. The study was designed as a factorial randomization for patients with stage III or IV disease. The details of the study are illustrated in Figure 1. Patients were eligible for study entry if they were aged <21 years of age at diagnosis; had biopsy-proven, previously untreated HB; and had adequate organ function documented at the time of study enrollment. The protocol was approved by the National Cancer Institute and by the individual institutional review board requirements of the participating COG-affiliated institutions. Informed consent was obtained for all patients before study entry. Randomization to carboplatin/cisplatin (CC) was suspended in January 2002, because the projected improvement in long-term outcome associated with CC was excluded statistically as a possible outcome of this trial.8 Subsequently, all patients were assigned to receive combined cisplatin, 5-fluorouracil, and vincristine (C5V) with or without amifostine.

Figure 1.

This chart illustrates the treatment schema for patients on Children's Oncology Group study P9645. CDDP indicates cisplatin; 5-FU, 5-fluorouracil; VCR, vincristine; AMI, amifostine; CARBO, carboplatin.


Patients were treated according to the schema illustrated in Figure 1. Patients with tumors of pure fetal histology were observed after undergoing surgical resection without further therapy. Patients who had stage I (nonpure fetal histology) and stage II disease received 4 cycles of C5V with or without amifostine. Patients who had stage III and stage IV disease were randomized to receive 6 cycles of either C5V or CC with or without amifostine. Each cycle of C5V consisted of intravenous (iv) CDDP (100 mg/m2 or 3 mg/kg for patients <1 year) administered over 4 hours, followed by iv hydration on Day 1, 5-FU (600 mg/m2 iv push) on Day 3 and VCR (1.5 mg/m2 iv push) on Days 3, 10, and 17. Each cycle of CC was comprised of carboplatin at a dose of 700 mg/m2 given iv over 1 hour (23 mg/kg for patients who weighed <10 kg) on Day 0 (560 mg/m2 or 18.5 mg/kg for patients who weighed <10 kg after 2 cycles) followed by cisplatin at a dose of 100 mg/m2 or 3 mg/kg for patients aged <1 year on Day 14 administered as described above. Granulocyte–colony-stimulating factor was given after each CC cycle. Amifostine was given at a dose of 740 mg/m2 over 15 minutes immediately before cisplatin. The total intended platinum dose for patients with stage I (nonpure fetal) and stage II disease was 400 mg/m2 (12 mg/kg for patients aged <1 year) of cisplatin. The total intended platinum dose for patients with stage III and IV disease was 600 mg/m2 of cisplatin (18 mg/kg for patients aged <1 year) and 3640 mg/m2 of carboplatin (120 mg/kg for patients who weighed <10 kg). Patients who had a glomerular filtration rate with a glomerular filtration rate <100 mL/minute/1.73 m2 were supposed to have their carboplatin dose calculated using the Calvert formula to achieve an area under the curve of 6. Patients were re-evaluated at the end of the initial chemotherapy phase of 4 cycles. Patients who had unresectable disease at that time were considered treatment failures, whereas patients who underwent tumor resection received 2 more cycles of the therapy to which they initially had been randomized.

Evaluation of Therapy

Physical examination, blood counts, serum alpha-fetoprotein (AFP) levels, appropriate imaging studies (including computed tomography scans of the chest and either a computed tomography scan or a magnetic resonance image of the abdomen), and assessment of audiologic function (audiogram or auditory brainstem responses) were performed before therapy. Subsequent examinations and AFP assays were done before every additional cycle of chemotherapy. AFP levels were supposed to be monitored monthly for 6 months, then every 2 months until 2 years off therapy, then every 3 months until 4 years off therapy, and then yearly. Imaging studies were repeated after Cycles 2, 4, and 6 and then at 2 months, 4 months, 6 months, 12 months, 18 months, and 24 months off therapy. Audiologic examination was repeated after the fourth cycle of chemotherapy, at the end of therapy, and yearly after therapy.

Toxicity of Treatment

The individual incidents of various toxicities were graded on a scale from 1 to 4 according to National Cancer Institute Common Toxicity Criteria version 2.0 (CTC). Limits for toxicity grades were dependent on both patient age and the particular organ system involved. The specific toxicity scales used in this study are available from the National Cancer Institute Cancer Therapy Evaluation Program website. Patients on regimen C5V who had delays in therapy >1 week or who experienced fever and neutropenia were allowed to receive granulocyte–colony-stimulating factor (5 μg/kg daily) beginning 24 hours after the 5-fluorouracil dose. Patients who experienced grade 4 neurotoxicity had their vincristine withheld until symptoms resolved; then, their therapy was reinstituted at half the initial dose. No specific modifications in therapy were made for any decrease in hearing.

Statistical Design and Analysis

Study design

Patients were randomized after the initial surgical procedure. Patients were stratified according to disease stage (stage I or II vs stage III or IV). Patients who had stage I/II disease were assigned randomly to receive or not receive amifostine. Patients who had stage III/IV disease were assigned randomly to receive C5V or CC and to receive or not receive amifostine. The primary outcome measure for the effect of chemotherapy and amifostine randomization was event-free survival. The intention of the study was to enroll patients for 5.5 years and to follow the last patient for 3 years after enrollment.

In November 2003, an analysis of toxicity was performed using toxicity data provided by the routine data collection done by institutional investigators according to the CTC criteria. At that time, the rate of grade 3 or 4 ototoxicity according to institutional reporting was 4%. Data from other authors9, 10 using information derived from expert review of serial audiograms demonstrated that quantifying ototoxicity from the CTC criteria substantially underestimated the true incidence of significant hearing loss. In response to these observations, we conducted a retrospective assessment of hearing loss in a subgroup of patients enrolled on the study (the “evaluation cohort”). A blinded, detailed review of ototoxicity using results from audiograms that were obtained as part of toxicity evaluations, as directed by the protocol, was performed by 1 of the authors (K.W.C.), who compared auditory brainstem responses (ABRs) and/or audiogram data (“hearing evaluation”) obtained before the start of treatment with data obtained after all therapy was completed. The evaluation cohort was comprised of any eligible patient who was enrolled on P9645 up to March 2003 who received cisplatin and who 1) had completed protocol therapy as of March 1, 2003; 2) who was alive; and 3) who had not experienced disease progression or a second malignant neoplasm while on protocol therapy. All data current to March 31, 2003 were used to establish which patients fit the criteria for inclusion in the evaluation cohort.

If >1 post-therapy hearing evaluation was available, then the evaluation that was done closest to March 31, 2003 was used. Modified Brock criteria11 (Table 1) were used to assign a rating of hearing loss for each ear.

Table 1. Modified Brock Criteria for the Classification of Hearing Loss
Grade of Hearing LossDescriptionPotential Clinical Effects on Hearing
0≤20 dB at 1, 2, and 4 kHzNone
1a>40 dB at any frequency from 6 kHz to 12 kHzMeasurable
1b>20 dB but ≤ 40 dB at any frequency from 3 kHz to 5 kHzMeasurable
2a>40 dB at any frequency from 3 kHz to 5 kHzNoticeable
2b>20 dB but ≤40 dB at 2 kHzNoticeable
3>40 dB at 2 kHzCorrectable with hearing aids
4>40 dB at 1 kHzSpeech comprehension deficits even with hearing aids

Analytic methods

An analysis was conducted using the lowest grade of hearing loss observed in either ear (“best ear”). The randomized assignment to receive amifostine was stratified by disease stage, and most patients in Stratum 2 also were randomized to receive 1 of 2 chemotherapy regimens. To account for these stratification factors, a log-linear model was used to assess whether significant hearing loss, defined as the presence of grade ≥2A hearing loss in the patient's best ear, was associated with the randomized amifostine assignment after adjustment for stage (stages I and II vs stages III and IV) or treatment regimen (CC vs C5V).12

Other toxicities

Each course of protocol therapy that was received by a randomized patient who had enrolled by the cutoff date was evaluated by the institutional investigator for the presence of all toxicities and was reported using CTC version 2 criteria. The incidence of each type of toxicity was calculated as the number of courses during which the toxicity was noted divided by the number of courses administered. These rates were compared across randomized regimens using the exact conditional test of proportions.13

Outcome definitions

To assess the relation between amifostine and outcome, patients who were randomized and who enrolled by the cutoff date were considered. Event-free survival was defined as the period from the date chemotherapy was started until evidence of an event (progressive disease, death, diagnosis of a second malignant neoplasm) or date of last contact, whichever occurred first. Patients who did not experience an event were censored on the date of last contact.

Life-table estimates were calculated according to the method of Kaplan and Meier,14 and the standard deviation of the Kaplan-Meier estimate of the survivor function at selected points was calculated using the Greenwood formula.14 For treatment comparisons, outcome was assigned to the randomized treatment, regardless of the therapy received.


Two hundred eighty-nine patients were enrolled on P9645. A detailed description of the results of this trial are published elsewhere.15 Of this total, 120 patients comprised the evaluation cohort, as described earlier (Fig. 2). Audiologic data were interpreted by a pediatric otologist (K.W.C.) and were categorized as evaluable if there was enough information to determine accurately whether or not ototoxicity had occurred. Eighty-two patients had sufficient audiologic data that made them fully evaluable for an analysis of ototoxicity. The demographic and treatment data of the evaluation cohort are detailed in Table 2.

Figure 2.

The evaluation cohort is shown. PFH indicates pure fetal histology.

Table 2. Characteristics of the 82 Patients Evaluable for Ototoxicity Treated on Children's Oncology Group Study P9645 According to Randomized Treatment Regimen
VariableNo. of Patients (%)Total
  1. CARBO indicates carboplatin; CDDP, cisplatin; 5-FU, 5-fluorouracil; VCR, vincristine.

Median age at enrollment [range], y1 [0-11]1 [0-6] 
 Boy6 (14)38 (86)44 (54)
 Girl12 (32)26 (68)38 (46)
 I/II0 (0)21 (100)21 (26)
 III/IV18 (30)43 (70)61 (74)
Randomized to receive amifostine
 Yes7 (19)30 (81)37 (45)
 No11 (24)34 (76)45 (55)

Amifostine did not prevent ototoxicity. Thirty-one of 82 evaluable patients (38%) had noticeable hearing loss (grade ≥ 2A). Two of 21 patients with stage I/II disease (10%) had grade 2A hearing loss, whereas 29 of 61 patients with stage III/IV disease (48%) had grade ≥2A hearing loss (Table 3). Six patients (7%) had correctable hearing deficits with hearing aids, and 2 patients (2%) had hearing deficits that could not be corrected completely even with hearing aids. There was no relation between noticeable hearing loss and amifostine assignment after adjustment for disease stage and chemotherapy treatment arm (P = .68) (Table 4). Specifically, amifostine did not prevent hearing loss of any grade. Chemotherapy assignment was not related to hearing loss (P = .75), but patients who had stage III/IV disease were more likely to have experienced hearing loss than patients who had stage I/II disease (P = .002). Patients with stage III/IV disease were to receive 2 more cycles of chemotherapy than patients with stage I/II disease.

Table 3. Hearing Loss Among Patients With Hepatoblastoma Treated on Children's Oncology Group Study P9645 According to Stage of Disease
Grade of Hearing LossNo. of Patients
Stage I/IIStage III/IVTotal
Table 4. The Effect of Amifostine on Hearing Loss Among Patients With Hepatoblastoma Treated on Children's Oncology Group Study P9645 According to Treatment Regimen*
Grade of Hearing LossNo. of Patients
C5VC5V/AmifostineCARBO/CDDPCARBO/CDDP AmifostineTotal
  • CV5 indicates combined cisplatin, 5-fluorouracil, and vincristine; CARBO, carboplatin; CDDP, cisplatin.

  • *

    P = .68 in a comparison between patients who were randomized to receive amifostine and patients who were randomized not to receive amifostine.


Of the 38 patients who were excluded from the audiologic analysis, 15 did not have a post-treatment evaluation, and 11 did not have data of sufficient quality to evaluate. Twelve patients had detectable hearing loss but had conductive rather than sensorineural hearing loss, which prevented attribution of the loss to protocol therapy and, thus, made them inevaluable for therapy-related ototoxicity.

The vast majority of patients had behavioral audiometric data. In total, 44 patients had hearing evaluated by ABR. However, only tone-burst threshold ABRs were considered evaluable and were performed in 11 patients; whereas 25 patients were not included in the analysis, because they were evaluated using nonfrequency-specific, click-evoked ABRs, which do not specifically test for high frequency hearing loss from ototoxicity. Eight patients who had nonthreshold ABRs, which typically are used to test neural latencies, also were excluded from the analysis as invalid clinical assessments of ototoxicity.

The frequencies of renal toxicities (proteinuria, hemoglobinuria, hypokalemia, hyponatremia, and hypomagnesiumia) and bone marrow toxicities (need for transfusion, thrombocytopenia, neutropenia, and anemia) were not different for patients who did or did not receive amifostine (Table 5). However, the use of amifostine was associated with a higher incidence of hypocalcemia (5% vs 0.5%; P = .00006). The use of amifostine was not related significantly to the risk of an adverse event (Fig. 3).

Figure 3.

Event-free survival of 184 patients who enrolled before March 31, 2003 and received chemotherapy by stratum of amifostine (AMI) randomization is shown. The adjusted P value for the risk of an adverse event associated with amifostine was P = .22.

Table 5. Treatment-Associated Toxicity Among Patients With Hepatoblastoma Treated on Children's Oncology Group Study P9645
Type of Grade 3 or 4 ToxicityNo. of Courses With Toxicity Present (%)P
AmifostineNo Amifostine
  • *

    Hemoglobin level <8 g/dL, a neutrophil count <1000/mm3, and a platelet count <50,000/mm3.

Hemoglobin*95 (27)99 (25).62
Neutropenia*165 (47)185 (47).94
Thrombocytopenia*46 (13)45 (11).58
Platelet transfusion requirements27 (8)26 (7).67
Packed erythrocyte transfusion requirements80 (23)97 (25).55
Hypocalcemia19 (5)2 (0.5).00006
Proteinuria0 (0)0 (0)1.0


Previous data suggested that amifostine could provide protection against treatment-related toxicities associated with the administration of cisplatin and provided the background rationale for the current trial.6, 7, 16, 17 This prospective, randomized trial failed to yield any evidence that amifostine could protect any degree of hearing loss, because 38% of patients experienced noticeable hearing loss, and 57% of patients experienced any measurable decrement in hearing. This is important, in that even mild hearing loss may have a significant impact on children with HB who are at a crucial age of cognitive development. These results are similar to those reported in several other pediatric series. A report by Marina et al revealed significant hearing loss in 18 of 24 patients with germ cell tumors who received high-dose cisplatin and amifostine.9 Fisher et al administered amifostine before cisplatin to 11 patients with medulloblastoma, and ototoxicity developed in 7 of 9 evaluable patients.18 Petrilli and coworkers randomized 39 patients with osteosarcoma to receive amifostine before cisplatin therapy, and observed no difference in the incidence of hearing toxicity.19 However, a recent publication by Fouladi et al described children with average-risk medulloblastoma who had a reduced risk of hearing loss in the group of patients who received amifostine.20 It is interesting to note that all of those patients had all been irradiated previously and had received amifostine both before and in the middle of the cisplatin infusion. The disparity in the results from these trials and others may be explained by differences in several factors, including patient age, individual and cumulative platinum dosage, timing and dosage of amifostine, tumor site, and potential interaction with other treatment modalities. Other agents, such as sodium thiosuflate and n-acetylcysteine, have been and continue to be evaluated as potential otoprotective compounds.

Historic data demonstrate that anywhere from 5% to 75% of patients may experience hearing loss after treatment with cisplatin.9, 10 Audiologic toxicity collected using the usual clinical trials mechanism, by the submission of grading information by the institutional investigator using CTC criteria, underestimated the true incidence of noticeable significant hearing loss (4% vs 38%) in this population, as demonstrated in the current analysis. In addition, although the analysis of audiologic data was possible only for 82 of 120 (68%) of potentially evaluable patients because of incomplete data, the evaluated cohort is a representative group that justifies the conclusion that amifostine was not otoprotective for this group of patients. Specific guidelines on how to administer audiograms or Brainstem Auditory Evoked Response tests were not part of the protocol.

Therefore, future pediatric trials that attempt to evaluate ototoxicity must be written with more stringent criteria for exactly which tests should be used and exactly how they should be performed. A separate article is in preparation as part of this endeavor that will detail pitfalls in data collection from this and other cooperative group studies and will make recommendations for more specific criteria to be included in future protocols. From the current experience, we suggest that the inclusion of otolaryngologists, audiologists, or others with the appropriate training and expertise is crucial in both study design and analysis so that the resulting data are both valid and reliable.

In the current study, the cytoprotective agent amifostine was administered in an effort to decrease platinum-induced toxicity without adversely affecting event-free or overall survival. However, in the doses and schedule used, amifostine did not protect against the ototoxicity associated with platinum agents. In addition, renal and bone marrow toxicity also were not ameliorated by the use of amifostine. The use of amifostine concomitantly with platinum therapy did not have an impact on the outcomes of children with HB (Fig. 3). Amifostine was associated with a much higher incidence of hypocalcemia, which is a well known side effect of this agent.


We are grateful to B. Greffe, L. C. Bowman, K. Newman, R. Womer, and R. P. Castleberry for their contributions.

Conflict of Interest Disclosures

Supported by U10 CA98543 & U10 CA98413 (Children's Oncology Group) and U10CA029139 (Pediatric Oncology Group) from the National Cancer Institute, National Institutes of Health, Bethesda, Md. Dr. Zappo has acted as a consultant or in an advisory role for Ziopharm.