Ototoxicity of Long-Term α-Difluoromethylornithine for Skin Cancer Prevention

Abstract

Objective.

Evaluate the effects of α-difluoromethylornithine (DFMO) on hearing thresholds as part of a randomized, double-blind, placebo-controlled trial.

Methods.

Subjects were randomized and assigned to the control (placebo) or experimental (DFMO) group. DFMO or placebo were administered orally (500 mg/m²/day) for up to 5 years.

Results.

Subjects taking DFMO had, on average, increased hearing thresholds from baseline across the frequency range compared to subjects in the control group. Statistical analysis revealed this was significant in the lower frequency range.

Conclusions.

This randomized controlled trial revealed the presence of increased hearing thresholds associated with long-term DFMO use. As a whole, DFMO may help prevent and treat certain types of cancers; however, it can result in some degree of hearing loss even when administered at low doses. This study further highlights the importance of closely monitoring hearing thresholds in subjects taking DFMO.

Introduction:

α-difluoromethylornithine (DFMO) inhibits polyamine synthesis, which have been shown to play a role in cellular proliferation and carcinogenesis. Inhibition of polyamine synthesis using DFMO, alone or in combination with other agents, has long been investigated as a possible chemopreventative and treatment avenue for various types of cancers.^1–22 Numerous clinical trials have evaluated the effects of DFMO as a chemopreventative agent^10–14,20 or treatment^15–19 for certain types of cancers. While the results from these trials are encouraging, a number of these reports have found hearing loss as the dose limiting effect of DFMO in their study.^{12–19,23,24} Fortunately, hearing loss associated with DFMO has been reversible in all reports with the exception of one.²⁵

A number of studies have evaluated the dose and ototoxic effects of DFMO on hearing thresholds^{11–19,23,24,26–28}; however, none have investigated the long-term effects of daily DFMO administration. In this phase III randomized, double-blind, placebo-controlled trial, we compared the effects of prolonged administration of DFMO and placebo for up to 5 years. 291 participants with a prior history of non-melanoma skin cancer (NMSC) were given an oral dose (500 mg/m²/day) of DFMO or placebo. Analysis was carried out to determine if low-dose administration of DFMO for prolonged periods of time resulted in ototoxicity, and if so, to what extent. Details and analysis of the potential of DFMO to prevent new NMSCs in these subjects was previously reported, with mild ototoxicity described.²⁹ Herein, we sought to further evaluate the ototoxic effects of DFMO on hearing thresholds.

Materials and Methods:

Study Design:

334 subjects were initially enrolled in the study (Fig. 1). To improve compliance, a one-month placebo run-in phase was completed for all qualifying participants. In this run-in phase, patients were given a placebo drug for 4 weeks to assess for compliance. Patients who were compliant, having taken more than 80% of the placebo and returned to the clinic, were then randomized to receive DFMO or placebo for up to 5 years. 291 individuals met the minimum requirements of the run-in phase and were subsequently randomized to the experimental or control group. Baseline hearing thresholds were established for all participants with the exception of one. Subjects were seen every 6 months for follow-up and compliance assessment.

Figure 1.

Study design and subject characteristics. Of the 334 subjects initially enrolled in the study, 291 subjects were randomized to the placebo or experimental group. Hearing loss data was available on 250 subjects. DFMO - α-difluoromethylornithine. BSA - body surface area.

Participants:

Participation in the clinical trial was open to individuals in the hospitals and clinics surrounding the Madison, Wisconsin area. All participants in this study gave their informed consent per institutional, state and federal standards, and the study protocol was approved by the University of Wisconsin Health Sciences Institutional Review Board and registered in ClinicalTrials.gov (NCT00005884).

Eligible participants were individuals: 1) with a prior history of basal or squamous cell carcinoma (stage 0–2); 2) older than 21 years of age; 3) with a status of 0 or 1 on the Eastern Cooperative Oncology Group performance status; 4) > 4 weeks from any major surgery or chemotherapy, radiotherapy or hormonal therapy for cancer; and 5) with adequate organ function defined as having a WBC of >=3,500/mm^3, platelets of >=100,000 mm^3, hemoglobin of >=11.0 g/dL, bilirubin of <=2 mg/dL, aspartate aminotransferase of <3x normal, and creatine of < 2 mg/dL. In addition, participants must not have used topical medications for the treatment of skin cancer for a minimum of four weeks prior to beginning study treatment.

Individuals with a pre-existing hearing loss (defined as significant clinical hearing loss or requiring the use of a hearing aid) were excluded from the trial, in addition to pregnant or lactating females. Women of childbearing age underwent a pregnancy test prior to beginning treatment in addition to agreeing to the use of an adequate form of contraception while participating in the study. Other exclusion criteria included individuals: 1) with a family history of early retinal blindness or ornithine diaminotransferase deficiency; 2) who were bone marrow transplant recipients; 3) receiving cytotoxic chemotherapy, tamoxifen or hormonal therapy for cancer treatment/prophylaxis; and 4) taking anti-seizure medication or corticosteroids.

Subject characteristics for the subgroup of interest are summarized in Fig. 1. The mean age at enrollment was 60.7 years of age, with a median age of 61.5 years. 154 participants were male and 96 participants female. Overall compliance with the study drug was exceptional, with a 93.4±11.7% mean (±S.D.) or a median of 98.5% for the DFMO group and 94.1±9.8% mean (±S.D.) or a median of 98.2% for the placebo group.

Audiometry:

Audiometry was based on the recommendations for cancer prevention trials.³⁰ Tympanometry and audiometric behavioral thresholds were obtained at octave intervals between 250 and 8,000 Hz prior to the start of treatment. Audiograms were performed at 1 year and at the end of the study. Audiometric data was measured at follow-up appointments only if a change in hearing was suspected by the participant. If a significant threshold shift ≥ 15 dB at 2 or more frequencies occurred, treatment was suspended. If toxicity resolved within 4 weeks, where hearing thresholds returned to baseline levels, participants were considered for re-treatment at half the original dose. If toxicity reoccurred, the treatment was discontinued permanently.

Study Treatment:

DFMO and placebo used in this study were supplied by ILEX™ Oncology Services, Inc. All staff and participants were blinded to the study medication, with the exception of the dispensing pharmacists and statisticians.

Statistical Considerations:

Behavioral thresholds were averaged between the two ears for each audiogram consistent with other studies.^11,31–33 Changes in behavioral thresholds were compared to the baseline audiogram obtained prior to the start of treatment. Two definitions of hearing loss were used for assessment purposes. A type I hearing loss (definition A) is defined as a ≥ 30 dB air conduction threshold change in the summation of three adjacent frequencies. A type II hearing loss (definition B) is defined as an air conduction threshold shift by ≥ 15 dB at two or more frequencies. Comparison of hearing loss between the placebo and DFMO groups were made using chi-square test. In order to examine the effects of age and gender on the effect of treatment on hearing loss, relative risk regression models were fit with type I and II hearing loss as dependent variables, examining gender and age as possible effect modifiers.

Results:

Type I and Type II Hearing Loss

291 subjects were randomized to the placebo or DFMO group. Hearing loss data was available on 250 subjects; 117 in the DFMO group and 133 in the placebo group (Fig. 1). Type I and type II hearing losses were observed in both the DFMO and placebo group (Table 1). In total, 47 subjects experienced either a type I and/or a type II hearing loss confirmed by audiometric testing during the course of the study; 28 (24%) in the DFMO group and 19 (14%) in the placebo group (p=0.051). Type I hearing loss was experienced by 45 individuals; 27 (23%) in the DFMO group and 18 (14%) in the placebo group (p=0.05). Type II hearing loss was experienced by 24 participants; 14 (12%) in the DFMO group and 10 (8%) in the placebo group (p=0.23). Both types of hearing loss were experienced by 22 participants; 13 (11%) in the DFMO group and 9 (7%) in the placebo group (p=0.23).

Table 1.

Hearing Loss for Each Group (placebo vs. α-difluoromethylornithine (DFMO)).

Hearing Loss	Placebo (n=133)	DFMO (n=117)	p-value
Type I and II	9 (7%)	13 (11%)	0.230
Type I	18 (14%)	27 (23%)	0.050
Type II	10 (8%)	14 (12%)	0.230
Type I or II	19 (14%)	28 (24%)	0.051

Age was a predictor of hearing loss for type I and trended for type II hearing loss (p=0.003 and p=0.054, respectively); treatment showed a trend for type I hearing loss (p=0.054), but not for type II (p=0.24), consistent with unadjusted comparisons. In a multivariate regression for type I hearing loss, age remained significant and treatment still showed a trend (p=0.002 and 0.07). Gender was not a significant predictor when evaluating the effect of DFMO on hearing loss. Interactions between age and treatment and gender and treatment were evaluated and not found to be significant.

Thresholds by Frequency

Overall hearing thresholds were not significantly different between the experimental and control group at the start of the study, one year and at the end of study (Table 2). In general, hearing thresholds worsened over the course of this study in both the placebo and DFMO groups, especially in the higher frequencies (Fig. 2). Individuals in the DFMO group consistently presented greater threshold shifts across the frequency range, which was significant from 250–2,000 Hz and trended for 4000 Hz at one year; and trended for 250 Hz and was significant for 500 Hz at the end of study (Table 3). For the most part, the differences decreased from one year to the end of study analysis. To examine the possibility of bias due to differential dropout, a last observation carried forward (LOCF) analysis was performed where the last non-missing post-baseline observed value was used to fill the missing end of study value. In the LOCF analysis, the treatment groups significantly differed at 250 Hz and 1,000 Hz, whereas they did not, for the end of study analysis. The analyses did not differ otherwise (Table 3).

Table 2.

Audiometric Thresholds at Baseline, 1 Year, End of Study and End of Study (last observation carried forward, LOCF). DFMO - α-difluoromethylornithine. Std - standard deviation.

	Frequency (Hz)	Threshold					Difference between Treatments
		DFMO		Placebo
Time		Mean	(Std)	Mean	(Std)	p-value	Mean	(Std)
Baseline (n=144 DFMO; n=146 Placebo)	250	13.9	(6.8)	14.8	(8.5)	0.34	−0.9	(7.7)
	500	11.8	(7.2)	13.3	(8.2)	0.097	−1.5	(7.8)
	1000	11.8	(8.1)	12.5	(8.6)	0.48	−0.7	(8.4)
	2000	16.0	(12.2)	15.3	(11.7)	0.62	0.7	(11.9)
	4000	31.9	(20.9)	29.7	(20.4)	0.38	2.1	(20.7)
	8000	41.9	(23.7)	41.6	(24.0)	0.91	0.3	(23.9)
One Year (n=112 DFMO; n=127 Placebo)	250	14.9	(6.9)	14.6	(8.0)	0.74	0.3	(7.5)
	500	12.0	(7.3)	11.9	(8.3)	0.87	0.2	(7.8)
	1000	12.9	(8.9)	12.0	(9.4)	0.42	1.0	(9.2)
	2000	16.5	(13.5)	14.5	(12.5)	0.22	2.1	(13.0)
	4000	34.6	(21.6)	30.7	(20.8)	0.16	3.9	(21.2)
	8000	44.0	(24.3)	41.4	(23.3)	0.40	2.6	(23.8)
End of Study (n=79 DFMO; n=106 Placebo)	250	13.9	(8.1)	12.9	(8.6)	0.43	1.0	(8.4)
	500	12.3	(8.1)	11.9	(9.0)	0.77	0.4	(8.7)
	1000	13.4	(8.5)	12.9	(10.6)	0.74	0.5	(9.8)
	2000	18.6	(14.7)	16.2	(11.5)	0.21	2.4	(13.0)
	4000	36.5	(21.9)	33.2	(21.1)	0.30	3.3	(21.4)
	8000	46.0	(26.0)	43.0	(25.1)	0.43	3.0	(25.5)
End of Study (LOCF) (n=115 DFMO; n=133 Placebo)	250	14.6	(7.9)	13.7	(9.0)	0.40	0.9	(8.5)
	500	12.7	(8.3)	12.3	(8.9)	0.66	0.5	(8.6)
	1000	13.9	(9.4)	13.3	(10.5)	0.59	0.7	(10.0)
	2000	18.6	(14.2)	17.3	(12.7)	0.45	1.3	(13.4)
	4000	38.0	(21.2)	34.0	(21.5)	0.15	3.9	(21.4)
	8000	47.3	(24.6)	43.8	(24.7)	0.26	3.5	(24.6)

Figure 2.

Hearing Thresholds by Frequency at Baseline, 1 Year, End of Study and End of Study (last observation carried forward, LOCF). Hearing tresholds worsened over the course of this study in both the placebo and DFMO groups. DFMO - α-difluoromethylornithine. The error bars represent 95% confidence intervals.

Table 3:

Change in Audiometric Thresholds from Baseline at 1 Year, End of Study and End of Study (last observation carried forward, LOCF). DFMO - α-difluoromethylornithine. Std - standard deviation.

Time	Frequency (Hz)	Change From Baseline							Difference between Treatments
		DFMO			Placebo
		%	Mean	(Std)	%	Mean	(Std)	p-value	Mean	(Std)
One Year (n=112 DFMO; n=127 Placebo)	250	13.2	1.7	(5.1)	−0.4	−0.1	(5.8)	0.011	1.8	(5.5)
	500	12.1	1.3	(5.4)	−9.9	−1.3	(4.9)	<.001	2.6	(5.1)
	1000	13.3	1.5	(4.9)	−3.3	−0.4	(4.0)	0.001	1.9	(4.4)
	2000	4.2	0.7	(4.9)	−3.7	−0.6	(4.5)	0.045	1.2	(4.7)
	4000	6.8	2.2	(6.1)	2.6	0.8	(5.2)	0.051	1.4	(5.6)
	8000	3.0	1.3	(5.2)	1.3	0.5	(5.1)	0.276	0.7	(5.2)
End of Study (n=79 DFMO; n=106 Placebo)	250	3.1	0.4	(5.6)	−9.0	−1.3	(6.2)	0.057	1.7	(5.9)
	500	12.1	1.3	(5.5)	−8.2	−1.1	(4.8)	0.002	2.4	(5.1)
	1000	22.0	2.4	(5.2)	9.4	1.1	(6.7)	0.156	1.3	(6.1)
	2000	19.8	3.1	(5.7)	23.6	3.1	(6.4)	0.982	0.0	(6.1)
	4000	19.5	5.9	(7.0)	21.2	5.8	(6.3)	0.881	0.1	(6.6)
	8000	14.9	5.9	(8.2)	10.9	4.2	(7.1)	0.127	1.7	(7.6)
End of Study (LOCF) (n=115 DFMO; n=133 Placebo)	250	8.6	1.2	(5.5)	−7.7	−1.1	(6.2)	0.002	2.3	(5.9)
	500	16.3	1.8	(5.7)	−8.4	−1.1	(5.0)	<.001	2.9	(5.3)
	1000	22.1	2.5	(5.7)	6.3	0.8	(6.4)	0.026	1.7	(6.1)
	2000	16.7	2.7	(5.4)	15.3	2.3	(6.3)	0.636	0.4	(5.9)
	4000	16.9	5.5	(6.9)	15.7	4.6	(6.5)	0.305	0.9	(6.7)
	8000	10.7	4.6	(7.6)	8.1	3.3	(7.1)	0.177	1.3	(7.4)

Recovery of Hearing Thresholds

Audiometry was conducted on subjects to evaluate hearing recovery following significant threshold shifts; however, not all subjects completed a follow-up audiogram to fully evaluate this. For individuals with a type II hearing loss or an increase in severity of tinnitus by one toxicity grade, or other symptom that was possibly treatment related, treatment was suspended for 4 weeks. If hearing thresholds, tinnitus, or other symptom returned to baseline level (within 10 dB of pre-study for audiogram), participants could resume treatment at half the original dose. In total 31 participants, 15 from the experimental group and 16 from the placebo group, experienced a type II hearing loss, tinnitus, or other possibly treatment-related symptom and had to suspend treatment. Of these, 29 individuals were able to resume treatment after 4 weeks, 13 (87%) from the DFMO group and 16 (100%) from the placebo group. In individuals resuming treatment at half dose, type II/type I/any hearing loss reoccurred in 6/13/13 individuals respectively (4/8/8 from the DFMO group and 2/5/5 from the placebo group). In participants with recurrent hearing loss by definition B (type II), treatment was promptly and permanently discontinued.

For participants considered to have hearing loss by definition A (type I) during the study, 10 (22%) of 45 were considered to be recovered at the end of the study; 2 (11%) of 18 in the placebo group and 8 (30%) of 27 in the DFMO group. For those considered to have hearing loss by definition B (type II) during study, 6 (21%) of 24 were considered to be recovered at the end of the study; 1 (10%) of 10 in the placebo group and 5 (36%) of 14 in the DFMO group. There was no significant difference between the placebo and DFMO groups for these measures.

Discussion:

DFMO has long been considered a promising candidate to help reduce and treat various types of cancers; however, hearing loss has emerged as the dose limiting effect. The objective of our study was to evaluate the ototoxic effects, from long-term administration of low-dose DFMO in a population of individuals with previous basal or squamous cell carcinoma. This study represents one of the few randomized blinded placebo controlled studies that provide prospective longitudinal audiometrics in a highly selected patient population.

The average age of subjects at the start of the treatment was 61 years. Predictably, high-frequency sensorineural hearing loss associated with presbycusis was present in many study participants prior to beginning treatment. In addition, changes in audiometric thresholds relative to initial baseline evaluations worsened over the course of this study for both control and experimental groups, consistent with age-related hearing loss over time³⁴. This change in threshold was expected and consistent with normal age-related hearing loss; however, it was noticeably more substantial in participants in the experimental group.

In general, average threshold shifts were consistently greater in the experimental group for all frequencies tested. On average, the threshold shits are greatest in the higher frequency range. However, the difference between treatment groups tended to be greater at the lower frequencies; the threshold shifts were only significant from 250–2000 Hz at 1 year, and at 500 Hz at the end of study. This is consistent with findings previously reported by Pasic et al.¹³ In the LOCF analysis, the treatment groups significantly differ at 250 Hz and 1000 Hz, in addition to 500 Hz, compared to the end of study analysis. These disparities might be attributed to the decrease in differences between treatment groups at the end of the study, as the result of subjects having their doses halved as necessary.

Overall, the lack of significance at the higher frequencies may be attributed to the greater variability observed amongst the subjects, which was not unexpected given the age of the participants. While these numbers were not significant, it is plausible that DFMO does affect the higher frequency region of the cochlea (base) with greater capacity than the lower frequencies given the increased threshold shift observed in this region. Animal studies support this observation, where DFMO toxicity has been reported to affect the basal high frequency region of the cochlea most^35–37. Collectively, average threshold shifts were greater in the experimental group relative to the control group across the frequency range; however, additional investigation with a larger study population is warranted to determine the extent of this threshold shift.

Sufficient data was available for 250 participants to carry out statistical analysis; 117 in the DFMO group and 133 in the placebo group. In total, 47 subjects experienced a type I and/or type II hearing loss described as a ≥ 30 dB change in the summation of 3 adjacent frequencies or an air conduction threshold shift by ≥ 15 dB at 2 or more frequencies, respectively. The number of type I and type II hearing losses were consistently greater for participants in the DFMO group for all conditions, representative of DFMO toxicity. Of importance, a majority of participants taking DFMO did not experience hearing loss. Based on this information, it is plausible that other unknown variables, possibly genetic, make some individuals taking DFMO to be more susceptible to hearing loss than others. Additional analysis of this might yield insightful results and could provide a novel standard for recommending DFMO to individuals in the future.

Subject age and gender were evaluated to determine the effect, if any, DFMO had on hearing thresholds. Age and male gender have previously been reported to adversely affect hearing²⁴. In this study, age had an overall effect on hearing loss, while gender did not. In addition, there was no interaction between age and treatment effect. That is to say, we did not detect a difference in agent effect by participant age. This discrepancy may be caused by the smaller sample size employed in the previous study. Moreover, the findings in this study are consistent with another phase II clinical trial¹³. While age and gender did not appear to influence the effect DFMO had on hearing levels, older participants and participants in the DFMO group were more likely to experience a type I hearing loss. It appears that age, to some extent, had some effect on hearing thresholds on individuals in the treatment group. Further evaluation with larger sample sizes are necessary to confirm that age and gender are not major factors when determining if DFMO is suitable for a particular candidate or for predicting plausible side effects.

The overarching goal of this clinical trial was to determine the efficacy of DFMO in preventing new non-melanoma skin cancers²⁹. Participants taking daily DFMO had a non-significant reduction (p=0.069) in new non-melanoma skin cancer (NMSC). While the observed difference in the incidence of new cancer cases was non-significant between the control and experimental groups, the potential of DFMO to prevent new skin cancers should be explored further. It is possible that DFMO, alone or in combination with another agent, may yield significant results in preventing and treating various types of cancer.

Of importance, the average threshold shifts were consistently greater in the experimental group for all frequencies tested; however, this shift was limited to a few decibels. This small difference in threshold shift across the frequency range should minimally interfere with overall hearing abilities when compared to a greater threshold difference. As such, the ototoxicity observed from long-term administration of low-dose DFMO may be of minor concern relative to the greater potential of preventing and treating various types of cancer, on balance.

Conclusion:

This phase III randomized double-blind placebo-controlled clinical trial found DFMO administered long term at 500 mg/m²/day effects hearing thresholds across the frequency range, with significant threshold shifts from 250–1,000 Hz. Given that threshold shifts have been found to correlate with DFMO dose, further investigation is warranted to determine if a lower dose of DFMO would have a less significant effect on hearing thresholds. While DFMO may help prevent the incidence of certain types of cancers, it can result in some degree of hearing loss even when administered at low doses. To date, there are no clinical trials that report of a non-ototoxic dose of DFMO. Because of this, we recommend all individuals taking DFMO undergo serial audiometric evaluations, regardless of the dose, to monitor and preserve hearing when possible.