Ethics approval was granted by the University College London Research Ethics Committee 12241/001. Informed consent was obtained from all participants.

A state-transition model, following the ISPOR-SMDM Best Practice Guidelines, was created using Microsoft Excel (Redmond, Wash) to assess the potential costs and effects of using novel regenerative hearing therapeutics in adults, 50 or older, with ARHL (Roberts et al., 2012). This was compared to the current standard of care, including hearing aids and cochlear implants. Figure 1 shows a representation of the patient’s pathway modeled in the study. Supplemental Digital Content (SDC) 1 contains an in-depth description of hearing loss classification using pure-tone averages (PTA) (Mathers et al., 2000).

The model adopts a cycle length of 1 year and spans the patient’s lifetime until death given the life-long costs and effects of hearing loss. The model is constructed from a healthcare perspective of the National Health Service (NHS) in the United Kingdom and evaluates direct medical costs. A willingness to pay threshold of £20,000 per quality-adjusted life year (QALY) is used to assess the cost-effectiveness (CE) of the intervention (NICE, 2017). Both costs and outcomes are discounted at a 3.5% per annum rate to comply with NICE guidelines (National Institute for Health and Clinical Excellence, 2008). NICE stands for the National Institute for Health and Care Excellence. It is an independent organization which evaluates and provides recommendations of which drugs and treatments are available on the NHS in England. Other key model assumptions can be found in SDC 2. See SDC 3 for a complete list of abbreviations used in the text along with verbal descriptors.

The model was validated using the AdViSHE validation assessment tool (Vemer et al., 2016). The conceptual model, input data, and model outcomes were tested on face and operational validity by consulting professional stakeholders (n = 24) from the pharmaceutical and biotechnology industry, national hearing charities, otolaryngology, audiology, discovery science, and research funding bodies. The model outcomes were cross validated with relevant literature. No other health economic models on ARHL were found for cross-validation. Additionally, the model was verified for inconsistencies by two independent modeling experts.

With no regenerative therapeutic for ARHL having been approved for clinical use, we used a hypothetical regenerative therapy to model different efficacy scenarios. Our model focused primarily on regenerative therapeutics and was not targeted toward a specific gene, cell, molecular therapy or method of delivery. In order to estimate the maximum potential benefit of the therapeutic (headroom), the base case scenario assumed a 100% adherence, uptake and efficacy of the therapy with zero costs. This meant that patients with any form of hearing loss recovered to normal hearing after therapy administration. It also assumed that rates of hearing loss progression in subsequent years were unaffected, meaning that patients could once again develop age-related hearing loss. The model assumed that all patients were eligible for both existing and novel strategies and could receive either a hearing aid, a cochlear implant, or the novel therapy as they developed hearing loss in the model.

Table 4 summarizes the results of the headroom analysis and shows that total costs and QALYs per patient in the standard care pathway are £4,462 [£3,262–£5,663] and 15.59 [15.09–16.09], respectively. The total costs and QALYs per patient for a perfect novel hearing therapeutic strategy are £11 [£5–£17] and 16.37 [15.67–17.06], respectively. This yields potential savings of £4,451 [£3,254–£5,648] and QALY gains of 0.78 [0.37–1.19] per patient. The iNMB of a perfect, zero cost, novel regenerative hearing therapeutic in ARHL is £20,017 [£11,299–£28,737].

The results demonstrated that compared to the headroom scenario (scenario 1), changing the recovery rates by only one (scenario 2) or two (scenario 3) hearing health states lowers the iNMB to £7,521 [£4,884–£10,158] and £9,265 [£5,861–£12,670], respectively (Table 4). Since the assumed drug cost is zero due to the headroom scenario, limiting treatment to patients with more severe degrees of hearing loss (scenarios 4–6) decreased the iNMB, owing to the fact that only a subgroup of the model’s population had more severe forms of hearing loss. This resulted in only a proportion of the total population being treated in this scenario which decreased the overall QALY gains and resulted in less cost savings.

The threshold analyses illustrate the ceiling prices of the novel regenerative hearing therapeutics with differing rates of effectiveness and hearing loss severities (Figures 2A–D). The lines in the graphs represent an iNMB of £0, identifying (1) the potential maximum price for each level of effectiveness, (2) the 95% confidence intervals, and (3) the average number of treatments given per person depending on the efficacy of the therapy. For example, if treatment is restricted to patients with moderate hearing loss or worse (Figure 2B), the ceiling price at an efficacy of 50% will be around £27,500/per person. See SDC 9 for a graphical comparison of threshold analyses results.

The headroom scenario yields an iNMB of £20,017 [£11,299–£28,737] compared to the current standard of care and show that novel regenerative hearing therapeutics have potential room for improvement in terms of both improved patient outcomes with QALY gains of 0.78 [0.37–1.19] and cost savings of £4,451 [£3,254–£5,648] per patient. Scenario analyses demonstrate that the iNMB increases with more effective therapeutics and more widespread eligibility criteria. Threshold analysis suggests that the ceiling price of the therapeutic is greater with severe forms of hearing loss compared to less severe states. Both scenario and threshold analyses demonstrate wide confidence intervals which reflects the uncertainty surrounding results.

This study shows that there is significant potential room for improvement in the current standard care pathway for patients with ARHL and that novel regenerative hearing therapeutics could become cost-effective in the NHS. This model can be used by both industry and policy makers to evaluate: (1) the maximum price of a novel regenerative therapeutic in ARHL for different levels of effectiveness, (2) the minimum effectiveness required at each pricing target for the therapeutic to remain cost-effective, and (3) the impact of limiting treatment to certain subgroups of patients with hearing loss. Exploring the ceiling price of novel therapeutics at maximum efficacy will also allow pharmaceutical companies to perform feasibility estimates on the potential to recoup research and development (R&D) costs (IJzerman and Steuten, 2011; Markiewicz et al., 2016).

Given that one in five people in the United Kingdom are estimated to be affected by a hearing disorder by 2035, means that the implementation of novel hearing therapeutics will bring about substantial change to health systems (No author list, 2016). The development of this model, before a therapeutic has entered the market, will facilitate informed decision making and will increase the likelihood of developing cost-effective novel regenerative hearing therapeutics. As these novel hearing therapeutics enter into clinical trial, this research will help commissioners and policy makers to make difficult decisions on which therapeutics provide value for money are worth funding, thereby paving the way for revolutionary and urgently needed hearing therapeutics.

Despite the robustness of our results to a variety of probabilistic sensitivity analyses, this study is subject to limitations; the first of which arises from our limited scientific understanding of ARHL. In addition to ARHL not being well understood, its association and interrelationship with age related neurocognitive decline further complicates our understanding. ARHL is not a single disease entity but a symptom of a range of underlying disease mechanisms and etiologies, resulting in an equal range of potential therapeutic targets within the inner ear and auditory pathways (Yamasoba et al., 2013; Nakagawa, 2014). As such there will be no single cure for hearing loss; in order to develop targeted therapeutics for hearing disorders in specific patient populations more detailed pheno-and genotyping of hearing loss patients will be required (Yang C. H. et al., 2015; Le Prell et al., 2016). For our model, this means that our results likely represent an overestimation of the target population with ARHL that will be eligible to receive and benefit from regenerative hearing therapeutics in the future. Despite these limitations, our model reflects our current understanding and uses the best available data on progression and utilities for ARHL.

Another limitation is that we did not include drug safety profiles for novel hearing therapeutics due to their hypothetical nature. This model assumes that along with being perfectly effective, the novel hearing therapies are free of adverse events which overestimates their potential added value. Finally, this study focused on direct medical costs incurred by the NHS and did not include indirect medical costs associated with hearing loss in either the standard of care or novel hearing therapeutic arms (NICE, 2017). This has likely led to an underestimation of the costs that NICE requires for decision making. Additionally, the PSA’s accuracy is inherently limited by the nature of modeling a hypothetical novel hearing therapeutic and by arbitrary variations in costs set at 10%.

Consultation with our panel of expert stakeholders revealed strong support to move away from the classic classification of hearing loss using pure-tone averages to a classification scheme that incorporates functional impairment. Hearing in Noise Testing (HINT) was proposed as a potential solution moving forward. In order to construct new models using HINT, as opposed to pure-tone averages, further research will be required to delineate both ARHL transition probabilities and utilities in identified HINT states.

A growing body of evidence suggests that the use of EQ-5Ds, as recommended by NICE, may not be appropriate for all health conditions such as hearing disorders and vision loss due to the poor responsiveness of the measure to detect change (Longworth et al., 2014). Our review of the literature highlights the lack of consensus amongst studies on the best utility estimates to employ but did identify the HUI-3 as the most commonly used measure in health economic evaluations relating to hearing loss and was therefore employed in our model (Longworth et al., 2014; Yang Y. et al., 2015). To enhance comparability across studies, either the use of HUI-3s should be standardized or an EQ-5D ‘bolt-on’ for hearing should be incorporated to improve its sensitivity (Yang Y. et al., 2015).

Finally, this study not only has the capacity to inform current R&D decisions, but also to accelerate decisions in later stages of development once better therapeutic efficacy estimates have been established. As an increasing number of these therapeutics enter into clinical trials, it will be important to incorporate new evidence as it becomes available. This will improve the quality and reliability of the results, allowing for lifecycle HTA of the therapeutics. It will also be vital to conduct similar early health technology assessments in other therapeutic areas of hearing loss such as sudden sensorineural hearing loss and noise and drug induced hearing loss as the field expands (Hartz and John, 2008; Markiewicz et al., 2016; Mandavia et al., 2020).