The Cost-Effectiveness of Bimodal Stimulation Compared to Unilateral and Bilateral 1 Cochlear Implant Use in Adults With Bilateral Severe to Profound Deafness

Objectives : An increasing number of severe-profoundly deaf adult unilateral cochlear implant 26 (CI) users receive bimodal stimulation; that is, they use a conventional acoustic hearing aid 27 (HA) in their non-implanted ear. The combination of electric and contralateral acoustic hearing 28 provides additional benefits to hearing and also to general health-related quality of life 29 compared to unilateral CI use. Bilateral CI is a treatment alternative to both unilateral CI and 30 bimodal stimulation in some healthcare systems. The objective of this study was to conduct an 31 economic evaluation of bimodal stimulation compared to other management options for adults 32 with bilateral severe to profound deafness. 33 Design : The economic evaluation took the form of a cost-utility analysis and compared 34 bimodal stimulation (CI+HA) to two treatment alternatives: unilateral and bilateral CI. The 35 analysis used a public healthcare system perspective based on data from the United Kingdom 36 (UK) and the United States (US). Costs and health benefits were identified for both alternatives 37 and estimated across a patient’s lifetime using Markov state transition models. Utilities were 38 based on Health Utilities Index (HUI3) estimates and health outcomes were expressed in 39 Quality Adjusted Life Years (QALYs). The results were presented using the Incremental Cost- 40 Effectiveness Ratio (ICER) and the Net Monetary Benefit approach to determine the cost- 41 effectiveness of bimodal stimulation. Probabilistic sensitivity analyses explored the degree of 42 overall uncertainty using Monte Carlo simulation. Deterministic sensitivity analyses and 43 Analysis of Covariance identified parameters to which the model was most sensitive; i.e. whose 44 values had a strong influence on the intervention that was determined to be most cost-effective. 45 A Value Of Information analysis was performed to determine the potential value to be gained 46 from additional research on bimodal stimulation. 47 Results : The base case model showed that bimodal stimulation was the most cost-effective 48 treatment option with a decision certainty of 72% and 67% in the UK and US, respectively.

Despite producing more QALYs than either unilateral CI or bimodal stimulation, bilateral CI 50 was found not to be cost-effective because it was associated with excessive costs. Compared 51 to unilateral CI, the increased costs of bimodal stimulation were outweighed by the gain in 52 quality of life. Bimodal stimulation was found to cost an extra £174 per person in the UK ($937 53 in the US) and yielded an additional 0.114 QALYs compared to unilateral CI, resulting in an 54 ICER of £1,521 per QALY gained in the UK ($8,192/QALY in the US). The most influential 55 variable was the utility gained from the simultaneous use of both devices (CI+HA) compared 56 to Unilateral CI. The value of further research was £4,383,922 at £20,000/QALY ($86,955,460 57 at $50,000/QALY in the US). 58

Conclusions:
This study provides evidence of the most cost-effective treatment alternative for 59 adults with bilateral severe to profound deafness from publicly-funded healthcare perspectives 60 of the UK and US. Bimodal stimulation was found to be more cost-effective than unilateral 61 and bilateral CI across a wide range of willingness-to-pay thresholds. If there is scope for future 62 research, conducting interventional designs to obtain utilities for bimodal stimulation 63 compared to unilateral CI would reduce decision uncertainty considerably. 64

INTRODUCTION 66
Cochlear Implants (CI) were formerly considered to be suitable for severe to profoundly deaf 67 patients who could not benefit from acoustic amplification via conventional hearing aids ( (assuming 70% of individuals used an acoustic hearing aid in their contralateral ear) to 116 conventional best practice (some patients used HAs and some did not). Although they did 117 account for the cost of the contralateral HA (£100 on average) and its replacement over time 118 (every 5 years), they were unable to identify reliable published estimates of the health benefits 119 ('utility gain') from bimodal stimulation and therefore only included the incurred HA costs in 120 their model. 121 Goman (2014) assessed the minimum utility gain required for bimodal stimulation to be cost-122 effective compared to a unilateral CI in adults. The study considered the additional costs 123 associated with bimodal stimulation including hearing aid appointments (assessment, fitting 124 etc.) and rehabilitation (aftercare, repairs etc.). The study explored four different scenarios 125 therefore to: (i) conduct a cost-utility analysis of bimodal stimulation compared to unilateral 140 and bilateral cochlear implantation; (ii) explore how small changes in the health benefits 141 associated with bimodal stimulation could impact the conclusions of the economic assessment; 142 and (iii) assess how sensitive the results are to assumptions around how long severe to 143 profoundly deaf adults will continue to use a contralateral HA after they receive their CI. 144

Model Description 170
A decision analytic model was constructed and was based on a decision tree (Fig. 1) and 171 Markov models using the cohort simulation approach, which follows a cohort of patients as a All patients were first assessed for cochlear implant candidacy and a proportion were assumed 176 to be ineligible for implantation. Although the modelling of those patients who are not eligible 177 for implantation is identical across all the treatment alternatives considered in the current study, 178 their inclusion in the model is necessary as the proportion of patients in a population who can 179 access a treatment can affect how cost effective it is. 180 All patients eligible for CI underwent surgery to implant the electrode array ('internal 181 component'). These patients then entered a three-state Markov model (Fig. 2) including one 182 which denotes the use of CI(s) and two absorbent states (i.e. states from which they could not 183 return to being CI users) representing the non-use of CI(s) and death. Adults in the state 184 denoting use of CI(s) subsequently entered a second Markov model comprised of four states 185 that described the success of implantation surgery and function of the device. 186 The structure of the model was identical for all three treatment options but the model for 187 bilateral CI assumed patients were implanted simultaneously in both ears and thus included the 188 cost of two cochlear implants but one surgery. The model for bimodal stimulation accounted 189 for extra costs and benefits from the additional use of the contralateral HA over unilateral CI 190 but also assumed that not all patients will be willing or able to use a contralateral HA. The 191 proportion of bimodal users was taken from a large-scale cross-sectional UK survey of 359 192 unilateral cochlear implant recipients that reported a percentage of contralateral HA use of 193 45.4% (Fielden et al. 2016a(Fielden et al. , 2017. 194 The 'working' state assumed that the fitted CI(s) was functioning and there were no adverse 195 effects. It was assumed that all patients experiencing a failure of the sound processor (external 196 failure) needed a replacement processor. In the case of an internal failure or a major 197 complication, patients were assumed to require an operation for re-implantation, while a 198 proportion of those patients would have the implantable component extracted. For bilateral CI,199 it was assumed that those who required an extraction continued as unilateral CI users with 200 associated benefits gained and costs incurred. The non-use state reflected the results of CI 201 extraction in the case of unilateral CI users and also voluntary permanent non-use. 202 The patients who were ineligible for CI entered a two-state Markov model with states 'alive' 203 and 'death'. The alive state represented the non-surgical management of severe to profound 204 deafness in which a proportion of adults were assumed to benefit from (and therefore use) HAs. 205 Of those, a proportion were assumed to use two HAs and the remainder a unilateral HA (Bond 206

Transition Probabilities 217
The probabilities used in the model are listed in Table 1. The probabilities related to cochlear 218 reported using a contralateral HA but using it at the same time as their CI. The incremental 240 utility gain associated with bimodal stimulation was evaluated by comparing the utility weights 241 of the bimodal group to the unilateral CI group using non-parametric analyses (Mann Whitney 242 U test) after accounting for differences in the time since implantation. Bias-corrected 243 confidence intervals for the mean utility weights for the unilateral and bimodal groups were 244 computed using bootstrapping (Davison & Hinkley, 1997). unilateral or bilateral implantation. The utility increment associated with bilateral CI over 254 unilateral CI from Summerfield et al. (2002Summerfield et al. ( , 2006 was used (0.031) both because it was 255 observed using the HUI3, the same instrument used in the current study to estimate the 256 increment from bimodal stimulation, because it was found consistently across two studies using 257 contrasting estimation methods, and because it approximated the average value reported across 258 all four studies. The final utility weights are shown in Table 2. 259

Resource Use 260
Direct costs of the hearing aid and the cochlear implant were calculated using the most suitable 261 and up-to-date unit costs ( Compared to unilateral CI, bimodal stimulation was associated with additional costs related to 267 the contralateral hearing aid. Although a proportion of the cohort of severe to profoundly deaf 268 adults were assumed to already use HAs before implantation (Fig. 1), the model assumed that 269 additional appointments were provided to each bimodally-aided patient following implantation 270 in order to ensure the two devices (CI+HA) were optimized to work together. It was assumed 271 that there was only one follow-up visit related to the HA given that these patients were not new 272 HA users. It was assumed that the HAs were replaced every 5 years ( demonstrates the proportion of adult bimodal users that discontinue hearing aid use) and 301 assumed that 39% of users would cease HA use after five years. 302 There is also published evidence supporting the use of age-adjusted utilities as the utility of a 303 normal-hearing person diminishes over time (Bond et al. 2009). Therefore, a secondary 304 analysis was conducted using age-dependent utilities to prevent the overestimation of quality 305 of life for which a scaling factor that reduced utilities as a function of age was extracted from 306

Bond et al. (2009). 307
Uncertainty 308 A univariate sensitivity analysis was conducted by varying the value of each parameter over a 309 plausible range while all other parameters were held constant (Claxton 2008). Each parameter 310 was varied between the 2.5% and 97.5% percentile values derived using its confidence interval or standard error (Tables 1-3). The results were visualised using a tornado plot in which the 312 effect of varying each parameter on the main output of the model (the INMB value) were 313 plotted for each parameter. 314 The overall level of uncertainty in the model was quantified by conducting a Probabilistic 315 Sensitivity Analysis (PSA) (Claxton et al. 2005). Instead of each parameter being represented 316 by a single value, each parameter was represented by a probability distribution that expressed 317 the likely range of values the parameter could take. Parameters that referred to probabilities 318 were represented by beta distributions (Table 1), and both utilities ( Table 2) and costs (Table  319 3) were represented by gamma distributions 1 . Monte Carlo Simulation was used to run the 320 years was chosen as a time horizon over which the benefits of an optimal decision could be 353 expected to accrue, but not so long as to mean that some of the treatment options or assumptions 354 in the model may no longer be applicable.

RESULTS 356
Base case results are summarized in Table 4. Bimodal stimulation generates an ICER of £1,521 357 per QALY compared to unilateral cochlear implantation. This ICER is below the £20,000 358 threshold adopted by NICE in the UK, suggesting that bimodal stimulation is more cost-359 effective than unilateral CI. Bilateral CI is deemed not to be cost-effective as it generates an 360 The first scenario analysis re-considered the cost-effectiveness of bimodal stimulation 375 assuming that approximately 39% of bimodal users would discontinue the use of the HA 376 voluntarily (Fig. 3, 'Stop being bimodal'). The incremental NMB was smaller than the base 377 of 83 years), and the time horizon adopted for the analysis (a lifetime horizon was adopted 394 following the approach taken by NICE when formulating guidance on cochlear implants). 395 The results of the univariate sensitivity analyses are presented in the left panels of Figure 4. 396 The key parameters shown in the tornado plot were identified using ANCOVA, which 397 indicated that the utility parameters had the highest impact on the variance of the model outputs, The model outputs were insensitive to the cost of the acoustic hearing aid in both countries. 408 Similarly, the appointment costs related to hearing aids (assessment, fitting & follow-up) did 409 not influence the economic outputs to the extent needed to affect the conclusions of the base 410 case analysis (Fig. 4, left panels). The univariate sensitivity analysis suggested that the cost 411 parameters in the US analysis were slightly more influential than the UK analysis, although the For willingness-to-pay thresholds of £20,000/QALY and £30,000/QALY the probability of 438 bimodal stimulation being the most cost-effective alternative was 72% and 73%, respectively. 439 In the US analysis, bimodal stimulation was cost-effective with a 67% and 59% certainty at a 440 $50,000/QALY and $100,000/QALY threshold, respectively (Table 4). 441 The only previous study that has considered the cost-effectiveness of bimodal stimulation 483 estimated that a utility gain of at least 0.0109 was required to make bimodal stimulation cost-484 effective. The current study obtained HUI3 data from 91 adults managed at two large CI 485 services in the UK and estimated a utility gain of 0.032 by comparing a bimodally-aided group 486 with a unilateral CI group while controlling for years of experience with CI use (Table 2,  after approximately 30 years of device use. Thus, even under the assumption that utilities 493 diminish with age, bimodally-aided adults will get benefit well into their working life. The 494 basic implication from the age-adjusted utilities analysis is that as patients get to an advanced 495 age their health looks increasingly similar regardless of the hearing devices they use; i.e. 496 unilateral CI or CI+HA. 497 The utility gain for bimodal stimulation used in the present analysis was obtained from UK 498 patients who had received no rehabilitation specific to bimodal stimulation as it is not routinely 499 provided in the UK (Fielden & Kitterick 2016). This fact underpinned the assumption that the 500 only additional costs for bimodal stimulation associated with that utility gain were the costs 501 required to ensure the hearing aid was up to date and maintained over time, which is also in 502 line with the assumptions made by Bond et al. (2009). Even if we used higher cost estimates 503 as Goman (2014) reported, who assumed a greater level of extra rehabilitation and aftercare in 504 the bimodal group, the final decision on the cost-effectiveness of bimodal stimulation remains 505 the same. The overall costs for bimodal stimulation were also higher in the US vs the UK model 506 in the present evaluation, resulting in an additional cost of $937 per person for bimodal users, 507 but bimodal stimulation still remained cost-effective. Thus, the current results suggest that 508 bimodal stimulation would be affordable under a public healthcare system in the US. The 509 provision of bimodal stimulation in the US not only has the potential to increase patient benefit, 510 but also reduce inequalities within the healthcare system as contralateral acoustic HAs are not 511 provided through the public health system in the US, but instead are purchased privately. 512 Bainbridge and Ramachandran (2014) reported that the prevalence of hearing aid use among 513 older adults was 28% to 66% higher in those with higher incomes compared to the adults on 514 the lower end of the income-to-poverty distribution. It is possible that similar differences could 515 arise in contralateral HA use among adult CI users in the US. 516 The EVPI depicts the upper price that a healthcare system should be willing to spend for 517 obtaining additional evidence. Performing more trials and conducting further research provides 518 more accuracy around the input parameters and resolves part of the uncertainty around the final 519 outcome of the model. It would appear feasible for randomised controlled trials or other types 520 of formal clinical evaluations to be conducted within these funding limits (£4,383,922 at 521 £20,000/QALY; $86,955,460 at $50,000/QALY). If there is scope for further investment, it 522 should be where the uncertainty is highest. Results from the sensitivity analyses suggest utility 523 weights as the key driver of parameter uncertainty. Future research could address this issue by 524 conducting interventional designs (e.g. randomized controlled trials) to obtain utility weights 525 for bimodal stimulation compared to unilateral CI, rather than rely on data from observational 526 studies on which the current utility weights are based. The lack of health utility data in the field 527 of cochlear implantation in general is an important evidence gap to address given the important 528 role that economic evaluations play in determining whether such a low-volume high-cost 529 intervention is good value for money from the payer's perspective. 530 Economic evaluations typically use generic health instruments to obtain data on health-related 531 quality of life; that is, instruments that are by design relevant to a wide range of health 532 conditions. Although the EQ-5D (Brooks, 1996; The Euroqol Group, 1990)  hearing. Such an eventuality would seem possible as the candidacy criteria adopted in UK are 573 some of the most restrictive in the world (Vickers et al. 2016). In the US, providing HAs 574 through the public healthcare system might increase the rate of contralateral HA use among CI 575 users without bimodal stimulation necessarily being offered as a distinct treatment option. effective than it currently appears to be, it would necessitate audiologists having to manage an 578 even greater number of bimodally-aided patients in the future. This growing trend places 579 additional emphasis on the need to develop guidance around maintaining both devices 580 simultaneously, to develop enhanced fitting procedures, and to identify which patients have the 581 capacity to derive benefit from a contralateral HA.              The warranty period of the external part of a CI *Inflated to 2015/16 prices. †Variance of costs set to 1/5 of the mean value following Summerfield et al. (2010).