Holoclar for treating limbal stem cell deficiency after eye burns

4.6 The primary outcome of the pivotal HLSTM01 study was treatment success, defined as stable corneal epithelium without significant recurrence of neovascularisation 12 months after treatment. The main secondary outcomes included symptom resolution (pain, burning and photophobia), inflammation, neovascularisation, visual acuity, number of successful keratoplasties and safety.

4.7 The results from HLSTM01 showed that 72.1% of patients had a successful Holoclar transplant. For the comparators, transplant success varied between 60.0% and 100.0%. Given its concerns with the quality of evidence, the committee asked the clinical experts if these success rates were likely to be reflective of those in clinical practice. The experts stated that the results for Holoclar were plausible, but that the results for the comparators should be interpreted with more caution. They considered the studies of conjunctival limbal autograft to be poorly conducted and biased, and the rates to be overestimated (in their experience, success rates were closer to 50%). The committee concluded that the evidence showed Holoclar to be an effective treatment in terms of the outcomes described in section 4.6. However, there was uncertainty about the comparator success rates and the effectiveness of Holoclar compared with them.

4.8 The company presented 2 separate models: a 'unilateral' and a 'bilateral' model. The models were similar, with both consisting of a decision tree followed by a Markov model, but the bilateral model assumed that patients had treatment in both eyes (and so included an additional year without treatment before treating the second eye). The evidence review group (ERG) considered the model to be reasonably well constructed. However, the committee noted that the company had not actually presented separate results for unilateral and bilateral populations. The 2 models did not, in fact, distinguish between unilateral and bilateral disease; both models included exactly the same population (that is, both included patients with unilateral and bilateral disease). Instead, the main difference was that only 1 eye was treated in the unilateral model, whereas both eyes were treated in the bilateral model. The committee agreed that it would be more informative to consider the company's unilateral model as the '1-eye treatment' model and the bilateral model as the '2-eye treatment' model. The committee concluded that the model structures were acceptable for its decision-making. It would take into account treatment between 1 eye and 2 eyes but it would be difficult to distinguish between unilateral and bilateral populations in the cost-effectiveness modelling when making its final recommendations.

4.9 The company had deviated from the reference case by using a discount rate for future costs and benefits of 1.5%, rather than 3.5% as outlined in NICE's guide to the methods of technology appraisal. The company justified this by explaining that it believed Holoclar to have a prolonged effect (over 30 years), and that it could return patients to a high utility state. The committee was aware that the NICE methods guide states that when appraising treatments that 'restore people who would otherwise die or have severely impaired life to full or near full health, and when this is sustained over a very long period (normally at least 30 years)', non-reference case rates may be considered. However, the committee noted that it is rarely considered appropriate to change the discount rate and that LSCD was very different from the fatal and near-fatal conditions implied by the methods guide. The committee therefore concluded that the company should have used a 3.5% discount rate.

4.10 The committee noted that if treatment with Holoclar were successful at 1 year, it remained successful for the lifetime of the model. When questioned about the plausibility of this assumption, the clinical experts stated that there currently was not enough evidence to support strong assumptions about the long-term effectiveness of Holoclar. However, the transplant being successful at 1 year would demonstrate that the patient's body has regenerated the cornea 3 to 4 times, suggesting there may be some scientific merit to this model assumption. The committee accepted the company assumption about long-term success in the model, but agreed that this was subject to a high level of uncertainty that could increase the incremental cost-effectiveness ratio (ICER).

4.11 The committee noted that the clinical-effectiveness assumptions in the model were taken from HLSTM01 for Holoclar, and from the literature for the comparators. For Holoclar, the committee had previously heard that the estimated transplant success rates were plausible. However, for the comparators, the committee noted that the company had pooled the data, despite the company and ERG agreeing that pooling these data would not be appropriate. The company had presented a limited exploration of different success rates for the comparators, but these scenarios included changes in other assumptions and so did not explore the effect of this rate change alone. Furthermore, only 1 alternative rate for each comparator had been explored. The committee agreed that it would have been more useful to explore a range of different success rates (between 50% and 80%) because there was no comparative evidence for Holoclar and the clinical experts considered the comparator success rates to be overestimated. The committee concluded that there was a substantial level of uncertainty in the clinical-effectiveness assumptions in the company's model.

4.12 In response to consultation, the company provided additional sensitivity analyses which included the committee's preferred assumptions (a discount rate of 3.5%, a utility value of 0.84 and a utility decrement for disfigurement of 0.140). The company varied the probability of initial transplant success and long-term probability of transplant failure for each comparator separately. The sensitivity analyses showed that the ICERs for Holoclar were sensitive to the annual failure probability and initial success rates for the comparator treatments; a higher probability of transplant success increased the ICERs. The committee recalled that the company's model may overestimate comparator success rates (section 4.7). It concluded that lower transplant success rates would decrease the ICER, but agreed that this was subject to a high level of uncertainty.

4.13 The committee was aware that conjunctival limbal autograft needs a substantial amount of tissue from the donor eye. It queried whether the potentially negative effect on the donor eye had been captured, and heard from the company, ERG and clinical experts that it had not. The committee noted further that the effect on the donor in the event of a transplant from a living, relative donor had also not been captured. It assumed that if donor disutility had been taken into account, the ICER would likely decrease. The committee agreed that it would have preferred to have seen analyses including the effect on the donor eye when comparing with conjunctival limbal autograft and the effect on the donor when comparing with conjunctival limbal allograft from a living, relative donor. The committee noted that the utility values used in the model were extraordinarily low: the values were far lower than any used in previous appraisals for eye treatments, with some lower than those for people in the last 3 months of life having palliative treatment for various cancers. The committee noted that the main reason for these low values appeared to be the utility decrement of 0.318 applied to patients experiencing disfigurement. This was over 100-times higher than the utility decrement applied to people experiencing any pain, burning or photophobia (the highest decrement was 0.019). The committee highlighted that company's 0.318 decrement was derived from non-reference case methods, which were likely to generate exaggerated results. The committee asked the clinical experts about the plausibility of these values. The clinical experts found the low overall utility difficult to quantify, but stated that LSCD has a long-term substantial negative effect on quality of life. For the high utility decrement for disfigurement, the committee heard varied reports about the relative importance of disfigurement. The clinical experts stated that for people with unilateral disease, contrary to conventional wisdom, visual acuity was not the most important outcome: because of its lasting and significant effect on their day-to-day lives, these people often prioritise disfigurement. For people with bilateral disease, disfigurement would be less of a priority than visual acuity because of the risk of complete loss of sight. The committee agreed that it was difficult to resolve this inconsistency in the relative importance of disfigurement. Furthermore, this was not the approach taken in the model, with utility values applying equally irrespective of whether disease was unilateral or bilateral. The committee agreed that the utility values used were implausibly low, noting that the ERG had used alternative values. For visual acuity, the ERG used a range with a more plausible maximum value of 0.861 (rather than the company's assumption of 0.706). For disfigurement, it used a decrement of 0.140 (rather than the company's assumption of 0.318), using cataracts as a proxy. The ERG noted that the decrement associated with cataracts also includes a loss of utility from both disfigurement and vision loss. The committee recognised that cataract disutilities were used as a proxy for disfigurement although uncertainty remained in the utility values, the ERG's values were a more realistic reflection of the impact on quality of life.

4.14 The committee noted that the cost of autologous serum eye drops were a substantial driver of cost-effectiveness results in the model. The company had assumed in the model that patients having the comparators needed these drops after treatment and for flare-ups, but patients having Holoclar did not. The committee heard from the clinical experts that autologous serum eye drops are essential after treatment with the comparators, but less so with Holoclar because the stem cells are cultivated in a laboratory and not on the patient's eye. However, some clinicians may still choose to use them with Holoclar. The clinical experts also stated that drops would rarely be used for flare-ups because they take up to 6 weeks to prepare. Although alternative eye drops are available that would be ready for immediate use (allogenic serum eye drops), the committee heard from the ERG that using these drops would not make much difference to the model results because they were similarly priced. The committee agreed that the costs of eye drops for Holoclar had been underestimated in the model, but that the effect of this was difficult to quantify. It also agreed that eye drops after treatment were more necessary for the comparators than for Holoclar. It therefore concluded that it could cautiously accept the company's assumptions about eye drops, but remained aware that any increase in the use of eye drops would make Holoclar less cost effective.

4.15 The committee considered the estimates of cost effectiveness calculated by the company. The base-case results were:

4.16 The committee was aware that there were uncertainties in the assumptions about the comparators' long-term success rates and the use of eye drops. It noted that the success rates for conjunctival limbal allograft from a living, related donor and keratolimbal allograft in particular were likely to have been overestimated. Furthermore, if the effect on the donor were taken into account, the ICERs for Holoclar would likely decrease and fall within the range of £20,000 to £30,000 per QALY gained. Given the patient need (section 4.1) and innovative nature of the treatment (section 4.17), the committee agreed that it would pragmatically accept this as a demonstration of cost effectiveness. The committee noted that Holoclar was only cost effective for treating 1 eye. It recalled that there was almost no evidence concerning Holoclar's clinical effectiveness for treating both eyes, and agreed that it could not recommend Holoclar for routine use in the NHS to treat both eyes. However, it agreed that its recommendations for treating a single eye should apply equally to unilateral and bilateral disease. The committee therefore concluded that it could recommend Holoclar as a cost-effective use of NHS resources in people with moderate to severe LSCD after eye burns, only if it is used to treat 1 eye and they have already had a conjunctival limbal autograft or there is not enough tissue for a conjunctival limbal autograft, or it is contraindicated, and the company provides it with the discount agreed in the patient access scheme.

4.17 The committee took several factors into account when considering if Holoclar could be recommended to treat both eyes. Firstly, the clinical experts explained that usually only the worst eye would be treated in practice. Secondly, the committee discussed the very limited clinical evidence for this group; only 1 patient with bilateral disease had treatment in both eyes, which the committee considered to be uninformative. Thirdly, the ERG's clinical experts stated that there are plausible reasons why Holoclar would not be as effective when used in both eyes. Specifically, the biopsy must be taken from a damaged eye where it may be more difficult to locate and extract healthy limbal cells, and it may not be possible to take the same number of biopsies from a damaged eye in the case of transplant failure. Fourthly, the company's economic model analysis showed that treating both eyes gave ICERs for Holoclar that were too high for it to be considered a cost-effective use of NHS resources. The 2-eye economic model also included several limitations such as not accounting for differential effectiveness of a second Holoclar treatment and using a population of both unilateral and bilateral patients. Finally, the committee considered that for a patient with bilateral disease, there are treatment options available in current clinical practice to treat the second eye (conjunctival limbal allograft and keratolimbal allograft). Taking all of this into account, the committee recommended Holoclar for treating both eyes only in the context of clinical research.

4.18 The committee recommended that further research should be designed to generate robust evidence of the clinical- and cost-effectiveness of Holoclar in treating both eyes in people who do not have enough tissue for a conjunctival limbal autograft. The study should recruit people with bilateral disease and evaluate the effectiveness of treatment in both eyes. Outcomes should include transplant success and assessments of health-related quality of life using a generic preference-based measure.