Ranibizumab for treating diabetic macular oedema

The Committee considered the evidence presented by the manufacturer on the clinical effectiveness of ranibizumab. It agreed that, of the 4 RCTs identified, it was appropriate to concentrate on the 2 larger, more directly relevant trials, RESTORE and DRCR.net. It noted that both trials were judged to be of high methodological quality by the manufacturer, and that the ERG agreed with this assessment. The Committee was aware that the studies were of relatively short duration, providing treatment for up to 3 years, and did not include long-term follow-up. The Committee concluded that the quality of the clinical-effectiveness evidence for ranibizumab was acceptable.

The Committee noted that, based on the results of the RESTORE study, ranibizumab (alone or in combination with laser photocoagulation) is associated with immediate and sustained gains in visual acuity in the treated eye, whereas improvement with laser photocoagulation alone is significantly less marked. The Committee understood from the clinical trial evidence that adding laser photocoagulation to ranibizumab does not appear to provide any additional improvement in vision over 2 years. However, it was aware of the clinical specialists' belief that laser photocoagulation is more likely than ranibizumab to have long-term benefits, and that this could reduce the number of injections of ranibizumab needed after 2 years, although this cannot be confirmed using current evidence. The Committee concluded that, when compared with laser photocoagulation alone, treatment regimens that include ranibizumab are effective in improving visual acuity in the treated eye over 2 years, but that there is no evidence of additional benefit in adding laser photocoagulation to ranibizumab. The Committee was aware that this is inconsistent with the expectations of the clinical specialists.

The Committee noted that initial subgroup analyses in both RESTORE and DRCR.net had shown little evidence of differences in clinical effectiveness by subgroup. It understood from the manufacturer's evidence that, in RESTORE, gains in visual acuity associated with ranibizumab were greatest in participants with thicker retinas and more severe visual impairment at baseline. In particular, it noted the manufacturer's suggestion that ranibizumab could be expected to have a superior relative effect among people with thicker retinas (see section 3.30). The Committee was aware that the manufacturer had provided evidence in its rapid review submission to confirm this subgroup effect by testing for interaction between retinal thickness and treatment allocation. The Committee noted that the manufacturer's revised subgroup analyses according to retinal thickness relied on different categories of retinal thickness to those presented in NICE technology appraisal guidance 237 and also differed from those specified in the methods for subgroup analyses of the RESTORE study. The Committee also noted that changing these categories led to different results. However, the Committee acknowledged the clinical specialists' suggestion that laser photocoagulation would be less clinically effective in people with a thicker retina. Therefore, the Committee concluded that it had received sufficient evidence of biological plausibility for a clinically relevant subgroup in which ranibizumab has a significantly greater relative effect.

The Committee considered the generalisability of the results of the RESTORE study to people with diabetic macular oedema in clinical practice. The Committee heard from the clinical specialists that glycaemic control as reflected by HbA1c varies more in clinical practice than in trials, because the RESTORE trial excluded people with HbA1c values of 10% or more. The Committee also noted the comments on the use and effects of laser photocoagulation from the clinical specialists (see section 4.10), and discussed whether the trials accurately reflect the way in which ranibizumab would be combined with laser therapy in clinical practice. The Committee concluded that the generalisability of trial results to the population with diabetic macular oedema seen in clinical practice is uncertain. The Committee also expressed concerns about whether the proportion of people who were treated in both eyes in the RESTORE study reflected clinical practice. The Committee concluded that the lack of evidence on vision in both eyes, as presented by the manufacturer in its original submission for NICE technology appraisal guidance 237, increased its uncertainty about how the effects of ranibizumab demonstrated in the trials would translate into benefits for people with diabetic macular oedema in clinical practice.

The Committee discussed the appropriate approach for determining the cost effectiveness of ranibizumab. The Committee noted that the manufacturer's original base case and all additional original analyses performed by the manufacturer and the ERG suggest that combination therapy is more expensive and little or no more effective than ranibizumab alone. As a consequence, the Committee concluded that ranibizumab and laser photocoagulation as part of a simultaneous treatment strategy could not be recommended as an effective use of NHS resources. However, the Committee also heard from the clinical specialists that people currently treated with ranibizumab are likely to have had laser photocoagulation, which the clinical specialists believed is more likely to be associated with a long-term decrease in the recurrence of diabetic macular oedema than treatment with ranibizumab. The Committee agreed with the ERG's suggestion and consultation comments from retinal specialists that it is possible that ranibizumab and laser photocoagulation in a sequential treatment algorithm could provide a valuable treatment option. However, because no evidence to support this was available, the Committee concluded that it could not make separate recommendations on the sequential use of ranibizumab and laser photocoagulation.

When it reviewed the manufacturer's revised model submitted in NICE technology appraisal guidance 237, the Committee concluded that the model did not reflect likely clinical practice in at least 6 respects, as described in section 3.46. The manufacturer addressed these issues in its rapid review submission, and offered a revised patient access scheme (see sections 3.47 to 3.49).

The Committee discussed the manufacturer's approach to estimating the cost effectiveness of treating both eyes, in which it multiplied the ICERs generated in its better-seeing eye model by a factor of 1.5. The Committee was aware that this approach was consistent with that adopted by the Committee in NICE's technology appraisal guidance on ranibizumab and pegaptanib for the treatment of age-related macular degeneration and that the same approach was used by the Committee when estimating the most plausible ICER in NICE technology appraisal guidance 237. The Committee also considered the new, alternative approach taken by the ERG in its 6 scenario analyses, which made explicit assumptions about the impact on costs and outcomes associated with treating only the worse-seeing eye, only the better-seeing eye or both eyes. The Committee noted that the ERG's 6 scenario analyses varied both the health-related quality of life impact of changes in the vision of the worse-seeing eye and the resultant QALYs associated with treatment of the worse-seeing eye or both eyes. The Committee agreed that, without available data on health-related quality of life associated with vision in both eyes, these scenarios fully explored the impact of treating both eyes on the relative cost effectiveness of ranibizumab. The Committee noted that, although there is little evidence of the impact of vision in the worse-seeing eye on health-related quality of life, the Brown study suggested that among people who had good vision in their better-seeing eye, the worse-seeing-eye contributed little to health-related quality of life. The Committee therefore considered scenario analysis 3 to be consistent with previous appraisals, which suggested that changes in vision for people treated in their worse-seeing eye had 30% of the health-related quality of life impact of the same change in vision from treating the better-seeing eye. In response to the rapid review appraisal consultation document, the Royal College of Ophthalmologists commented that the ERG's approach seemed logical, but that scenario 4 might be more appropriate. However, in the absence of new empirical evidence to suggest otherwise, the Committee accepted that scenario 3 reasonably reflected the clinical situation for people with diabetic macular oedema. The Committee concluded that the ERG's more comprehensive approach to modelling treatment of both eyes was likely to be more valid than multiplying the ICER associated with treating the better-seeing eye by 1.5 because it explored more explicitly the impact on costs and outcomes associated with treating only 1 or both eyes.

The Committee considered the cost-effectiveness analyses presented by the manufacturer in its response to the appraisal consultation document for the rapid review of NICE technology appraisal guidance 237, which adopted the approach taken by the ERG (see sections 3.58 to 3.60). The Committee discussed the manufacturer's approach to estimating the proportion of people who would be treated in the better-seeing eye only, worse-seeing eye only or both eyes. The Committee noted that, as part of these 3 new analyses, the manufacturer presented data on the proportion of patients in RESTORE whom the manufacturer considered as having the same vision in both eyes at the start of treatment. The Committee was surprised that the manufacturer had not presented these data earlier in the appraisal. However, the Committee acknowledged that, although the manufacturer had originally proposed that patients would need treatment in only 1 eye despite bilateral disease, it had subsequently attempted to consider the cost effectiveness of treating both eyes in its rapid review submission. The Committee heard from the manufacturer that it considered the same-seeing eye pertinent only after the Committee had expressed its preference for the ERG's approach to estimating the costs and QALYs associated with treating both eyes. The Committee also heard from the manufacturer that it chose its definition of a same-seeing eye from a single study, that it was aware that other definitions existed, and that it had not taken a systematic approach to assessing other possible definitions. The Committee noted that the ICERs did not vary substantially between the 3 analyses presented by the manufacturer. The Committee concluded, given its concerns around the definition of same-seeing eyes, that analysis 2 (which suggested that the same-seeing eye has a visual acuity within 1 letter of the other eye) was the most plausible of the 3 new analyses presented by the manufacturer.

The Committee considered the utility values that quantified the changes in health-related quality of life attributed to vision in the manufacturer's new model in its rapid review submission. It was aware that the manufacturer used utility values estimated from the study by Czoski-Murray et al., in which members of the UK general public valued levels of visual impairment simulated by wearing vision-worsening contact lenses in both eyes. The Committee noted that this study resulted in a broader range of utility values between best and worst health states in the model than those used in the manufacturer's original submission in NICE technology appraisal guidance 237, which were estimated from the RESTORE study and which the Committee considered to be unrealistically large (see section 3.26). The Committee also noted that participants in the Czoski-Murray study wore the lenses simulating bilateral visual impairment for a short period of time (between 1.5 and 2 hours). Therefore, the Committee considered that the participants may have overstated the detrimental impact on health-related quality of life of visual impairment in both eyes because they had little time to adjust to it. The Committee was aware that the Brown study identified by the ERG measured health-related quality of life directly from patients with impaired vision in at least 1 eye, and that this produced a narrower range of utility values than the study by Czoski-Murray et al. In consideration of a comment from the manufacturer who suggested that the utility values from the Brown study were not in line with the NICE reference case, the Committee noted that neither set of utility values was in line with the NICE reference case for measuring and valuing health effects. The Committee concluded that there was uncertainty about which utility data were most appropriate to include in the model. However, the Committee agreed that, in the absence of further evidence, it was reasonable to assume that the range of utility values would probably lie somewhere in between those estimated from the Czoski-Murray and Brown studies.

The Committee discussed the manufacturer's assumptions about how often people would receive ranibizumab in clinical practice. The Committee noted that, on the basis of the results from an extension to the RESTORE study available at the time of the rapid review submission, the manufacturer assumed that people would need 4 injections in year 2 and 3 injections in year 3 and no more injections from year 4 onwards. The Committee commented that it was unlikely that people who received 3 injections in year 3 would receive no further injections in year 4, especially if vision was assumed to remain stable during this period. The Committee was aware that some consultees had suggested that people with diabetic macular oedema would need more frequent treatment with ranibizumab than was assumed by the manufacturer. The Committee also noted that uncertainty remained about whether people would need ranibizumab beyond 4 years and, if they did, what the costs of ongoing treatment would be. However, the Committee acknowledged that the manufacturer had attempted to address this uncertainty by conducting a threshold analysis to assess the maximum number of injections per person that could be administered while maintaining an ICER below £30,000 per QALY gained. The Committee was aware that these analyses allowed for only 4 additional injections in the first 3 years of the model. The Committee concluded that, without longer-term clinical data, significant uncertainty remained about the number of ranibizumab injections that people with diabetic macular oedema are likely to need.

The Committee considered the manufacturer's assumptions about how vision in the treated eye improved and deteriorated beyond the third year of the model, when the model assumes that ranibizumab treatment finishes. It understood that, although vision deteriorated over time in the model, it did so at the same rate in people previously treated with ranibizumab as in people who had previously been treated with laser photocoagulation. The Committee was aware of the opinion of clinical specialists that the most important effect of ranibizumab is to reduce the permeability of blood vessels and oedema in the eye, and heard from the clinical specialists that it is implausible that this effect would persist in the long term. By contrast, the benefits to vision from laser photocoagulation, although not as great as those of ranibizumab, are believed to last longer. The Committee noted that the manufacturer reduced uncertainty about the projected effects of ranibizumab treatment by following the approach discussed by the Committee in NICE technology appraisal guidance 237 and adopting a 10-year time horizon. The Committee also noted that this approach was consistent with previous appraisals. The Committee was aware of the new clinical evidence submitted by consultees in their response to the rapid review appraisal consultation document. The Committee understood that the consideration of such new clinical evidence on the long-term clinical benefits of the comparator treatment laser photocoagulation is beyond the remit of a rapid review, and would require a full review of the appraisal. Therefore the Committee concluded that, although significant uncertainty remains about the long-term benefit of ranibizumab treatment, compared with the manufacturer's original submission, the rapid review model more accurately reflects the duration of benefit that could be expected from treatment with ranibizumab.

The Committee considered the manufacturer's assumptions about the number of treatment and monitoring visits for people treated with ranibizumab and those treated with laser photocoagulation. The Committee was aware that in its original submission in NICE technology appraisal guidance 237, the manufacturer had assumed that a visit for treatment with ranibizumab would double as a monitoring visit and that people treated with laser photocoagulation would need a separate monitoring visit. The Committee was unaware of any clinical evidence to justify this difference, and the manufacturer had not explained the difference in its original submission or in consultation comments. The Committee noted that the manufacturer had addressed this issue in its rapid review submission by assuming that a treatment visit for people receiving ranibizumab or laser photocoagulation doubles as a monitoring visit. The Committee concluded that, compared with the manufacturer's original submission, the rapid review model provided a more plausible reflection of the number of treatment and monitoring visits that people receiving ranibizumab treatment or laser photocoagulation would need.

The Committee considered whether the revised base-case model applies to the population with diabetic macular oedema in England and Wales. It noted the clinical specialists' advice that glycaemic control as reflected by HbA1c is likely to be worse in clinical practice than in the RESTORE trial, which excluded people with HbA1c levels of 10% or more (see section 4.12). The Committee observed that a subgroup analysis provided as part of the manufacturer's original submission for NICE technology appraisal guidance 237 suggested that restricting the analysis to people with good glycaemic control (HbA1c less than 8%) produced a much lower ICER than the ICER based on the group of people with poor control (HbA1c 8% or more; see section 3.18). The Committee noted that the manufacturer did not provide further analyses for these subgroups in its rapid review submission because of the relatively small sample sizes, which may have resulted in a small number of people in the extreme health states influencing the results. The Committee was aware of the new clinical evidence submitted by the manufacturer in response to the consultation on the rapid review appraisal consultation document. The Committee understood that submission of such further evidence would not normally be expected in the context of a rapid review, and accepted that this evidence could not be considered without formal review by the ERG. The Committee acknowledged that the issue of glycaemic control had been considered in NICE technology appraisal guidance 237, and that the Committee's considerations had been upheld at appeal. Based on the evidence provided in the manufacturer's original submission, the Committee agreed that uncertainty remained about the cost effectiveness of ranibizumab in people with poorer glycaemic control. Therefore, the Committee concluded that the manufacturer's model would probably generate a higher ICER if it was more reflective of the population seen in routine clinical practice.

The Committee discussed what could be considered as the most plausible ICERs. In the Committee's view, ICERs reflecting the possibility of treating both eyes were the most useful starting points for considering the cost effectiveness of ranibizumab for treating diabetic macular oedema. The Committee was aware that the manufacturer's rapid review base-case model produced an ICER of £21,200 per QALY gained for treating both eyes by multiplying the ICER for the better-seeing eye model by a factor of 1.5. The Committee agreed that this ICER was from a model that relied on a more plausible set of assumptions than those used in the manufacturer's original submission for NICE technology appraisal guidance 237. However, the Committee also acknowledged the ERG's technically more comprehensive approach of accounting for treatment in both eyes and noted that the manufacturer acknowledged the advantages of this approach. The Committee noted that this approach was subsequently adopted by the manufacturer in its response to the rapid review appraisal consultation document and led to ICERs in the range of £24,600 to £31,600 per QALY gained depending on the utility values used in the model for the Committee's preferred analysis. The Committee agreed that these ICERs would increase if the model accounted for people needing more than 4 treatments with ranibizumab beyond the third year, if people who had laser photocoagulation maintained any improvements in vision after treatment for longer than people treated with ranibizumab, and if the model better reflected the population with poorer glycaemic control seen in routine clinical practice. The Committee concluded that the most plausible ICER was likely to be above £30,000 per QALY gained, and that it therefore could not recommend ranibizumab as an effective use of NHS resources for the treatment of all people with diabetic macular oedema.

The Committee discussed whether it had received evidence of any groups of people for whom ranibizumab could be considered an effective use of NHS resources. It noted that, in its rapid review submission, the manufacturer provided additional subgroup analyses that showed that ranibizumab has a lower ICER in people with thicker retinas (central retinal thickness of 400 micrometres or more) than in people with thinner retinas (central retinal thickness of less than 400 micrometres) at the start of treatment. The Committee recognised the clinical plausibility of a greater relative efficacy of ranibizumab in people with a central retinal thickness of 400 micrometres or more, because it understood that laser photocoagulation may be less effective when used on a thicker retina. The Committee noted that the manufacturer had presented statistical evidence of greater clinical effectiveness in this predefined group. The Committee also noted that the manufacturer had reduced the impact of small sample sizes, which had been raised as a concern in NICE technology appraisal guidance 237, by combining groups of people with thinner retinas (central retinal thickness less than 300 micrometres and 300 to 400 micrometres) into 1 larger group (people with central retinal thickness greater than 400 micrometres). This also produced more plausible cost-effectiveness results across the 2 groups. The Committee also acknowledged that the manufacturer had adequately accounted for differences in costs and outcomes for these subgroups by making adjustments to subgroup‑specific parameters for other important model inputs. The Committee therefore concluded that the manufacturer had provided robust evidence demonstrating a subgroup effect in favour of people with thicker retinas.

The Committee considered the most plausible ICERs for people with thicker retinas (central retinal thickness of 400 micrometres or more) at the start of treatment. The Committee noted that the manufacturer's rapid review model produced an ICER of £13,300 per QALY gained for treating both eyes in this group by multiplying the ICER for the better-seeing eye model by a factor of 1.5. The Committee agreed that this ICER would increase if the model accounted for people needing more frequent treatment with ranibizumab beyond the third year, if people who had laser photocoagulation maintained any improvements in vision after treatment for longer than people who had ranibizumab, if the model better reflected the population with poorer glycaemic control seen in routine clinical practice, and if people with thicker retinas had higher rates of mortality than people with thinner retinas. The Committee also noted that neither the ERG nor the manufacturer provided exploratory scenario analyses for people with central retinal thickness of 400 micrometres or more. However, the Committee agreed that the ICER would likely increase if the ERG's approach of adapting the manufacturer's model to consider treating both eyes was used along with the Committee's preferred assumptions. The Committee therefore concluded that the most plausible ICER for the subgroup of people with thicker retinas was likely to be higher than the manufacturer's estimate, but would be under £25,000 per QALY gained. The Committee also considered the manufacturer's suggestion in response to the rapid review appraisal consultation document that ranibizumab would be cost effective for the whole population. However, the Committee noted that the ICER for treating people with thinner retinas (central retinal thickness of less than 400 micrometres) was £43,300 per QALY gained in the manufacturer's rapid review model and that ranibizumab would therefore not be an effective use of NHS resources in this group. Therefore the Committee recommended ranibizumab as an option for treating visual impairment due to diabetic macular oedema only if the eye has a central retinal thickness of 400 micrometres or more at the start of treatment.

The Committee noted that the scope for the appraisal included bevacizumab as a comparator. It was also aware of its previous conclusions on bevacizumab in NICE technology appraisal guidance 237, that ranibizumab at that time did not represent an effective use of NHS resources when compared with laser photocoagulation, and that therefore the Committee did not believe that considering evidence for bevacizumab would have altered its decision. However, because the Committee had concluded after the rapid review that ranibizumab represents a cost-effective use of NHS resources when compared with laser photocoagulation for people with a central retinal thickness of 400 micrometres or more, it discussed the comparison of ranibizumab with bevacizumab. The Committee noted and reviewed information from the regulatory authorities to prescribers in which the use of bevacizumab as an intravitreal injection in people with diabetic macular oedema is considered 'unlicensed'. Also, the Committee heard conflicting evidence about the extent to which bevacizumab is currently used to treat diabetic macular oedema in England and Wales. It concluded that bevacizumab is adopted by some clinicians and funded by some NHS trusts, but is not in use throughout the NHS. The Committee was aware that some consultees and commentators supported a comparison with bevacizumab and others opposed it. The Committee discussed whether a cost-effectiveness analysis of ranibizumab compared with bevacizumab was possible. The Committee recognised that a formal comparison of the 2 drugs would need evidence not only of all aspects of clinical effectiveness and safety, but also of the costs associated with preparing and administering bevacizumab, including the dose and number of injections needed. The Committee agreed that such evidence, in particular about the balance of harms and benefits associated with bevacizumab, was not readily available for people with diabetic macular oedema. The Committee also noted that it was unaware of any evidence of the effectiveness of intravitreal bevacizumab compared with ranibizumab in the subgroup of patients with thicker retinas. The Committee agreed that, taking into account all these uncertainties, it could not consider a comparison of ranibizumab with bevacizumab. The Committee also concluded that further research directly comparing the clinical and cost effectiveness of ranibizumab and bevacizumab in people with diabetic macular oedema would reduce some of these uncertainties.

The Committee discussed whether ranibizumab should be considered an innovative treatment. It considered that, in terms of both pharmacological progress and potential benefits for people with diabetic macular oedema, the development of the anti-angiogenic drugs pegaptanib sodium and bevacizumab preceded that of ranibizumab. Therefore, the Committee concluded that ranibizumab itself could not properly be considered to provide distinctive pharmacological innovation. The Committee further noted that the analyses of the incremental health benefit of ranibizumab were based on a comparison with laser photocoagulation, and that the Committee had not been alerted to any benefits that were not already captured in the QALY measure. The Committee was also aware that, before NICE technology appraisal guidance 237 was published, the Committee's conclusions on innovation as described above were upheld by the Appeal Panel. It therefore concluded that the incremental value of ranibizumab for people with diabetic macular oedema had been appropriately captured.

The Committee discussed the proposed date for review of the guidance. The Committee was aware of the emerging evidence on the effectiveness and safety of bevacizumab as a treatment option for diabetic macular oedema, including work undertaken by NICE's Decision Support Unit and ongoing clinical trials comparing bevacizumab with ranibizumab in diabetic macular oedema and other eye diseases. The Committee was also aware that additional clinical data, including 3-year results from the DRCR.net study, had become available since the publication of NICE technology appraisal guidance 237, but that these data could not be considered as part of the rapid review process. The Committee heard that some commentators suggested that the proposed date for review should be earlier than February 2016, because the guidance would exclude ranibizumab as a treatment option for a significant proportion of people with diabetic macular oedema. Therefore, the Committee agreed that the proposed date for review of the guidance should be brought forward to February 2015.

The Committee discussed whether NICE's duties under the equalities legislation required it to alter or add to its preliminary recommendations in any way. During the scoping phase of the appraisal, NICE had received evidence that some people in full-time residential care had restricted access to treatment for diabetic macular oedema. However, consultees suggested that the national screening programmes for diabetic retinopathy in England and Wales has reduced this inequality across the NHS. In submissions, the Committee had been made aware that there is a higher prevalence of diabetes in people of South Asian, African and African–Caribbean family origin and that, among people with diabetes, sight-threatening eye disease is more common in people of African and African–Caribbean family origin than in white Europeans. However, the Committee agreed that this was an issue that could not be addressed in a technology appraisal.