3 The company's submission

The Appraisal Committee (section 7) considered evidence submitted by the company on aflibercept and a review of this submission by the Evidence Review Group (ERG; section 8).

Clinical effectiveness

3.1 The main sources of evidence presented in the company's submission came from 2 ongoing phase‑III trials: VIVID and VISTA. VISTA (n=466) is a double‑blind, randomised (1:1:1) active‑controlled superiority study carried out at 54 sites in the USA. VIVID (n=406) is an ongoing prospective, randomised, double‑blind, active‑controlled superiority study carried out at 73 sites across Japan, Europe and Australia. Both trials administered once‑monthly intravitreal doses of 2 mg aflibercept for 5 months followed by either aflibercept 2 mg every 4 weeks (2Q4) or aflibercept 2 mg every 8 weeks (2Q8) with laser photocoagulation.

3.2 The primary outcome in the trials was the mean change from baseline to 52 weeks in best corrected visual acuity (BCVA), based on the Early Treatment of Diabetic Retinopathy Study (ETDRS) letter score, in eyes with diabetic macular oedema (DMO) involving the centre of the macula, for aflibercept compared with laser photocoagulation. The results showed a statistically significant improvement in BCVA with aflibercept compared with laser photocoagulation in both VISTA and VIVID. The mean treatment difference for aflibercept compared with laser in the 2Q8 group of VISTA was 12.19 (97.5% confidence interval [CI] 9.35 to 15.04) and in VIVID 9.05 (97.5% CI 6.35 to 11.76).

3.3 The secondary outcomes of the trials included: the proportion of patients gaining 10 or more ETDRS letters and 15 or more ETDRS letters from baseline to week 52; the mean change in central retinal thickness (CRT) from baseline to week 52, as assessed on ocular coherence tomography; vision‑related quality of life (assessed by the National Eye Institute Visual Functioning Questionnaire‑25 [NEI VFQ‑25]); and quality of life (assessed by the EuroQol‑5 dimension health questionnaire [EQ‑5D]). The results showed a statistically significant improvement for outcomes in both trials in people having aflibercept compared with laser. For the NEI‑VFQ‑25, the VISTA trial showed a statistically significant advantage with aflibercept 2Q4 (but not with aflibercept 2Q8) on the Near Activities subscale compared with laser.

3.4 The company carried out subgroup analyses of the secondary endpoints, including for baseline CRT (less than 400 micrometres or 400 micrometres or more) and previous cataract surgery (presence of pseudophakic lens). The results of these analyses were submitted as academic in confidence.

3.5 The company completed further analyses on efficacy outcomes, which included losing 5 or more, 10 or more and 15 or more ETDRS letters from baseline. The results at week 52 in the 2Q4 and 2Q8 groups in both trials showed a smaller proportion of patients in the aflibercept group losing 5 or more, 10 or more and 15 or more ETDRS letters compared with the laser group.

3.6 The company undertook a meta‑analysis of the results from the VISTA and VIVID trials for some outcomes using both fixed and random effects models. The meta‑analysis comparisons were made exclusively between aflibercept 2Q8 and laser because this is the dose that has a marketing authorisation in the UK. The results from the meta‑analysis showed a greater gain in mean BCVA from baseline to 12 months with aflibercept, when compared with laser. The results also indicated that a higher proportion of patients treated with aflibercept achieved a gain of 10 or more ETDRS letters or 15 or more ETDRS letters, from baseline to 12 months, when compared with laser photocoagulation, and that a lower proportion of patients treated with aflibercept lost 15 or more ETDRS letters or 10 or more ETDRS letters, from baseline to 12 months, when compared with laser treatment. All of these results were statistically significant.

3.7 The company collected EQ‑5D health‑related quality‑of‑life data during the pivotal trials at baseline, week 24 and week 52. A regression analysis was done to estimate the relationship between BCVA (in both eyes) and quality of life. Patients completed the EQ‑5D questionnaires and the quality‑of‑life estimates were based on a general population tariff. The mean total change score from baseline to 52 weeks in VIVID, the European arm of the study, was provided by the company as academic in confidence in the submission and therefore cannot be presented.

3.8 The company presented the safety data from the VISTA and VIVID trials, which showed that aflibercept had a favourable safety profile at 2 years in people with DMO.

3.9 The company did a systematic review to identify studies for inclusion in the network meta‑analysis of aflibercept 2Q8 with ranibizumab. The company used the pooled estimates from the meta‑analyses for aflibercept 2Q8 with laser. The indirect comparisons showed statistically significant improvement in the BCVA mean change from baseline in favour of anti‑vascular endothelial growth factor (VEGF) treatments (both aflibercept and ranibizumab) compared with laser. Results of the network meta‑analysis showed a statistically significant improvement in visual acuity as measured by BCVA mean change from baseline and loss of 10 or more ETDRS letters for aflibercept 2Q8 compared with ranibizumab. There was no significant difference between aflibercept and ranibizumab for alternative visual acuity outcomes (gain of 15 or more ETDRS letters, loss of 15 or more ETDRS letters and gain of 10 or more ETDRS letters) or safety outcomes. An indirect comparison with the pivotal MEAD study for dexamethasone was not possible because there was no common comparator. The company stated that a comparison of aflibercept with fluocinolone acetonide was not possible because there was no common comparator for an indirect analysis.

Cost effectiveness

Company's original submission

3.10 The company provided a bilateral vision, state transition Markov model in which each eye was in 1 of 8 possible health states. The 2 worst health states, HS7 and HS8, represented blindness. The baseline age of patients in the model was 63 years and the proportion of women patients was 42.1%. These values were based on the population enrolled in VIVID and VISTA. The proportion of fellow eye involvement at baseline (46.5%) was drawn from expert opinion rather than the VIVID and VISTA trial data. The starting vision health state distributions for the study eye and the fellow eye were estimated from the baseline characteristics of participants in the integrated VIVID and VISTA trial analyses.

3.11 The company used the results of the VISTA and VIVID trials to inform laser efficacy in the model (BCVA based on gaining or losing 10 or 15 ETDRS letters). Aflibercept and ranibizumab efficacy in the model were based on the probabilities of gaining or losing 10 or 15 ETDRS letters, which were estimated by applying the relative risks calculated as part of the network meta‑analysis. Dexamethasone efficacy was based on the probability of gaining 10 ETDRS letters, which was estimated by applying the relative risks from the indirect comparison of aflibercept with dexamethasone using the PLACID study. For the comparison of aflibercept with fluocinolone acetonide, rates of improvement were taken directly from the FAME trial (that compared fluocinolone acetonide with sham fluocinolone acetonide); rates of worsening were assumed to be the same as those of laser from the VISTA and VIVID trials.

3.12 The health states in the company's economic model were defined by vision in both eyes and therefore health state utilities (and hence quality‑adjusted life years [QALYs]) account for the better seeing eye and the worse seeing eye. This approach needed 36 utility values to account for every possible combination of the better seeing eye and worse seeing eye. The company used 4 sources of health‑related quality‑of‑life data in its cost‑effectiveness analyses. For its base‑case analyses, the company used utility values from Czoski‑Murray et al. (2009). The company stated that they had used these values because they had been accepted by the Appraisal Committee during the appraisal of other technologies for DMO (NICE's technology appraisal guidance on ranibizumab for treating diabetic macular oedema and fluocinolone acetonide intravitreal implant for treating chronic diabetic macular oedema after an inadequate response to prior therapy). The company also used utility values from the EQ‑5D data collected in the pivotal trials (submitted as academic in confidence), and Brown (1999) and Brown (2000) in the sensitivity analyses (see tables 1 and 2). The company's submission provided details of an ordinary least squares analysis of the pooled VIVID and VISTA EQ‑5D data. This regresses quality of life on the BCVA logarithms of the better seeing eye and of the worse seeing eye. The utility values from Czoski‑Murray et al. (2009), Brown (1999) and Brown (2000) apply to the BCVA in both eyes (bilateral), therefore the company applied a 30% utility decrement to the better seeing eye to estimate the utility of each corresponding health state in the worse seeing eye (resulting in a proportional decrement of 23%). The company assumed a constant utility in each health state meaning utility changes were only in relation to BCVA and not the duration spent in the health state. For adverse events, disutilities were applied for cataract, endophthalmitis, retinal detachment, glaucoma, vitreous haemorrhage and raised intraocular pressure.

Table 1 BCVA quality of life: values for the better seeing eye

	Czoski‑Murray et al. (2009)	Brown (1999)	Brown (2000)
HS1	0.856	0.839	0.890
HS2	0.764	0.839	0.890
HS3	0.690	0.783	0.810
HS4	0.617	0.783	0.690
HS5	0.543	0.732	0.558
HS6	0.469	0.681	0.545
HS7	0.396	0.630	0.520
HS8	0.263	0.579	0.460

Table 2 BCVA quality of life: values for the worse seeing eye

	Czoski‑Murray et al. (2009)	Brown (1999)	Brown (2000)
HS1	0.856	0.839	0.890
HS2	0.835	0.839	0.890
HS3	0.818	0.826	0.872
HS4	0.801	0.826	0.844
HS5	0.784	0.814	0.813
HS6	0.767	0.803	0.810
HS7	0.750	0.791	0.805
HS8	0.720	0.779	0.791

3.13 In the company's cost‑effectiveness analyses the unit cost of treatment with aflibercept was modelled using the confidential patient access scheme. Adverse event costs were taken from NHS reference costs. In addition, an average annual cost of blindness of £6448 was obtained from the literature, and updated for inflation. This cost was applied to both eyes in the HS7 or HS8 health states. The company modelled 8 aflibercept treatments in the first year, in line with the marketing authorisation and assumed all treatment visits would double as monitoring visits. People in the ranibizumab arm were assumed to receive 7.93 ranibizumab treatments in the first year along with 12 monitoring visits. The company acknowledged that the summary of product characteristics for ranibizumab had recently changed to reduce the number of monitoring visits needed in the first year. This change was not included in the model because it was not considered by the company to be established practice in England.

3.14 The company's base‑case incremental cost‑effectiveness ratios (ICERs) reported that aflibercept (all ICERs reported in the company's analysis used the confidential patient access scheme price for aflibercept) dominated (is more effective and less costly than) laser and ranibizumab (when the list price of ranibizumab was used). The company explored the effect of different patient access scheme discounts on the list price of ranibizumab. The results showed that the ICER for aflibercept remained under £20,000 per QALY gained up to a ranibizumab price discount of 70%.

3.15 The company's scenario analyses compared aflibercept with dexamethasone and fluocinolone acetonide. These analyses showed that aflibercept (using the confidential patient access scheme price) dominated both dexamethasone and fluocinolone acetonide. The company undertook exploratory analyses comparing aflibercept with fluocinolone acetonide in a subgroup of patients with pseudophakic lenses. In this comparison, probabilities of gaining 10 and 15 ETDRS letters were obtained from the pseudophakic subgroup in VIVID and VISTA and from the FAME trial (gaining 15 or more ETDRS letters for fluocinolone acetonide). The company presented the results of the exploratory analyses using both the list price of fluocinolone acetonide and applying various discounts (0% to 100%). Aflibercept continued to dominate fluocinolone acetonide in all of the subgroup analyses.

3.16 The company performed a scenario analysis for a subgroup of patients with CRT of 400 micrometres or more, using the probabilities of gaining and losing ETDRS letters for the subgroup of patients in the VIVID and VISTA trials. The results showed that aflibercept dominated both laser and ranibizumab.

Evidence Review Group comments on the company's original submission

3.17 The ERG commented that the main entry criterion for VIVID and VISTA was a CRT of 1 micrometre in the central retina (defined as clinically significant macular oedema). The ERG stated that this is usually determined by ocular coherence tomography, but the company did not mention its use in the submission. Therefore, at entry, patients may or may not have fulfilled the standard definition of clinically significant macular oedema. However, clinically significant macular oedema was used as the re‑treatment criterion for laser photocoagulation therapy. The ERG stated that it could be argued that the initial laser treatment was not based on the presence of clinically significant macular oedema, whereas the re‑treatments were; the rationale for this was unclear to the ERG. The ERG noted that it was also not specified whether fluorescein angiography was done before laser treatment to guide the laser (as recommended by the ETDRS).

3.18 The ERG noted that patients in VIVID had a significantly higher mean CRT than patients in VISTA for the laser and aflibercept 2Q8 groups. The ERG commented that this may be important because there is evidence that the clinical effectiveness of anti‑VEGF treatment for DMO varies according to baseline CRT measurements. The ERG noted that more eyes in VISTA had previous anti‑VEGF treatment than eyes in VIVID (42.9% compared with 8.9%, respectively). The ERG also noted that about half of the patients in VISTA had also had previous laser photocoagulation treatment in the study eye. The ERG commented that the mean HbA1c across VISTA and VIVID was 7.6 to 7.9, which is lower than most people seen in clinical practice in England, who often have HbA1c levels over 8 or 9. Therefore, it is possible that aflibercept may be less effective in clinical practice than in the results of the pivotal trials. The ERG considered that the integrated analysis was not appropriate because the VISTA and VIVID trials differed significantly in the proportion of patients who had previous anti‑VEGF treatment and in the mean CRT.

3.19 The ERG identified aspects of the company's original base‑case model that involved errors in the data analysis and also identified parameter values, which it preferred. These are listed below:

Revising the number of aflibercept injections in the first year from 8.0 to 8.50. The ERG noted that the dosing specified in the summary of product characteristics for aflibercept suggested that it could be applied every 4 weeks, which would result in a mean number of 8.50 injections in the first year.
Revising the number of aflibercept injections in year 2 from 4.0 to 5.45. The ERG noted that the company had assumed an equal number of injections in the second year for ranibizumab and aflibercept. The ERG obtained the mean number of injections reported in VISTA and VIVID (5.45) and used this in its revised analysis.
Revising the number of ranibizumab injections in the first year to 7.93. The ERG noted that the number of injections of ranibizumab may have been overestimated by the company. The ERG noted that the company combined the mean number of injections reported in the RESTORE and REVEAL trials with the median number from the DRCR.net trial. The ERG commented that it may not be appropriate to combine the values in this way.
The number of monitoring visits for ranibizumab (12) in the first year may have been overestimated because the recently revised summary of product characteristics for ranibizumab removes the need for additional hospital monitoring visits in the first year of treatment.
Correcting the cost of blindness in the company model and applying a discount of 3.5%. The ERG commented that the method used in the company's model to inform the cost of blindness (£6448) was overestimated because it used the annual amount monthly and it was not discounted.
The company used utility values from Czoski‑Murray (2009) in its base‑case analyses. For these values to fit the better seeing eye and worse seeing eye states separately, the company has allowed for a proportion of the better seeing eye utility impact for a given change in the health state to apply to the same change in the health state of the worse seeing eye. The ERG noted a discrepancy in the decrement applied to the worse seeing eye, which was stated as 30% in the company's submission but resulted in a proportional difference of 23%.
Not treating eyes in HS7 and HS8 health states (blindness) during the maintenance phase, as in the company's base case.

3.20 The results of the ERG's base‑case analysis comparing aflibercept (using the confidential patient access scheme discount price) with laser reported an ICER of £33,921 per QALY gained (incremental QALYs 0.381; incremental costs cannot be reported because these were considered commercial in confidence).

3.21 The results of the ERG's base‑case analysis comparing aflibercept with ranibizumab showed that aflibercept dominated ranibizumab when the list price of ranibizumab was used. The ERG presented the results over various ranibizumab patient access scheme discounts. The ICERs ranged from dominant (0% discount) to £111,215 per QALY gained (100% discount).

3.22 The ERG also did sensitivity analyses. The parameters that were changed included:

exploring the use of the VIVID and VISTA EQ‑5D utility data using the ordinary least squares, random effects and generalised estimating equation models
exploring the health‑related quality‑of‑life values from Brown (1999) and Brown (2000)
excluding the REVEAL trial relative risks (trial was based on a predominantly Asian population)
exploring the effect of reducing the proportion of people in the model who were blind and needed residential care from 30% to 20% (that reduced the annual average cost of blindness from £7429 to £5640).

3.23 The results of the ERG sensitivity analyses for aflibercept compared with laser showed ICERs above £30,000 per QALY for all parameters with the exception of using the Brown (2000) values (ICER £29,915 per QALY gained). The results of the ERG sensitivity analyses for aflibercept compared with ranibizumab (over various patient access scheme discounts) showed ICERs that ranged from aflibercept dominating ranibizumab to ICERs of up to £1,260,695 per QALY gained (100% ranibizumab discount using the EQ‑5D generalised estimating equation analysis). The ERG noted that in these analyses the choice of quality‑of‑life values had the biggest effect on the ICER.

3.24 The ERG did an additional cost‑effectiveness analysis to examine the CRT subgroups for aflibercept compared with laser (less than 400 micrometres or 400 micrometres or more). The ERG used the company's post hoc analysis of CRT subgroups to calculate the relative risks from the VIVID and VISTA trials of aflibercept compared with laser in gaining or losing 10 or 15 ETDRS letters. The ERG used these relative risks to derive probabilities to recalculate the ICERs for the two subgroups using the ERG's base case. The results of the cost‑effectiveness subgroup analysis showed the ICER for aflibercept compared with laser of £21,958 per QALY gained in the CRT 400 micrometres or more group and £49,421 per QALY gained in the CRT less than 400 micrometres group.

New evidence submitted by the company following consultation on the appraisal consultation document

3.25 The company was granted permission to provide a new cost‑effectiveness analysis for the comparison of aflibercept with laser in the whole trial population. The new analysis used the ERG's original base‑case analysis (see section 3.19), which was discussed by the Committee and documented in the appraisal consultation document (ACD). The company provided new evidence (see section 3.26) and further rationale (see section 3.27) to support reinstating its original base‑case assumptions for some of the parameters.

3.26 The company presented new evidence to support its view that the ERG had overestimated the number of aflibercept injections in year 1 and year 2 and underestimated the cost of a laser administration visit in its base‑case analysis. The company's new evidence included:

An online survey of 10 ophthalmologists to establish the mean number of injections and monitoring visits in each year of treatment. The results of the survey showed the average number of injections of aflibercept were 6.5 and 4.1 in years 1 and 2 respectively. The company commented that this was lower than that estimated in the VIVID and VISTA trials and supports the hypothesis that a similar number of injections should be assumed for aflibercept and ranibizumab in years 1 and 2. The results for the number of monitoring visits of aflibercept showed a mean of 5.6 in year 1 and 4.2 in year 2.
An online survey of 34 ophthalmologists to establish the mean time taken for a laser visit compared with an intravitreal injection visit. The results of the survey showed the mean time spent with a patient for a laser visit was 23.7 minutes compared with 22.30 minutes for an intravitreal injection visit. The company commented that this supports increasing the cost of a laser administration visit to equal the cost of an intravitreal injection visit.

3.27 The company's new analysis included the following amendments to the ERG's original base‑case analysis:

Increased aflibercept injections in year 1 from 8.50 to 8.55 and decreased the injections in year 2 from 5.45 to 4.0. The company also assumed the same number of injections in year 1 and year 2 for aflibercept and ranibizumab. The company stated that the estimated number of aflibercept injections in year 1 and year 2, used in the ERG's base‑case analysis, did not reflect the summary of product characteristics for aflibercept, or clinical opinion (obtained from the online survey of 10 UK ophthalmologists; see section 3.26). The company also stated that the online survey supported the assumption that a similar number of injections should be assumed for ranibizumab and aflibercept in year 1 and year 2.
Increased cost of a laser administration from £139 to £256. The company stated that based on the results of the online survey of 34 ophthalmologists (see section 3.26), it was appropriate to assume at least a similar administration cost for injection and laser visits. The company acknowledged however, that it was aware that the actual cost varied across England.
Increased laser monitoring visits from 4.00 to 12.00 in the first year to equal the number of laser visits in VISTA and VIVID. The company highlighted that the number of laser administrations in VIVID, VISTA and the ranibizumab trials was based on monthly monitoring visits because this was when the decision to administer an injection was undertaken.
Used 46% instead of 85% for the rate for fellow eye involvement at baseline. The company commented that the rate of fellow eye involvement assumed at baseline and accepted by the committee in NICE's technology appraisal guidance on ranibizumab for treating diabetic macular oedema was 35%. It also commented that it was aware of several epidemiological studies that suggested the rate of clinically significant DMO is lower than any other form of DMO in the UK.
Used an alternative mortality rate from Preis et al. (2006). The company stated that the ERG had used a mortality rate of 2.45 based on Mulnier et al. (2006), which was used in other NICE technology appraisals of treatments for eye conditions including NICE's technology appraisal guidance on ranibizumab for treating diabetic macular oedema. The company commented that the Preis. et al. study was a more recent study than Mulnier. et al., but it acknowledged that both studies had strengths and limitations.
Used patient‑level data from the VIVID and VISTA trials to inform the transition probabilities for aflibercept. The company commented that the ERG had stated in its original report that the use of patient‑level data would be more appropriate. The company noted that no patient‑level data was available for ranibizumab.

3.28 The company's new economic analysis reduced the ICER for aflibercept (with the confidential patient access scheme applied) compared with laser in the whole trial population from £33,921 per QALY gained (see section 3.20) to £21,718 per QALY gained.

Evidence Review Group's response to the company's new evidence and analysis following consultation

3.29 The ERG reviewed the new evidence presented by the company for the comparison of aflibercept with laser in the whole trial population (see section 3.26) and the company's amendments to the ERG's original base case (see section 3.27).

Decreased aflibercept injections in year 1. The ERG questioned the validity of the results from the online survey of 10 ophthalmologists, which suggested a mean of 6.5 aflibercept injections in year 1. The ERG commented that the results suggested that the ERG had either misinterpreted the summary of product characteristics for aflibercept or that the ophthalmologists surveyed anticipated a higher discontinuation rate than was observed in VIVID and VISTA. The ERG highlighted that based on the wording of the summary of product characteristics for aflibercept, calendar‑month dosing would imply 8 administrations in the first year while 4‑weekly dosing would imply 8.55 (that was in line with the mean number of administrations during VIVID and VISTA).
Decreased aflibercept injections in year 2. The ERG stated that the amendment for the number of aflibercept injections in year 2 encompasses 2 changes: i) equivalence of treatment numbers with ranibizumab and ii) applying the number of ranibizumab injections from the RESTORE extension study rather than the higher number of aflibercept injections from the VIVID and VISTA trials. The ERG commented that in its base‑case analysis, the number of aflibercept and ranibizumab injections in year 2 was calculated from the mean number of injections given in VIVID and VISTA and in the RESTORE extension study. The ERG acknowledged that the number of aflibercept injections in VIVID and VISTA may be protocol driven, but stated that the clinical effectiveness evidence for aflibercept relates to the dosing frequency used in VIVID and VISTA. Therefore, the estimate for the number of injections of aflibercept from these trials should be used in the base case because of their alignment to the dosing frequency. The ERG commented that the concerns it had about the validity of the results from the 10 UK ophthalmologists for the number of aflibercept injections in year 1 also raised concerns about the reliability of the mean estimate of 4.0 aflibercept injections for the second year of the online survey. However, the ERG acknowledged that it had questioned to what degree the dosing of aflibercept might be protocol driven in its original report, and therefore the ERG applied 4.0 aflibercept injections in year 2 within a sensitivity analyses.
Cost of a laser administration: The ERG noted the results of the online survey of 34 ophthalmologists to establish the mean time taken for a laser visit compared with an intravitreal injection visit. The ERG accepted that it was plausible to assume the same administration cost for laser as for intravitreal injections. It noted that in the new company analysis the figure used was higher than the cost of an injection (£256). It therefore corrected this to £196 in its revised analysis.
Mortality multiplier: The ERG‑preferred mortality multiplier from Mulnier et al. (2006; a multiplier for people with diabetes compared with the general population) was combined with a mortality multiplier from Hirai et al. (2008; a mortality multiplier for people with diabetes and DMO compared with people with diabetes and without DMO). The ERG explained that if the mortality multiplier from Preis et al. (2005) was combined with the mortality multiplier from Hirai et al., it would result in a similar overall mortality rate to that used in the ERG's original base‑case analysis.
Fellow eye involvement: The ERG stated that reducing the fellow eye involvement at baseline by approximately half (from 85% to 46%) would have no overall impact on the results because the model is designed to assume only 50% of fellow eyes get treated. The ERG explored the impact on the ICER in a sensitivity analysis of a 46% fellow eye involvement with 50% on treatment and with 100% on treatment.
Laser monitoring visits in year 1: The ERG stated that increasing the number of laser monitoring visits in year 1 from 4 to 12 was inappropriate because current professional guidance suggests no more than 4 monitoring visits per year.
Use of patient‑level trial data: The ERG noted that the use of patient‑level data from VIVID and VISTA to inform the transition probabilities could improve the ICER for aflibercept compared with laser. However, because these data were not used in the company's submission, the ERG was unclear as to how it was generated and utilised in the company's model.

3.30 The ERG redid its original base‑case analysis for the comparison of aflibercept with laser using an equivalent cost of a laser administration with an intravitreal injection thereby increasing the cost of a laser administration from £139 to £194 (corrected by the ERG from £256 in the company's new evidence and analysis; see section 3.27). The results of the revised analysis showed the ICER for aflibercept (incorporating the confidential patient access scheme) compared with laser in the whole trial population reduced from £33,921 to £33,123. The ERG also provided a revised subgroup analysis for people with a CRT of less than 400 micrometres and for people with a CRT of 400 micrometres or more. The ERG's revised base cases for people with a CRT less than 400 micrometres was £48,255 per QALY gained and for people with a CRT of 400 micrometres or more was £21,442 per QALY gained.

3.31 The ERG conducted a sensitivity analysis on the parameters that it considered important to further explore for the comparison of aflibercept with laser in the whole trial population (section 3.29). The assumptions explored were:

equivalent cost for laser and intravitreal injection visits
4.0 aflibercept injections in year 2
2.3 aflibercept injections in year 3
46% fellow eye involvement at baseline with 100% on treatment
46% fellow eye involvement at baseline with 50% on treatment
using the EQ‑5D random effects quality‑of‑life data from VIVID and VISTA
using the EQ‑5D general estimating equation quality‑of‑life data from VIVID and VISTA.

3.32 The results of the ERG's sensitivity analysis for aflibercept (incorporating the confidential patient access scheme) compared with laser in the whole trial population showed the ICER ranged from £30,793 for 4.0 aflibercept injections in year 2 to £114,463 per QALY gained using the trial EQ‑5D trial data.

3.33 The ERG conducted the sensitivity analysis for the 2 subgroups: the CRT less than 400 micrometres and the CRT 400 micrometres or more. The results of the sensitivity analyses in the CRT less than 400 micrometres group showed the ICERs ranged from £44,883 per QALY gained for 4.0 aflibercept injections in year 2 to £185,829 per QALY gained using the EQ‑5D trial data. In the CRT 400 micrometres or more group, the ICERs ranged from £19,925 per QALY gained for 4.0 aflibercept injections in year 2 to £78,268 per QALY gained using the EQ‑5D trial data.

3.34 Full details of all the evidence are available.