3 The manufacturer's submission

The Appraisal Committee (section 6) considered evidence submitted by the manufacturer of ocriplasmin and a review of this submission by the Evidence Review Group (ERG; section 7).

Clinical effectiveness

Background to the clinical evidence

3.1 The manufacturer's systematic literature review identified 2 randomised controlled trials that were relevant to the decision problem: TG‑MV‑006 and TG‑MV‑007. The data from these trials were assessed individually and as an integrated dataset. Three non‑randomised controlled trials (TG‑MV‑001, TG‑MV‑008 and TG‑MV‑010) provided relevant safety and pharmacokinetic data.

3.2 TG‑MV‑006 was a randomised, placebo‑controlled, double‑blind trial conducted in the USA. People with vitreomacular traction were randomised to receive either a single injection of ocriplasmin (n=219), or a placebo injection of saline (n=107), and were followed up over 6 months. Inclusion criteria included a best corrected visual acuity (BCVA) of 20/25 or worse in the study eye, and 20/800 or better in the non‑study eye. Exclusion criteria included a macular hole larger than 400 micrometres and previous vitrectomy in the study eye. At baseline 27.3% had a full‑thickness (stage II) macular hole, 72.7% had vitreomacular traction (which could include a stage I macular hole), 37.1% had an epiretinal membrane, and 79.4% had an expected need for a vitrectomy. The mean BCVA score at baseline was 64.8 (standard deviation 10.53).

3.3 TG‑MV‑007 was a randomised, placebo‑controlled, double‑blind trial conducted in the USA and Europe (including the UK). People with vitreomacular traction were randomised to receive either a single injection of ocriplasmin (n=245), or a placebo injection of saline (n=81), and were followed up over 6 months. Inclusion criteria included a BCVA of 20/25 or worse in the study eye and 20/800 or better in the non‑study eye. Exclusion criteria included a macular hole larger than 400 micrometres and previous vitrectomy in the study eye. At baseline, 19.6% had a full‑thickness (stage II) macular hole, 80.4% had vitreomacular traction (which could include a stage I macular hole), 40.2% had an epiretinal membrane, and 88.7% had an expected need for a vitrectomy. The mean BCVA score at baseline was 63.8 (standard deviation 13.20).

3.4 The primary outcome of TG‑MV‑006 and TG‑MV‑007 was the proportion of patients with non‑surgical resolution of focal vitreomacular adhesion at day 28 post‑injection, as determined by masked central reading centre optical coherence tomography (OCT) evaluation. Secondary outcomes included the proportion of patients with total posterior vitreous detachment (PVD) at day 28, proportion of full‑thickness (stage II) macular holes that closed without vitrectomy, proportion of patients not needing vitrectomy, an improvement of at least 2 or 3 lines in BCVA without need for vitrectomy, improvement in mean BCVA, and improvement in the 25‑item Visual Function Questionnaire (VFQ‑25). Safety outcomes included adverse events, with special attention to ocular events, such as worsening visual acuity (VA), worsening macular oedema, vitreous haemorrhage, retinal tear, retinal detachment, increase in ocular inflammation and intraocular pressure increases.

Clinical trial results

3.5 The manufacturer presented the whole population results from TG‑MV‑006 and TG‑MV‑007 (see 3.6), as well as the following subgroups:

  • VMT without ERM (vitreomacular traction without an epiretinal membrane); this included people with a stage I macular hole (see 3.7)

  • VMT with ERM (vitreomacular traction with an epiretinal membrane); this included people with a stage I macular hole (see 3.8)

  • VMT with MH (vitreomacular traction with a stage II macular hole); this included people with an epiretinal membrane (see 3.9).

3.6 Data from both the clinical trials and the integrated analyses were presented for the whole population. The proportion of patients with vitreomacular traction resolution was statistically significantly greater in the ocriplasmin arm than the placebo arm for both trials and in the integrated analyses for the whole population (TG‑MV‑006: 27.9% and 13.1% respectively, 95% confidence interval [CI] 6.0 to 23.5, p=0.003; TG‑MV‑007: 25.3% and 6.2% respectively, 95% CI 11.6 to 26.7, p<0.001; integrated analysis: 26.5% and 10.1% respectively, 95% CI 10.5 to 22.3, p<0.001). The proportion of patients with total PVD by day 28 was statistically significantly greater in the ocriplasmin than the placebo arm for both trials in the whole population (TG‑MV‑006: 16.4% and 6.5% respectively, 95% CI 3.1 to 16.7, p=0.0014; TG‑MV‑007: 10.6% and 0% respectively, 95% CI 6.8 to 14.5, p<0.001).

3.7 The manufacturer presented data on the VMT without ERM subgroup. The integrated analyses showed that the proportion of patients with vitreomacular traction resolution or total PVD by day 28 was statistically significantly greater in the ocriplasmin arm than the placebo arm (vitreomacular traction resolution: placebo 7.7%, ocriplasmin 29.8%, p<0.001; total PVD: placebo 2.6%, ocriplasmin 17.0%, p<0.001). Further secondary outcomes did not show a statistically significant difference between the treatment arms but did favour ocriplasmin (proportion of patients who received a vitrectomy by month 6: placebo 15.4%, ocriplasmin 8.0%, p=0.091; mean change in VA from baseline at day 28: placebo 2.5 letters, ocriplasmin 2.6 letters, p=0.890; mean change in VA from baseline at month 6: placebo 2.8 letters, ocriplasmin 3.1 letters, p=0.728). At month 6 more patients treated with ocriplasmin than placebo had gained letters, on the Early Treatment Diabetic Retinopathy Study (ETDRS) scale (at least 10 letters gained: placebo 15.4%, ocriplasmin 25.5%, p=0.051; at least 15 letters gained: placebo 5.1%, ocriplasmin 10.1%, p=0.140. However, more patients treated with ocriplasmin had lost letters (at least 10 letters lost: placebo 3.8%, ocriplasmin 6.9%, p=0.381; at least 15 letters lost: placebo 1.3%, ocriplasmin 5.9%, p=0.121).

3.8 The manufacturer presented data on the VMT with ERM subgroup. The integrated analyses showed that there was no statistically significant difference between the placebo and ocriplasmin arm for any of the outcomes, and that the differences were small (proportion of patients with vitreomacular traction resolution by day 28: placebo 1.6%, ocriplasmin 7.8%, p=0.085; proportion of patients with total PVD by day 28: placebo 1.6%, ocriplasmin 3.6%, p=0.444; proportion of patients who received a vitrectomy by month 6: placebo 17.5%, ocriplasmin 11.4%, p=0.231; mean change in VA from baseline at day 28: placebo 2.7 letters, ocriplasmin 1.9 letters, p=0.325; mean change in VA from baseline at month 6: placebo 2.3 letters, ocriplasmin 2.0 letters p=0.685). At month 6 more patients treated with ocriplasmin than placebo had gained letters. However, more patients treated with ocriplasmin had lost letters (at least 10 letters gained: placebo 9.5%, ocriplasmin 20.5%, p=0.058; at least 15 letters gained: placebo 3.2%, ocriplasmin 4.8%, p=0.625; at least 10 letters lost: placebo 1.6%, ocriplasmin 8.4%, p=0.052; at least 15 letters lost: placebo 0%, ocriplasmin 4.8%, p=0.069).

3.9 The manufacturer presented data on the VMT with MH subgroup. The integrated analyses showed that the proportions of patients with vitreomacular traction resolution, total PVD or macular hole closure without vitrectomy by day 28, or macular hole closure without vitrectomy by month 6, were statistically significantly greater in the ocriplasmin arm than the placebo arm (vitreomacular traction resolution: placebo 25.5%, ocriplasmin 50.0%, p=0.006; total PVD: placebo 8.5%, ocriplasmin 22.6%, p=0.033; macular hole closure at day 28: placebo 10.6%, ocriplasmin 40.6%, p<0.001; macular hole closure at month 6: placebo 17.0%, ocriplasmin 40.6%, p<0.001). Further secondary outcomes did not show a statistically significant difference between the treatment arms but did favour ocriplasmin (proportion of patients who received vitrectomy by month 6: placebo 57.4%, ocriplasmin 44.3%, p=0.157; mean change in VA from baseline at day 28: placebo 3.0 letters, ocriplasmin 3.9 letters, p=0.691; mean change in VA from baseline at month 6: placebo 2.3 letters, ocriplasmin 6.8 letters, p=0.126). At month 6, the proportion of patients gaining letters and patients losing letters favoured ocriplasmin (at least 10 letters gained: placebo 30.4%, ocriplasmin 44.3%, p=0.104; at least 15 letters gained: placebo 13.0%, ocriplasmin 27.4%, p=0.063; at least 10 letters lost: placebo 15.2%, ocriplasmin 8.5%, p=0.233; at least 15 letters lost: placebo 10.9%, ocriplasmin 6.6%, p=0.421).

Health‑related quality-of-life results

3.10 Health‑related quality-of-life data were collected during the TG‑MV‑006 and TG‑MV‑007 trials using the VFQ‑25. The VFQ‑25 has 12 subscales that measure the influence of visual disability and visual symptoms on generic health domains such as emotional wellbeing, social functioning, sense of independence, and other task‑oriented domains related to daily visual functioning (such as driving). The manufacturer presented the integrated analyses of these data for the whole population. In the ocriplasmin group, mean increases (representing improvements) were observed across all of the subscale and composite scores at month 6, and were numerically better than the placebo arm. A clinically meaningful improvement in the VFQ‑25 composite score (minimum clinically important difference: 3.6) was observed in a significantly larger proportion of the ocriplasmin group (35.9%) than the placebo group (22.7%, p=0.0016). The mean change from baseline in VFQ‑25 composite score at 6 months was significantly greater in the ocriplasmin group (3.4) than the placebo group (0.7, p=0.007). The general vision subscale score improved more in the ocriplasmin group than the placebo group (placebo 2.1, ocriplasmin 6.1, p=0.024). This score was affected by the primary efficacy outcome (vitreomacular adhesion resolution): in the ocriplasmin group, mean improvement from baseline in the general vision subscale score was 8.4 for patients who had vitreomacular adhesion resolution, and 5.3 for those who did not. The mean improvement from baseline in the composite score was 5.7 among patients who had vitreomacular adhesion resolution and 2.6 for those who did not. The manufacturer presented a meta‑analysis that showed statistically significant improvements in VFQ‑25 outcomes (expressed as a difference in means) with ocriplasmin in general vision, distance activities, dependency, and composite scores.

Adverse events of treatment

3.11 The manufacturer presented pooled adverse event data from 7 clinical trials because TG‑MV‑006 and TG‑MV‑007 were not powered to detect significant differences in treatment‑related adverse events. The pooled trials were TG‑MV‑001, TG‑MV‑002, TG‑MV‑003, TG‑MV‑004, TG‑MV‑006, TG‑MV‑007 and TG‑MV‑010. The adverse events were mainly ocular events. The incidence of serious adverse events was similar between placebo (TG‑MV‑006 and TG‑MV‑007 12.8%, all studies 13.8%) and ocriplasmin (TG‑MV‑006 and TG‑MV‑007 13.3%, all studies 13.5%). The only suspected adverse drug reaction that appeared to be dose dependent was vitreous floaters. The most common serious adverse event reported in the study eye was macular hole, which was less common in the ocriplasmin group than in the placebo group (4.7% and 6.5% respectively in all studies combined). The manufacturer stated that most adverse events were non‑serious, mild in intensity, and resolved, and therefore were not considered to be clinically significant.

Cost effectiveness

3.12 The cost‑effectiveness evidence presented by the manufacturer consisted of a systematic literature review and a de novo model. The literature review was to identify all existing studies of the cost effectiveness of any intervention in patients with vitreomacular traction (including vitreomacular traction associated with a macular hole or epiretinal membrane). The systematic review did not identify any relevant cost‑effectiveness studies, so the manufacturer submitted a de novo economic analysis that assessed the cost effectiveness of ocriplasmin compared with 'watch and wait' in patients with VMT with ERM, VMT without ERM, and VMT with MH.

Model structure

3.13 The manufacturer's model had 2 components: a short‑term decision tree and a long‑term extrapolation Markov model. The short‑term decision tree covered the first 6 months of treatment, and was predominantly based on clinical trial data pooled from TG‑MV‑006 and TG‑MV‑007. It had monthly cycles. The short‑term decision tree determined the starting position of patients in the long‑term extrapolation Markov model, which started at 6 months post‑treatment and then had a lifetime time horizon. The Markov model applied 3-monthly cycles for the first 2–5 years. After 2 years (or 5 years in the sensitivity analysis) annual cycles were used, therefore assuming events such as vitrectomy and spontaneous resolution would occur before this.

3.14 The manufacturer's model covered the population stated in the scope and the marketing authorisation (that is, adults with vitreomacular traction, including when associated with macular hole of diameter less than or equal to 400 microns). However, the manufacturer modelled 3 different subgroups of this population, which were defined by baseline characteristics: VMT without ERM, VMT with ERM (the VMT subgroups included patients with stage I macular holes), and VMT with MH (macular holes were stage II; this group included people with an epiretinal membrane). There were 2 short‑term decision tree models used, one for the VMT with MH group and one for the other 2 subgroups. The structure of the long‑term extrapolation Markov model was the same for all 3 subgroups. The manufacturer's short‑term decision tree for patients with VMT (with or without ERM) had the following decision nodes: non‑surgical vitreomacular traction resolution (at day 28 or month 6), visual acuity health state 1–6 (see 3.15), vitrectomy, macular hole at month 6. The manufacturer's short‑term decision tree for patients with VMT with MH had the following decision nodes: non‑surgical macular hole resolution (at day 28 or month 6), vitrectomy (first or second), macular hole closed, non‑surgical vitreomacular traction resolution at month 6. The manufacturer assumed that macular hole closure would lead to vitreomacular traction resolution. Patients finished the decision tree in 1 of 6 different health states, which were in the long‑term Markov model. These health states were: resolved (vitreomacular traction and macular hole), vitreomacular traction unresolved without macular hole, vitreomacular traction unresolved with macular hole, vitreomacular traction resolved with macular hole (no vitrectomies), vitreomacular traction resolved with macular hole (1 vitrectomy) or vitreomacular traction resolved with macular hole (2 vitrectomies). The manufacturer's long‑term Markov model also had a 'death' state. Patients transitioned to another health state as a result of the following events: spontaneous vitreomacular traction and/or macular hole resolution, resolution of vitreomacular traction and/or macular hole through vitrectomy, failure of vitrectomy to resolve macular hole, spontaneous development of a macular hole, or death (see 3.16 and 3.19 for model transitions). There were 2 assumptions in the model that were based only on the manufacturer's clinical expert opinion:

  • There is a maximum of 2 vitrectomies. This is based on clinical expert opinion that the probability of having a third vitrectomy was very low.

  • Vitrectomies are 100% effective at treating vitreomacular traction, and therefore a second one is only used to close a persistent macular hole.

3.15 Within each health state of the long‑term extrapolation Markov model (other than death) there were 6 sub‑states that represented levels of visual acuity, called the vision health states. Patients could move in any direction through the vision health states because they could improve or deteriorate. These health states were determined by the patients' BCVA in terms of ETDRS letters read: VA1: 76–100 letters, VA2: 66–75 letters, VA3: 56–65 letters, VA4: 46–55 letters, VA5: 36–45 letters, VA6: 0–35 letters.

Model transitions

3.16 The transitions between health states in the manufacturer's decision tree and Markov models were calculated from the integrated phase III trial data (from TG‑MV‑006/007), with the following exceptions: probability of opting for a second vitrectomy (75%), probability of success of second vitrectomy (50% of success rate of first vitrectomy [probability of macular hole closure post‑vitrectomy was estimated as 82%]). Each of these were based on manufacturer expert opinion.

3.17 The mortality rates applied in the manufacturer's model were based on VA state. For VA1–5, the mortality rate of the general population was used, taken from England and Wales interim life tables 2008–2010, from the Office of National Statistics. These were weighted according to sex. For a best‑seeing eye VA6 score, which represented severe visual impairment, the manufacturer used a mortality hazard rate of 1.54 from a US study (Christ et al. 2008).

3.18 As described in 3.15, the manufacturer's Markov model included visual acuity states within each health state. At the start of the model, the distribution of patients across the visual acuity states was estimated using an ordered logit model based on trial data (TG‑MV‑006 and TG‑MV‑007). This was assumed to be the same for both treatment arms (ocriplasmin and placebo). The presence of a macular hole affected the visual acuity state distribution, with more patients with a macular hole having VA3–VA6 scores than those without.

3.19 Patients could move between any visual acuity state (representing improving or declining visual acuity) within the model. The transition probabilities between visual acuity states were assumed to be different depending on whether vitreomacular adhesion/traction was resolved or remained unresolved, and it was assumed that the presence of a macular hole did not affect the rate of visual acuity change. The transition probabilities between visual acuity states were based on estimates from the literature that relate to changes over time in the general population for patients with resolved vitreomacular traction, and in patients with persistent vitreomacular traction for unresolved vitreomacular traction. The manufacturer's submission recognised that after specific events there would be an initial change in vision that would be different from changes observed in the general population over time. The manufacturer therefore included within the model visual acuity transitions after a specific event for 1 cycle. The events that were modelled were vitreomacular traction resolution only, macular hole closure only, vitreomacular traction resolution and macular hole closure and vitreomacular traction progression to macular hole. The corresponding transition probabilities were calculated from the integrated phase III clinical trial data (TG‑MV‑006 and TG‑MV‑007) using an ordered logit model (except for the macular hole opening, for which calibration was used).

Utility values and adverse events

3.20 The manufacturer applied utility values to each of the visual acuity states in the model. These were derived from a study that used contact lenses to simulate the effects of visual impairment caused by age‑related macular degeneration. The study grouped patients into 4 categories, according to their best‑seeing eye, and a time trade‑off instrument was used to elicit utilities from the general UK population. Because the manufacturer's model used 6 visual acuity states, the utility values from the study were adapted to fit the 6 visual acuity states in the model. In addition, because the study was based on the best‑seeing eye only, a matrix was developed to account for the visual acuity state of both the best- and worst-seeing eye. The utility for each visual acuity combination (for example, VA1 and VA1, or VA1 and VA2) was then estimated.

3.21 The adverse events included in the manufacturer's model were modelled as occurring post‑vitrectomy or post‑ocriplasmin. These included retinal tear (13.23% and 0.22%, respectively), retinal detachment (13.23% and 0.43% respectively), elevated intraocular pressure (26.46% and 2.37% respectively) and vitreous haemorrhage (3.31% and 0.22% respectively). The probabilities of these events occurring were estimated using the integrated clinical trials data (TG‑MV‑006 and TG‑MV‑007).

3.22 Many patients develop cataracts after vitrectomy, therefore the manufacturer's model included a 96% probability (determined from published data) of developing cataracts after having a vitrectomy. The proportions of people in each subgroup who could develop cataracts (who have not had previous cataract surgery) were 59% for VMT with ERM, 63.9% for VMT without ERM and 79.1% for VMT with MH.

3.23 The manufacturer accounted for adverse events by applying disutility values. A disutility for metamorphopsia (0.017), which has a significant impact on quality of life, was derived from the literature and included in the manufacturer's model. Disutility values derived from the literature were also applied to each of the adverse events captured in the model (see 3.21): retinal detachment (0.13 for 1 month), vitreous haemorrhage (0.02 for 1 month) and cataract (0.14 for 6 months). Disutility values for retinal tear and increased intraocular pressure were not identified in the literature and therefore no disutility was applied to these. Disutilities were all applied during 1 cycle, with an exception made for metamorphopsia, which persisted until vitreomacular traction resolved. Disutilities were normalised to a 3‑month cycle length. Vitrectomy surgery is associated with a post‑surgery reduction in health‑related quality of life which was accounted for in the model by transitioning to the visual acuity state VA6 for 2 weeks (rather than applying a disutility value).

Costs applied in the model

3.24 The manufacturer's model included the following costs: ocriplasmin (£2500), ocriplasmin administration (£117), surgery (vitrectomy [£2191] and cataract [£851]), follow‑up visits (£80), OCT (£54.29), annual cost of blindness (£6496), and costs of adverse events (retinal detachment [£2012], retinal tears [£424], increased intraocular pressure [£40.65], vitreous haemorrhage [£1852]). The manufacturer used NHS reference costs to estimate the cost of vitrectomy surgery, cataract surgery, administration of ocriplasmin, follow‑up visits, retinal detachment, and vitreous haemorrhage. The annual cost of blindness was estimated from the Personal Social Services Research Unit (PSSRU) costs of Health and Social Care and accounts for residential care (£18,191 for 30% of blind patients), community care (£8195 for 6% of blind patients), depression (£539 for 39% of blind patients), and hip replacement (£6728 for 5% of blind patients). The cost of OCT was taken from the literature, and the costs of retinal tears and increased ocular pressure were taken from NICE submissions. The model base case assumes 1 follow‑up visit per 3 months, 1 OCT per 3 months, 4 post‑vitrectomy follow‑up visits, 4 OCTs post‑vitrectomy, 2 follow‑up visits post‑ocriplasmin injection and 1 OCT post‑ocriplasmin injection. These visit estimates are based on clinical expert advice.

Manufacturer's base‑case incremental cost‑effectiveness ratio (ICER), sensitivity and scenario analyses

3.25 The manufacturer presented a base‑case ICER for each patient subgroup. The ICER for ocriplasmin compared with 'watch and wait' in the VMT without ERM subgroup was £18,481 per quality‑adjusted life year (QALY) gained (incremental cost: £1880.67, incremental QALY: 0.1018), for VMT with ERM it was £67,119 per QALY gained (incremental cost: £2487.13, incremental QALY: 0.0371) and for VMT with MH it was £21,593 per QALY gained (incremental cost: £1752.90, incremental QALY: 0.0812). The manufacturer estimated that the probability of ocriplasmin being cost effective, if the maximum acceptable ICER was £20,000 or £30,000 per QALY gained, compared with 'watch and wait' was: 51% and 80% respectively for VMT without ERM; 0% and 2% respectively for VMT with ERM; and 46% and 72% respectively for VMT with MH.

3.26 The manufacturer conducted univariate sensitivity analyses for each of the subgroups. For the VMT without ERM and VMT with ERM subgroups the model outcomes were most sensitive to the inputs determining non‑surgical resolution of vitreomacular traction at 6 months and 28 days. The QALY discount rate was also an important driver for these subgroups. For the VMT with MH subgroup, the model outcomes were most sensitive to the inputs that determined non‑surgical macular hole closure, cataract disutility and the chance of macular hole closure post‑vitrectomy.

3.27 The manufacturer also conducted scenario analyses to investigate the following:

  • The time limit of vitrectomy. It was assumed that vitrectomies would occur within 2 years in the base case and therefore the model applied a 3‑month cycle length for 2 years and an annual cycle length thereafter. The 3‑month cycles enable rapid changes in visual acuity in response to vitrectomy. By changing the length of time the model was running at 3‑monthly cycles from 2 years in the base case to 1 or 5 years, the manufacturer could investigate the impact of changing the time period for vitrectomy.

  • The impact of treating patients with a macular hole earlier than usual with a vitrectomy by assessing the cost effectiveness at day 28 of ocriplasmin compared with vitrectomy.

  • Accounting for the lack of mortality in the decision tree part of the model by doubling mortality in the first year of the Markov model.

  • Using visual acuity state transitions derived from a study by Van der Pols et al. (2000) on British patients, rather than Finnish patients as in the base case.

  • Applying the same rate of visual acuity decline whether vitreomacular traction was resolved or not, by using the transition rates from Laitinen et al. (2005) for all visual acuity health states.

  • Using spontaneous vitreomacular traction resolution rates from the literature rather than the clinical trial.

  • Applying utility values derived from patients with age‑related macular degeneration in the US, rather than utilities derived from the general UK population as in the base case.

  • The impact of modelling the best or worst‑seeing eye only, rather than accounting for both eyes.

  • Applying an alternative metamorphopsia disutility value of 0.14, derived using the EQ‑5D, rather than 0.017 from the literature as used in the base case.

3.28 The different scenarios had different impacts on the 3 subgroups. Two scenarios increased the ICER the most. These increased the ICER substantially from the manufacturer's base case for the VMT without ERM and VMT with ERM subgroups and resulted in small increases for the VMT with MH group:

  • making the long‑term vision transition rates equal whether vitreomacular traction had resolved or not

    • VMT without ERM: £44,489 per QALY gained (incremental cost £2235, incremental QALY 0.050)

    • VMT with ERM: £142,347 per QALY gained (incremental cost £2599, incremental QALY 0.018)

    • VMT with MH: £21,723 per QALY gained (incremental cost £1754, incremental QALY 0.081)

  • using a 16.5% rate of spontaneous resolution

    • VMT without ERM: £67,320 per QALY gained (incremental cost £2257, incremental QALY 0.034)

    • VMT with ERM: £230,656 per QALY gained (incremental cost £2575, incremental QALY 0.011)

    • VMT with MH: £21,615 per QALY gained (incremental cost £1753, incremental QALY 0.081).

3.29 Applying a different metamorphopsia disutility value reduced the ICER substantially from the manufacturer's base case in all 3 subgroups:

  • VMT without ERM: £12,190 per QALY gained (incremental cost £1881, incremental QALY 0.154)

  • VMT with ERM £42,388 per QALY gained (incremental cost £2487, incremental QALY 0.059)

  • VMT with MH £17,837 per QALY gained (incremental cost £1753, incremental QALY 0.098).

3.30 Modelling only the best‑seeing eye reduced the ICER and modelling only the worst‑seeing eye increased the ICER in all 3 subgroups, with a large impact for both analyses in the VMT with ERM subgroup. Using utility values derived from UK patients increased the ICER for all 3 subgroups, with a substantial increase in the VMT with ERM subgroup. Changing the time limit of vitrectomy had little impact on the VMT with MH subgroup but for both the VMT without ERM and VMT with ERM subgroups using 1 year substantially decreased the ICER, and using 5 years substantially increased the ICER. The other scenario analyses had only a small impact on the ICERs.

Evidence Review Group comments

Clinical effectiveness

3.31 The ERG reviewed the manufacturer's literature review and considered that the manufacturer was likely to have identified all the randomised controlled trial evidence relevant to the decision problem. The ERG reviewed the designs of TG‑MV‑006 and TG‑MV‑007. It noted that the patients in the placebo group had undergone an injection, which was an invasive procedure, rather than having been initially observed without treatment, as is typical in UK clinical practice.

3.32 The ERG noted that the visual acuity of patients enrolled on the trials, with the exception of those with a macular hole, was better than would be seen in clinical practice for patients with vitreomacular traction and/or macular hole who would normally be offered vitrectomy. The ERG identified this as a limitation of the data because efficacy may be affected by disease severity.

3.33 The ERG noted that the primary outcome of non‑surgical resolution of vitreomacular adhesion is a surrogate outcome for preventing deteriorating vision, which can result from untreated and progressive vitreomacular traction. The ERG commented that there is limited evidence on the validity of non‑surgical resolution of vitreomacular adhesion as a surrogate for preventing deteriorating visual function.

3.34 The ERG noted that adverse events reported in the pooled results of 7 completed clinical trials (the 'safety set') were consistent with the vitreolytic activity of the drug or method of administration and most were mild, or moderate and transient. However, the ERG commented that none of the safety set of trials, which ocriplasmin's safety profile is based on, was designed to assess safety outcomes with sufficient power to detect differences in incidence rates. This safety set can only detect adverse events with an incidence greater than 0.4% because of the uneven randomisation ratios.

Cost effectiveness

3.35 The ERG considered the modelling approach presented by the manufacturer to be appropriate and stated that it enabled important anatomical and visual outcomes to be simultaneously captured in the short and long term. The ERG reviewed the model, and considered many of the approaches, assumptions and data sources applied by the manufacturer to be reasonable.

3.36 The ERG identified some areas of uncertainty in the manufacturer's model, which it investigated, if it was possible, in exploratory analyses. Three scenarios had a large impact on the ICER:

  • applying a different rate of probability for cataracts (see 3.38)

  • enabling cataract and vitrectomy surgery to be done simultaneously (see 3.39)

  • applying a different macular hole vitrectomy success rate (see 3.40).

Evidence Review Group's exploratory analyses

3.37 Most of the ERG's exploratory analyses did not affect the manufacturer's base-case ICERs (see 3.25) for the subgroups substantially. The resulting ICERs were in the following ranges:

  • VMT without ERM:

    • base case: £18,481 per QALY gained

    • range: £17,733–£18,986 per QALY gained

  • VMT with ERM:

    • base case: £67,119 per QALY gained

    • range: £64,331–£67,666 per QALY gained

  • VMT with MH:

    • base case: £21,593 per QALY gained

    • range: £20,551–£22,985 per QALY gained.

3.38 The ERG explored the rate of cataracts after vitrectomy. The ERG noted that a study of the National Ophthalmology Database identified that 64.6% of eyes that had a macular hole operation (without combined or previous cataract surgery) needed lens removal within 1 year of vitrectomy, which was much lower than the 96% applied in the model by the manufacturer (to non‑pseudophakic eyes only). Applying this new rate increased the ICER of each subgroup by at least £1000 (VMT without ERM: from £18,481 to £19,858 per QALY gained [incremental cost £1899, incremental QALY 0.096]; VMT with ERM: £67,119 to £71,737 per QALY gained [incremental cost £2494, incremental QALY 0.035]; VMT with MH: from £21,593 to £28,289 per QALY gained [incremental cost £1806, incremental QALY 0.064]). The ERG conducted a sensitivity analysis around this and applied an 88.8% and 92.0% probability of cataract. This increased the manufacturer's base‑case ICER, but to a lesser extent than applying the 64.6% probability. Applying an 88.8% probability of cataract increased the manufacturer's base‑case ICER to: VMT without ERM: £18,782 per QALY gained (incremental cost £1885, incremental QALY 0.100); VMT with ERM: £68,127 per QALY gained (incremental cost £2489, incremental QALY 0.037); VMT with MH: £22,863 per QALY gained (incremental cost £1765, incremental QALY 0.077). Applying a 92.0% probability of cataract increased the manufacturer's base‑case ICER to: VMT without ERM: £18,647 per QALY gained (incremental cost £1883, incremental QALY 0.101); VMT with ERM: £67,675 per QALY gained (incremental cost £2488, incremental QALY 0.037); VMT with MH: £22,283 per QALY gained (incremental cost £1760, incremental QALY 0.079).

3.39 The ERG explored the impact of combining vitrectomy and cataract surgery. The ERG received advice from clinical specialists highlighting that, as a result of the high incidence of cataract formation after vitrectomy, lens removal is frequently combined with vitrectomy as a preventative measure. The ERG noted that this was not accounted for in the manufacturer's model. Furthermore, the ERG noted that Jackson et al. (2013) report that 40.5% of patients undergoing macular hole vitrectomy had combined lens removal. Therefore, to investigate the potential impact of combined surgery on the manufacturer's ICERs, the ERG further reduced the probability of cataract surgery by 40.5%. The ICER of each subgroup increased (VMT without ERM: from £18,481 to £20,212 per QALY gained [incremental cost £1904, incremental QALY 0.094]; VMT with ERM: £67,119 to £72,929 per QALY gained [incremental cost £2496, incremental QALY 0.034]; VMT with MH: from £21,593 to £30,458 per QALY gained [incremental cost £1818, incremental QALY 0.060]).

3.40 The ERG explored the success rate of macular hole vitrectomy in the manufacturer's model. The ERG noted that 82% of vitrectomies to treat macular hole are assumed to be successful (based on trial data) in the manufacturer's base‑case economic analyses. However, expert clinical opinion highlighted that, in patients with a macular hole of 400 micrometres or less, success after vitrectomy involving internal limiting membrane peeling is over 90%. Therefore the ERG carried out a sensitivity analysis assuming 95.8% of macular hole vitrectomies are successful. The ICER associated with each subgroup increased, most notably in patients with a macular hole (VMT without ERM: from £18,481 to £19,250 per QALY gained [incremental cost £1911, incremental QALY 0.099]; VMT with ERM: from £67,119 to £69,588 per QALY gained [incremental cost £2501, incremental QALY 0.036]; VMT with MH: from £21,593 to £26,854 per QALY gained [incremental cost £1847, incremental QALY 0.069]).

3.41 The ERG estimated a revised (deterministic) base‑case ICER for each subgroup that took into account all of the exploratory analyses detailed in 3.37–3.40, and used a probability of cataract of 64.6%. The resulting ICERs were:

  • VMT without ERM £20,861 per QALY gained (incremental costs £2082, incremental QALY 0.100)

  • VMT with ERM £69,694 per QALY gained (incremental costs £2568, incremental QALY 0.037)

  • VMT with MH £56,137 per QALY gained (incremental costs £2132, incremental QALY 0.038).

3.42 The ERG commented that there were other areas of uncertainty that could affect the ICER, including:

  • The uncertainty associated with the clinical data because patients receiving a placebo injection had been used to represent the outcomes of 'watch and wait' patients. The ERG concluded that this was likely to bias against ocriplasmin, and that the ICER would be expected to decrease (in all subgroups) if this was addressed.

  • The health states modelled did not include epiretinal membrane. The ERG stated that this would affect only the VMT with ERM subgroup but would be likely to increase the ICER in this subgroup.

  • Differences in the results and baseline characteristics within and between the relevant clinical trials (TG‑MV‑006 and TG‑MV‑007). The ERG concluded that any bias was likely to be against ocriplasmin and therefore the ICER would be expected to decrease if baseline characteristics for these subgroups were balanced.

  • The manufacturer conservatively assuming long‑term vision decline was the same for all patients with unresolved vitreomacular traction, regardless of whether they had a macular hole or not. The ERG stated that the ICER would be expected to decrease if visual decline was different for vitreomacular traction patients with or without a macular hole but that the impact was not quantifiable for any of the subgroups.

  • The manufacturer assuming the quality of life impact of metamorphopsia is the same for both vitreomacular traction and macular hole, and applies to patients whose vitreomacular traction is unresolved, or whose macular hole is open. The ERG considered these assumptions to be potentially inaccurate because the patient population in Fukeda et al. (2009) continued to have symptoms of metamorphopsia after vitrectomy to close the macular hole. This suggests the possibility of metamorphopsia in patients with resolved vitreomacular traction, which clinical specialist advice highlighted may be a result of a persistent epiretinal membrane (not accounted for in the manufacturer's model). The ERG concluded that the impact of accounting for metamorphopsia in patients with resolved vitreomacular traction was not quantifiable but that any bias was likely to be small and against ocriplasmin. The ICER would be expected to decrease if a lower disutility for metamorphopsia was applied in patients with vitreomacular traction alone (no macular hole).

  • The fact that the manufacturer provided no rationale for not including an increased mortality risk in patients with 'some' visual impairment. The ERG anticipated that the impact on the ICER of incorporating an increased mortality risk for patients with visual impairment in their worst‑seeing eye was likely to be small, and that the direction of any bias was unclear.

3.43 The ERG reviewed the scenario presented by the manufacturer in which long‑term vision outcomes were assumed to be equivalent, whether vitreomacular traction was resolved or unresolved. It noted that this increased the ICER substantially (see 3.28). The ERG agreed with the manufacturer and stated that, based on the ERG's clinical specialist opinion, this scenario was unlikely.

3.44 Full details of all the evidence are in the manufacturer's submission and the ERG report.

  • National Institute for Health and Care Excellence (NICE)