Note from the National Guideline Clearinghouse (NGC): The National Institute for Health and Care Excellence (NICE) commissioned an independent academic centre to perform a systematic literature review on the technology considered in this appraisal and prepare an assessment report. The Evidence Review Group (ERG) report for this technology appraisal was prepared by Aberdeen Health technology Assessment (HTA) Group (see the "Availability of Companion Documents" field).
The manufacturer assessed the quality of all included studies: the two aflibercept randomised controlled trials (RCTs) and the 10 RCTs involving either ranibizumab or aflibercept, which informed the network meta-analysis. The methods used for quality assessment were considered adequate by the ERG.
The methodological quality of the VIEW 1 and VIEW 2 trials was good. Methods used to achieve randomisation were adequate and sequence allocation was concealed using a central interactive voice response system. Randomisation appears to have been successful, and there was not any imbalance between groups in terms of sociodemographic factors at baseline. All patients were masked (blind) to treatment status, and masking was maintained in the aflibercept 2 mg every 8 weeks (Q8) arm by giving sham injections on alternate months. The only study personnel unmasked to treatment status were those involved in the preparation and injection of the study drug. All personnel involved with outcome measurement and assessment were masked. The ERG considers the masking strategies of the VIEW trials appropriate. Although the manufacturer conducted per protocol analysis, the ERG does not consider that this is likely to increase the risk of bias as, for non-inferiority trials, use of the full analysis set is generally not considered to be conservative.
The quality of the other trials included in the indirect analysis was mixed. The report from the Kleijnen Systematic Reviews group, which accompanied the manufacturer's submission, highlighted particular concern with the CATT, DETAIL and MOON trials. The ERG shares this concern over the potential risk of bias of these trials.
The ERG performed a quality assessment of the manufacturer's systematic review using the York Centre for Reviews and Dissemination (CRD) criteria (see Table 4 of the ERG report). The quality of the systematic review was good, and the ERG has no major concerns in any of the quality areas.
No meta-analysis of the VIEW 1 and VIEW 2 trials was presented by the manufacturer in their submission. They stated that this was because the two trials were similarly designed so that their data could be pooled directly. The meta-analysis of VIEW 1 and VIEW 2 was, however, presented within the Kleijnen systematic review.
Indirect Comparison – Summary of Results
The treatment regimen for ranibizumab used in VIEW 1 and VIEW 2 was fixed dose but the manufacturer states that in clinical practice a 'treat to target' approach is used. This involves monthly treatment until the patient's visual acuity (VA) is stable for 3 consecutive months, with re-treatment in a similar way upon loss of VA (with minimum of 2 injections). Therefore the manufacturer commissioned a systematic review by Kleijnen Systematic Reviews group (included in the reference pack of the current submission) to identify studies which included this alternative and then undertook an indirect comparison of the data to compare fixed dose aflibercept (AFB 2 mg Q8) compared with ranibizumab 0.5 mg in a 'reactive dosing' or 'treat as needed' regimen. This type of dosing is referred to as 'pro re nata' (PRN).
The Kleijnen Systematic Reviews group produced three networks for consideration: at 6, 12, and 24 months. However, as no trials reported aflibercept results at 6 months, this network is not discussed further. Figures 1 and 2 of the ERG report display the networks for 12 months and 24 months, respectively.
The Kleijnen Systematic Reviews group undertook three types of indirect comparison: simple Bucher analysis, frequentist network analysis (using STATA) and Bayesian network analysis (using WinBUGS) and the results are now summarised.
See Section 4 of the ERG report for more information on the Kleijnen Systematic Reviews group report, and additional work carried out by ERG.
Summary and Critique of Manufacturer's Submitted Economic Evaluation by the ERG
The manufacturer economic evaluation of aflibercept for the treatment of wet age-related macular degeneration is based on a de novo economic model (Markov model) as none of the cost-effectiveness studies identified by the systematic literature review addressed the decision problem. The de novo economic model developed by the manufacturer is a two-eye model based on the appraisal scope.
See Section 5.2, "NICE Reference Case Checklist," in the ERG report (see the "Availability of Companion Documents" field).
The model structure is best thought of as a one eye model, with the facility for the development of 2nd eye involvement and the application of some costs and benefits to any 2nd eye involvement. Because of this, the one eye model will be described in detail followed by a description of the modelling of 2nd eye involvement. The ERG has major concerns about the modelling of 2nd eye involvement.
Visual acuity is banded into five health states, with these mostly being 15 letters wide, with the additional health state of death. See Table 23 of the ERG report for visual acuity bands within the model.
Patients begin the model with wet age-related macular degeneration in their 1st eye, and it is assumed that there is no wet age-related macular degeneration and no visual impairment (NVI) in their 2nd eye.
The baseline distribution between the visual acuity bands for the 1st eye is taken from the screening visit of the aflibercept arm of the VIEW 2 trial.
To model the aflibercept arm for year 1, the proportions of patients:
- Gaining more than 30 letters
- Gaining 15 to 30 letters
- Remaining within 15 letters
- Losing 15 to 30 letters
- Losing more than 30 letters
Between baseline and year 1 are applied to the baseline patient distribution, with the assumption that these proportions apply equally to each health state. To model the aflibercept arm for year 2, the proportions of patients gaining and losing letters between year 1 and year 2 are applied to the estimated patient distribution at year 1. As these are last observation carried forward (LOCF) distributions a proportion of patients are modelled as discontinuing and moving onto best supportive care (BSC) and a proportion of patients are modelled as dying.
The modelling of ranibizumab PRN in year 1 and year 2 follows the same logic, only with the year 1 proportions gaining letters for aflibercept being conditioned by the relative risk of gaining letters to provide estimates for ranibizumab, and the proportion remaining stable being similarly conditioned by the relative risk of maintaining letters. These relative risks are drawn from the Kleijnen systematic review 12 month analyses. The year 1 to year 2 proportions gaining letters and maintaining letters for aflibercept are also conditioned by relative risks, these relative risks are drawn from the Kleijnen systematic review 24 month analyses.
For years 3 to 5 patients are assumed to remain on treatment and have stable visual acuity. But patients on treatment may exit the one eye on treatment model to BSC due to discontinuations. From the start of the third year patients may also develop 2nd eye involvement. From year 6 all patients are assumed to cease treatment and move onto BSC. BSC is associated with a steady loss of visual acuity over time.
A range of univariate sensitivity analyses were presented around resource use, coupled with two sensitivity analyses around the relative risk of improving vision in year 1 and the relative risk of maintaining vision in year 2.
It may have been more sensible to have performed the sensitivity analyses varying the proportion receiving therapy in one stop clinics across all 5 years of treatment, rather than varying it for individual years.
Model Validation and Face Validity Check
Model validation is limited to comparing the modelled distribution of visual acuity for the study eye for the aflibercept 2 mg Q8 arm with the trial data at baseline, 52 weeks and 96 weeks. The manufacturer response to ERG clarification questions provides the distributions of the treated eyes within the trials, though note that this reporting has switched to the Safety Analysis Set.
See Section 5 of the ERG report (see the "Availability of Companion Documents" field) for more information on cost-effectiveness analysis.