Article Text
Abstract
Aims To project the number of cases with age-related macular degeneration (AMD) and the numbers with attributable sight loss in the UK in 2010–2020, taking into account the expected beneficial effect of the new anti-vascular endothelial growth factor (VEGF) therapies.
Methods A ‘system dynamics’ approach was used in constructing the model to simulate the dynamics of the disease in large populations. The model computed the pool of affected cases over the simulation period, taking into account the expected demographic changes. Other determinants taken into account included: prevalence; incidence; mortality; and the expected efficacy and coverage of anti-VEGF treatment.
Results In the UK, 608 213 persons in 2010 are estimated to have AMD, and this is expected to increase to 755 867 by the end of the decade. Numbers with sight loss from AMD are expected to rise from 223 224 in 2010 to 291 982 by 2020. Cases with sight loss due to neovascular AMD are expected to increase from 145 697 to 189 890 by the end of the decade.
Conclusions The model predicts that the beneficial effects of the treatment would be outweighed by the strong anticipated demographic ‘ageing’ effect. This reaffirms the importance of continuing efforts to develop more effective and more broadly applicable therapies for AMD.
- Macular degeneration
- prevalence
- reduced vision
- population dynamics
- Computerised models
- macula
- epidemiology
Statistics from Altmetric.com
- Macular degeneration
- prevalence
- reduced vision
- population dynamics
- Computerised models
- macula
- epidemiology
Introduction
Age-related maculopathy (ARM), a chronic degenerative disease, is strongly age-related, affecting persons 50 years or older. In its early stages (early ARM), the clinical features are indistinct soft drusen and disturbance of the retinal epithelial pigment in the macular area, generally without disruption of visual acuity. Progression of the disease to later stages leads to age-related macular degeneration (AMD), one of the most common causes of blindness in the older population.1–3 AMD is manifest in two main forms: ‘wet’ AMD (neovascularisation) and ‘dry’ AMD (geographic atrophy). In the latter condition, progression to serious sight loss tends to be slower but irreversible; the former is characterised by more rapid deterioration of visual acuity but can be amenable to anti-vascular endothelial growth factor (VEGF) therapy. About 65–70% of AMD cases have the ‘wet’ form in at least one eye. Anti-VEGF treatment is already expected to make a substantial contribution to the reduction of blindness from AMD, with an accompanying impact on the structure of service delivery. Ophthalmologists deal with these changes and expectations within the service commissioning process and research prioritisation agenda, and these require updated estimates of prevalence in the UK population.
It is unlikely that new prevalence surveys will be funded in the present economic climate. Building a knowledge base of methods for estimating numbers likely to be affected by sight-threatening disease in the coming decade is an objective of the professional and voluntary society partners within the UK Vision Strategy. In line with this objective this paper reports on a method of modelling the prevalence of AMD for the decade 2010–2020. The resulting prevalence estimates, shaped by a set of expectations of the impact of the anti VEGF treatment, are reported. The study reported here used derivations from best available epidemiological data for construction of a computer model to simulate the population dynamics of AMD, taking into account the expected demographic changes over the simulation period (2010–2020). The estimates included the number of cases progressing to serious sight loss in relation to the expected efficacy and coverage of treatment by an anti-VEGF agent (ranibizumab), recently approved by the National Institute for Health and Clinical Excellence (NICE).4
This paper reports on an aspect of work commissioned by the Royal National Institute of Blind People in 2008 as the lead for research in the UK Vision Strategy. The objectives of the commissioned work were to estimate the current and future prevalence of the main eye disorders, the expected numbers with sight loss in the UK and the costs to society (not reported here). The interlinked epidemiological and economic findings provide estimates available to inform the UK Vision Strategy.
Methods
Overview of the epidemiological model
A ‘system dynamics’ 5–8 approach was used in constructing the model to simulate the dynamics of the eye disorder in large populations. The model computed estimates of the pool of cases in the population (numbers with AMD and with sight loss due to AMD) as the simulation years progressed, taking into account the expected demographic changes. Other important determinants taken into account included: prevalence; incidence; mortality; and the expected treatment efficacy and coverage. ‘Level’ variables representing the population, and the pools of individuals affected by the eye disorder and by the consequent sight loss, were used. These were interconnected by ‘flow paths’ allowing flow in and out of the pools. ‘Rate’ variables acting as ‘taps’ controlled the rates of flow. ‘Auxiliary’ variables representing influence factors (determinants) opened or closed the ‘Rate taps’ (increased or decreased the rates of flow). The model ‘monitored’ and recorded the changing pools over the simulation period, as new cases were allowed to flow in according to incidence, and existing cases to flow out through mortality or through treatments leading to gains in visual acuity.
The ‘system’ being modelled comprised the UK population, and the healthcare and related resources therein. When the model was run for the specified simulation period, the size of the pools changed with passing time, according to the population dynamics of the eye disorder and the care facility in the system. Snapshots of the main pools were outputs that gave summary results of the simulation at selected calendar year points (midyear).
Our preferred choice of modelling software was the programming facility of Excel using Visual Basic code.
Input data required for the model
Population projections
The latest projections of population estimates for the years 2010 to 2020 were obtained from the Government Actuary's Department (GAD). These give estimated numbers by 1-year and 5-year age classes. Projection for the relevant age groups re shown in table 1.
Mortality rates
Age- and sex-specific mortality rates by calendar year were those used by GAD for the population projections. For the model computations, the rates were converted to conditional survival probabilities over the projection period.
Prevalence proportions for AMD
The following sources were considered for selection of the most appropriate prevalence data to be used as inputs for the model: the European Eye Study (EUREYE) 9 (multicentre study in seven European countries); the Eye Disease Prevalence Research Group 200410 (meta-analysis of large population-based studies); Evans et al (2004) (a population-based cross-sectional study2; Owen et al (2003) (meta-analysis)11; and the Rotterdam Study (1995).12 For AMD prevalence, the estimates from the Eye Disease Prevalence Research Group were used. These were similar to the estimates from the European Eye Study, but included younger age groups. The study focused on obtaining robust estimates for neovascular AMD (NV-AMD) and the ‘dry’ form of AMD, and on balance, was considered more suitable for our purpose. In both studies (and in others, eg, Owen et al11), NV-AMD and geographic atrophy were not mutually exclusive categories, in so far as the geographic atrophy group included some persons with NV-AMD in the fellow eye. All persons classified as NV-AMD had the disorder in at least one eye, some having the ‘dry’ form in the fellow eye. Since estimates were also given for ‘Any AMD’, ie, the ‘dry’ and/or ‘wet’ forms, we were able to obtain the prevalence for two mutually exclusive categories: ie, (1) NV-AMD in one or both eyes, and (2) only ‘dry’ AMD in one or both eyes. The data sources used as inputs to the model were based on studies of predominantly white populations, and did not include black African-Caribbean people. The data were considered suitable for application, without adjustment for any ethnicity effect, to the UK where the African-Caribbean ethnic groups constitute only a small proportion (about 2%) of the population aged 60 years or older. Risk adjustment (even if large) for ethnicity would not have altered the overall ‘un-adjusted’ estimates materially.
Sight loss attributable to AMD
For proportions with sight loss due to AMD, we used the estimated of Evans et al2 and the Owen et al.11 The Owen estimates covered all the desired age groups, but did not give estimates for the required levels of sight loss (<6/12–6/60 and <6/60). The estimates were for visual acuities of 6/18 to >6/60, 6/60 to 3/60, and poorer than 3/60. The Evans estimates did give the proportions for the required levels of sight loss (<6/12–6/60 and poorer than 6/60), but did not cover the younger age groups. The relative proportions Owen/Evans for the common age groups were used to derive proportions for the required levels of sight loss for the younger age groups. This adjustment did not affect the estimates for ‘any sight loss’ (<6/12). The prevalence proportions for sight loss due to ‘wet’ AMD were not applied directly in the simulation model for estimations beyond mid 2008, as they were based on the pre anti-VEGF era. Appropriately modified incidence figures were used instead, as described below.
Incidence
Estimates were computed for the effective age-specific cumulative incidence of the disorder at issue over a calendar period, for birth cohorts, by sex (and ethnicity when necessary). The method used was based on the procedures described by Elandt-Johnson and Johnson (1980) for the estimation of incidence onset distribution from prevalence data.13 The calculations used the prevalence figures derived earlier, mortality rates and population projections (which included net migration figures). The mortality rates were those used by the GAD to make the population projections. The incidence figures derived for sight loss due to ‘wet’ AMD were regarded as historical, since the source data were from pre anti-VEGF therapy era and could not be applied directly in the simulation. The model modified the historic incidence to reflect the reduced risk of sight loss due to ranibizumab therapy, and applied the modified incidence to the case population expected to receive the treatment.
Ranibizumab treatment efficacy
Published data from the Anti-VEGF antibody for the treatment of predominantly classic choroidal neovascularization in age-related macular degeneration (ANCHOR)14 and minimally classic/occult trial of the anti-VEGF antibody ranibizumab in the treatment of neovascular age-related macular degeneration (MARINA)15 studies were used to derive values for the treatment effect: relative risk (RR) of progression to blindness, and of the proportion among the treated who gain visual acuity of 15 or more letters (3+ Snellen lines). The derived values were 0.17 and 0.25, respectively. In the model, the reduced risk of progression was allowed to persist for 2 years from start of treatment and gradually return to baseline in 5 years.
Treatment eligibility and coverage
The assumptions for the model concerning eligibility and indications for ranibizumab treatment in NV-AMD were based on the guidelines in the report by National Institute for Health and Clinical Excellence4 and the Royal College of Ophthalmologists clinician's guide.16
Model assumptions for treatment
It was anticipated that 75% of cases of NV-AMD with corrected visual acuity of <6/12–6/60 in the better-seeing eye would be eligible (clinically suitable) for treatment. Among cases with visual acuity of <6/60 in the better-seeing eye, 10% would be eligible for treatment. Treatment coverage was assumed at 75% of the eligible cases. The model also used two variations of this assumption: 90% treated to reflect improvements in treatment coverage and 50% treated to reflect possible limitations in access to treatment or patient concerns. Ranibizumab treatment was anticipated to comprise, on average, eight injections in the first year, followed by six injections in the second year of treatment.4
Case definitions
In reporting the results the following case definitions are used. Clinical definitions of neovascular ‘wet’ AMD (NV-AMD) and geographic atrophy (GA-AMD) were as defined by the International ARM Study Group,17 and as used by the Eye Disease Prevalence Research Group (EDPRG) study.10 In our report, a NV-AMD case indicates presence of ‘wet’ AMD in one or both eyes, that is in at least one eye regardless of the status of the fellow eye. A case of GA-AMD indicates geographic atrophy in one or both eyes, with no NV-AMD in either eye. Accordingly, persons with NV-AMD in one eye and GA-AMD in the fellow eye are counted as NV-AMD cases and are not included in the GA-AMD group. Levels of visual impairment for the person relate to the best corrected visual acuity in the better-seeing eye. Sight loss (SL) indicates visual acuity of poorer than 6/12, visual impairment (VI) indicates visual acuity of <6/12–6/60 and blindness (BL) indicates visual acuity of <6/60, all in the better-seeing eye.
Results
The UK population aged 50 years and older projected to the years 2010, 2015 and 2020 are shown in table 1. For the UK, 414,561 persons in 2010 are estimated to have NV-AMD and this is expected to increase to 515 509 by the end of the decade. GA-AMD, which at present is irremediable, will be present in 193 652 persons in 2010 and is estimated to increase to 240 358 by the year 2020 (table 2).
Considering the numbers in the UK going into sight loss from both types of AMD, 132 970 are likely to have visual impairment and an additional 90 254 are expected to be blind in 2010. By the end of the decade, the expected numbers will be 171 530 visually impaired persons and 120 452 blind. These estimates are made under the assumption that the anti-VEGF treatment for NV-AMD covers 75% of those eligible.
The projected numbers with sight loss attributable to NV-AMD are shown in table 3. The findings indicate an increase over the decade resulting in substantial numbers remaining with sight loss due to NV-AMD by the year 2020, even with an assumed treatment coverage of 90% (about 144 000 in 2010 and 188 000 by 2020). This reflects the strong effect of the anticipated increase in the elderly population outweighing the beneficial effects of ranibizumab therapy.
Projections for the devolved countries within the UK are shown in table 4, in a Supplemental File.
Discussion
The projected rise in prevalence of AMD over the decade was in line with expectations, since the strong association of AMD with advancing age is well recognised, as is the demographic trend of the ‘ageing population’. The model estimated absence of NV-AMD in about 32% of all AMD cases (table 3). These patients had geographic atrophy only, not amenable to anti-VEGF therapy. Intuitively, a larger impact of treatment, however, was expected on the prevalence of sight loss due to AMD than the estimates given by the model. The projections suggest that the prevalence (numbers) of sight loss from AMD should continue to rise throughout the decade in spite of ranibizumab therapy. It seems that the beneficial effects of the treatment in reducing the risk and prevalence of sight loss would be outweighed by the strong anticipated demographic effect. This reaffirms the importance of continuing efforts to develop more effective and more broadly applicable therapies for AMD.
A recent US study (reported 2009) also predicts large increases in number of cases with visual impairment and blindness attributable to AMD over the next 40 years regardless of the treatment steps taken, with virtually all of the increases being attributable to the ageing of the US population.18 Compared with no treatment, the US model predicted that universal use of anti-VEGF therapies would result in only a small (16%) reduction in cases of visual impairment and blindness in patients with NV-AMD, pointing out that the model assumed anti-VEGF use only to treat sub-foveal choroidal neovascularisation after that condition had led to initial visual loss in at least one eye. By contrast, prophylaxis with vitamin supplementation (vitamins C and E) together with β-carotene and zinc applicable to a much wider case population of early AMD was predicted to have a much larger beneficial effect (subject to high compliance) in terms of the future number of cases with sight loss.
Bevacizumab (Avastin), a possible alternative currently used (off-label) by a number of ophthalmologists, has the same mechanism of action as ranibizumab but is about 50 times less costly per injection. The importance of considering pharmacoeconomics in the management of NV-AMD are discussed in a recent editorial.19 There are no adequate Phase III clinical trial data on effectiveness of Avastin compared with ranibizumab for treatment of NV-AMD; however, the 2-year National Eye Institute (of the US National Institutes of Health) clinical trials comparing the two treatments is expected to report in 2011. Assuming a similar effectiveness to ranibizumab, this alternative to ranibizumab would be far more cost-effective.
References
Supplementary materials
Web Only Data
Files in this Data Supplement:
Footnotes
Funding The study was funded by the Royal National Institute of Blind People (RNIB).
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- At a glance