Discussion

We have applied findings from a recent systematic review 4 based on 31 populations of European ancestry with a wide age range (50–97 years) (including 18 populations from Western Europe with four from the UK23–26) to the population of the UK to model the number of prevalent cases of late AMD, GA and NVAMD. We estimate that there are currently 513 000 (95% Cr I 363 000 to 699 000) prevalent cases of late AMD, 276 000 cases of GA (95% CrI 188 000 to 396 000) and 263 000 cases of NVAMD (95% CrI 185 000 to 361 000). With knowledge of the probability of death by age, and assuming similar mortality in those with and without AMD, the annual number of incident cases is 71 200 (95% CrI 41 100 to 120 900), 43 700 (95% CrI 27 000 to 71 200) and 39 700 (95% CrI 23 700 to 68 300), respectively. While women have slightly higher age-specific prevalence rates of late and NVAMD, it is the much greater number of older women in the UK that results in the gender difference in the number of prevalent and incident cases (figures 1 and 3). This explains the perceived view of an abundance of older women presenting to medical retina clinics with AMD.

How do these numbers compare with previous estimates? Numbers from this study are similar to, but more precise than, our earlier estimates of prevalent cases of 191 000 (95% CI 171 000 to 310 000) with GA and 271 000 (95% CI 179 000 to 405 000) with NVAMD, which combine to give 462 000 with late AMD within the present study's CrIs. Our CrIs also include recent point estimates3 derived by applying findings from another review27 to the UK population (although our NVAMD estimates are lower). There have also been a number of attempts to estimate incidence of NVAMD in the UK in order to plan treatment demand and costs.28 We are not aware of any UK estimates for GA incidence, probably because this is less of a priority given lack of effective treatment (although the need for low vision services remains). Initial estimates of NVAMD requiring photodynamic therapy in England and Wales suggested 5000 patients per year, estimated from either blind and partial sight registrations10 or from two non-UK prospective cohort studies.11 29 30 With the introduction of newer therapies for NVAMD,5 potentially suitable for classic and occult types of choroidal neovascularisation, estimates of the number eligible for treatment have increased from 7000 to 26 000 patients per year.28 The source of these estimates is unclear,28 but Novartis Pharmaceuticals UK also suggested 26 000 new cases of wet AMD in the UK per year, reported as being calculated from our earlier review of prevalence.2 31 This figure is commensurate with estimates that there are 13 000 to 37 000 incident cases of NVAMD in England and Wales per year,28 based on two prospective cohort studies in the Netherlands32 and Australia.8 Our prediction of incidence is higher, but our CrIs are wide and include many of the previously reported estimates (including the popular figure of 26 000 annual cases of NVAMD). Our overall and age-specific annual incidence rates (Web table 2) also appear to be higher than those calculated33 from ‘first eye’ cumulative AMD incidence in other studies.6–9 The exponential rise in prevalence and incidence rates with age (with prevalence rates of late AMD quadrupling per decade)4 makes direct comparisons with estimates from individual studies difficult because they are heavily dependent on the age distribution of the sample at baseline and period of follow-up. Representation of data from older people (especially those aged 80 years or more) is needed to avoid underestimation of population numbers with late stage AMD (while recognising that the number with or developing AMD in a population will ultimately fall with age due to increased mortality). Unfortunately, any single prospective study has limited information on incidence at older ages when AMD is particularly prevalent. Although there are some studies with long durations of follow-up (up to 15 years),8 13 14 results are generally reported as 5- or 10-year incidence rates making it difficult to determine the annual incidence without knowledge of the change in pattern of incidence with time/age. Therefore, estimating incidence from prevalence, particularly at older ages, is an efficient approach and likely to be more accurate in the absence of large studies with follow-up at all ages (particularly at older ages where follow-up is more problematic).

Our numbers are calculated from either eye prevalence rates.4 While this might overestimate any potential visual loss associated with the condition (especially when both eyes are used), this is appropriate when estimating potential therapeutic costs as bilateral disease is less common, and presence of any disease may warrant treatment. Using ‘either eye’ case definitions will result in some individuals being counted in GA and NVAMD groups, but this number is likely to be small. While this may partly explain why the estimated number with GA and NVAMD combined is higher than those estimated with late AMD, a more plausible explanation is that the estimates of prevalence of late AMD are more precise as they are estimated from a meta-analysis of a larger number of studies (31 studies) compared with those for GA and NVAMD (23 studies). Similar to earlier work estimating incidence of glaucoma from age-specific prevalence,22 our calculations assume that: (i) AMD is life-long after diagnosis, (ii) overall mortality risks are the same in those with and without AMD and (iii) AMD is stable in the population (with risk factors for the condition remaining the same). The first assumption is correct, there is no consistent evidence to argue against the second34–36 and we have found no strong evidence in our meta-analysis that late AMD prevalence has changed over time,4 although a recent study suggested that the prevalence of AMD (including early and late forms) may have fallen,37 attributed to improved diets and reductions in smoking prevalence.38 The estimation of incidence for older age groups is the most plausible explanation for our higher numbers, with the peak in number of prevalent and incident cases being just below 90 years of age. The noticeable kink in the decline beyond 90 years of age reflects the sharp decline in births during World War I (1914–1918), followed by the rapid postwar baby boom (approximately 1920).39 While our estimates of NVAMD incidence are high, this may not reflect the number in need of treatment. The type of NVAMD on presentation cannot be estimated from these data, that is, classic or occult, although limited evidence from other sources suggests a ratio of 2:1.11 Moreover, older patients may be too frail to benefit/undergo treatment.

In the absence of any large, geographically representative population study of AMD in the UK, this review provides the best estimates of the number of prevalent and incident cases of late AMD, GA and NVAMD in the largely white older population of the UK. While most of the current older UK population are white (96% white population in England and Wales aged ≥65 years),40 this is likely to diminish in the future. However, evidence suggests similar or lower rates of AMD in populations of non-European ancestry.27 41 Our estimates are based on pooled findings from studies carried out in white populations from North America, Australia and Europe with shared ancestry (including recent23 and historical studies in the UK24–26). All studies contributed to the prevalence estimates used to calculate incidence. We adjusted for differences between studies in methodology which may have biased rate estimates24–26 (such as no fundus imaging and study-specific as opposed to internationally recognised disease definitions).15 16 Projections over the next decade suggest that the number of prevalent cases of late AMD will increase steadily by a third by 2020 due to population ageing. These evidence based estimates can be used to help plan social and healthcare provision for the present and the future.