Six-Year Incidence and Progression of Age-Related Macular Degeneration in Kenya

This population-based cohort study estimates the 6-year cumulative incidence and progression of age-related macular degeneration in older adults (≥50 years of age) in Nakuru, Kenya.

A ge-related macular degeneration (AMD) is a progressive degenerative disease that affects the central retina and is highly associated with age. 1 Advanced AMD, including geographic atrophy (late dry) and neovascular AMD (wet), leads to central vision loss.In the early dry form of the disease, deposits known as drusen are layered between the retina and choroid, and subtypes of drusen (based on size and morphologic features) form part of the more detailed classifications.Although AMD is a leading cause of visual impairment and blindness in populations living in high-income countries, 2 there is a paucity of available data from low-and middle-income countries, 3 including sub-Saharan Africa.However, a systematic review 4 found that overall posteriorsegment disease is a common cause of visual impairment in sub-Saharan Africa, and a survey 5 in Kenya found that 1 in 10 persons 50 years and older had signs of AMD.
Estimation of the incidence and progression of AMD and associated sight loss is important for planning of services.Treatment of neovascular AMD is currently possible in wellestablished health care systems but infrequently available in low-and middle-income countries.It is therefore important to be able to identify individuals at high risk for AMD to consider targeted approaches for prevention and/or treatment.Furthermore, rehabilitation services need to be planned for individuals developing visual loss as a result of AMD.Unfortunately, data to plan these services are currently lacking.The incidence of AMD has been investigated in 7 cohort studies of eye disease worldwide, [6][7][8][9][10][11][12][13][14][15][16] with no data from the African continent.There are large variations in the prevalence, phenotypes, and incidence of AMD in different populations, [6][7][8][9][10][11][12][13][14][15][16] making extrapolation of findings from studies in other regions of the world to an African setting difficult.The aims of the current study were to estimate the 6-year cumulative incidence of AMD in Nakuru, Kenya, and to identify risk factors for incident disease.

Methods
We studied a population-based cohort with 6-year follow-up of 4414 participants who had a complete assessment.Random cluster sampling with probability proportionate to size procedures was used to select a representative, crosssectional sample of adults 50 years and older from January 26, 2007, through November 11, 2008.A 6-year follow-up was undertaken from January 7, 2013, through March 12,  2014.The following examination protocols were implemented at baseline and follow-up, with detailed methods available elsewhere 17 and in the eMethods in the Supplement.The London School of Hygiene & Tropical Medicine Ethics Committee and the African Medical Research Foundation granted ethical approval for the study, which was also approved by the provincial medical officer for Nakuru County.Written approval was sought from the administrative heads in each cluster, usually the village chief.All participants gave written or thumbprint consent to participate.People requiring medical treatments were referred to the appropriate center.All data were deidentified.

Ophthalmic and General Examination
All participants underwent logMAR visual acuity testing on each eye separately and corrected visual acuity when less than 20/40 Snellen equivalent.Detailed interviews were undertaken in the local language on demographic details, information on risk factors, socioeconomic status, and full medical history.A nurse recorded the blood pressure, weight, height, and waist and hip circumferences.Participants had 2 nonstereoscopic, digital, 45°fundus photographs (1 disc and 1 macula centered) taken per eye by an ophthalmic clinical officer.Digital images were graded at an approved grading center.The senior grader (N.S.) graded all images for the presence of AMD.All eyes classified as having late-stage AMD were adjudicated by the Moorfields Eye Hospital Reading Centre clinician (T.P.).The adjudicator (T.P.) also graded 5% of randomly selected images to ensure quality control.

Definitions of AMD Used
A modified version of the international classification and grading system for age-related maculopathy and AMD was used for image grading at baseline and follow-up. 18Drusen were categorized based on size, uniformity of color, and margins.Patients were classified into hard or soft drusen categories: small drusen (<63 μm) were considered to be hard.Large drusen with a uniform density, sharp margins, and a nodular surface texture were placed in the soft distinct category, whereas those without sharp margins were classified as indistinct.When endstage disease was apparent, patients were classified as having geographic atrophy in the presence of well-demarcated regions with diameters greater than 175 μm, within which large choroidal vessels were clearly visible to the atrophy of the overlying choriocapillaris and retinal pigment epithelium.Neovascular AMD was graded as present when exudative features, such as serous fluid, hemorrhage, lipid exudates, or fibrosis, were seen to be originating primarily from the subretinal, pigment, and epithelial tissue layers.
Case definitions were based on the eye with more severe status if both eyes were gradable and on the gradable eye if only one was gradable.Early AMD was defined as the presence of large, soft drusen and pigmentation greater than 63 μm, and late AMD was defined as the presence of geographic atrophy or neovascular AMD.
Incident AMD was defined on the basis of the absence of AMD features at baseline on retinal images and the subse-quent presence of these features at follow-up.Incident late AMD was defined as the combination of no or early AMD at baseline and signs of late AMD at follow-up.

Dealing With Loss to Follow-up
Logistic regressions corrected for the survey design were used to calculate P values to assess differences between participants seen and lost to follow-up and those known to have died.An inverse probability weighting (IPW) model was used to allow estimation of cumulative incidence while accounting for those lost to follow-up.Those who had died between baseline and follow-up were excluded from the analysis.Multivariable logistic regression was used to identify independent baseline covariates associated with loss to follow-up.Covariates for which there was evidence of association with the outcome (P < .10)were kept in a multivariable model.Individuals without a complete set of the baseline covariates included in the final multivariable model were excluded from any estimations based on the weighted analysis.From this final model, the probability of being followed up was estimated based on the presence or absence of each of these baseline covariates.The inverse of this probability formed the weighting to be applied to account for those lost to follow-up.
The final step was to exclude those individuals lost to follow-up from the analysis and apply the IPW to account for those lost to follow-up.A sensitivity analysis for this approach involved a complete records analysis (ie, including only individuals who had complete records for outcome and all variables in the analysis).

Cumulative Incidence Estimation
The 6-year cumulative incidence of AMD was estimated by dividing the total (weighted) number of individuals who were clas-sified as having AMD at follow-up by the (weighted) number of individuals who were AMD free at baseline and examined at follow-up.The 6-year cumulative incidence was then used to estimate the expected number of new AMD cases per year.The size of the at-risk population in Kenya was estimated using the baseline prevalence of AMD from this cohort and the 2015 Kenyan population estimates for those 50 years or older.The 6-year incidence was then multiplied by this at-risk population and divided by 6, with the assumption that cumulative incidence was constant over time.Annual cumulative incidence was also estimated separately for men and women and in 10-year age categories (50-59, 60-69, 70-79, and ≥80 years).The incidence of progression from early to late AMD was calculated by examining participants with early AMD at baseline who were followed up and had a valid AMD status at follow-up.

Assessing Risk Factors Associated With AMD Incidence
The age-adjusted association between AMD incidence and each covariate was estimated using a Poisson regression model.A multivariable model was created with backward stepwise selection using the likelihood ratio test and a threshold of 2-tailed P < .05for retention of a variable in the model.

Results
At baseline, 4414 participants had a complete assessment, of whom 3304 (74.9%) had an AMD assessment from retinal imaging (Figure).Of these participants, 404 (12.2%) had AMD at baseline, with 366 (90.6%) having early AMD and 38 (9.4%) having late AMD.An additional 2900 participants did not have AMD at baseline and were therefore at risk for developing AMD at follow-up.1. Nonparticipants were divided into those who had died and those who lived but did not attend the examination clinic (eg, because of mass displacement in the period of postelection violence after the baseline study period) and/or those without a valid AMD assessment (eg, cataract obstructing a view of the retina).Compared with those followed up, participants who had died during follow-up were older and more likely to be female, have lower educational level, have higher systolic blood pressure, and have diabetes but lower body mass index.Compared with participants seen, those lost to follow-up were less likely to be Kikuyu or Kalenjin speakers, had lower levels of education, and were more likely to be from urban areas and from the highest or low- a Data are presented as number (percentage) of participants unless otherwise indicated.b Participants not followed up were alive, unknown, or had AMD status missing.c P < .05for association between the baseline characteristic and the odds of dying during the follow-up period among all participants identified as non-AMD at baseline, excluding the group who were not followed up.d Calculated as weight in kilograms divided by height in meters squared.e P < .05for association between the baseline characteristic and the odds of being followed up among all participants identified as non-AMD at baseline and not known to be deceased at follow-up.
est socioeconomic quartile.Individuals seen at follow-up were less likely to have impaired vision at baseline (48 [3.7%]) compared with those who died before follow-up (30 [13.3%]) or those not seen at follow-up (97 [7.0%]).In total, 1453 persons (50.1%) at risk for AMD were followed up after 6 years (Figure ), and 1282 had data on AMD at follow-up.Of these, 202 developed early AMD, and no participants developed late AMD.The 6-year cumulative incidence of early AMD, after taking account of loss to follow-up by using IPW, was 164.2 per 1000 persons (95% CI, 136.7-195.9per 1000 persons).
In addition, 366 participants with early AMD were at risk of progressing to late AMD at follow-up, of whom 173 were followed up and 142 had a valid AMD assessment (25 of 31 who did not have an AMD assessment had a lens opacity that obscured the retinal images).Two individuals with early AMD from the 142 at risk had developed late AMD at follow-up (Figure ), giving a 6-year cumulative incidence of progression from early to late AMD of 24.5 per 1000 persons (95% CI, 5.0-111.7 per 1000 persons).
Of the 38 individuals with late AMD at baseline, 17 (44.7%)were followed up, 5 (13.2%) died, and 16 (42.1%)were not located for follow-up.Of the 17 individuals who were followed up, 4 (23.5%) did not have a valid AMD assessment (because of obstructing lens opacities), 11 (64.7%)remained classified as having late AMD, and 2 (11.8%) had a critical eye that was difficult to grade because of image quality but most likely had stable end-stage AMD.The visual status at baseline and follow-up is given in eTable 1 in the Supplement.
On the basis of extrapolations of these results to census data and population estimates in 2015 (assuming incident cases annually is proportional to the cumulative incidence), we estimate that, in Kenya, 103 070 new cases of AMD (at any severity) develop every year in persons 50 years or older, of whom 65 720 (63.8%) are women (Table 3).
Multivariable analysis of factors associated with incident AMD indicated an increasing incidence of AMD with older age (P for trend = .02),female sex (P = .001),and diabetes (P = .04)(Table 4).
Of the 234 individuals in the cohort who developed incident vision impairment, 162 (69.2%) had an available AMD assessment at follow-up.Of the 162 individuals, 52 (32.1%) had AMD, and 3 of these patients were classified as blind.It was not possible to infer whether vision loss was attributable solely to AMD or a combination of other ocular comorbidities.Change in vision category from baseline to follow-up in all those with  a valid AMD status at baseline and follow-up is given in eTable 3 in the Supplement.
A total of 202 participants in the cohort developed incident AMD, of whom 192 (95.0%) had normal vision at baseline.A total of 27 (14.1%) of the 192 developed visual impairment by follow-up.A total of 1080 participants did not develop AMD, with 1040 (96.3%) in this group having normal vision at baseline.Of these 1040 individuals, 83 (8.0%) developed vision impairment.

Discussion
The Nakuru Eye Disease Cohort Study provides longitudinal data on AMD in sub-Saharan Africa from a population-based cohort.Although there is 50% loss to follow-up and few cases of late AMD (resulting in wide CIs for that outcome), there are limited data on these outcomes from this region.With those caveats in mind, during 6 years, 1 in 6 adults 50 years or older developed early manifestations of AMD, with women having a higher incidence than men.Increasing age was strongly related to the prevalence and incidence of AMD.Most incident cases of AMD were defined on the basis of the development of large drusen (>64 μm).Late AMD was infrequent at baseline, and consistent with this pattern, only 2 cases of incident late AMD were found at follow-up.Both incident cases developed in individuals with early AMD at baseline, and no case was identified that progressed from no AMD to late AMD.
Our data estimate a higher incidence of AMD than other (non-African) cohort studies of eye disease (eTable 4 in the Supplement).A likely explanation is that the Nakuru cohort includes only persons 50 years or older, similar to the next 2 highest cumulative annual incidence estimates, which were also from samples of older individuals in Copenhagen 6 and Reykjavik. 11Furthermore, in the Nakuru study, there was a higher incidence of early AMD and a lower incidence of late AMD compared with other populations.][7][8][9][10][11][12][13][14][15][16] The 2 participants who progressed from early to late AMD were older than 80 years; the reduced number of individuals in this age group in the Nakuru cohort may explain the low incidence of late AMD because age is the leading risk factor for incident AMD.
A high prevalence of early AMD at baseline and a high incidence of early AMD at follow-up may suggest that the population under investigation has a higher risk of developing AMD.The relatively high prevalence and incidence may possibly be attributed to greater UV light exposure, earlier biological aging, greater genetic predisposition, or greater susceptibility to inflammatory processes, which have been attributed to AMD. 19 The proportion of persons with vision loss attributable to AMD is relatively low because overall vision loss primarily attributable to conditions such as cataract is largely under control in more developed health care systems.Of those with AMD, 14% developed vision impairment during the 6-year study period compared with 9% during 14 years in a well-developed health care system in Copenhagen.However, vision impairment cannot be attributed to AMD alone.

Strengths and Limitations
This study was conducted under challenging circumstances with limited infrastructure.It provides independently graded, digital image-based analysis of AMD in an East African cohort that is diverse and largely representative of the population from which it was sampled.Detailed ophthalmic, demographic, and anthropometric assessment of each participant has enabled valuable analyses of associations and risk factors.
Despite these strengths, limitations of the current study may have contributed to an underestimation of the true incidence of AMD and AMD lesions.First, the main limitation of our study is the large loss of participants at followup.Of the 4414 persons who participated in the baseline study, only 2171 persons participated in the follow-up examination, of whom 1424 had gradable images at baseline and follow-up, with most being excluded from analysis as a result of camera failure at baseline, lack of electricity access, and/or ungradable images attributable to media opacities.Only those who had gradable images at both time points were included in the analysis.A complete case record analysis was conducted without weighting for loss to follow-up (eTable 5 in the Supplement), with results similar to those using the IPW modeling (Table 2), with a possible underestimation of the incidence in women when loss to follow-up is not taken into account.
Second, changes in procedures between the baseline and follow-up examination (different retinal camera) may have introduced bias.Change in cameras may have caused images with different color profiles and saturation levels, resulting in different abilities to detect AMD features (eg, drusen and pigment).Furthermore, the lack of stereophotographs meant cases of retinal elevation may have been overlooked.These factors may have resulted in bias toward an underestimation of the incidence of subtle early AMD lesions, such as small drusen, or an overestimation of pigmentation attributed to AMD.Comparison images between cameras of the same study participants at baseline and follow-up are given in eTable 6 in the Supplement.Overall image quality was considered to be equivalent at the 2 time points in those with clear media (ie, no lens or corneal opacity).
Third, survival bias may have caused an underestimation of the true incidence of late AMD if those who died before the follow-up had experienced advanced AMD lesions after the first examination, and it is possible that those with worsening disease were more likely to attend follow-up visits.The low prevalence of late AMD at baseline and low incidence can, in part, be attributed to a lack of older individuals in this study population, with an expected shorter life span than other populations with data on AMD incidence.Classification bias may also have contributed to the estimates, and histologic studies would be required to confirm whether the manifestations being attributed to AMD are consistent with those in other populations.

Conclusions
We estimate that, in Kenya, more than 100 000 new cases of AMD, mainly early AMD, will develop every year in persons 50 years and older, although a 50% loss to follow-up and wide CIs for progression to late AMD limit definitive conclu-sions from these findings.The AMD in this population was found to be phenotypically different from that in a prior study. 5However, because the relatively high incidence was restricted to occurrence of early AMD, the high incidence of early AMD may not have major implications for clinical practice given the low number of individuals with associated visual loss.

Table 1 .
Baseline Characteristics of All 2900 Individuals With No AMD at Baseline According to Availability of AMD Status at 6-Year Follow-up a

Table 3 .
Extrapolated Number of New Adults 50 Years and Older in Kenya Developing Age-Related Macular Degeneration per Year a

Table 2 .
Age-and Sex-Specific 6-Year Cumulative Incidence of Age-Related Macular Degeneration Among the Nakuru Eye Disease Cohort Study Participants jamaophthalmology.com (Reprinted) JAMA Ophthalmology June 2017 Volume 135, Number 6 635 Downloaded From: by a Queen's University Belfast User on 10/23/2017

Table 4 .
Age-Adjusted and Multivariable Analysis of Baseline Covariables and Incident AMD in the Nakuru Eye Disease Cohort Study Sample of 1282 Individuals For multivariable analysis, an initial model was fitted that included those variables associated with outcome in age-adjusted analysis (using a Wald test threshold P < .05 to indicate association).A backward stepwise approach was then applied to obtain a final multivariable model, removing variables with P > .05one by one.d Calculated as weight in kilograms divided by height in meters squared.
a At risk indicates no AMD at baseline.bIncident AMD indicates early or late AMD at follow-up.c