The majority of data on AMD available globally is from population-based studies undertaken in white and Asian populations [5]–[13], and few data are from peoples of African descent. The data that do exist for individuals of African descent are largely from studies undertaken in populations living outside of the African continent [14],[15]. It is presumed that AMD is rare in Africans; however, in the last 10 y, African population-based studies have suggested that posterior segment eye diseases are highly prevalent, and this group of disorders, which includes AMD, diabetic retinopathy, and glaucoma, has been highlighted as either the leading or second leading cause of blindness in surveys undertaken in Cameroon [16], Tanzania [17], Kenya [18], Rwanda [19], Zanzibar [20], and Guinea [21]. These studies, however, did not assess AMD as a specific entity and did not use digital retinal photography. Moreover, comparative data for whites and Africans living in the same geographical area (Baltimore, Maryland, US) have suggested differing predispositions towards AMD, with possible genetically protective factors for AMD progression in individuals of African descent compared to their white counterparts [22]. Population-based evidence for African populations living in Africa on the prevalence of the disease, and of the risk factors for AMD, is therefore important.

3,387 (77.3%) participants underwent retinal photography. An image for grading for AMD in at least one eye was available for 3,304 individuals (75.4%) (image group). 3,274 individuals (74.7%) had both methods of screening (Figure 1).

Compared to people in the SLB group, those who had images taken were more likely to be male, younger, urban residents, not Kikuyu, not diabetic, and not visually impaired (Tables 1 and S1). A correction factor was necessary because there were significant differences in the characteristics of the SLB-only group compared to the image group, and so it was not possible to generalise the results from the image group to the whole population.

The prevalences of drusen and pigmentary irregularities as observed on the fundus images, by gender and age, are shown in Table 2. Note that varying features of AMD, e.g., drusen and pigmentation, can co-exist in a single eye and in both eyes and therefore one individual may be listed under more than one AMD characteristic. However, the prevalence of AMD in the population is based on the grading of AMD in a person. Drusen were present in a large proportion of the population. The most common type of drusen encountered in all ages was small, hard drusen, <63 µm, which were present in 59.1% of the study population. Large, soft drusen, which are considered to be indicative of early AMD, were present in 9.4% of the population. There were significant age trends (X² trend test, p<0.001), with increased prevalence of all drusen and all pigmentary changes from age 50 y to age 79 y. The gender difference was less strong, though drusen and pigmentary changes were more common in women than men. The overall prevalence of retinal pigment abnormalities was 4.8% (95% CI, 3.7–6.1). Increased pigment was seen more frequently than depigmentation in all age groups, and prevalence increased from 1.6% in the lowest age groups to 7.2% in those aged 80 y or more. The difference in prevalence of pigment in men and women was not significant (p=0.66).

Neovascular AMD was more common (0.9%) than geographic atrophy (0.5%) (Table 3). There were significant age trends for both, with geographic atrophy being prevalent in only 0.3% of those in their 50 s and increasing to 2.0% in those age 80 y and over.

The prevalence of all stages of AMD was lower when SLB grading was used than when grading was from retinal images (Table 4). The total prevalence of AMD from SLB was 7.4% (early AMD, 6.7%, and late AMD, 0.7%), while the prevalence of AMD by retinal image grading was 12.4% (early AMD, 11.2%, and late AMD, 1.2%). A smaller proportion of participants were classified as ungradable by SLB (30; 1.6%) than by retinal image grading (83; 2.6%). A correction factor of 1.7 for total AMD needs to be applied for those who did not have retinal imaging (considered the gold standard) to get the true prevalence estimate of AMD in the SLB group.

Sensitivity and specificity analysis for the detection of early AMD and late AMD by SLB grading versus retinal imaging, in those individuals who had both SLB and image grading available, showed that SLB grading had poor sensitivity (early, 21.3%, and late, 36.8%) and good specificity (early, 95.2%, and late, 99.9%).

A total of 404 (366 early AMD and 38 late AMD) cases were confirmed by images, while another 85 cases (75 early AMD and ten late AMD) were detected in the group that received SLB only. Combining all cases gives a total of 489 cases, or a prevalence of 11.3% for AMD (435 [10.2%] early AMD and 48 [1.1%] late AMD) in this population. However, since SLB underdiagnosed AMD by a factor of 1.7, if this factor is applied to the SLB prevalences, an adjusted prevalence of 12.7% (11.4% for early AMD and 1.3% for late AMD) is reached (Table 5).

When extrapolating these data to the entire Kenyan population based on data from 2007, we estimated that there are 283,900 to 362,800 people over 50 y in Kenya with early AMD and 25,200 to 50,500 with late AMD.

Age/sex-adjusted analyses show that only age and gender were significantly associated with early AMD, with those affected more likely to be female and with prevalence increasing with each decade of age (Table 6).

Modelling age as a continuous variable did not alter the findings (Table S2).

Age/sex-adjusted analyses showed that only age was significantly associated with late AMD, with increased late AMD prevalence with every decade after 50 y (Table 7). All other variables showed no association.

Of the 487 people with any grade of AMD (diagnosed by SLB or retinal images), a total of 137 (28.1%) were visually impaired, including 12 blind people (2.5%; 95% CI, 1.3–4.8), four with severe visual impairment (0.8%; 0.3–2.2), and 82 with moderate visual impairment (16.8%; 13.4–20.9). 350 (71.9%; 7.0–76.4) of those with AMD had normal vision (Table 8). Among the 669 people with visual impairment in the entire Nakuru study, 137 (20.5%) had features of AMD, either exclusively or in combination with other pathology.

Very few data exist on the prevalence and causes of AMD in Africa, and to our knowledge, this is the only population-based study in Africa using an internationally recognised grading system and digital retinal photographs. Although data from Rapid Assessment of Avoidable Blindness surveys exist, these cluster AMD with other posterior segment eye diseases, and so no direct comparisons can be made with the findings from this study. A Nigerian survey used similar methodology, including a population-based approach and fundus photographs; however, retinal imaging was performed only in individuals with a visual acuity of ≤6/12 [26]. In the present study, 75.1% of individuals identified as having AMD had an acuity of 6/12 or greater.

Only one other prospective study of AMD in Africa was found in the peer-reviewed literature. A hospital-based study in South Africa that looked specifically at AMD in Africans was published in 1978 [31]. The study participants were aged 50 y and older, as in this study, and were examined by indirect ophthalmoscopy as well as photography. Higher levels of “senile macular degeneration”, as AMD was then termed, were reported than in this study, affecting 17.4% of participants in the study. It is likely that the hospital population sampled was not representative of the general population, as enrolled participants had attended the hospital eye clinic of their own volition and therefore were more likely to have had symptomatic vision loss than the general population. Typically, the demographic of individuals attending a hospital clinic also differs from the general population in terms of SES and life expectancy. No population-based studies specifically reporting AMD in Africa have been published, to our knowledge.

The prevalence of AMD in this study was also lower than that in a study in the Caribbean, where a 28.7% prevalence of AMD was found [32]. In general, the prevalences in our study are similar to or higher than those documented for Hispanic and Asian populations (range 7.1%–13.6%) [13],[33]–[36], but lower than those found in white populations (range 9.3%–43.1%) [36]–[39].

Direct comparison between studies is not appropriate because of the different grading systems and diagnostic techniques used.

A strength of this study was the high response rate (88.1%). Despite the sophisticated equipment used in the examinations, people were not transported to a fixed examination site, but instead the examination site moved from cluster to cluster. A large, representative population-based sample was examined by SLB and retinal imaging. Another strength of the study is that the same experienced ophthalmologist (W. M.) was present throughout the study and examined all participants. However, a lack of stable electricity supply resulted in the number of people who had retinal images being reduced in some clusters.

A limitation of the study was our not having been able to obtain retinal images for all study participants. This is in large part due to the logistical constraints of performing electricity-dependent examinations in a setting where electricity supply cannot be guaranteed. Univariable analyses comparing those with gradable images (n=3,304) and those without (n=83) found significant differences. Those with no gradable images were more likely to be older, have poor vision, and have a cataract, thus the prevalence for this population may be slightly underestimated. The difference between the groups is likely due to a lack of stable electricity supply in the more rural clusters, where participants were demographically different.

When disease estimates using both methods were compared, SLB was found to have consistently under-diagnosed AMD. An analysis of the false negatives for late AMD revealed that lesions were noted but were not called neovascular AMD; instead they were called macular scars. This may be a reflection of general practice in Africa, where it is often repeated in residency courses that “wet AMD” does not occur in Africans, as has been asserted in several studies [40],[41]. This leads to late AMD being placed far down in the differential diagnosis for macular scars.

There was also discrepancy in identifying pigmentary lesions of early AMD. Retinal imaging identified many more hyperpigmented changes at the macula than did SLB. There are large natural variations in retinal pigmentation, resulting in colour differences between individuals. Such variation can tend to mask the more subtle variation between the important lesion types [42]. Studies have shown that the macular pigment density in other population groups is significantly lower than in African individuals [43]. The study ophthalmologist's perception of what constitutes increased pigment and what is normal background pigment in an African eye, in comparison to the reading centre's criteria, may have led to differences in classification.

Vision is measured based on the person's better eye, whereas AMD affects both eyes, and therefore a disparity between late AMD and poor vision can be seen. For example, there were 38 participants in the present study with late AMD, but only 16 with AMD and a visual acuity <6/60.

Disease subgroups included limited sample sizes, particularly for advanced AMD and blindness, which should be noted when interpreting the results.

Data collection began in 2007 and was based on electoral roll data from 2006. The population demographic is likely to have changed in the time to publication, and given population growth and increased survival, the estimated national prevalence of AMD could underestimate current prevalence.

Of note, location information collected from participants in this study will allow for incidence studies to be carried out in the future, thus providing new insights into the natural progression and incidence of AMD in this population.

Table S1

Table S2