Four-compartment model for incident cataracts.
Prevalence of cataract by age for 4 eastern Africa sites. VA indicates visual acuity.
Incidence of cataract by age for 4 eastern Africa sites, calculated for the midpoint between prevalence data points. VA indicates visual acuity.
Lewallen S, Williams TD, Dray A, Stock BC, Mathenge W, Oye J, Nkurikiye J, Kimani K, Müller A, Courtright P. Estimating Incidence of Vision-Reducing Cataract in AfricaA New Model With Implications for Program Targets. Arch Ophthalmol. 2010;128(12):1584–1589. doi:10.1001/archophthalmol.2010.307
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
To estimate the incidence of vision-reducing cataract in sub-Saharan Africa and use these data to calculate cataract surgical rates (CSR) needed to eliminate blindness and visual impairment due to cataract.
Using data from recent population-based, standardized, rapid-assessment surveys, we calculated the age-specific prevalence of cataract (including operated and unoperated eyes) from surveys in 7 “districts” across Africa. This was done at 3 levels of visual acuity. Then we used the age-specific prevalence data to develop a model to estimate age-specific incidence at different visual acuities, taking into account differences in mortality rates between those with cataract compared with those without. The model included development of opacity in the first eye and second eye of people older than 50 years. The incidence data were used to calculate target cataract surgical rates.
Incidence and CSR needs varied significantly in different sites and were lower in some than expected. Cataract surgical rates may depend on genetic, environmental, or cultural variations and will vary with population structure, which is not uniform across Africa.
Africa should not be viewed as homogeneous in terms of cataract incidence or CSR needed. These CSR calculations should be useful for more appropriate planning of human resources and service delivery on the continent. The methodology can be applied to other population-based data as they become available to determine appropriate CSR targets.
For many years, the magnitude of vision loss and blindness in Africa has been estimated using a limited number of population-based surveys. It is usually assumed that about 1% of the population is blind, with half of this due to cataract,1 although some recent data indicate that the prevalence of blindness may not be this high.2- 4
Cataract remains the leading cause of blindness in Africa; therefore, planning for its treatment is a priority of the VISION 2020 initiative to eliminate avoidable blindness by the year 2020. The VISION 2020 strategy includes planning by administrative districts of around 1 million people. Planning requires realistic target setting—how many cataract operations need to be done to eliminate blindness or visual impairment? The cataract surgical rate (CSR), the number of operations done per million persons, is a convenient indicator for planning and monitoring.5 However, estimating what the CSR needs to be requires one to take into account a number of factors and to make assumptions. Key factors that determine what the CSR to eliminate visual disability from cataract needs to be for any population of 1 million people include (1) the age structure or proportion of the population that is old enough to be at significant risk of cataract-related vision loss (for convenience, this is often assumed to be those aged 50 years and older); (2) the visual acuity threshold at which cataract is operated on, ie, how blindness or visual disability is defined (this factor might be determined by policy [rationing or case selection] or by the demands of the population); and (3) biologic or environmental factors that determine incidence of cataract in a population. For example, cataract occurs at a younger age and has a higher incidence among Indians than among Europeans.6
The Rapid Assessment of Avoidable Blindness (RAAB) survey uses a population-proportional-to-size sampling technique to select a representative group of people aged 50 years or older in a district of 1 to 2 million people. Each person receives a standard eye examination, including visual acuity measurement in each eye, measured in 6 categories (≥6/18, <6/18 but ≥6/60, <6/60 but ≥3/60, <3/60 but ≥1/60, light perception, or no light perception). If presenting acuity of an eye is less than 6/18, it is remeasured with a pinhole as a crude determination of refractive error. The lens of the eye is evaluated by examining the red reflex with distant direct ophthalmoscopy in a dark or shaded area without pharmacologic dilation and graded as normal, obvious lens opacity, aphakia, or pseudophakia. The age of the patient at the time of surgery is recorded for pseudophakic or aphakic patients. A main cause of presenting visual acuity of less than 6/18 is assigned to each eye, and the pupil is dilated at the discretion of the examiner.
We used data from RAAB surveys in Africa to model the epidemiology of visually significant cataract and to estimate the incidence of cataract causing loss of visual acuity at different levels. Then we used the incidence data to calculate target cataract surgical rates required for different visual acuity levels.
Researchers who have carried out published RAABs in Africa were contacted to obtain the raw data from the surveys in Microsoft Office Excel form (Microsoft, Redmond, Washington). Data were imported into the free software program, language R (R Foundation for Statistical Computing, Vienna, Austria) and Mathematica version 18.104.22.168 (Wolfram Research, Champaign, Illinois) for analysis.
Our interest was in the prevalence of visually significant lens opacity, whether the eye had been operated on or not. This included eyes that were aphakic, pseudophakic, or had a lens opacity causing vision loss at a given level. In this study, we refer to this condition as cataract, recognizing that it includes eyes with lens opacity as well as aphakic and pseudophakic eyes. Patients with 1 or both eyes having vision loss due to cataract with best-corrected visual acuity of less than 6/18 (or <6/60 or <3/60 in subsequent iterations) or eyes that had been operated on for cataract were identified using the appropriate codes from the RAAB forms. For the calculations at visual acuity of less than 6/18, we assumed that eyes that received surgery had a visual acuity of less than 6/18 at time of surgery. For calculations with cataract at visual acuity of less than 6/60, we included eyes with cataract (without surgery) causing acuity of less than 6/60 and assumed that eyes that received surgery had visual acuity of less than 6/60 at the time of operation. Similar procedures were used for cataract at a visual acuity of less than 3/60.
First, we calculated the age-specific prevalence (in 5-year age groups) of unilateral and bilateral cataract separately, designating each individual as unilateral, bilateral, or cataract free. The individual's age at the time of examination was used. We used eight 5-year age groups beginning at 50 years, with the oldest age group consisting of individuals aged 85 years and older. In cases in which both eyes were affected but operated on at different times we used the most recent (older) age of operation for the person. If one eye had cataract and the other eye was operated on (designated as bilateral), we used the age of examination for the person.
To estimate incidence, we extended the method described by Podgor in 1986.7 Podgor's method allows one to estimate the age-specific incidence of a single disease based on known age-specific prevalence and mortality rates. The method consists of a 3-state model in which all people are described as healthy, diseased, or dead. Then, known values (prevalence and probability of death) and an assumption that the system is closed can be used to estimate the unknown quantity of interest (incidence). In the case of cataract, we wished to model the incidence of both first and second eye cataracts. The two conditions are obviously related and cannot be modeled as separate diseases; hence, a 4-compartment extension of Podgor's method was necessary (Figure 1).
We defined 4 states: cataract-free, unilateral cataract, bilateral cataract, and dead. Like Podgor, we assumed that mortality is higher in those with cataract than in those without. We assumed that everyone who develops cataract first develops opacity in one eye, and then may develop cataract in the second eye at any later time. These transitions may both occur in the same 5-year period but need not. Starting with the known prevalence of unilateral and bilateral cataract at age x (P0U,P0B) and, assuming that mortality is exponentially distributed, we calculated what the first and second eye cataract development rates (λ2, λ4) had to be over a 5-year interval to produce the known prevalence at age x + 5 (P1U,P1B). Figure 1 shows this 4-state model.
The pool of people who are cataract-free at time X1 would be those who were cataract-free at time X0 and neither died nor developed cataract in the first eye, described as N1(1 − P1U − P1B) = N0(1 − P0U − P0B) A, where N0 and N1 are the total population at X0 and X1, respectively, and A is the probability that a person who is cataract-free at time X0 stays alive and cataract-free until time X1.
The pool of people with unilateral cataract (first eye) at time X1 would be those who had unilateral cataract at time X0 and survived without developing a second eye cataract, plus those who were cataract free at time X0 but developed cataract in 1 eye only. This gives us N1P1U = N0(1 − P0U − P0B) B + N0P0UC, where B is the probability that an initially cataract-free person stays alive but develops cataract in 1 eye only and C is the probability that a person with unilateral cataract at time X0 will survive without developing bilateral cataract.
Finally, the pool of people with bilateral cataract (second eye affected) at time X1 has 3 sources: cataract-free individuals at time X0 who develop unilateral, then bilateral cataract; those with unilateral cataract at time X0 who survive and develop bilateral cataract; and those with bilateral cataract at time X0 who survive. This gives us N1P1B = N0(1 − P0U − P0B) D + N0P0UE + N0P0BF, where D is the probability that a person who is cataract free at X0 develops unilateral, then bilateral cataract by X1; E is the probability that a person with unilateral cataract at X0 survives and develops bilateral cataract by X1; and F is the probability that a person with bilateral cataract at X0 survives until X1.
The probabilities A through F can be described by an exponential model in terms of death rates λ1, λ3, λ5 (which can be calculated from death probabilities), and λ2 and λ4, which are unknown first- and second-eye cataract incidence rates.
The age-specific probability of death during the interval for each country was taken from life tables published by the World Health Organization.8 We used the average of these for the African countries from which we had data. The 5-year death rate, λ1, for people who are cataract free was calculated from the 5-year probability of death, nqX, as follows: λ1 = −ln(1 − nqX). We assumed that the death rate for people with cataract in 1 or both eyes would be 1.5 times that of those without cataract. That is, we let λ3 = λ5 = 1.5λ1.
Manipulating the 3 equations above, we can eliminate the total number of people at each time to yield 2 equations:
Substituting the values for A through F gives us:
With 2 equations and 2 unknowns, we can solve for λ2 and λ4 using prevalence data from each site.
The λ2 value we find in this way can be interpreted directly as 5-year, first-eye cataract incidence. To find the 5-year, second-eye incidence from λ4 (incidence of second eye cataract among those who already have cataract in 1 eye), we multiply by the prevalence of unilateral cataract averaged over the initial and final age periods. To calculate the overall incidence in the survey population (aged ≥50 years), we multiplied the age-specific incidence by the proportion of population in each age group. We divided the incidence rate per 5 years by 5 to arrive at an annual incidence for both first eye cataract and second eye cataract.
To calculate target CSRs, we assumed, as have previous models, that operating on the incident cataracts each year would gradually result in a steady-state situation in which visual loss from cataract would be eliminated in 5 to 10 years. Target CSRs to operate on first or second eye incident cataracts were calculated for a hypothetical population of 1 million, of which 10% of the population was older than 50 years. A final adjustment was needed to take into account the proportion of the population older than 50 years, if it is not 10%. The CSR with X % aged 50 years or older is a simple proportion: CSR10%/10 = CSRX%/ X.
Raw data from 5 published RAAB surveys in sub-Saharan Africa (Kilimanjaro, Tanzania; Kericho, Kenya; Nakuru, Kenya; Western Region, Rwanda; and The Gambia)2- 4,9,10 and 2 unpublished surveys (Koulikor, Mali, and Eritrea) were obtained. Teams carrying out these RAABs were all trained by persons who attended a standardization workshop.
The prevalence and incidence of cataract causing vision loss in each site at different levels are shown in Table 1 and Table 2, respectively. Table 3 shows the CSR required to operate on first and second eyes at each visual acuity level, calculated for districts where 10% of the population is older than 50 years. Table 4 shows how CSR varies as the proportion of the population older than 50 years changes, using incidence rates from Kilimanjaro, Tanzania as an example.
Prevalence and incidence at each visual acuity level increased with age according to the model as shown in Figures 2 and Figure 3 respectively, where we have pooled data from the 4 closely situated sites in eastern Africa.
Of those who already had unilateral cataract, the probability of the second eye developing cataract within 5 years was 30% to 70% among the sites and tended to increase with age. This was true at all 3 visual acuity levels.
This model could also be used to determine the CSR needed to operate on a particular case mix, for example, to operate on 1 eye of each person with bilateral cataract at 3/60, a second eye of half of these, and half of the eyes at the 6/60 level. However, this is not a particularly useful approach to planning because programs do not set these sorts of restrictions and the case mix is the result of a complex mix of factors involving demands by patients as well as practices and outcomes of the surgeon.
This model provides a method to estimate cataract incidence from age-specific cataract prevalence. The incidence rates can be used to plan target CSRs needed to eliminate visual impairment from cataract at different visual acuity levels.
There are a number of limitations to the model, starting with the RAAB examination of the lens, which is based only on the red reflex and does not use any of the accepted cataract grading schemes. We only included eyes with cataract if the main cause of visual loss was considered to be lens opacity, similar to the criteria used to determine whether an eye is suitable for cataract surgery in these settings. Therefore, eyes with cataract and obvious vision-reducing posterior pole pathology would not be included as cataracts. This will result in underestimation of the true magnitude of cataract, but we wanted to use the data to estimate the number of surgical procedures that needed to be done. If the examination teams at some sites were more careful than others to exclude posterior pole disease (by dilating all pupils, for example) the systematic error introduced would affect relative prevalence among the sites. This would be especially problematic if there were more posterior pole disease in one site compared with another. Therefore, caution must be taken in making comparisons between sites.
The need to make assumptions about the visual acuity of eyes preoperatively is also a limitation. The estimates at visual acuity of less than 6/18 are likely to be most accurate because a negligible number of cataracts in Africa are operated on at a visual acuity better than this. Estimates of incidence (and hence CSRs) for less than 6/60 and less than 3/60 visual acuity are too high because an unknown proportion of cases that received surgery would have had visual acuity better than this at the time of operation.
We assumed a mortality rate of 1.5 times in those with cataract compared with those without. This is a question that has received considerable study; although many studies include important methodological flaws, most evidence suggests that there is increased mortality among those who develop cataract (whether operated on or not) compared with those who do not.11 Minassian et al12 assumed a difference of 1.25 times in their model, while others have suggested larger differences.11 Had we assumed mortality to be 1.75 times higher, rather than 1.5 times, it would increase the incidence by around 12%; assuming 1.25 times higher results in a decreased incidence of roughly 12% compared with our calculation.
Despite the limitations, there is new and valuable information derived from this analysis. A strength of the study is the fact that the RAAB surveys were all conducted according to a standard methodology, justifying some comparisons among them. We found in all sites that the incidence of cataract at visual acuity less than 6/18 was just less than twice the incidence at 6/60, which is consistent with other modeling of cataract.13
Our findings are also consistent with those from a small incidence study in Uganda14 that reported an estimated 24 872 new cases of people older than 55 years with visual acuity less than 6/18 each year, of which 41% (10 197) would be due to cataract. Assuming that 6% (1 476 000) of the population of 24.6 million at that time would have been older than 55 years, the incidence of bilateral cataract causing visual acuity of less than 6/18 in those older than 55 years would be 10 197/1 476 000 = 0.007. In comparison, our model predicted the incidence of bilateral cataract at visual acuity of less than 6/18 in the population older than 50 years to be 0.006, 0.007, and 0.008 in sites in neighboring Rwanda, Tanzania, and Kenya, respectively.
For planning purposes, 2 extreme CSR rates can be considered. These are the number of surgical procedures needed to operate on 1 eye of each person with bilateral cataract at visual acuity of 3/60 (corresponding to the number of surgical procedures that would be necessary to eliminate cataract blindness in people) and the number of surgical procedures needed to operate on all eyes with cataract at visual acuity of less than 6/18. These 2 rates define a range within which the CSR would need to fall, depending on demands of the population and on surgical practices. In reality, programs in Africa rarely, if ever, restrict surgery to 1 eye per patient, and targets will probably be set to reach all or most eyes at a particular level of visual acuity, depending on the surgeons' practices. If the quality of surgery is good, we should expect that the case mix will change over time, tending toward demand for more surgery at better preoperative visual acuities. This is already happening in urban settings in Africa.
The age structure of various districts in Africa will make a large difference in the CSR needed. Although census statistics indicate that, in many African countries, around 8% to 10% of the population is older than 50 years,15 in a district of 1 million persons, there may be considerably more variation than this. For example, we have noted that in the Kilimanjaro region of Tanzania, the percentage of people older than 50 years is about 13%, or twice that of Nakuru District in Kenya.4 All other things being equal, this translates into a need for a CSR twice as high.
Although potentially owing to systematic differences in survey techniques, as mentioned above, it is interesting that there is fairly large variation in incidence rates (and hence CSRs) across these 7 sites. The CSRs among the 4 closely situated sites in eastern Africa are only one-third to one-half of the values in the sites in Mali and Eritrea at each visual acuity level. If these differences are real, they could be owing to genetic variation resulting from migrations across the continent, environmental factors including climate and ecology, cultural factors such as diet and lifestyle, or other factors. The magnitude of the variation has a large effect on CSR targets and deserves further study. For planning in districts in which no incidence data are available, data from a neighboring site would probably be the most useful.
According to these estimates, the CSR needed to eliminate blindness (visual acuity <3/60) due to cataract in some sites in Africa are considerably lower than the 2000 often suggested as a target. The current analysis suggests that the possibility of operating on all eyes with less than 6/60 visual acuity or even less than 6/18 is not unrealistic in parts of Africa. If we are to try to prevent blindness from cataract as opposed to curing it, we should prepare to offer surgery to everyone with visually impairing cataract who wants it rather than waiting until they become blind. This means we must put a much stronger emphasis on offering surgery earlier and achieving good results than has been done in the past. We must base our planning for human resource development on the evidence and use effective strategies to enable patients to access surgical services when they feel the need for better visual acuity rather than when they reach some arbitrary threshold determined by the health system.
In summary, we used population-based survey data collected in a standard fashion to calculate age-specific prevalence of vision-reducing cataract, then designed a model to estimate the incidence. Incidence data were used to estimate the CSR needed to eliminate visual impairment at different levels of visual acuity. This methodology could be used for analysis of future RAAB or other population-based surveys to provide better CSR targets for planning.
Correspondence: Susan Lewallen, MD, Kilimanjaro Centre for Community Ophthalmology, Good Samaritan Foundation, Tumaini University, PO Box 2254, Moshi, Tanzania (email@example.com).
Submitted for Publication: November 20, 2009; final revision received March 26, 2010; accepted April 3, 2010.
Financial Disclosure: None reported.
Funding/Support: Partial funding for this study was provided by the Fred Hollows Foundation and the Jonsson Fund.