In 2001, a meta analysis revealed that women constituted almost two-thirds of the world’s blind.1 The major factor in the poorest countries, where cataract is the main cause of blindness, is that women do not receive cataract surgery at the same rates as men. This was further demonstrated in a report of studies from low-income countries published up to 2000.2

The cataract surgical coverage (CSC) indicates the number of people (or eyes) that have received cataract surgery divided by the number that “require” surgery. Ideally, it would be near 100%. The CSC is a useful indicator of service provision, but often a difficult one to produce, as it requires population-based data. CSC may be calculated at different visual acuity levels. In this paper and our previous work,2 we focused on the cataract surgical coverage for people with blindness or severe visual impairment (<6/60 presenting visual acuity in the best eye). CSC estimates in industrialised countries, in contrast, would focus on coverage in people with much less visual impairment.

Since the studies analysed in the previous report were completed, the VISION 2020 initiative has been launched, with emphasis on increasing the numbers of people receiving cataract surgery, particularly in low- and middle-income countries. We searched the published literature for reports of CSC published since 2000 with the goal of comparing rates in men and women; we report the findings here.

METHODS

In September 2008 we conducted an electronic search of three databases (Medline, Embase, Cochrane (OVID—Comprising Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and Database of Abstracts of Reviews of Effectiveness)) using specific terms related to cataract surgery limited by date (2000 to 2008) and terms for developing countries. A keyword search of the defined impact factor cataract surgical coverage (CSC) was run in the same three databases. Additionally, we did a general search for cataract surgical coverage through Google. The authors searched their personal libraries and professional networks for additional articles. Two reviewers (SL and PC) reviewed the titles or abstracts from all of the above and selected reports to retrieve in full and review. From this group, reports were excluded if: (1) they were not population-based, (2) they included <1000 adults or (3) they did not present cataract surgical coverage by sex (at any visual acuity level) or the raw data necessary to calculate this. Authors of studies were contacted when possible to request additional data.

Most reports do not present the raw data needed in a meta-analysis to generate a Peto odds ratio with confidence intervals. Thus, we used several methods to combine the data.

Method 1: calculation of the weighted mean of the odds ratios

For each study the respective CSC for men and women was used to calculate an OR using the formula OR = CSC_m×(100−CSC_f)/((100−CSC_m) × CSC_f). Then, the average (mean) for these ORs was calculated, weighted by overall study size.

Method 2: calculation of a Peto OR with 95% confidence intervals and relative risk

For this we needed to know the number of operated and unoperated cataract patients of each sex. If CSC with 95% CI was presented, we worked backward from the lower CI to estimate the denominator (N), which equals the total number of operated people plus the total number unoperated at the appropriate visual acuity level. N = 1.962×P(1−P)/(P−lower CI), where P = CSC/100. Then, with N and the CSC, we calculated the number of operated people using the formula CSC = operated people/N and calculated the number unoperated (N-operated).

For studies without 95% CI, but providing the number of people or eyes operated and unoperated, we used the latter, and also calculated the 95% CI for the CSC. After testing for heterogeneity, we entered the number of operated and unoperated for each study by sex into NCSS 2007 (Number Cruncher Statistical System, Kaysville, Utah) to generate an overall Peto odds ratio with confidence intervals.

Method 3: calculation of the risk difference (RD) for men and women

In each study, we calculated the RD. These were combined with NCSS 2007, weighted by N (number of operated plus unoperated people, as calculated above) to produce an overall RD. We also generated an overall relative risk and used this to estimate the population attributable risk percentage (PAR) using the formula PAR = b(relative risk−1)/(b (relative risk−1)+1)×100%3, where b = proportion of adult population that is female, estimated at 55%.

RESULTS

The initial search found 164 unique publications, of which 354^–38 were selected after a review of titles and abstracts. Among these, the following were rejected: seven due to absence of sex disaggregated data,6 24^–29 two from special leprosy villages,22 23 one due to size,33 two which were not population-based,32 39 two in Chinese,31 36 two37 38 because the data had been included in a separate report already included9 and one with internal data inconsistencies.30 Additional sex data supplied by the authors for one study later allowed this to be included.6 In addition, three studies known to the authors from the grey literature were included.10 40 41 Thus, there were 22 reports included in the analysis. These are reported as 23 separate surveys, shown in table 1.

The CSC rates (with 95% CI when possible) and the risk differences in table 1 show that, in all but two4 12 surveys, the CSC was higher among men than women.

Method 1 produced the ORs shown in table 1. Weighting the ORs by the size of the survey gives an overall weighted mean OR of 1.57.

Method 2 was used with 18 surveys (those with CIs). A test for heterogeneity of ORs produced Cochran Q = 18.95 (p = 0.33), so we used a random effects model. The overall Peto OR was 1.71 (95% CI 1.48 to 1.97). One study accounted for 24% of the analysis,9 but removing this did not significantly change the results. Figure 1 shows a Forest plot of the ORs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Forest plot of odds ratios.

Method 3 was used with 18 surveys. Heterogeneity testing of the RDs produced Cochran Q = 21.31, p = 0.213, and a random effects model produced an overall risk difference 0.116 (95% CI 0.083 to 0.150). The population attributable risk percentage was 10.8%.

DISCUSSION

VISION 2020 programmes in the low- and middle-income countries focus on reducing the number of people blind or visually impaired from cataract. Although previous research showed the need to increase utilisation of cataract services by women, only a few programmes report efforts to meet women’s needs,42 43 and some blindness surveys still do not report important indicators by sex.

CSC is one of the most meaningful measures of progress in cataract service provision. Statistically, the rates for cataract surgical coverage are significantly higher (at the α = 0.05 level) in men compared with women in only two of the surveys listed in table 1.4 12 Many of the surveys analysed here used the methodologies of the Rapid Assessment of Avoidable Blindness (RAAB) or Rapid Assessment of Cataract Surgical Services (RACSS). These are epidemiologically sound and invaluable for planning; however, they are intentionally designed to provide specific information as efficiently as possible and are generally not large enough to demonstrate statistically significant differences in CSC between men and women. This does not mean that meaningful differences do not exist, as this meta-analysis demonstrates. The pattern for rates to be higher in males than females is clear, and this has implications on a national and global level.

The simplest measure of attributable risk is the risk difference, while the population attributable risk percentage is a more refined attempt to extrapolate to larger populations, taking into account the proportion of individuals in the population with the risk factor (being female in this case). The overall RD of 0.116 and the PAR of 10.8% may both be interpreted as indicating that blindness and severe visual impairment from cataract could be reduced by around 11% in the low- and middle-income countries if women were to receive cataract surgery at the same rate as men. Interestingly, this is close to the estimated 12% excess cataract blindness due to gender inequity that we calculated in the earlier study, before the launch of the VISION 2020 initiative.

There are potential biases to this analysis. The surveys included in this systematic review may or may not represent all of the low- and middle-income countries, although, to our knowledge, there is no rationale for study-site selection based upon perceived gender inequity. Ideally, we would have had raw data from each study to use, rather than working backward from CSC, some of which were age-adjusted and others not. Finally, although we used CSC at 6/60 by person for most studies, this was not available for all.

One cannot directly compare the CSC between studies reported at different visual acuity levels, or when eyes are used rather than people, nor ORs derived from these CSC rates. However, within any study, the same definitions are used for male and female, so the risk difference may be a better way to combine the data than using the ORs. In fact, all methods used here lead to a similar conclusion: there is still significant imbalance in the CSC for men and women. Additional focus by all partners engaged in provision of eye services is needed to bring cataract surgical services to women in low- and middle-income countries.