Authors: Malhotra S, Prasad M, Vashist P, Kalaivani M, Gupta SK.

Geographical coverage: India

Sector: Burden of disease

Sub-sector: Epidemiology

Equity focus: Older adults

Study population: Adults aged 50+

Review type: Other review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis method: Not applicable

Background: The National Programme for Control of Blindness and Visual Impairment in India modified the definition of blindness in 2017 in line with the internationally accepted definition. Now, the definition of blindness is visual acuity 3/60 in the better eye with available correction 3/60 in the better eye. From the late 1970s till early 2017, the definition of blindness was visual acuity 6/60 in the better eye (or with available correction).

Objectives: To compute pooled estimates of blindness in India among adults aged 50 years and above by using recent and old definitions of blindness.

Main findings:

Authors found there is a decrease in the prevalence of blindness in India using recent and previous definitions and a declining trend over time.

A total of 19 publications were included in this review, encompassing 211,502 participants. One of the published studies reported two estimates which were for two different periods, and four published reports were also included in the review. For the recent definition of blindness, 16 papers and reports provided relevant data and contributed to the pooled estimate. For the previous definition of blindness, the corresponding number was 18. There was complete inter-observer agreement for the full-text screening.

Most of the studies were conducted in southern India, particularly in rural areas. The mean age of the study participants ranged from 53.8 to 70 years. The response rate was satisfactory in all studies that reported it.

All but one study clearly mentioned a sampling methodology, keeping them at low risk of selection bias. All but one study calculated a minimum sample size a priori. Apart from four studies, all reported a satisfactory response rate ranging from 80.2% to 96.3%. CIs for estimates were mentioned by 15 of 17 published studies. Only one published report among the four mentioned the CI along with the point estimate.

The pooled prevalence (95% confidence interval [CI]) obtained for recent and previous definitions of blindness in India was found to be 6.11% (5.07%-7.14%) and 9.91% (8.57%-11.25%), respectively. The stratified pooled prevalence (95% CI) from rapid surveys was 4.81% (3.26%-6.35%) and 4.68% (2.91%-6.46%) for studies published during 1995-2005 and 2006-17, respectively, using the new definition. The corresponding figures for comprehensive surveys were 9.22% (95% CI 6.48%-11.96%) for the period 1995-2005 and 3.81% (95% CI 2.76%-4.84%) for the period 2006-17.

Authors found high level of heterogeneity in the meta-analysis (I²=98.8%). On exclusion of the two studies with the outlying estimates from the analysis, the I² statistic did not change (I² 97.9). Heterogeneity remained on stratification by year of publication and type of survey. Further, meta-regression analysis was done to identify the sources of between-study heterogeneity in the pooled prevalence estimates considering the sample size, rural location, state of study and year of publication, and none of the variables were statistically significant.

Authors indicated that the limitations of their study include a high level of observed heterogeneity, which could not be explained by either exclusion of studies with outlying estimates or by meta-regression of independent variables on the univariable model.

This study should be followed by an assessment of cause-specific prevalence of blindness and vision impairment, to enable policy-makers to guide allocation of resources and plan health services. There is a need for standardisation of the definition of blindness and vision impairment and more nationally representative surveys than have been done so far for comparison with international studies and worldwide reporting.

Methodology:

The study included population-based surveys conducted in India from 1990 to 2017, reporting prevalence estimates for individuals aged 50 and above, with blindness measured according to established definitions. Exclusions were made for editorials, letters, news, reviews, expert opinions, case reports, studies without original data, studies on specific disease groups (e.g., leprosy or diabetes), and those reporting blindness due to specific causes. In cases of data duplication, the most recent publication was used. The year 1990 was chosen as the starting point due to its significance in policy changes.

PubMed, Web of Science, Google Scholar and TRIP (Turning Research into Practice) databases were searched up to December 2017 to identify all relevant studies. Reference lists of retrieved articles and pertinent reviews were also searched for relevant articles. No language restrictions were imposed.

Titles and abstracts were screened in duplicate by two reviewers independently, and full-texts of articles that either reviewer considered potentially eligible were obtained. Similarly, data was abstracted by two reviewers independently and risk of bias was assessed using the CASP (Critical Appraisal Skills Programme) checklist.

A study used the random-effects model in Stata 12.0 to calculate pooled prevalence estimates for blindness, considering studies published from 1990-2017. The estimates were stratified by survey type and publication year. Forest plots displayed prevalence with a 95% confidence interval. The I² statistic quantified heterogeneity, and potential reporting bias and small study effects were detected using funnel plots and the Egger method. Sensitivity analysis investigated the effect of two studies with large prevalence. Meta-regression was performed to investigate heterogeneity using variables like rural/urban residence, sample size, study location, and publication year.

Applicability/external validity: Authors identify a high degree of heterogeneity in results of different studies, which they did not manage to resolve through subgroup analysis of excluding certain studies. This may have implications for the applicability of their estimate for India as a whole.

Geographic focus: The authors emphasize the importance of comparing the prevalence of blindness in India with other Southeast Asian countries and low-to-middle-income countries. However, they acknowledge that such comparisons may be challenging due to variations in the age groups sampled. Despite this, they highlight that recent findings from Pakistan, Nepal, Malaysia, and Oman show similarities to the results obtained in this review.

Summary of quality assessment:

The approaches used to identify, include and critically appraise studies were generally sound, with two authors undertaking all key tasks and a search that was inclusive in terms of time period and language. However, there is no evidence of attempts being made to include unpublished material. While the approach to the analysis of the data, by meta-analysis, was also robust, the authors did not attempt to analyse the data separately based on the quality ratings assigned to different studies. We also note the high level of heterogeneity observed, despite a very thorough analysis plan to detect the reasons for this. For these reasons, we attributed medium confidence in the findings of this study.