Sir,

We read with interest the article by You et al describing the prevalence of epiretinal membranes (ERM) in the Beijing Eye Study (BES).1 While varying prevalence rates of ERM have been reported in different ethnic groups, with rates from 6 to 11.8% in whites, 18.5% in Hispanics, and 4% in Japanese,2, 3 the BES reported the lowest prevalence rate in the literature (2.2%). Interestingly, the prevalence estimates were equal for both cellophane macular reflex (1.8%) and premacular fibrosis (1.8%). Previous studies, however, show that cellophane macular reflex, an earlier form of ERM, is invariably more common than premacular fibrosis, a later stage of ERM. Additionally, the data presented for the associations are inadequate for readers to examine possible reasons for the lower prevalence. For example, there are no point estimates (i.e., odds ratios) accompanying the P-values and 95% confidence intervals. It is also unclear whether the associations were adjusted for age. All reported associations therefore can only be interpreted as unadjusted, which are not helpful in understanding risk factors associated with age-related conditions.

Nonetheless, we offer several possibilities for why their findings contrast with existing epidemiological data. First, grading of ERM was based on non-stereoscopic retinal photographs in the BES, while most other studies used stereoscopic retinal photographs. The use of non-stereoscopic photographs may miss subtle retinal abnormalities, such as early age-related macular degeneration4 and ERM.2 The BES has also previously reported low rates of age-related macular degeneration in their sample.5 Second, while the authors described that ERM assessment was performed by a trained grader, actual reliability of the grading process was not defined. Some information regarding intra- and inter-grader variability would be helpful. Third, nuclear cataract was highly prevalent (82%) in the BES. ERM rates may be lower in eyes with nuclear cataract owing to increased difficulty in detection of this lesion.4 However, the authors claim that 98.6% of the sample had gradable retinal photographs. All these factors are potential sources of ascertainment errors that could lead to underestimation of ERM in the BES. Further studies are needed to provide clear understanding of possible racial/ethnic differences in the epidemiology of ERM.