Abstract
Purpose :
Prediction of the likelihood of MR for any given age/RE combination is challenging because of the shortage of population based studies and, within those, low sample sizes in the age/RE cells. Here, we combine a previous analysis of MR across RE and a large sample of high myopes categorized according to age and myopic maculopathy (MM) to make predictions for risk of MR.
Methods :
A relationship for odds ratios (OR) of developing MR by RE was obtained from a previous analysis using 3 population-based (PB) studies from different countries (Brennan, ARVO 2014). Data for the relation between MR and age were drawn from Shih et al (BJO 2006), which provided a table of MM gradings (according to Avila et al Ophthalmol, 1984) versus age for 552 highly myopic (RE ≤ -6.00D). A best fit sigmoid function relating age to prevalence of MM, defined as M3 or greater, was calculated from this data. The OR for age and RE were then multiplied together to obtain estimates as a function of age and RE. Various assumptions are inherent in the model; in particular, it is assumed that age and RE operate independently on the risk of developing MR, that OR remain constant across ethnicity, geography and classification, and that OR with age for MM also apply to MR.
Results :
The following equation for OR of developing MR according to age and RE resulted:
OR = exp ((ln(age/60.45)/0.134) + RE*0.63)
A constant k can then be applied to adjust predicted relative risk to match limited data sets. For example, using a value of 1.32E-03 for k allows fitting of the data to one of the 3 PB studies linking MR to RE (see figure- broken line represents extrapolation beyond data). The model predicts the following risk for a set of REs at age 65 (versus estimates from Vongphanit et al, Ophthalmol, 2002): at -2.00D; 0.8% (0.7%): -4.00D; 2.7% (3.0%): -6.00D; 9.0% (11.4%): -8.00D; 25.9% (28.6%): -10.00D; 55.3% (52,4%).
Conclusions :
This model represents an initial attempt to provide risk estimates of developing MR for given age-RE combinations. As standardization of diagnosis takes place and more data are collected, the value of k can be refined and assumptions inherent in the model can be tested.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.