Retinopathy of prematurity: the influence of gestational age and retinal maturity on the statistical behavior of risk factors.

The clinical and laboratory data on 140 premature infants, 74 cases with retinopathy of prematurity (ROP) and 66 control cases without ROP, were correlated with ROP grades of increasing severity. By using multiple linear regression (MLR), it is shown that for predicting ROP grades the importance of certain variables varies considerably depending on the gestational age. Below 32 weeks of gestation, acidosis, hyperoxemia, gestational age, pathologic paCO2 levels, and multiple birth are strong regressors. Above 31 weeks, the most important regressors are multiple birth and acidosis, while gestational age and duration of FiO2 greater than 0.4 are much less influential. Blood transfusions and artificial ventilation do not seem important in the MLR of either group. When ROP is regressed on the number of paO2 values above 100 torr and gestational age or on the number of paCO2 values above 50 torr and gestational age, the regression coefficients of these variables drop to near zero at a gestational age of about 32 weeks. This loss of weight of the two variables paO2 and paCO2 with increasing gestational age coincides with a comparable drop of the relative incidence of ROP and the relative incidence of immature retinal vessels. It is hypothesized that it is the proportion of infants with an immature retinal vasculature in populations of given gestational ages rather than the gestational age itself which is responsible for the widely varying importance of certain factors during the development of ROP.