UPenn  Vision Research Center Biostatistics Example Projects

Development of ROP Prediction Model:
Retinopathy of Prematurity (ROP) is a leading cause blindness in children. Risk of ROP is assessed primarily by birth weight (BW) and gestational age (GA). The preterm babies with BW <1501 grams or GA<32 weeks undergo physically stressful, resourceintensive, serial diagnostic eye exams, but these exams have very low predictive value. However, slower than expected postnatal increase weight has been found to be associated with higher risk of severe ROP (Figure 1). Working with vision core investigators Drs. Gil Binebaum and Graham Quinn, we performed the secondary analysis of data from a clinical trial “The Premature Infants in Need of Transfusion Study”. From this data analysis, we developed the following ROP prediction model that used the postnatal weight growth, BW and GA.
Probability of severe ROP = 1/(1 + exp (−risk score)), where risk score = (−1.66) + (3.24 if GA = 22–24) + (2.25 if GA = 25) + (1.99 if GA = 26) + (0.78 if GA = 27) + (1.48 if GA = 28) + [(−0.0021) * (BW)] + [(−0.0139) * (weight gain rate)].
The above prediction equation was represented graphically through a nomogram (Figure 2). Our ROP prediction model could have reduced the need for examinations by 30%, while still identifying all infants requiring laser surgery. This work was published in Pediatrics 2011; 127(3): e607–e614.
We recently applied this prediction model to a cohort more representative of current screening guidelines, and updated the prediction model. The work will soon be published in Archives Ophthalmology.
Built on these pilot works, we developed a grant proposal “Postnatal Growth and Retinopathy of Prematurity (GROP) Studies”, to develop and validate ROP prediction model, and evaluate its relative costeffectiveness using postnatal weight gain to identify infants who are likely to develop severe ROP.
Figure 1: Mean weight gain rate (grams per day) from previous week and 95% CI for infants who did and did not develop severe ROP.
Figure 2: Sample nomogram to determine risk of ROP in infants with various GAs at birth. A straight line is drawn between the values for BW and daily weight gain rate. The intersection of this line with the gray auxiliary axis is then connected to the value for GA. The intersection of this second line with the probability line provides the predicted probability of severe ROP. If the risk is >0.085, eye examinations are indicated.