Modified WINROP algorithm may increase sensitivity for detecting high-grade ROP

The algorithm can be used to complement the current screening practice, not replace it.

Issue: June 10, 2015

A specific subgroup of neonates may be at risk of being missed by WINROP, Jennifer Jung, MD, said in a presentation at the American Association for Pediatric Ophthalmology and Strabismus annual meeting.

The original WINROP algorithm, an online surveillance system for detecting the likelihood of development of retinopathy of prematurity, considered postnatal measurements and insulin-like growth factor 1 (IGF-1) levels. The algorithm was later simplified to use postnatal weight gain as a surrogate measure for a slower than expected rise in serum IGF-1 because those levels can be difficult to follow, Jung said. Therefore, actual weight gain that is less than expected weight gain triggers an alarm to indicate that the infant is at risk for development of sight-threatening ROP.

Current screening guidelines that were published in 2013 do not take into consideration postnatal factors, Jung said.

Chart review

In a retrospective chart review, Jung and colleagues evaluated 483 preterm infants who received ROP exams at two Colorado hospitals between 2008 and 2011.

Preterm infants’ gestational age, birth weight and weekly weights were entered into the WINROP website until an alarm was signaled or until the time of the first ROP exam. The infants were separated into an alarm group consisting of type 1 or 2 ROP and a no-alarm group with low-grade or no ROP.

ROP developed in 33.7% of infants, with 28 infants having type 1 ROP and 16 having type 2 ROP. All infants with type 1 and three infants with type 2 received laser treatment for ROP.

A WINROP alarm was triggered in 50% of infants, with a sensitivity of 81.8% and specificity of 53.3% for identifying infants with high-grade ROP.

Missed high-grade ROP

The algorithm missed eight infants with high-grade ROP. All were born at 27 weeks’ gestational age or less, with three infants having type 2 ROP and five infants having type 1 ROP. Six of the infants required treatment, while all eight infants had complicated neonatal courses.

The study evaluated these infants more thoroughly through additional weekly weight entries after the first ROP exam to determine whether the alarm status would change.

The alarm status changed in three of the eight infants, all of whom had the lowest weight percentiles among the group. The five infants whose alarm status did not change had weight percentiles of 80% or greater on the Fenton growth chart. Therefore, very preterm infants at high percentiles for weight may not be detected by WINROP for being at risk for development of high-grade ROP and may need special attention, Jung said.

With the additional weekly weight entries, the sensitivity increased to 88.6% and specificity remained at 53.3%.

“The conclusion of our study is that it confirms the findings of previous studies that the WINROP algorithm should complement rather than replace the current screening practice,” Jung said. “Another approach would be to modify the algorithm in consideration of these findings.”

Recommendations

In a discussion of Jung’s presentation, Gil Binenbaum, MD, MSCE, said that the sensitivity of the suggested system should equal the current system of screenings.

He explained that a few different models have been developed with different statistical approaches, and that regardless of which one proves to be the most accurate, he recommended some steps to ensure the sensitivity is high. First, researchers should use a larger data set to narrow the confidence interval around the point estimate of sensitivity, then evaluate the model in new infant groups, and finally combine the development and validation study cohorts into one group, so as to recalibrate the model to fit the even larger data set.

Additionally, Binenbaum recommended a national collaborative postnatal weight gain in ROP registry to best meet these goals in an ongoing fashion.

“And in doing so, hopefully, we would be able to meet the potential that’s been suggested by these studies,” he said. – by Kristie L. Kahl

For more information:

Gil Binenbaum, MD, MSCE, can be reached at Children’s Hospital of Philadelphia, Division of Ophthalmology, 9 MAIN, 34th St. and Civic Center Blvd., Philadelphia, PA 19104; email: binenbaum@email.chop.edu.
Jennifer Jung, MD, can be reached at Children’s Hospital Colorado, 1675 Aurora Court, F731, Aurora CO, 80045; email: jennifer.jung@ucdenver.edu.

Disclosure: Binenbaum reports he is a speaker for the National Institutes of Health, Bayer Pharmaceutical and the Pan American Ophthalmological Foundation. Jung reports no relevant financial disclosures.