MANUSCRIPT CITATION

Ying GS, Quinn GE, Wade KC, et al. Predictors for the Development of Referral-Warranted Retinopathy of Prematurity in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy (e-ROP) Study. JAMA Ophthalmology 2015; 133(3):304-311. PMID: 25521746

REVIEWED BY

Meera Ramakrishnan, BS
Perelman School of Medicine at the University of Pennsylvania
meerar@upenn.edu

John Flibotte, MD
Children’s Hospital of Philadelphia & Perelman School of Medicine at the University of Pennsylvania
flibottej@email.chop.edu

TYPE OF INVESTIGATION

Prognosis
Clinical prediction guide

QUESTION

In infants with gestational age 23-34 weeks and birth weight ≤1250g who receive serial exams for ROP, what demographic, medical, and ocular factors predict progression to referral-warranted ROP (RW-ROP)?

METHODS: 

• Design:  Secondary analysis of the e-ROP study (NCT01264276), a multicenter observational cohort study
• Allocation: Does not apply
• Blinding: Does not apply
• Follow-up period: Investigators performed standardized eye examinations on infants every other week until one of the following was noted: 1) mature retinal vessels; 2) immature zone II on 2 occasions at least 7 days apart; 3) ROP regressed or regressing on 2 visits at least 7 days apart; 4) treatment for severe ROP; or 5) infant reached 40 weeks PMA with no ROP or only stage 1 or 2 ROP.
• Setting: 12 large children’s centers in the US and one in Canada, between 2011 and 2014
• Patients: Cohort of the e-ROP study (NCT01264276)
  • Inclusion criteria: Premature infants with birth weights <1250g, admitted to NICU and expected to survive 28 days, had at least 2 serial ROP examinations.
  • Exclusion criteria: Known ocular abnormalities that prevent imaging of retina, RW-ROP identified at the first study-related eye examination, ROP that is already regressing or being treated.
• Intervention:  This is a non-interventional observational cohort study. However, in the context of the original e-ROP study, infants received standardized ophthalmologic examinations by study-certified ophthalmologists. Examinations that were not performed by these investigators were not included in the data set. Because the outcome of interest in the primary study was RW-ROP, the examiners also reported the presence or absence of this at the conclusion of the exam.
• Outcomes:
  • Primary outcome: The primary outcome of the e-ROP study was the sensitivity and specificity of detecting RW-ROP in either eye by trained non-physician readers evaluating digital retinal images. The gold standard for comparison was the diagnosis provided by an ophthalmologist who is experienced in ROP evaluation after direct retinal exam.
• Exposures: Investigators pre-defined the following predictors of RW-ROP:
  • Demographic: birth weight, gestational age, sex, race, ethnicity, singleton/ multiple birth
  • Medical: weight gain rate, relative weight gain rate, respiratory support
  • Ocular: characteristics of posterior pole vessels, quadrants with plus/ pre-plus, most dominant vascular feature, presence, stage and zone of ROP, and retinal hemorrhage
• Analysis and Sample Size: E-ROP investigators calculated that a sample size of 250 infants with RW-ROP was required to achieve the primary outcome of determining sensitivity and specificity for detecting RW-ROP. Assuming the RW-ROP rate of 14%, a total enrollment of 2000 infants was calculated in order to identify 250 patients with RW-ROP. For the present study, a subset of infants was chosen based on the criteria reviewed above. Of the 1284 infants enrolled in the e-ROP study, 979 infants (76.2%) had at least 2 study-related eye examinations prior to the development of RW-ROP and were included in this study. Of these infants, 149 (15.2%) developed RW-ROP. Univariate analyses were initially applied to select those predictors with p<0.10 for inclusion in subsequent multivariate analyses. Multivariate modeling was performed with backward selection of predictors that associated with RW-ROP with a p<0.05. Odds ratios were reported based on multivariate modeling and receiver operator characteristic curves (ROC) were generated and areas under the curve (AUCs) were calculated for several combinations of predictors to evaluate diagnostic accuracy. Finally, sensitivity and specificity of several cut points for the predicted probability of RW-ROP were calculated.

MAIN RESULTS:

In the univariate analysis, low birth weight (p <0.001), low gestation age (p<0.001), male sex (p=0.03), and non-black race (p = 0.02) were demographic predictors of RW-ROP. Ocular findings in the first study-related eye examination associated with RW-ROP were presence of preplus disease (p<0.001), number of quadrants with preplus disease (p<0.001), zone I incomplete vascularization (p=0.04), stage 1 or 2 ROP (p<0.001), zone II ROP (p<0.001) and the presence of retinal hemorrhage (p< 0.001). The need for respiratory support (p<0.001) or the absence of enteral feedings (p=0.004) preceding the first study-related eye examination was associated with high risk of RW-ROP, as well as a slower weight gain of <12g/d when compared to infants with weight gain of >18g/d. Relative weight gain was not associated with the risk of RW-ROP.

In the multivariate analysis, demographic and clinical factors at the first study-related eye examination that were independently associated with an increased risk of RW-ROP included male sex, non-black race, low BW (<500g), earlier GA (≤24 weeks), more quadrants with preplus disease, stage 2 ROP, presence of retinal hemorrhage, the need for controlled mechanical ventilator or high-frequency oscillatory ventilation, and weight gain <12g/d. The adjusted odds ratios and 95% confidence intervals for each of these factors are provided in Table 1.

Table 1:

When these significant predictive factors are used to create a multivariate prediction model, the AUC was 0.88 (95%CI 0.85-0.91), which is significantly better than the prediction by demographic characteristics alone (p<0.001) and also better than the prediction by demographics + ocular findings (p=0.005). When the predicted probability of 0.05 or greater was considered high risk, the sensitivity was 96% (95%CI 91.5-98.1) and specificity was 52.7% (95%CI, 49.2-56.0).

CONCLUSION:

This study’s prediction model could be used to identify lower- and higher-risk infants, and thus can be used to tailor how frequently an infant requires repeated eye examinations. This stratified approach based on an infant’s risk of developing RW-ROP can be used to reduce the burden of ROP clinical examinations while still capturing the high-risk infants to direct them towards timely treatment.

Visit Acta to access a pdf copy of this EBNEO commentary!

COMMENTARY:

ROP remains the most common cause of preventable childhood blindness in the world.¹ Currently, detection of ROP that requires treatment is achieved through serial retinal exams by skilled ophthalmologists, with 4.2% of infants needing treatment in a recently published cohort evaluating ROP screening guidelines in Sweden.² In a higher risk US population, up to 14% of infants developed severe ROP requiring treatment.³ Current AAP Guidelines recommend screening for ROP based on birth weight (BW) <1500 grams and gestational age (GA) less then 30 weeks with some allowance for including infants of older GA and greater BW based on clinical course.⁴ This approach, and recommendations in other countries that are similar, results in frequent ophthalmologic exams that are costly⁵ and lead to increased discomfort for patients in the NICU.

Ying et al. add to this body of literature in two unique ways: 1) their inclusion of ophthalmologic characteristics in determining risk profile; and, following from that, 2) their choice of referral-warranted ROP (RW-ROP) as their outcome. With this approach, the authors do not obviate initial ophthalmologic screening but may reduce repeat screening when considering the results of early exams. An additional strength of this study is the large number of infants and multi-center involvement.

This study has important limitations to immediate clinical application and the authors identify all of these. First, the study population was a higher risk subset of premature infants, with birth weight less than 1251 grams. The population typically screened per AAP guidelines (<1500 grams, <30 weeks)⁴ would presumably have a lower prevalence of disease; therefore, the positive predictive value of the proposed model would likely suffer and have a higher rate of screen positive infants who do not go on to have RW-ROP. Some infants in the original cohort were not included, as they had only one ROP examination. This may serve to reduce the severity of ROP, as some with only one examination had RW-ROP on initial examination. Finally, the initial examinations were not uniformly performed at the time recommended by AAP guidelines due to need to transfer infants into an e-ROP center. However, the authors point out that there were no major differences in findings between late and early exams in those infants who had 3 evaluations.

Overall, this is an interesting study that adds to the ongoing debate about the best approach to screening for ROP in preterm infants. The authors do not validate the predictive ability of this model in a separate or subset population and this will be required before clinical application. It will be helpful to quantify the reduction of repeated screening examinations in at-risk populations to determine the true impact of this model.

REFERENCES

1. Shah PK, Prabhu V, Karandikar SS, Ranjan R, Narendran V, Kalpana N. Retinopathy of prematurity: Past, present and future. World journal of clinical pediatrics 2016;5:35-46.
2. Holmstrom G, Hellstrom A, Jakobsson P, Lundgren P, Tornqvist K, Wallin A. Evaluation of new guidelines for ROP screening in Sweden using SWEDROP – a national quality register. Acta ophthalmologica 2015;93:265-8.
3. Kennedy KA, Wrage LA, Higgins RD, et al. Evaluating retinopathy of prematurity screening guidelines for 24- to 27-week gestational age infants. Journal of perinatology : official journal of the California Perinatal Association 2014;34:311-8.
4. Fierson WM. Screening examination of premature infants for retinopathy of prematurity. Pediatrics 2013;131:189-95.
5. van den Akker-van Marle ME, van Sorge AJ, Schalij-Delfos NE. Cost and effects of risk factor guided screening strategies for retinopathy of prematurity for different treatment strategies. Acta ophthalmologica 2015;93:706-12.
6. Hutchinson AK, Melia M, Yang MB, VanderVeen DK, Wilson LB, Lambert SR. Clinical Models and Algorithms for the Prediction of Retinopathy of Prematurity: A Report by the American Academy of Ophthalmology. Ophthalmology 2016;123:804-16.