R etinopathy of prematurity (ROP) remains the leading cause of preventable childhood blindness in the United States. According to the Early Treatment for Retinopathy of Prematurity (ETROP) study, type 1 ROP is a high-risk form of disease that requires urgent treatment to reduce the risk of retinal detachment and vision loss.

A recent population-based analysis from the Postnatal Growth and Retinopathy of Prematurity (G-ROP 1 and 2) studies of premature infants undergoing ROP screening estimated that approximately 6% develop type 1 ROP. Several pioneering studies have defined the current standards for ROP management, including the Cryotherapy for Retinopathy of Prematurity (CRYO-ROP), ETROP, and Bevacizumab Eliminates the Angiogenic Threat of Retinopathy of Prematurity (BEAT-ROP).

These landmark clinical trials have primarily focused on structural outcomes, disease regression, or safety at predefined endpoints. While the efficacy of treatments such as laser photocoagulation and intravitreal anti-VEGF injections is well established, there remains a need for clinical data that examines the timing and patterns of retreatment following these interventions on a more granular level. The recently published TWO-ROP and FIRST-ROP algorithms ^, have demonstrated the potential to risk-stratify the frequency and timing of ROP screening based on individual gestational age and birthweight. Building on this concept, we hypothesized that these same baseline birth factors could be risk variables for post-treatment outcomes, including retreatments. However, there has been a paucity of data examining post-treatment ROP outcomes as a function of baseline gestational age and birthweight. Therefore, we employed the risk cohort definitions from the TWO-ROP and FIRST-ROP studies for analysis of retreatment outcomes.

In clinical practice, treatment is often performed by specialists at referral centers and identifying lower-risk infants (those meeting screening criteria for FIRST- or TWO-ROP algorithms) who may be safely transferred back to local care could improve efficiency and resource allocation. To our knowledge, no prior studies have examined retreatment patterns across these neonatal risk cohorts nor compared early (≤30 days) versus late (31-90 days) retreatment periods. By analyzing these post-treatment intervals, we aim to identify opportunities to refine follow-up protocols and characterize how the timing and frequency of retreatment vary after two common contemporary treatment modalities (primary bevacizumab versus laser therapy) and across distinct neonatal risk groups.

This investigation was determined to be exempt by the Institutional Review Board of the Boston Children’s Hospital and was conducted in adherence with the Declaration of Helsinki. This is a retrospective consecutive, comparative clinical study of patients who were treated for ROP at three different tertiary-level care neonatal intensive care units (NICUs) in the greater Boston area from January 2017 to August 2023. A single group of pediatric ophthalmologists and pediatric retina specialists performed all screenings, treatments, and encounters. Demographic and clinical variables were collected, including gestational age (GA), birth weight (BW), ROP stage, zone, presence of plus disease, and treatment characteristics. Postmenstrual age was defined as the sum of chronological age and gestational age. Patients were stratified into two treatment groups based on the sequence and type of therapy received: primary laser group vs primary standard-dose bevacizumab group. We defined microprematurity as those infants born at 24 to 26 weeks’ GA and/or with a BW between 600 and 799 grams, and nanoprematurity as those born at <24 weeks’ GA and/or with a BW <600 grams. Infants meeting the FIRST-ROP screening criteria were defined as those not meeting criteria for microprematurity or nanoprematurity (ie, born at ≥27 weeks’ gestational age and ≥800 grams birth weight), whereas the TWO-ROP screening criteria were defined as infants meeting one or none of the current U.S. screening thresholds (GA ≤ 30 weeks or BW ≤ 1500 grams). ^, Cases of treatment-warranted ROP (TW-ROP) were identified based on the ETROP criteria for type 1 ROP, which include: (1) zone I with plus disease, regardless of the ROP stage; (2) zone I, stage 3 without plus disease; and (3) zone II, stage 2 or 3 with plus disease. In addition, high-risk type 2 ROP cases treated at the discretion of the attending ophthalmologist were also considered to have TW-ROP. As per ICROP-3, we defined reactivation as the recurrence of acute-phase features of ROP.

We evaluated retreatment—defined as a second intervention for TW-ROP following initial therapy—within the first 12 weeks after initial treatment (not including persistent avascular retina). Retreatment timing was further categorized as: (1) early retreatment, occurring within 1 month (30 days) of initial therapy, and (2) late retreatment, occurring between one and 3 months after initial therapy. Although there is no clear definition in the literature for what constitutes early versus late retreatment in ROP, the one-month interval was chosen based on adult retinal literature for anti-VEGF therapy, where studies support monthly dosing in conditions, such as neovascular age-related macular degeneration and diabetic macular edema, ^, reflecting a waning treatment effect over approximately 1 month. In addition, the one-month interval was selected based on previous ROP studies using bevacizumab to help differentiate early versus late disease reactivation and potential treatment failure. ^, Lastly, this interval was also chosen in the context of clinical care and referral-based follow-up. This may represent a sufficiently meaningful period to assess early treatment response while allowing potential transition of care back to referring centers when clinically appropriate, balancing monitoring needs with logistical considerations for families.

In this cohort, bevacizumab injections were typically preferred for more severely ill infants or those with posterior disease (zone I or posterior zone II). We excluded patients without sufficient clinical information or follow-up period (<3 months after first therapy) to evaluate early and late retreatment outcomes, those with primary treatment with an anti-VEGFs other than bevacizumab (including ranibizumab and aflibercept, as these agents have different pharmacokinetics) and those with very-low dose bevacizumab (0.016, 0.008, 0.004, and 0.002 mg as part of a clinical trial within our institution; as these doses are not commonly used and although effective, can lead to earlier reactivation. All the standard doses used within our institution (0.125, 0.25, 0.375, and 0.50 mg) were included; cases with unknown dosing were also retained to minimize risk of selection bias.

Descriptive statistics were calculated, including median with interquartile range (IQR) for continuous variables and frequency with percentages for categorical variables. The Wilcoxon rank-sum test was used to compare continuous variables between groups, and Chi-square tests were used to compare proportions. We used mixed effects logistic regression models to evaluate the occurrence, timing, and factors associated with retreatment. The main predictors analyzed were primary treatment group (laser vs bevacizumab), aggressive ROP status (AROP, defined as type 1 ROP in zone 1 or posterior zone 2 requiring treatment before 35 weeks PMA), gestational age, and birth weight. Each model included the primary predictor of interest and was adjusted for gestational age (GA) and aggressive ROP (AROP) status at first treatment, except when GA or birth weight (BW) served as the main predictors; in those cases, models were adjusted for AROP and treatment group only, given the high collinearity between GA and BW. To account for intereye correlation, patient eye was included as a random effect in all models. Survival analysis was conducted to evaluate time to retreatment, and results were visualized using a Kaplan–Meier survival plot (survival package used for time-to-event analysis). Descriptive analysis and mixed models were performed using SPSS 29 Inc and survival analysis using R, version 4.4.1 (R Core Team, 2024). Statistical significance was set at a value of P <.05 on two-sided tests.