Identification of mechanical ventilation-related risk factors for retinopathy of prematurity in preterm infants

Introduction

Background

Retinopathy of prematurity ROP is a proliferative vascular disease of the immature retina and one of the leading preventable causes of childhood blindness worldwide (1,2). The pooled global prevalence of ROP is approximately 31.9% (highest in lower-middle-income countries), with severe ROP occurring in 7.5% of screened preterm infants (highest in high-income countries) (1). This translates to an estimated 32,300 preterm infants annually who develop irreversible vision impairment, with nearly 20,000 cases resulting in severe visual loss or blindness (3).

Rationale and knowledge gap

Among all clinical variables examined in ROP research, gestational age (GA) and oxygen therapy have consistently emerged as the two dominant risk factors, outperforming other predictors in multivariate models across diverse study populations (4-7). Other important risk factors include inflammatory morbidities such as sepsis and NEC, as well as blood transfusions (8-10). This dominance creates a fundamental methodological challenge: virtually all other clinical variables—including surfactant administration, patent ductus arteriosus (PDA) treatment, and intraventricular hemorrhage (IVH)—tend to lose statistical significance once GA and oxygen exposure are controlled for. As a result, identifying independent risk factors beyond these two predictors has proven exceptionally difficult, leaving a substantial gap in actionable knowledge for ROP risk stratification.

Critically, prior studies have almost universally treated oxygen therapy as a single, aggregate exposure variable—operationalized as total duration of supplemental oxygen use, fraction of inspired oxygen (FiO₂) levels at a single time point, or simply the requirement for any form of respiratory support (4-6,11). This approach, while methodologically straightforward, conflates fundamentally different modalities that vary substantially in their physiological effects on oxygenation dynamics. Invasive mechanical ventilation, high-frequency oscillatory ventilation (HFOV), nasal continuous positive airway pressure (NCPAP), and high-flow nasal cannula (HFNC) each deliver oxygen and airway pressure through distinct mechanisms and may therefore exert different influences on retinal oxygenation and vascular development.

In the neonatal intensive care unit (NICU), various non-invasive respiratory support strategies are employed in preterm infants, including NCPAP, HFNC, and non-invasive neurally adjusted ventilatory assist (NIV-NAVA) (12,13). Non-invasive support is preferred over invasive ventilation due to well-established benefits in reducing the risk of bronchopulmonary dysplasia (BPD) and mortality (14). However, a key physiological distinction exists between these two approaches: during invasive ventilation, tidal volumes are tightly controlled and relatively constant, whereas during NCPAP, infants breathe spontaneously against a continuous positive pressure, resulting in inherently greater moment-to-moment variability in tidal volume, respiratory effort, and effective oxygenation—driven primarily by apneic events and loss of functional residual capacity (FRC) rather than changes in delivered FiO₂.

Despite the pathophysiological plausibility that NCPAP-associated oxygenation variability may influence ROP risk, whether the duration of specific non-invasive support modalities is independently associated with ROP after controlling for GA has not been directly investigated. This study, therefore, aimed to investigate whether the mode and duration of respiratory support—beyond total oxygen exposure—are associated with ROP risk in extremely preterm infants.

Objective

This study aimed to investigate the association between ROP and specific modes and parameters of mechanical ventilation in preterm infants with a GA of less than 28 weeks. Specifically, we sought to determine whether individual respiratory support modalities (conventional ventilation, HFOV, NCPAP, HFNC, NIV-NAVA) and time-weighted average (TWA) ventilator settings [FiO₂, mean airway pressure (MAP), positive end-expiratory pressure (PEEP), and peak inspiratory pressure (PIP)] at defined postnatal time points are independently associated with ROP after adjustment for GA. By moving beyond the aggregate concept of “oxygen therapy” and disaggregating respiratory support into modality-specific durations examined in a GA-adjusted framework, this study aimed to provide more granular evidence to guide respiratory management strategies in the NICU to prevent ROP. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0278/rc).

Methods

Study population

Medical records were reviewed for 83 newborns who underwent ROP screening at the Jeonbuk National University Hospital (JBUH) NICU between January 2014 and December 2024. Inclusion criteria required a GA of less than 28 weeks; exclusion criteria included delivery at an external hospital (n=6), ocular hemorrhage (n=1), and congenital ocular conditions such as Goldenhar syndrome (n=1). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This single-center, retrospective cohort study received formal approval from the Institutional Review Board of JBUH, which waived the requirement for patient consent (No. CUH 2025-04-038).

The NICU at JBUH follows a protocol prioritizing non-invasive respiratory support. For preterm infants born before 25 weeks’ gestation, invasive mechanical ventilation is initiated first, with early extubation determined by respiratory pattern stability. For infants born between 25 and 28 weeks’ gestation, non-invasive ventilation (NIV) [non-invasive positive pressure ventilation (NIPPV), NIV-NAVA, NCPAP, or HFNC] is the primary mode when spontaneous breathing is confirmed. The selection of non-invasive modality was based on clinical assessment of each infant’s respiratory effort, stability, and response to support.

Data collection

Demographic and clinical data were extracted from the electronic medical record (EMR), specifically from NICU discharge summaries and vital signs sheets. Variables included sex, GA, birth weight (BW), delivery mode, plurality, and 5-minute Apgar scores. Clinical complications recorded included the highest ROP stage, number of surfactant doses, PDA treatment, respiratory distress syndrome (RDS) status, vasopressor use within the first week of life, massive pulmonary hemorrhage, air leak syndrome, necrotizing enterocolitis (NEC), spontaneous bowel perforation, and culture-proven sepsis. The duration of each respiratory support modality during hospitalization was also extracted, including conventional ventilation, HFOV, NIV-NAVA, NCPAP, HFNC, and supplemental oxygen (Table 1).

Definition of ROP

ROP was defined based on formal ophthalmic diagnostic records (15). Preterm infants were classified into two groups—ROP and non-ROP—to analyze clinical predictors associated with disease occurrence. Of the 83 infants, 52 (62.7%) were diagnosed with ROP: stage 1 in 7 (8.4%), stage 2 in 20 (24.1%), and stage 3 in 25 (30.1%). No cases of stage 4 or stage 5 ROP were observed in this cohort.

TWA mechanical ventilation variables

Respiratory therapy-related variables were extracted from biological signal records. Respiratory status was assessed at postnatal days 1, 3, 7, and 10—time points encompassing the early hyperoxic phase (Phase 1) of ROP pathogenesis, during which vascular arrest is primarily driven by high oxygen concentrations—including FiO₂, MAP, PIP, and PEEP. Outlier handling excluded FiO₂ values below 21%, MAP values below 4, PIP values exceeding 50, and PEEP values exceeding 12. Missing values were imputed using linear interpolation to preserve the continuous nature of vital sign data (16). As of 06:00, the period from midnight to 05:59 was defined as the previous medical day.

Missing values were classified into three types: complete missing (no measurements during the entire monitoring period), complete measurement (continuously measured), and partial missing (intermittent). For missing values at the start of a medical day, forward imputation was applied; for missing values at the end, backward imputation was applied. When MAP, PIP, and PEEP values were not recorded—reflecting periods when mechanical ventilation was not being applied—values were set to 0, as airway pressure is absent during non-ventilated periods. When FiO₂ was not recorded, a value of 21% was assigned to represent spontaneous breathing of room air without supplemental oxygen. The TWA FiO₂ was calculated by dividing the sum of each FiO₂ value multiplied by its corresponding time interval by the total measurement time (16). TWA MAP was computed analogously. The respiratory severity score (RSS) was calculated as (TWA FiO₂/100) × TWA MAP, and the oxygenation saturation index (OSI) was calculated as (TWA FiO₂ × TWA MAP × 100)/TWA SpO₂ (17,18). FiO₂ values, recorded as percentages, were divided by 100 to convert to fractional form prior to RSS and OSI calculation. SpO₂ was also processed as a TWA using the same method applied to FiO₂ and MAP (Table 2). The proportion of infants with periods of no recorded values was 0% for SpO₂, 8.4% for PEEP and PIP, 15.7% for FiO₂, and 28.9% for MAP. These periods reflect clinical states in which the respective support was not being administered, rather than true missing data, and were handled using the imputation strategy described above.

Statistical analysis

To compare baseline characteristics between the ROP and non-ROP groups, the Mann-Whitney U test or Student’s t-test was used for continuous variables, as appropriate, and the Chi-squared test for categorical variables. Normal distribution was assessed using the Shapiro-Wilk test. Results are presented as median (interquartile range) or mean ± standard deviation (SD) for continuous variables, as appropriate, and as number (%) for categorical variables.

Logistic regression analyses were performed to evaluate associations between respiratory-related variables and ROP occurrence. Given the limited sample size and the potential for small-sample bias, Firth’s penalized logistic regression was used for all regression analyses to obtain bias-reduced estimates. Unadjusted associations were first assessed using univariate Firth logistic regression models. GA-adjusted analyses were subsequently performed by including each respiratory variable and GA simultaneously in separate Firth logistic regression models, with GA as the sole covariate in all adjusted models.

Results are reported with odds ratios (ORs), 95% confidence intervals (CIs), and exact P values in accordance with Statistical Analyses and Methods in the Published Literature (SAMPL) guidelines. Data preprocessing was conducted in Python (version 3.12), and statistical analyses were performed in R (version 4.5) using the ‘logistf’ package. All tests were two-sided; P<0.05 was considered statistically significant.

Given the retrospective and exploratory nature of this study, a formal a priori sample size calculation was not performed. The analyses should therefore be considered hypothesis-generating, and the observed effect sizes will inform power calculations for future prospective validation studies.

The per-day OR assumes a linear relationship between NCPAP duration and the log-odds of ROP across the observed distribution; to address this limitation, a sensitivity analysis was performed using tertile-based categorization of NCPAP duration (0–2.5, 2.5–23.0, and >23.0 days). Categorization using arbitrary cut-off values may introduce information loss; therefore, the continuous per-day OR was retained as the primary analytical approach, with the categorical analysis presented in Tables S1,S2.

In accordance with Dammann et al. (19), we recognize that P values alone do not establish risk factors; accordingly, effect size estimation and clinical contextualization are emphasized alongside statistical significance throughout this manuscript.

To minimize selection bias, all consecutive eligible infants admitted during the study period were included without further selection. Information bias was mitigated by extracting respiratory variables from continuous EMR vital sign records rather than manual chart review. Residual confounding due to unmeasured variables remains an inherent limitation of the retrospective design.

Results

Demographic and clinical characteristics

All infants were followed throughout NICU hospitalization until discharge or death. The median length of hospital stay was 93 (82.5–128) days in the ROP group and 72 (64–94.5) days in the non-ROP group. A total of 83 infants were included in the analysis, of whom 52 (62.7%) developed ROP (Figure 1). Infants in the ROP group had significantly lower GA [25 (24.0–26.2) vs. 26 (26.0–27.0) weeks, P=0.003] and BW (826±170 vs. 953±220 g, P=0.008) compared with the non-ROP group. Multiple gestation was more frequent in the ROP group (P=0.045). No significant differences were observed in sex distribution, delivery mode, 5-minute Apgar score, surfactant administration, PDA treatment, or other neonatal morbidities between the two groups. There were no in-hospital deaths among the 83 enrolled infants during the study period; all infants survived to discharge (Table 3).

Figure 1 Study flow diagram. Of 91 preterm infants with gestational age <28 weeks admitted to the NICU at Jeonbuk National University Hospital between January 2014 and December 2024 who underwent ROP screening, 52 (62.7%) were diagnosed with ROP. Exclusion criteria included delivery at an external hospital (n=6), ocular hemorrhage (n=1), and congenital ocular disease (n=1). GA, gestational age; NICU, neonatal intensive care unit; ROP, retinopathy of prematurity.

Mechanical ventilation factors

There was no statistically significant difference in the duration of conventional mechanical ventilation between the ROP and non-ROP groups [14.5 (2.0–38.8) vs. 3.0 (1.0–23.0) days, P=0.18]. Similarly, the duration of HFOV use did not differ between groups [0.0 (0.0–8.8) vs. 0.0 (0.0–4.0) days, P=0.65].

Among non-invasive respiratory therapies, NCPAP duration was longer in the ROP group than in the non-ROP group [14.5 (2.8–25.2) vs. 8.0 (2.5–13.5) days, P=0.08]. In contrast, HFNC duration and NIV-NAVA duration did not differ significantly between groups. Supplemental oxygen duration was also comparable. Regarding respiratory severity indicators, no statistically significant differences were observed between groups on day 1 for any parameter. Weighted PEEP was significantly higher in the non-ROP group on days 7 and 10 [day 7: 3.9 (3.0–5.0) vs. 3.1 (1.5–3.9), P=0.01; day 10: 3.7 (2.4–4.7) vs. 3.2 (1.3–3.8), P=0.04]. Weighted OSI values did not differ significantly between the ROP and non-ROP groups at any timepoint (days 1, 3, 7, and 10; all P>0.05). All other respiratory indicators showed no statistically significant between-group differences at any time point (Table 4).

Risk factors for ROP

Unadjusted logistic regression analysis identified GA as a significant protective factor (OR =0.54; 95% CI: 0.35–0.80; P<0.001) and NCPAP duration as a significant risk factor (OR =1.04; 95% CI: 1.01–1.09; P=0.01) for ROP occurrence. Among time-weighted ventilator settings, weighted PEEP at days 7 and 10 was significantly associated with reduced ROP risk (day 7: OR =0.73; 95% CI: 0.54–0.94; P=0.02; day 10: OR =0.76; 95% CI: 0.57–0.98; P=0.03). No other respiratory therapy duration or daily respiratory indicator showed a statistically significant association with ROP.

After adjustment for GA, only NCPAP duration remained a statistically significant, independently associated variable for ROP (OR =1.04; 95% CI: 1.00–1.08; P=0.04). All other variables, including weighted PEEP at days 7 and 10, lost statistical significance after GA adjustment (Table 5).

In a supplementary analysis additionally adjusted for sepsis, NCPAP duration remained a statistically significant, independently associated variable (OR =1.03; 95% CI: 1.00-1.08; P=0.04), supporting the robustness of the primary finding. As a secondary analysis, we repeated the logistic regression using treatment-requiring ROP (TR-ROP; stage 3, n=25) as the outcome, with non-TR-ROP infants (stage 0–2, n=58) as the reference group. NCPAP duration remained the sole independently associated variable for TR-ROP after GA adjustment (OR =1.03; 95% CI: 1.00–1.07; P=0.03), consistent with the primary analysis. None of the other variables were significantly associated with TR-ROP (Tables S3,S4).

Discussion

Key findings

This study used TWA analyses of continuous EMR records to examine how specific ventilation modes affect ROP risk in infants born before 28 weeks. Our principal finding is striking: even after strictly accounting for GA, NCPAP duration emerged as the sole independently associated variable with ROP (OR =1.04; 95% CI: 1.00–1.08; P=0.04). In practical terms, every additional day an infant remains on NCPAP translates to a 4% rise in the odds of developing ROP.

To contextualize this effect size, an infant requiring 30 days of NCPAP would carry an estimated OR approximately 2.19 times higher (1.04²⁰) than one requiring 10 days. Given that the ROP group in our cohort had a median NCPAP duration of 14.5 days, this represents a clinically substantial doubling of risk within a very realistic timeframe for this GA group.

Strengths and limitations

To our knowledge, this study is among the first to disaggregate mechanical ventilation into modality-specific duration variables, apply TWA analyses from continuous EMR records, and demonstrate that NCPAP duration—but not invasive ventilation, HFOV, HFNC, or NIV-NAVA—independently predicts ROP after GA adjustment. The dissociation between NCPAP duration and TWA FiO₂ as predictors suggests that oxygenation variability during spontaneous breathing, rather than oxygen exposure per se, is the more proximal mechanism. Firth’s penalized logistic regression was employed to reduce small-sample bias.

This study has several limitations. First, as a single-center retrospective study with 83 infants, generalizability requires multicenter validation. Second, the primary outcome was any ROP occurrence rather than TR-ROP, which is a clinically more meaningful endpoint. Third, only GA was included as a covariate; other confounders (sepsis, blood transfusions, caffeine, steroids) were not simultaneously adjusted due to sample size constraints. Fourth, oxygenation variability was not directly measured; NCPAP duration is proposed as a surrogate, but this mechanistic link requires direct testing. Fifth, physiologically referenced default values (0 for pressure parameters, 21% for FiO₂) were assigned during unmonitored intervals, which may not capture all clinical scenarios. Sixth, the analysis was restricted to the first 10 postnatal days, capturing primarily Phase 1 of ROP pathogenesis; Phase 2 respiratory parameters were not assessed. Seventh, modalities were analyzed as independent exposures without accounting for their sequence or combination in clinical practice, which may introduce residual confounding. Additionally, data on blood transfusions and NEC were unavailable for regression analyses, and the small annual enrollment (7–8 infants/year) raises the possibility of department-specific ventilation practice effects. Respiratory parameters were assessed at only four predefined time points rather than continuously, and subtle temporal changes in clinical practice over the 11-year period cannot be fully excluded despite consistent SpO₂ targets (90–95%).

Comparison with similar research

Prior studies on ROP risk factors have consistently identified GA, BW, oxygen therapy duration, and mechanical ventilation use as major associated factors, but have treated these as aggregate exposures without disaggregating by modality (4-6). de las Rivas Ramírez et al. [2022] identified the need for and duration of oxygen therapy (particularly at 28 days of age), mechanical ventilation duration, and NIV requirement as independent risk factors for ROP development in multivariate analysis (4)—notably reporting the need for NIV as a categorical yes/no variable rather than examining modality-specific duration as a continuous predictor. Tapak et al. [2024] identified extensive oxygen therapy as a time-dependent risk factor influencing ROP onset and progression (6), further supporting the temporal dimension of oxygen exposure. Chaves-Samaniego et al. [2020] found that each additional day of mechanical ventilation was associated with an 8.1% increase in the risk of TR-ROP (11), which is quantitatively comparable to our finding of a 4% increase per additional day of NCPAP.

Explanations of findings

As in prior studies (4-6), GA was the strongest predictor of ROP in unadjusted analyses (OR =0.54; P<0.001), and other variables typically lost significance once it was controlled for. That NCPAP duration remained an independent predictor after adjustment for this dominant covariate is noteworthy, and the non-significant unadjusted comparison (P=0.08) versus the significant GA-adjusted result (OR =1.04; P=0.01) is consistent with a confounding suppression effect whereby GA masks the true NCPAP-ROP association in crude analyses. That NCPAP duration—but not mean FiO₂—was independently associated with ROP suggests the mechanism is not explained by cumulative oxygen delivery alone; although the adjusted CI (1.00–1.08) is marginal, the consistency of point estimates and biological plausibility of the NCPAP-oxygenation variability-ROP pathway support this as a signal warranting prospective validation.

This is mechanistically consistent with Lin et al. (20) and Kubota et al. (21), who demonstrated that FiO₂ coefficient of variation and SpO₂ fluctuation independently predicted severe ROP and TR-ROP, respectively. During NCPAP, spontaneous breathing produces greater SpO₂ oscillation than controlled invasive ventilation; amplified by apnea of prematurity, these fluctuations drive intermittent hypoxemia (IH) and rebound VEGF overexpression, promoting pathological neovascularization in Phase 2 of ROP (22-24). The absence of independent associations for HFNC and NIV-NAVA further supports oxygenation stability as the relevant variable. We acknowledge, however, that NCPAP duration may partly represent a surrogate marker of disease severity rather than a direct causal factor.

Implications and actions needed

Translating these findings into clinical action requires a balanced perspective. It would be premature to suggest shortening NCPAP duration solely to lower ROP risk; NCPAP remains a cornerstone of neonatal respiratory care with well-documented benefits for BPD prevention and survival (12-14), and weaning decisions must continue to be guided strictly by respiratory stability.

The primary clinical value lies in risk stratification. NCPAP duration is an objective, universally accessible EMR-derived metric: among infants with comparable GA, those with longer NCPAP courses represent a high-risk subgroup that warrants earlier or more frequent ophthalmological surveillance, particularly where specialized eye care is limited.

These findings also raise a secondary hypothesis worth prospective investigation: whether minimizing oxygenation fluctuations during NCPAP—via automated FiO₂ controllers, optimized alarm settings, SpO₂ histogram monitoring, or aggressive caffeine therapy to reduce apnea-related desaturations (22)—may reduce ROP risk beyond what mean FiO₂ targeting can achieve, given evidence that IH patterns rather than mean SpO₂ are primary drivers of ROP severity (23). Future prospective trials should consider nonlinear modeling approaches such as restricted cubic splines to more fully characterize the dose-response relationship between NCPAP duration and ROP risk.

Conclusions

NCPAP duration is independently associated with ROP in preterm infants with GA less than 28 weeks, with each additional day associated with a 4% increase in risk after GA adjustment. This association was independent of mean FiO₂, suggesting that oxygenation variability during spontaneous breathing—rather than oxygen quantity alone—may be a more proximate pathophysiological mediator; however, as this analysis was restricted to the first 10 postnatal days and SpO₂ variability was not directly measured, this interpretation applies primarily to Phase 1 of ROP pathogenesis and should be regarded as hypothesis-generating (20,23). NCPAP duration, as an objective EMR-derived metric, may serve as a practical tool for ROP risk stratification in the NICU. Prospective multicenter studies with direct measurement of oxygenation variability are needed to validate these findings.

Acknowledgments

The authors thank the NICU medical staff at Jeonbuk National University Hospital for their dedication to patient care.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0278/rc

Data Sharing Statement: Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0278/dss

Peer Review File: Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0278/prf

Funding: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government, Ministry of Science and ICT (MSIT) (No. RS-2025-00553891); and the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (No. RS-2025-02313278).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0278/coif). H.H.K. reports the funding from the National Research Foundation of Korea (NRF) grant funded by the Korean government, Ministry of Science and ICT (MSIT) (No. RS-2025-00553891); and the Korean Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (No. RS-2025-02313278). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Jeonbuk National University Hospital (No. CUH 2025-04-038). Individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. García H, Villasis-Keever MA, Zavala-Vargas G, et al. Global Prevalence and Severity of Retinopathy of Prematurity over the Last Four Decades (1985-2021): A Systematic Review and Meta-Analysis. Arch Med Res 2024;55:102967. [Crossref] [PubMed]
2. Hong EH, Shin YU, Cho H. Retinopathy of prematurity: a review of epidemiology and current treatment strategies. Clin Exp Pediatr 2022;65:115-26. [Crossref] [PubMed]
3. Blencowe H, Lawn JE, Vazquez T, et al. Preterm-associated visual impairment and estimates of retinopathy of prematurity at regional and global levels for 2010. Pediatr Res 2013;74:35-49. [Crossref] [PubMed]
4. de Las Rivas Ramírez N, Luque Aranda G, Rius Díaz F, et al. Risk factors associated with Retinopathy of Prematurity development and progression. Sci Rep 2022;12:21977. [Crossref] [PubMed]
5. Sankar BK, Amin H, Pappa P, et al. Risk Factors of Retinopathy of Prematurity: A Prospective Study. Indian J Public Health 2025;69:111-4. [Crossref] [PubMed]
6. Tapak L, Farahani LN, Taleghani NT, et al. Risk factors for the time to development of retinopathy of prematurity in premature infants in Iran: a machine learning approach. BMC Ophthalmol 2024;24:364. [Crossref] [PubMed]
7. Kim SJ, Port AD, Swan R, et al. Retinopathy of prematurity: a review of risk factors and their clinical significance. Surv Ophthalmol 2018;63:618-37. [Crossref] [PubMed]
8. Glaser K, Härtel C, Klingenberg C, et al. Neonatal Sepsis Episodes and Retinopathy of Prematurity in Very Preterm Infants. JAMA Netw Open 2024;7:e2423933. [Crossref] [PubMed]
9. Fundora JB, Binenbaum G, Tomlinson L, et al. Association of Surgical Necrotizing Enterocolitis and Its Timing with Retinopathy of Prematurity. Am J Perinatol 2023;40:1178-84. [Crossref] [PubMed]
10. Uberos J, Fernandez-Marin E, Campos-Martínez A, et al. Blood products transfusion and retinopathy of prematurity: A cohort study. Acta Ophthalmol 2023;101:e294-301. [Crossref] [PubMed]
11. Chaves-Samaniego MJ, García Castejón M, Chaves-Samaniego MC, et al. Risk Calculator for Retinopathy of Prematurity Requiring Treatment. Front Pediatr 2020;8:529639. [Crossref] [PubMed]
12. Lavizzari A, Zannin E, Klotz D, et al. State of the art on neonatal noninvasive respiratory support: How physiological and technological principles explain the clinical outcomes. Pediatr Pulmonol 2023;58:2442-55. [Crossref] [PubMed]
13. Shi Y, Muniraman H, Biniwale M, et al. A Review on Non-invasive Respiratory Support for Management of Respiratory Distress in Extremely Preterm Infants. Front Pediatr 2020;8:270. [Crossref] [PubMed]
14. Isayama T, Iwami H, McDonald S, et al. Association of Noninvasive Ventilation Strategies With Mortality and Bronchopulmonary Dysplasia Among Preterm Infants: A Systematic Review and Meta-analysis. JAMA 2016;316:611-24. [Crossref] [PubMed]
15. Chiang MF, Quinn GE, Fielder AR, et al. International classification of retinopathy of prematurity. Ophthalmology 2021;128:e51-68. [Crossref] [PubMed]
16. van Rossum MC, da Silva PMA, Wang Y, et al. Missing data imputation techniques for wireless continuous vital signs monitoring. J Clin Monit Comput 2023;37:1387-400. [Crossref] [PubMed]
17. Thurston SW, Harrington D, Mruzek DW, et al. Development of a long-term time-weighted exposure metric that accounts for missing data in the Seychelles Child Development Study. Neurotoxicology 2022;92:49-60. [Crossref] [PubMed]
18. Jung YH, Jang J, Kim HS, et al. Respiratory severity score as a predictive factor for severe bronchopulmonary dysplasia or death in extremely preterm infants. BMC Pediatr 2019;19:121. [Crossref] [PubMed]
19. Dammann O, Chui K, Stansfield BK. Retinopathy of prematurity and beyond: P values don’t make risk factors. Pediatr Res 2025;99:432-4.
20. Lin WC, Jordan BK, Scottoline B, et al. Oxygenation Fluctuations Associated with Severe Retinopathy of Prematurity: Insights from a Multimodal Deep Learning Approach. Ophthalmol Sci 2024;4:100417. [Crossref] [PubMed]
21. Kubota H, Fukushima Y, Kawasaki R, et al. Continuous oxygen saturation and risk of retinopathy of prematurity in a Japanese cohort. Br J Ophthalmol 2024;108:1275-80. [Crossref] [PubMed]
22. Thompson L, Werthammer JW, Gozal D. Apnea of Prematurity and Oxidative Stress: Potential Implications. Antioxidants (Basel) 2024;13:1304. [Crossref] [PubMed]
23. Rabienia Haratbar S, Chen L, Cheng Q, et al. The impact of intermittent hypoxemia on type 1 retinopathy of prematurity in preterm infants. Pediatr Res 2024;96:766-72. [Crossref] [PubMed]
24. Hartnett ME. Studies on the pathogenesis of avascular retina and neovascularization into the vitreous in peripheral severe retinopathy of prematurity (an american ophthalmological society thesis). Trans Am Ophthalmol Soc 2010;108:96-119.