Risk factors and prevalence of retinopathy of prematurity in very low birth weight preterm infants: a systematic review and meta-analysis

Abstract

Background

Retinopathy of prematurity (ROP) is a potentially blinding disorder affecting preterm infants, especially those with very low birth weight (<1,500 g). While advances in neonatal care have improved survival, the incidence of ROP remains high, and risk factors vary across studies. This meta-analysis aims to determine the prevalence of ROP in very low birth weight preterm infants and identify key risk factors to inform clinical screening and preventive strategies.

Methods

PubMed, Embase, Cochrane Library, and Web of Science databases were searched from database creation to January 7, 2026. Stata 15.0 was used to merge and analyze the data.

Results

Fourteen articles (including three case-control studies and 11 cohort studies) involving 12,666 newborns were finally included, and the results of meta-analysis found that blood transfusion [odds ratio (OR) =1.81, 95% confidence interval (CI): 1.37–2.39, P=0.001], low gestational age (OR =1.50, 95% CI: 1.21–1.86, P=0.001), low birth weight (OR =1.05, 95% CI: 1.02–1.08, P=0.02), sepsis (OR =1.47, 95% CI: 1.26–1.71, P=0.02), intraventricular hemorrhage (OR =1.87, 95% CI: 1.27–2.75, P=0.001), lung disease (OR =2.08, 95% CI: 1.37–3.17, P=0.001), necrotizing enterocolitis (OR =1.68, 95% CI: 1.16–2.41, P=0.001), and patent ductus arteriosus (OR =1.72, 95% CI: 1.31–2.25, P=0.003) were risk factors for ROP in very low birth weight preterm infants.

Conclusions

The study suggests ROP should be prevented early in very low birth weight preterm infants with a combination of blood transfusions, low gestational age, low gestational weight, sepsis, intraventricular hemorrhage, lung disease, necrotizing enterocolitis, and patent ductus arteriosus.

Highlight box.

Key findings

• This meta-analysis identified key risk factors for retinopathy of prematurity (ROP) in very low birth weight preterm infants, including blood transfusion, low gestational age, low birth weight, sepsis, intraventricular hemorrhage, lung disease, necrotizing enterocolitis, and patent ductus arteriosus.

What is known and what is new?

• ROP is a leading cause of blindness in preterm infants, particularly those with very low birth weight. Previous studies have suggested various risk factors, but no consensus on their prevalence and clinical significance has emerged.

• This study synthesizes data from 14 studies with 12,666 infants, confirming the association of key clinical factors with ROP, which could aid in refining screening protocols and prevention strategies.

What is the implication, and what should change now?

• Preventive strategies for ROP should prioritize infants with identified risk factors, particularly those with very low birth weight. By addressing these factors early, clinicians can potentially reduce the incidence of ROP and mitigate its long-term impact on vision in preterm infants.

Introduction

Retinopathy of prematurity (ROP), which primarily affects preterm and low-birth-weight infants, is characterized by the atypical development of retinal blood vessels and remains the leading cause of preventable blindness in children globally (1,2). With the advancement of perinatal medicine and the rise of neonatal intensive care units, the survival rate of preterm infants has increased significantly, which has led to the high prevalence of ROP (3,4). A 2025 meta-analysis shows regional variations, with higher rates in low-resource settings. While increasing survival rates contribute to more ROP cases, recent trends indicate stabilization in high-income countries, largely due to improved oxygen management (5). Although most ROP cases are mild or can regress spontaneously, severe forms may result in unilateral or bilateral blindness (6). The delayed detection and untimely intervention lead to irreversible ocular impairment and severe visual loss with the utmost impact on child development (7-9). ROP rates are higher in infants with very low birth weight, defined as those with a weight of less than 1,500 g at birth (10,11). Currently, recognized risk factors include gestational age, birth weight, and oxygen use, mainly due to limited retinal vascularization in preterm infants. Very low birth weight infants have immature retinal development and are susceptible to hyperoxia, an important driving factor in the process of vascular cell growth and arrest in stage 1. Even ordinary indoor air has a higher oxygen concentration than the uterine environment. More critically, supplemental oxygen administered to premature infants for respiratory distress may lead to abnormally high oxygen saturation. Excessive oxygen suppresses oxygen-regulated vascular growth factors, particularly erythropoietin and vascular endothelial growth factor (VEGF). This suppression ultimately halts retinal vascularization and may result in the loss of some existing retinal blood vessels (12). However, the exact risk factors for ROP development in very low birth weight infants remain controversial (13,14). The study aims to conduct a meta-analysis of the prevalence and risk factors of ROP in very low birth weight preterm infants, to improve the prognosis and quality of life of very low birth weight preterm infants. We present this article in accordance with the PRISMA reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-aw-785/rc).

Methods

The registration number was CRD42024519457.

Literature retrieval

PubMed, Embase, the Cochrane Library, and Web of Science databases were searched. The search cut-off date was set to January 7, 2026, without any language restrictions. The search approach utilized both subject headings and free text terms, concentrating on essential topics: ROP, infants with very low birth weight, and associated risk factors. For an in-depth description of the search methods, please see Table S1.

Inclusion and exclusion criteria

The criteria for inclusion in this study comprised preterm infants weighing less than 1,500 g at birth. The primary focus was on ROP, with the main outcome measured being the risk factors identified through multivariable logistic regression analysis. The secondary outcome addressed the prevalence of ROP. The research methodology incorporated either case-control or cohort studies.

Exclusion criteria included: conference abstracts, meta-analyses, research protocols, letters, duplicate publications, systematic reviews, studies lacking full text, and animal experiments. This rigorous screening process aims to ensure the validity and reliability of the study results, focusing on studies that are comprehensive and applicable to humans.

Data extraction

The literature review was diligently conducted by two independent evaluators. Initially, they screened titles and abstracts to identify relevant studies and then examined full texts to refine their selections according to predefined inclusion and exclusion criteria. When disagreements arose, they consulted with relevant experts to achieve a consensus. During the data extraction phase, the evaluators carefully gathered key indicators from the studies. Examples include author, year, country, study design, sample size, number of ROP cases, sex distribution, gestational age, birth weight, and statistical methods used. To ensure the reliability of their findings, all extracted data were cross-checked for consistency.

In this study, the Newcastle-Ottawa Scale (NOS) (15) was used to assess the quality of the literature reviewed. This scale assesses case-control or cohort studies based on three main criteria: the selection of the study population (up to 4 points), the comparability of the groups (2 points), and the accuracy of the exposure or outcome measurements (3 points), with a maximum score of 9 points. Studies scoring 4 points or less were categorized as low quality, those scoring between 5 and 6 points were classified as medium quality, and studies with scores of 7 or higher were considered high quality. Disagreements in quality assessment between researchers were resolved through discussion or by consulting a third party if necessary.

Statistical analysis

Odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) were pooled for each included study using Stata 15 software. First, we extracted the OR and its 95% CI for each study. To combine these OR values, a random effects model was used for aggregation, which was able to account for heterogeneity across studies and variability in effect sizes. OR and 95% CI were calculated for each study and combined into overall effect sizes. Model heterogeneity was assessed by the I² statistic. If the I² value is greater than 50%, high heterogeneity is considered, and the source of heterogeneity needs to be further explored. For high heterogeneity, we may perform sensitivity analyses to identify potential factors that may affect the pooled effect size. In addition, funnel plots and Egger’s test were used to assess the possibility of publication bias. If bias is present, it may affect the interpretation of the results. Finally, pooled effect values are reported as ORs with 95% CIs to facilitate interpretation of the results and statistical inference.

Results

Literature screening

As shown in Figure 1, we commenced with an initial tally of 1,188 articles, from which 426 duplicates were expunged. Further scrutiny involving the review of titles and abstracts resulted in the elimination of 742 articles [including irrelevance to the research question (n=300); duplicates not identified earlier (n=100); focus on a different population or condition (n=300); insufficient outcome data (n=42)]. Subsequent detailed reading of full texts led to the removal of an additional 6 articles, culminating in the inclusion of 14 articles in our study.

Figure 1.

Flowchart of literature retrieval.

Fourteen articles (16-29) comprised three case-control studies (16,23,26) and 11 cohort studies (17-22,24,25,27-29) involving a collective total of 12,666 newborns. The gestational age of these newborns at birth varied from 27 to 36 weeks, this study comprised two articles reporting ROP at stage 3 or above, five articles reporting ROP at stages 1 to 5, one article reporting ROP at stages 2 to 4, one article reporting ROP at stages 1 to 2, one article reporting ROP at stages 1 to 4, one article reporting ROP at stage 3, and one article reporting ROP at stages 1 to 3. The basic characteristics of these included articles are summarized in Table 1. Notably, each of these 14 articles achieved a score greater than 7 on the NOS, denoting them as high-quality sources. Details of their scoring can be found in (Tables S2,S3), providing a transparent view of the rigorous standards maintained throughout our research.

Table 1. Basic feature table.

Study	Year	Country	Study design	Stage of ROP	Sample size	Number of ROP	Gender (male/female)	Gestational age (weeks)	Birth weight (g)	Statistical methods
Celebi (26)	2014	Turkey	Case-control	≥Stage 3	235	173	106/129	27.08	858.31	Multivariate logistic regression
Darlow (16)	2005	Australia	Case-control	≥Stage 3	2,105	203	1,034/1,116	27	930	Multivariate logistic regression
Fortes (17)	2013	Brazil	Cohort study	Stage 1–5	157	58	92/65	28.3	844.04	Multivariate logistic regression
Fortes (18)	2009	Brazil	Cohort study	Stage 1–5	88	43	53/35	28.5	825.34	Multivariate logistic regression
Freitas (27)	2018	Brazil	Cohort study	Stage 1–3	520	196	300/220	<30	≤1,500	Multivariate logistic regression
Holmström (19)	1998	Sweden	Cohort study	Stage 3	202	81	96/106	28.9	≤1,500	Multivariate logistic regression
Huang (20)	2020	China	Cohort study	Not report	107	23	55/52	<32	≤1,500	Multivariate logistic regression
Hwang (21)	2015	Korea	Cohort study	Stage 1–4	2,009	686	NA	<32	≤1,500	Multivariate logistic regression
Mishra (28)	2019	India	Cohort study	Stage 1–5	361	43	NA	27.2	1,115	Multivariate logistic regression
Opara (22)	2020	USA	Cohort study	Stage 1–2	883	374	NA	28.65	≤1,500	Multivariate logistic regression
Sathar (23)	2018	India	Case-control	Not report	812	203	NA	30.59	1,280	Multivariate logistic regression
Shemesh (29)	2024	Israel	Cohort study	Stage 2–4	4,092	851	NA	≤29	≤1,500	Multivariate logistic regression
Wu (24)	2018	China	Cohort study	Stage 1–5	504	131	261/243	30.6	1,251.7	Multivariate logistic regression
Yang (25)	2011	China	Cohort study	Stage 1–5	216	99	NA	≤33	≤1,500	Multivariate logistic regression

NA, not applicable; ROP, retinopathy of prematurity.

Results of meta-analysis

Prevalence of ROP in very low birth weight preterm infants

This study included 14 publications on the prevalence of ROP in extremely low birth weight preterm infants. Heterogeneity testing (I²=99.3%, P=0.001) indicated substantial heterogeneity, prompting the use of a random-effects model for data analysis. Results (see Figure S1) indicate that the prevalence of ROP in extremely low birth weight preterm infants was 32% (95% CI: 24–40%). Subgroup analyses were also conducted based on ROP severity. Findings (Figure S2) revealed that the prevalence of severe ROP was 41%, with a 95% CI ranging from 1% to 83%; and the incidence of any ROP was 30% (95% CI: 12–47%). Due to high heterogeneity, sensitivity analyses were conducted using a “one-by-one exclusion method”. The results (see Figure S3) showed low sensitivity, indicating robust findings. Publication bias was assessed using the Egger test, which revealed no significant bias (P=0.12).

Blood transfusion

Six studies discussed blood transfusion, and the results of heterogeneity test (I²=15.6%, P=0.31) showed little heterogeneity, so the data were analyzed by random effects model. The analysis results (see Figure 2) showed that blood transfusion was a risk factor for ROP in very low birth weight preterm infants (OR =1.81, 95% CI: 1.37–2.39, P=0.001).

Figure 2.

Forest plot of blood transfusion meta-analysis. CI, confidence interval; OR, odds ratio.

Low gestational age

Eleven literatures mentioned low gestational weeks. The heterogeneity test results (I²=95.7%, P=0.001) indicated significant heterogeneity. A random-effects model was used for data analysis. The analysis results (see Figure 3) show that low gestational age is a risk factor for ROP in very low birth weight preterm infants (OR =1.67, 95% CI: 1.27–2.20, P=0.001). Due to the large heterogeneity, a sensitivity analysis was conducted using the one-by-one elimination method. The results (see Figure S4) indicated that the sensitivity of this indicator was relatively low and the analysis results were relatively stable.

Figure 3.

Forest plot of gestational age meta-analysis. CI, confidence interval; OR, odds ratio.

Low gestational weight

Twelve studies discussed low birth weight. The heterogeneity test results (I²=99.8%, P=0.001) indicated significant heterogeneity, and a random-effects model was used for data analysis. The analysis results (see Figure 4) indicate that low birth weight is a risk factor for ROP in very low birth weight preterm infants (OR =1.05, 95% CI: 1.02–1.08, P=0.02). Due to the high heterogeneity, a one-to-one elimination sensitivity analysis was conducted. The results (see Figure S5) indicated that the sensitivity of this indicator was relatively low and the analysis results were relatively stable.

Figure 4.

Forest plot of gestational weight at birth meta-analysis. CI, confidence interval; OR, odds ratio.

Sepsis

Sepsis was mentioned in six studies. The results of heterogeneity test (I²=12.9%, P=0.33) showed little heterogeneity, and the data were analyzed by random effects model. The analysis results (see Figure 5) showed that sepsis was a risk factor for ROP in very low birth weight preterm infants (OR =1.50, 95% CI: 1.21–1.86, P=0.02).

Figure 5.

Forest plot of sepsis meta-analysis. CI, confidence interval; OR, odds ratio.

Intraventricular hemorrhage

Intraventricular hemorrhage was mentioned in three studies, and the results of heterogeneity test (I²=0%, P=0.82) showed no heterogeneity, and the data were analyzed by random effects model. The results (see Figure 6) showed that intraventricular hemorrhage was a risk factor for ROP in very low birth weight preterm infants (OR =1.87, 95% CI: 1.27–2.75, P=0.002).

Figure 6.

Forest plot of intraventricular hemorrhage meta-analysis. CI, confidence interval; OR, odds ratio.

Pulmonary diseases

Eight articles mentioned lung diseases, and the results of heterogeneity test (I²=67.6%, P=0.003) showed moderate heterogeneity, and the data were analyzed by random effects model. The analysis results (see Figure 7) showed that lung disease was a risk factor for ROP in very low birth weight preterm infants (OR =2.08, 95% CI: 1.37–3.17, P=0.001). Because of the large heterogeneity, a sensitivity analysis with one-by-one exclusion was performed, and the results (Figure S6) showed that the sensitivity of this measure was low and the results were stable.

Figure 7.

Forest plot of pulmonary diseases meta-analysis. CI, confidence interval; OR, odds ratio.

Necrotizing enterocolitis

Necrotizing enterocolitis was mentioned in three studies. The results of heterogeneity test (I²=0%, P=0.77) showed no heterogeneity, and the data were analyzed by random effects model. The analysis results (see Figure 8) showed that necrotizing enterocolitis was a risk factor for ROP in very low birth weight preterm infants (OR =1.68, 95% CI: 1.16–2.41, P=0.001).

Figure 8.

Forest plot of necrotizing enterocolitis meta-analysis. CI, confidence interval; OR, odds ratio.

Patent ductus arteriosus

Patent ductus arteriosus was mentioned in three studies, and the results of heterogeneity test (I²=0%, P=0.82) showed no heterogeneity. The random effects model was used for data analysis. The analysis results (see Figure 9) showed that patent ductus arteriosus was a risk factor for ROP in very low birth weight preterm infants (OR =1.72, 95% CI: 1.31–2.25, P=0.003).

Figure 9.

Forest plot of patent ductus arteriosus meta-analysis. CI, confidence interval; OR, odds ratio.

Publication bias

The publication bias was evaluated using Funnel plot (Figures S7-S15) and Egger’s test for blood transfusion (Egger P=0.18), low gestational age (Egger P=0.59), low gestational weight (Egger P=0.82), sepsis (Egger P=0.09), and pulmonary disease (Egger P=0.22), suggesting that there was no publication bias.

Meta-regression

This study employed meta-regression to investigate sources of heterogeneity for prevalence, low gestational age, low gestational weight, and pulmonary diseases. Results (see Table S4) suggest that study design and gestational age may be sources of heterogeneity for prevalence, while stage of ROP and gestational age may be sources of heterogeneity for low gestational age.

Discussion

This study investigated the prevalence and risk factors of ROP in very low birth weight preterm infants and found that blood transfusion, low gestational age, low birth weight, sepsis, intraventricular hemorrhage, pulmonary disease, necrotizing enterocolitis and patent ductus arteriosus were risk factors for ROP.

Preterm infants, particularly those with very low birth weight, are more prone to anemia compared to those with higher gestational age and/or birth weight. Due to their minimal weight and low blood volume, these infants often require blood transfusions. Research by Prasad et al. (30) suggests that anemia could be an independent risk factor for the development of ROP. Similarly, El Emrani et al. (31) found that the frequency of blood transfusions within the first 30 days was linked to an increased risk of in very low birth weight infants. By around 6 months of gestational age, there is a significant proliferation of the retinal vasculature, extending to the oral serrata by 8 months, and by full term, the fundus retina is fully developed (32,33). The lesser the gestational age and the lower the birth weight, the less mature the temporal retinal capillary development, leading to more severe atrophy of these immature capillaries in the fundus. This atrophy promotes peripapillary zone formation and increases the likelihood of neoplastic capillary proliferation, subsequently raising the risk of ROP (34,35). This prevalence aligns with our findings, showing a prevalence rate of 32% (95% CI: 24–40%) in very low birth weight preterm infants. Therefore, prioritizing early screening and intervention for these infants is crucial to improving outcomes. Furthermore, systemic factors such as sepsis also play a significant role in the pathogenesis of ROP. Studies suggest that bacterial infections and the systemic inflammatory responses they trigger can lead to vascular damage in the retina. Microorganisms and their toxins may compromise blood vessel integrity, causing leukocyte adhesion, microthrombosis, and vascular occlusion. Additionally, inflammatory mediators and growth factors like interleukin-1β may enhance the activity of the hypoxia-inducible factor (HIF-1α pathway), exacerbating ROP conditions (36,37). Lung diseases requiring respiratory support, particularly those necessitating surfactant therapy, have also been identified as independent risk factors for ROP. Mechanical ventilation and the associated need for increased oxygen therapy can elevate ROP risks, as shown in studies linking high-dose indomethacin and surgical interventions for patent ductus arteriosus to severe ROP (38,39). A positive correlation between high-dose indomethacin and the development of severe ROP has been found, and patent ductus arteriosus is an independent risk factor for the development of ROP. Data from the Canadian Neonatal Analytic Network showed a higher incidence of severe ROP in infants treated with surgical ligation compared to those treated with medication (40). Children with necrotizing enterocolitis are critically ill and may require more aggressive respiratory support, including mechanical ventilation, increased oxygen exposure and fluctuating partial pressure of oxygen, all of which may increase the risk of ROP. Children with early-onset necrotizing enterocolitis may have more preterm complications than children with late-onset necrotizing enterocolitis and may require additional assisted oxygen exposure and respiratory requirements at the time of diagnosis. In addition to necrotizing enterocolitis, duration of exposure and disease severity may be associated with the development of ROP, which we sought to mitigate by controlling other comorbidities in preterm infants (41). Therefore, these findings suggest that early clinical intervention is needed to prevent the development of retinal disease of prematurity in very-low-birth-weight preterm infants with these risk factors.

In this study, particularly concerning analyses of gestational age at preterm birth, birth weight, and epidemiological estimates, considerable heterogeneity was observed (I²>95%). This finding indicates substantial variability between studies, potentially attributable to multiple factors. Firstly, differing definitions and criteria were employed across studies to assess gestational age at preterm birth and birth weight. Despite efforts to standardize these variables, the diversity of data may still account for significant heterogeneity. Secondly, the included studies differed in sample selection, geographical location, and population characteristics. Notably, significant variations in neonatal care, therapeutic interventions, and risk management may exist between studies from high- and low-income countries and regions. Finally, methodological heterogeneity represents another substantial contributor to the high overall heterogeneity. Although sensitivity analyses were conducted to assess result stability, the presence of such substantial heterogeneity may compromise the reliability of our conclusions. Future research should priorities greater consistency in study design and explore potential confounding factors to further validate the accuracy of these associations.

The study has several limitations. First, due to the small number of included studies, we could not perform subgroup analyses for indicators with higher heterogeneity, which may have enhanced the robustness of the conclusions. Second, the inherent heterogeneity of preterm infants in the study subjects increased the variability of the results, which may have affected the consistency and generalizability of the results. Finally, although ORs were adjusted to account for potential confounders, these adjustments may have introduced statistical error. However, the specific impact of adjustment on outcomes is unclear, and future studies are needed to further quantify this effect.

Clinical implications

This study indicates that very low birth weight preterm infants constitute a high-risk cohort for ROP, with susceptibility observed in infants characterized by low gestational age, low birth weight, blood transfusion history, and concomitant conditions such as sepsis, intraventricular hemorrhage, or pulmonary disease. Clinically, early screening and stratified management protocols should be established for these high-risk infants to ensure timely ophthalmic assessment and intervention, alongside enhanced prevention and management of associated complications. Furthermore, significant heterogeneity was observed across studies, potentially attributable to variations in regional healthcare standards, ROP diagnostic criteria, and follow-up protocols, necessitating cautious interpretation of findings. While broadly consistent with recent meta-analyses, future individual patient data (IPD) meta-analyses may yield more precise risk assessments. This could further optimize screening, intervention, and long-term follow-up protocols for high-risk infants, thereby improving visual outcomes and reducing the risk of lifelong visual impairment.

Conclusions

Based on the current study, we found that blood transfusion, low gestational age, low gestational weight, sepsis, intraventricular hemorrhage, pulmonary disease, necrotizing enterocolitis, and patent ductus arteriosus are associated with an increased risk for ROP among very low birth weight preterm infants. However, due to the limitations of the study, more studies are needed to support our conclusions.

Supplementary

The article’s supplementary files as

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.