Association of the C-reactive protein-to-albumin ratio with treatment-requiring retinopathy of prematurity in premature infants

Retinopathy of prematurity (ROP) is a multifactorial retinal vascular disorder in which prematurity, oxygen-related injury, oxidative stress, and inflammatory mechanisms contribute to abnormal retinal vascular development. Although gestational age (GA) and birth weight (BW) are established risk factors, biomarkers reflecting systemic inflammatory burden may provide additional information regarding the risk of treatment-requiring ROP. The C-reactive protein-to-albumin ratio (CAR) has been investigated as a marker of systemic inflammatory burden in several neonatal and systemic conditions, but evidence regarding its independent association with treatment-requiring ROP remains limited. Therefore, this study aimed to evaluate the association between CAR and treatment-requiring ROP and to determine whether CAR provides additional discriminative information beyond GA and BW. This retrospective study included 390 premature infants screened for ROP between January 2021 and December 2025. Infants were categorized as having no ROP ( n = 209), spontaneously regressed ROP ( n = 102), or treatment-requiring ROP ( n = 79). CAR was calculated from serum CRP and albumin measurements obtained at the time of the highest documented ROP stage or within the preceding two weeks. Associations with treatment-requiring ROP were evaluated using multivariable logistic regression. ROC analysis assessed the discriminative performance of GA, BW, CAR, and combined models. The incremental contribution of CAR was evaluated using DeLong’s test, Brier score, calibration analysis, and 1,000-bootstrap internal validation. Sensitivity analyses additionally adjusted for major neonatal confounders. CAR differed significantly among the three groups and was highest in infants with treatment-requiring ROP ( P < 0.001). Higher CAR was independently associated with treatment-requiring ROP after adjustment for GA and BW (OR = 1.46, 95% CI 1.29–1.65; P < 0.001). The AUCs for GA, BW, and CAR were 0.799, 0.761, and 0.712, respectively. The AUC increased from 0.801 for the GA + BW model to 0.927 after addition of CAR (ΔAUC = 0.127, 95% CI 0.073–0.181; P < 0.001). The GA + BW + CAR model showed a lower Brier score (0.074 vs. 0.122) and an optimism-corrected AUC of 0.926 after 1,000-bootstrap validation. CAR remained independently associated with treatment-requiring ROP after adjustment for postmenstrual age at CRP measurement, oxygen therapy duration, and mechanical ventilation duration (OR = 1.44, 95% CI 1.27–1.65; P < 0.001). Additional analyses including sepsis, bronchopulmonary dysplasia, necrotizing enterocolitis, intraventricular hemorrhage, patent ductus arteriosus, small-for-gestational-age status, and respiratory distress syndrome did not materially alter this association. Higher CAR was independently associated with treatment-requiring ROP and provided incremental predictive information beyond GA and BW. CAR may serve as a potential adjunctive biomarker of systemic inflammatory burden, although prospective multicenter studies with external validation are required before clinical implementation.