OBJECTIVE The study aimed 1) to evaluate the association between the presence or absence of umbilical cord arteritis (UCA) and the cord blood cytokine levels, and 2) morbidity and mortality of preterm neonates; and 3) to identify predictive markers for UCA of preterm neonates. METHODS In this single-center retrospective observational cohort study, preterm neonates born at gestational age (GA) < 36 weeks were categorized pathologically according to the severity of intrauterine inflammation; those without UCA as Group 1, those with UCA as Group 2, and those without any intrauterine inflammation as Group 3 (control), and subgroup analyses classified by their GA were performed. We compared morbidity and mortality, and eight representative cytokine levels in cord blood samples between the groups. Subsequently, receiver operating characteristics (ROC) curves for UCA diagnosis for each cytokine were created, and values of areas under the curve (AUC) were calculated to determine the optimal predictive markers. RESULTS In total, 105 patients (36, 58, and 11 in Groups 1, 2, and 3, respectively) were included. Multivariate logistic analysis revealed that patients with UCA had higher incidence of brain injury (Odds Ratio [OR] = 8.53, P = 0.0049, 95% Confidence Interval [CI]: 1.91 - 38.0), at term equivalent age in the subgroup analysis with GA < 32 weeks. Although the median value of cord blood granulocyte colony-stimulating factor (G-CSF) was significantly higher in Group 2 than in Group 1 or 3, only the G-CSF level was found to be high in the subgroup analysis with GA < 32 weeks. For UCA diagnosis, the AUC values of G-CSF were the highest among eight cytokines including interleukin 6 (IL-6). These findings were similar in the subgroup analysis with GA < 32 weeks. CONCLUSIONS Preterm neonates, especially born at GA < 32 week, had higher morbidity from　brain injury in the group with UCA. The cord blood G-CSF level was highly accurate for predicting UCA and could thus be used as an optimal biomarker.