The study is a prospective longitudinal cohort study with VLBW infants enrolled in the KNN. The KNN is an ongoing web-based nationwide registry database participated by 69 hospitals across South Korea [11,12]. The registry includes approximately 70% of all VLBW infants born in South Korea. The registry database collects not only prenatal and postnatal data but also post-discharge data, including the results of the BSID test at corrected ages of 18 to 24 months. Written informed consent was obtained from the parents upon enrollment at each participating hospital. From the KNN registry, we included VLBW infants who were born between January 2013 and December 2015 with birth weights of <1,500 g and underwent the BSID-III test at corrected ages of 18 to 24 months. The exclusion criteria were VLBW infants with major congenital malformations including genetic anomalies and malformation of a major organ system. VLBW infants with seizure and periventricular leukomalacia (PVL) and intraventricular hemorrhage (IVH) grade II or higher were also excluded.

The collected prenatal data included multiple gestations, normal pregnancy and in vitro fertilization and embryo transfer, gestational diabetes mellitus, pregnancy-induced hypertension, clinical and histological chorioamnionitis, administration and completion of antenatal corticosteroid therapy, delivery method, maternal age, and maternal and paternal educational level. The educational level of the parents was classified into two groups according to whether they have graduated from college or not.

The analyzed perinatal data included gestational age, sex, and body measurements (weight and head circumference) at birth. The variables in the delivery room included 1- and 5-minute Apgar scores and implementation of neonatal resuscitation. Respiratory and cardiac statuses in the neonatal intensive care unit (NICU) were investigated as the presence of respiratory distress syndrome and patent ductus arteriosus. Also, the use of steroid therapy and diagnosis and grade of BPD were considered. The diagnosis and severity of BPD were defined by the need for supplemental oxygen support at 28 days of postnatal age and 36 weeks of corrected age [13]. ROP was defined using the international committee criteria, and necrotizing enterocolitis (NEC) was defined as higher than stage II according to modified Bell’s criteria [14]. IVH grade I was defined according to the classification of Papile et al. [15] Culture-proven sepsis, duration (days) of total parenteral nutrition (TPN), and automated auditory brainstem response result were also included in the neonatal variables. The body measurements at the time of NICU discharge were considered.

In the registry, the BSID-II or BSID-III test was performed for VLBW infants depending on the routine procedure in the followup center. In the study, we included infants who underwent the BSID-III test at corrected ages of 18 to 24 months. Cerebral palsy (CP) was defined as a permanent disorder of movement and posture that causes limitations of physical activity that are attributed to nonprogressive disturbances in the fetal or infant brain and graded according to the Gross Motor Function Classification System of Palisano and associates for children aged ≤2 years [16]. Developmental delay was defined as a composite score of ≤85 (≤1 standard deviation [SD]) in each category [17]. Moderate delay was defined as a composite score of ≤70 (≤2 SD), and severe delay was defined as a composite score of ≤55 (≤3 SD).

The baseline demographic, anthropometric, and clinical characteristics of the study population were described using either the median (interquartile range [IQR]) or numbers of cases (percentage). For categorical variables, the chi-square and Fisher exact tests were used to compare the associations of variables between the normal development and developmental delay groups. The Mann–Whitney U and Wilcoxon rank-sum tests were used to compare continuous variables. To identify the quantitative relationships between variables, analysis was performed in two steps: a univariate analysis and then a multivariate analysis. The univariate logistic regression analysis was employed to identify the variables that showed an apparent association with neurodevelopmental impairment. Then, the multivariate logistic regression analysis was performed to identify the strong predictors for the adverse neurodevelopmental outcomes. In the multivariate analysis, gestational age and birth weight were included in the analysis assessing outcome prediction. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA), and associations were presented as odds ratio (OR) and 95% confidence interval. For all tests, P<0.05 was considered significant.

The selection process of the study population is summarized in Figure 1. During the study period, 5,899 VLBW infants were born and registered in the KNN registry. After discharge, 2,759 (47.4%) of the VLBW infants were followed up in outpatient clinics, and 1,053 VLBW infants (17.8%) underwent the BSID test at corrected ages of 18 to 24 months. Of the 1,053 VLBW infants, 383 (36.3%) underwent the BSID-III test. Of the 383 VLBW infants, 77 infants with abnormal findings in the neurological imaging tests (n=43), IVH grade II or higher (n=25), PVL (n=9), or major congenital malformation (n=8) were excluded. Eventually, 285 VLBW infants who underwent the BSID-III test at corrected ages of 18 to 24 months were included for analysis. The mean gestational age was 29 weeks (IQR, 27 to 30 weeks), and the mean birth weight was 1,130 g (IQR, 900 to 1,290 g) (Table 1). The mean gestational age and birth weight of the remaining group were 28.3 weeks (P=0.06) and 1,074 g (P=0.17), respectively, which did not show statistical significance. Furthermore, three of the enrolled VLBW infants were diagnosed with CP, and the mean discharge weight was 2,470 g (Table 2).

The incidence of developmental impairment was highest for language (20.7%, 59/285), followed by cognition (11.9%, 34/285) and motor function (11.2%, 32/285). The mean±SD developmental scores in the cognition, language, and motor scales were 99.99±13.59, 95.14±14.72, and 99.32±11.61, respectively. The average score was the lowest in the language group in both total study and delayed VLBW groups. Moreover, the incidence of moderate and severe delay was the highest in the language scale (Table 3).

In the univariate analysis, cognition delay was significantly associated with multiple factors. Lower 5-minute Apgar score, lower birth weight, moderate and severe BPD, IVH grade I, and longer TPN duration were associated with the poor cognitive outcome (Supplementary Table 1). However, in the multivariate models, none of the factors were significantly related. IVH grade I (P=0.079) and moderate and severe BPD (P=0.056) were highly suspected predictors, and they showed marginal associations (Table 4).

In terms of the language delay, birth weight and head circumference at the time of birth and discharge were significant factors associated with the outcome in the univariate analysis (Supplementary Table 2). Also, IVH grade I and paternal educational level were associated with language delay. In the multivariate analysis, the paternal educational level was the only related factor. As the paternal educational level increased, the incidence of language delay decreased (OR, 0.33; P<0.05) (Table 5).

In the univariate analysis of motor delay, birth weight, head circumference at birth, and 5-minute Apgar score were significant factors. Similar to other scales, moderate and severe BPD, IVH grade I, total TPN duration, and head circumference at discharge were the major risk factors (Supplementary Table 3). In the multivariate analysis, birth weight and head circumference at discharge were significant factors associated with motor delay (P<0.05) (Table 6).