Definitions of bronchopulmonary dysplasia and long-term outcomes of extremely preterm infants in Korean Neonatal Network

New definitions for bronchopulmonary dysplasia (BPD) have recently been suggested, and an accurate diagnosis, including severity classification with proper definition, is crucial to identify high-risk infants for appropriate interventions. To determine whether recently suggested BPD definitions can better predict long-term outcomes of BPD in extremely preterm infants (EPIs) than the original BPD definition, BPD was classified with severity 1, 2, and 3 using three different definitions: definition A (original), National Institute of Child Health and Human Development (NICHD) definition in 2001; definition B, the modified NICHD 2016 definition (graded by the oxygen concentration and the respiratory support at 36 weeks’ postmenstrual age [PMA]); and definition C, the modified Jensen 2019 definition (graded by the respiratory support at 36 weeks’ PMA). We evaluated 1050 EPIs using a national cohort. Whereas EPIs with grade 2 or 3 BPD as per definition A did not show any increase in the risk, EPIs with BPD diagnosed by definition B and C showed significantly increased risk for poor outcomes, such as respiratory mortality and morbidities, neurodevelopmental delay, and growth restriction at 18–24 months of corrected age. The recently suggested definition and severity grading better reflects long-term childhood morbidities than the original definition in EPIs.

www.nature.com/scientificreports/ L/min; grade 2, non-invasive ventilator [NIV] (including nasal cannula > 2 L/min); and grade 3, invasive ventilator [IV]). This definition could better predict long-term outcomes and identify infants at the greatest risk for poor outcomes 15,16 . However, a number of BPDs are still defined by the NICHD 2001 definition, so it is necessary to verify the validity of recently suggested definitions in a large cohort of EPI with the highest risk for the development of BPD. To better predict long-term outcomes, we compared long-term morbidities of preterm infants with BPD defined by recently suggested definitions with those of the original NICHD 2001 definition.
In this study, we redefined the cohort of EPIs who survived 36 weeks' PMA by three definitions of BPD: original NICHD 2001, modified NICHD 2016, and modified Jensen 2019. We then analyzed the risks for poor respiratory, neurodevelopmental, and growth outcomes at corrected age (CA) 18-24 months according to the severity of BPD according to each definition. (Table 1). A total of 1707 very low birth weight infants (VLBWIs) who were registered in the Korean Neonatal Network (KNN), were identified initially. Mean GA was 25.7 ± 1.2 weeks, and mean birth weight (BWt) was 875.5 ± 193.7 g. Their characteristics are shown in Supplementary Table 1.

Agreement between three different definitions of BPD (Supplementary
Adjusted odds ratios (AOR) [OR was adjusted for GA, SGA, intraventricular hemorrhage (IVH) (≥ grade 3), periventricular leukomalacia (PVL), sepsis, necrotizing enterocolitis (NEC) (≥ stage 2), and retinopathy of prematurity (ROP) (requiring surgery), which might affect mortality and long-term neurodevelopmental and growth outcomes] of mortality or morbidities due to respiratory causes compared to no BPD as per each definition were analyzed.
According to definition A, infants with BPD did not have a higher risk of mortality or morbidities due to respiratory causes than infants without BPD.
By definition B, when compared with infants without BPD, infants with grade 2 and 3 BPD had 2.28 times (95% confidence interval [CI] 1.25, 4.15) and 3.40 times (95% CI 1.98, 5.83) higher risk for mortality or morbidities due to respiratory causes, respectively. By definition C, when compared with infants without BPD, infants with grade 2 and 3 BPD had 2.10 times (95% CI 1.27, 3.48) and 4.41 times (95% CI 2.37, 8.23) higher risk for mortality or morbidities due to respiratory causes, respectively. The rates of neurodevelopmental impairment were approximately 50% in grade 3 BPD by definition A. However, neurodevelopmental impairment rates increased to approximately 50% in grade 2 BPD by definitions B and C, which were greater than 50% in grade 3 BPD by definitions B and C (Table 6).

Neurodevelopmental impairment (
According to definition A, infants with BPD did not have a higher risk for neurodevelopmental impairment than infants without BPD. As per definition B, infants with grade 3 BPD had 1.92 times (95% CI 1.16, 3.17) higher risk for mental developmental delay than infants without BPD. By definition C, when compared with infants without BPD, infants with grade 2 and 3 BPD had 1    Growth restriction (Table 7). A total of 391 infants (38.1%, 391/1027) had growth restriction at follow-up between 18 and 24 months of CA.
Growth restriction rates were approximately 50% in grade 3 BPD as per definition A. However, growth restriction rates increased to approximately 50% in grade 2 BPD by definitions B and C, which were greater than 50% in grade 3 BPD by definitions B and C ( Table 8).
As per definition A, infants with grade 3 BPD had 2.80 times (95% CI 1.15, 6.81) higher risk for growth restriction than infants without BPD. As per definition B, when compared with infants without BPD, infants with grade 2 and 3 BPD had 1.72 times (95% CI 1.13, 2.62) and 2.03 times (95% CI 1.35, 3.05) higher risk for growth restriction, respectively. As per definition C, when compared with infants without BPD, infants with grade 2 and 3 BPD had 1.65 times (95% CI 1.18, 2.30) and 2.39 times (95% CI 1.43, 4.01) higher risk for growth restriction, respectively.

Discussion
In the present study, we have clearly demonstrated that EPIs of 23-27 weeks' gestation with BPD, according to the original NICHD 2001 definition (definition A), did not show significantly increased risks for long-term respiratory mortality, morbidities, or neurodevelopmental delay. However, EPIs with BPD diagnosed as per the modified NICHD 2016 definition (definition B) and modified Jensen 2019 definition (definition C) showed significantly increased risks for poor long-term outcomes, such as respiratory mortality and morbidities, neurodevelopmental delay, and growth restriction at 18-24 months of CA.
It has been 20 years since the introduction of the NICHD 2001 definition, which relies heavily on the concentration and duration of oxygen. Although this definition is still one of the most widely used definitions, the definition of respiratory support at 36 weeks' PMA regardless of oxygen or respiratory support during the first 28 days of life has recently emerged as one of the most commonly used definitions 8,17 18 . We also have to consider the improved survival of EPIs with GA < 28 weeks who require oxygen in the first few weeks of life. BPD, as per definition A, was not associated with respiratory morbidity or neurodevelopmental impairment in this study. Only grade 3 BPD as per definition A was associated with growth restriction at 18-24 months of CA. Grade 1 BPD as per definition A with oxygen dependency during the first 28 days of life is not very important for these tiny preterm infants with a GA of 23-27 weeks. Almost all grade 1 BPD cases as per definition A were assigned to no BPD according to definitions B and C in this study. Over 90% of preterm infants with GA of 23-27 weeks had BPD according to definition A. According to a meta-analysis by Gou et al., oxygen dependency in the first 28 days of life (mild BPD by the NICHD 2001 definition) was not associated with cerebral palsy 10 . In addition, preterm infants with grade 1 BPD as per definition A did not increase the risk of long-term morbidities in this study, so there is no need to classify those patients as "BPD".
Definition B in this study is based on the use of oxygen and positive pressure only at 36 weeks' PMA 13 . Grades 2 and 3 BPD by definition B increased respiratory morbidities, motor developmental delay, and growth restriction in this study. However, the most important drawback of definition B is that it is too complicated for clinical application. Patients were assigned to different levels of grade according to the oxygen concentration, although the same flow rates were used by the NICHD 2016 definition. Therefore, the NICHD 2016 definition had to be modified to definition B because of the complexity of the subdivided oxygen concentration and flow rate.
Definition C in this study is based on the use of positive pressure at 36 weeks' PMA instead of supplemental oxygen, which was modified from the new definition provided by Jensen et al. in 2019. The Jensen 2019 definition predicted mortality or morbidities due to respiratory cause in 81% and mortality or neurodevelopmental impairment in 69% of preterm infants with GA < 32 weeks between 18 and 26 months of CA 14 . Grades 2 and 3 BPD, as per definition C, significantly increased respiratory morbidities, mental developmental delay, motor developmental delay, and growth restriction in this study. We found a stepwise increment in the risks for these long-term morbidities according to the severity of BPD defined as per definition C. Following definition C, more than four-fold higher risk of respiratory morbidities and more than two-fold higher risk of mental and motor developmental delay and growth restriction were found among infants with grade 3 BPD compared to those without BPD. These findings strongly support the clinical usefulness of the classification of BPD severity by definition C. In addition, definition C is very simple and easy to utilize clinically in most preterm infants. Table 8. Growth restriction defined as z-scores < − 1.28 (equivalent to < 10 th percentile) of weight, height, or head circumference at 18-24 months of corrected age. Values are expressed as number (%) or mean ± standard deviation. Wt, number of infants with z-scores < − 1.28 of weight at 18-24 months of corrected age; Ht, number of infants with z-scores < − 1.28 of height at 18-24 months of corrected age; HC, numbers of infants with z-scores < − 1.28 of head circumference at 18-24 months of corrected age; Growth restriction, numbers of infants with z-scores < − 1.28 of weight, height, or head circumference at 18-24 months of corrected age. www.nature.com/scientificreports/ Mortality or morbidities due to respiratory causes included readmission ≥ 3 times due to respiratory causes after discharge from the NICU prior to follow-up in this study. This definition was similar to those reported by the Canadian Neonatal Follow-Up Network 19 and Isayama et al. 20 . The definition of respiratory morbidities by Jensen et al. 14 was similar; however, they also included continued hospitalization due to respiratory causes at or beyond 50 weeks' PMA. In this study, 234 infants among 527 preterm infants with BPD (234/527 [44.4%], defined as B or C) were readmitted at least once during 18-24 months of CA. This result is similar to that of Thébaud et al., who reported that about 50% of preterm infants with BPD were readmitted during the first 2 years of life 21 .
Despite remarkable advances in perinatal and neonatal care and improved survival of EPIs, the incidence of BPD has not decreased or even increased 1,22 . The overall incidence of BPD diagnosed as requiring oxygen or respiratory support at 36 weeks' PMA was reported to be approximately 40% to 45% in preterm infants with GA < 29 weeks 23,24 . Walsh et al. reported that among 45 VLBWIs with BPD diagnosed with oxygen dependency at 36 weeks' PMA, 15 infants successfully ceased oxygen supplementation, as per the oxygen reduction test, and were hence classified as no BPD 25 . This result reflects that the need for oxygen is somewhat determined by the attending physician's care practice, not by physiologic assessment. Inter-center variability in oxygen supply can affect the incidence and severity of BPD. This is a limitation of the currently used definition of BPD, which is based on the use of oxygen. Therefore, definition C, based on the positive pressure instead of oxygen, is more reasonable than definitions A or B.
Jensen et al. suggested that the optimal BPD definition predicted respiratory morbidities better than neurodevelopmental impairment 14 . Similarly, Linsell et al. reported that these results are quite natural, because diverse factors, in addition to respiratory morbidities, contribute to adverse neurodevelopment outcomes in these very preterm infants 26 . Mental and motor developmental delay was increased stepwise according to the severity of BPD as per definition C in preterm infants with GA < 28 weeks in this study. Jensen et al. reported that death or neurodevelopmental impairment increased according to the severity of BPD in preterm infants with GA < 32 weeks (33% with no BPD, 46% with grade 1, 60% with grade 2, and 79% with grade 3 BPD) 14 . Moreover, prolonged respiratory support is associated with an increased risk of mortality or neurodevelopmental disability, which implies an important stepwise relationship between the severity of BPD and poor neurodevelopmental outcome 27 .
There are some limitations to the present study. First, since we used the KNN pre-set registry data, some modifications from the original description of the NICHD 2016 definition or Jensen 2019 definition in definitions B and C, respectively, were inevitable, which can cause some difficulty in directly comparing the results with other studies that used the original definitions. Second, there is heterogeneity in long-term neurodevelopmental studies conducted among enrolled infants because of the multi-center nature of the registry. We prepared composite data, to use as much data as possible and to minimize selection bias; nevertheless, bias may still persist because of inter-hospital variability or inter-test variability, including recall bias arising from parent questionnaires, such as Korean Developmental Screening Test for Infants and Children (K-DST) or Korean Ages & Stages Questionnaires (K-ASQ). Furthermore, follow-up rates were highly variable with inter-hospital variability. High NICU volume with greater number of personnel and resources, easy approaches to follow-up in clinics, younger GA (≤ 29 weeks), and BWt (≤ 1,160 g) were associated with higher follow-up rates 28 .
However, a strength of this study was the use of a nationwide, prospective cohort registry covering > 80% of VLBWIs in Korea with good data management 29 ; further, there was no racial difference, which can exclude various potential biases. Moreover, the subjects of this study were EPIs of 23-27 weeks of gestation, who are the most vulnerable and at high risk of BPD, and have poor respiratory, neurodevelopmental, and growth outcomes compared to more mature preterm infants.
In conclusion, EPIs with grade 3 BPD as per the modified NICHD 2016 definition had a 3.40-fold higher risk for mortality or morbidities due to respiratory cause, 1.92-fold higher risk for mental developmental delay, 1.84-fold higher risk for motor developmental delay, 2.68-fold higher risk for social developmental delay, and 2.03-fold higher risk for growth restriction than those without BPD. EPIs with severe BPD, dependent on IV (grade 3 BPD in modified Jensen 2019 definition), had a 4.41-fold higher risk for mortality or morbidities due to respiratory cause, a 2.45-fold higher risk for mental developmental delay, a 2.22-fold higher risk for motor developmental delay, and a 2.39-fold higher risk for growth restriction than those without BPD. Moderate or severe BPD defined as per the original NICHD 2001 BPD definition was no longer associated with an increased risk of long-term respiratory or neurodevelopmental poor outcome in EPIs. Therefore, more recently suggested definitions of BPD, at least in most immature infants at high risk of BPD, such as EPIs, need to be adopted., which mainly classify BPD by the type of respiratory support required, to identify more high-risk infants with poor long-term outcomes requiring early intervention.

Methods
The KNN is a nationwide, prospective cohort registry of VLBWIs admitted to the 70 participating NICUs, covering > 80% of VLBWIs in Korea since it was launched in 2013 29 . Data for a total of 2220 VLBWIs born between January 1, 2013 and December 31, 2015, with a GA of 23-27 weeks, who were registered in the KNN were collected. A total of 512 infants who died before 36 weeks PMA and one infant who died due to major congenital abnormalities were excluded. Therefore, a total of 1707 VLBWIs with GA of 23-27 weeks were registered, and among them, 1050 were enrolled in the present study who were followed up at 18-24 months of CA.
BPD was assessed at 36 weeks' PMA, and infants were classified according to the severity of BPD as no, grade 1, grade 2, and grade 3 BPD by definition. The risk of poor outcomes of respiratory, neurodevelopmental, and growth at 18-24 months of CA in infants with BPD were compared with those in infants without BPD for each definition. Definition A is the original NICHD 2001 definition 9 with the precondition of oxygen or respiratory support in at least the first 28 days of life and assessed at 36 weeks' PMA. Grade 1 (mild) BPD was defined as breathing room air, grade 2 (moderate) BPD as need for < 30% oxygen, and grade 3 (severe) BPD as need (2) patients with nasal cannula < 2 L/min were allocated to grade 1 in definition B; however, nasal cannula < 1 L/ min with FiO 2 > 0.7 were allocated to grade 2 according to the NICHD 2016 definition (Supplementary Table 3). Definition C is the modified Jensen 2019 definition 14 . There is no precondition of oxygen or respiratory support for at least the first 28 days of life, nor subdivision by oxygen concentration. BPD was graded by respiratory support assessed at 36 weeks' PMA. Grade 1 included nasal cannula < 2 L/min, grade 2 included NIV (included nasal cannula ≥ 2 L/min), and grade 3 included IV. The difference between definition C in this study and Jensen 2019 definition was NIV, which included nasal cannula ≥ 2 L/min in definition C and > 2 L/min in Jensen 2019 definition (Supplementary Table 4). Respiratory status, neurodevelopment, and growth were assessed at follow-up between 18 to 24 months of CA. Mortality or morbidities due to respiratory causes were defined as (1) death due to respiratory causes between 36 weeks' PMA and 18-24 months of CA; (2) readmission (≥ 3 times) due to respiratory cause after discharge from the NICU prior to follow-up; or (3) oxygen, mechanical ventilator, or tracheostomy at follow-up.
Neurodevelopmental assessments included composite scores on the Bayley Scales of Infant Development-Second Edition (BSID II) 30 , BSID III 31 , K-ASQ 32 , and K-DST 33 . Different neurodevelopmental assessment modalities were used with inter-hospital variability, including the BSID II, BSID III, K-ASQ, or K-DST, because of the multi-center registry nature of this study. K-ASQ and K-DST are questionnaires, not confirmatory tests. To use as much data as possible, in order to minimize selection bias, composite data of each domain in each modality were developed. BSID II has two domains (mental and psychomotor), and BSID III has three domains (cognitive, language, and motor). However, the K-ASQ has five domains, and the K-DST has six domains, including social development. Therefore, we developed the composite data as follows. (1) Mental domain in BSID II; (2) cognitive and language domains in BSID III; (3) communication and problem-solving domains in K-ASQ; and (4) cognition, language, and self-help domains in the K-DST were used to evaluate mental development. The motor domain in BSID II and BSID III, gross motor, and fine motor domains in the K-ASQ and K-DST were used to evaluate motor development. The personal-social domain in the K-ASQ and sociality domain in the K-DST were used to evaluate social development. The K-ASQ has been used in Korea since 2000 for developmental assessment of infants 34 , which is a revised Korean version of the ASQ developed in the US by Squires 32 . The K-ASQ includes five domains: gross motor, fine motor, communication, problem-solving, and personalsocial. The K-ASQ showed relatively high concurrent validity with DDST II, which can be used for screening and follow-up of developmental delay 35 . The K-DST has been used for developmental screening of infants and children in Korea since 2014 as part of a national health screening program by the Korean government. The K-DST includes six domains: gross motor, fine motor, cognition, language, sociality, and self-help. The specificity of the K-DST compared to the BSID II is > 70% in preterm infants, and it could be a useful screening tool for neurodevelopmental assessments 36 .
Mental developmental delay was defined as (1) a mental developmental index < 70 on the BSID II; (2) cognitive or language < 70 on the BSID III; (3) communication or problem-solving score less than the cut-off value on the K-ASQ; or (4) cognition, language, or self-help score is less than the cut-off value on the K-DST. Motor developmental delay was defined as (1) a psychomotor developmental index < 70 on the BSID II; (2) motor < 70 on the BSID III; (3) gross motor or fine motor score is less than the cut-off value on K-ASQ; (4) gross motor or fine motor score is less than the cut-off value on K-DST; or (5) Cerebral palsy defined as Gross Motor Functional Classification System ≥ 2 37 . Social developmental delay was defined as (1) a personal-social score less than the cut-off value on the K-ASQ, or (2) sociality is less than the cut-off value on the K-DST.
Growth restriction was defined as failure of catch-up growth, defined as z-scores < − 1.28 (equivalent to < 10th percentile) of weight, height, or head circumference according to the CA by the 2006 World Health Organization (WHO) Child Growth Standards 38 .
Statistical analysis. Data are presented as mean ± standard deviation for continuous variables and as number (percentage) for categorical variables. The chi-square test or Fisher's exact test was performed for categorical variables, and the t-test or Mann-Whitney U test was used for continuous variables, as appropriate. ORs were adjusted for GA, SGA, IVH (≥ grade 3), PVL, sepsis, NEC (≥ stage 2), and ROP (requiring surgery), which might affect mortality, long-term neurodevelopmental, and growth outcomes. AOR and 95% CI compared to no BPD according to the severity of BPD using multivariable logistic regression were presented to evaluate increased risks of respiratory morbidities, neurodevelopmental impairment, or growth restriction. In each model, the study outcome was considered as a binary dependent variable, and the BPD definition was considered as a categorical independent variable. All statistical analyses were performed using SAS (version 9.4; SAS Inc., Cary, NC, USA). Statistical significance was set at p < 0.05.