Validation of a New Classification for Severe Bronchopulmonary Dysplasia in Extremely Preterm Infants: Insights from a Large Japanese Cohort

Bronchopulmonary dysplasia (BPD) classification has evolved, with Jensen’s system – based on respiratory support at 36 weeks’ postmenstrual age (PMA) – gaining recognition for predicting early childhood morbidity [1‒4]. However, diagnosing and intervening before 36 weeks’ PMA remains challenging, and new respiratory methods may necessitate updates. An ideal classification should incorporate BPD etiopathogenesis, integrating endotypes and phenotypes [5], to enable earlier prediction and intervention.

In Japan, a unique BPD classification has been used since the 1990s, defined by postnatal day 28 and based on etiology and phenotype, including respiratory distress syndrome (RDS), intrauterine infection/inflammation, and chest X-ray (CXR) findings [6, 7]. With advances in survival rates for extremely preterm infants, respiratory care, and updated RDS criteria [8], there is increasing support for revising the Japanese classification [9] to improve early prediction and management of BPD and its long-term outcomes. Therefore, using data from the Neonatal Research Network, Japan (NRNJ), we aimed to (1) identify risk factors for severe BPD based on a recent scoping review (ScR) [10] and (2) assess whether these confirmed risk factors were independently and incrementally associated with severe BPD and long-term outcomes at 3 years, including home oxygen therapy (HOT) and neurodevelopmental impairments (NDIs).

This retrospective cohort study analyzed NRNJ data with approval from the Internal Review Board of Osaka Women’s and Children’s Hospital (approval number: 1104) and Ethics Committees of all participating institutions, including Kitasato University Hospital (approval number: C20-040). Informed consent was acquired via website opt-out, in line with the retrospective design.

The study included all extremely preterm infants born at <28 weeks’ gestational age (GA), registered with the NRNJ from 2003 to the most recent year with follow-up data. Infants with congenital anomalies, out-of-hospital births, perinatal deaths, or incomplete records were excluded.

Two BPD classifications were used: BPD at 28 days of age (BPD-28d) and at 36 weeks (BPD-36wks), defined as oxygen dependency and/or positive pressure on postnatal day 28 and at 36 weeks’ PMA, respectively. This dual approach aligns with the Japanese practice of classifying BPD at 28 days based on the three previously mentioned factors, with severity further assessed at 36 weeks’ PMA according to the 2001 NICHD definition [2]. Infants were further categorized into non-BPD, mild, moderate, and severe BPD. Severe BPD, defined as a need for an oxygen concentration ≥0.30 or mechanical ventilation, was used as the primary outcome.

Bubbly/cystic appearance on CXR (bubbly/cystic CXR) was defined as diffuse, streaky infiltrates with small cystic areas (diameter, 1–10 mm) [11‒14]. The diagnosis was established on postnatal day 28 by the attending neonatologists at each institution as a distinct X-ray finding was consistently assessed across all Japanese neonatal intensive care units.

The other study variables were defined using the NRNJ registration manual [15]. GA was assessed using ultrasonography early during pregnancy or from the last menstrual period. Small for gestational age (SGA) was defined as below the 10th percentile of sex-specific birth weight [16]. Histological chorioamnionitis (HCA) was diagnosed by institutional pathologists through pathology [17].

Respiratory and neurodevelopmental outcomes at 3 years of age included HOT, cerebral palsy (a non-progressive central nervous system disorder with abnormal muscle tone movement control) [18], and NDIs, defined as any of the following: cerebral palsy, visual impairment (loss of functional vision in one or both eyes), hearing impairment (requiring amplification), and significant developmental delay (a Kyoto Scale of Psychological Development developmental quotient [KSPD DQ] <70, equivalent to a Bayley III Cognitive score of <85) [19].

The primary outcome was to identify which risk factors from the ScR: GA, male sex, RDS, symptomatic patent ductus arteriosus (hsPDA), HCA, SGA, and bubbly/cystic CXR [10], were significantly associated with severe BPD, using the NRNJ database. The secondary outcome was to determine whether the combination of confirmed risk factors by primary outcome increased the prevalence of severe BPD. The tertiary outcome was to assess whether the confirmed risk factors by primary outcome were independently associated with the aforementioned long-term outcomes and whether their combination increased the prevalence of those outcomes.

All analyses were performed using JMP Pro 17 (SAS Institute Inc., Cary, NC, USA). Prevalence data were expressed as means and percentages with 95% confidence intervals (CIs). Significant differences were identified by non-overlapping 95% CIs (p < 0.05) [14, 20]. Non-normally distributed variables were reported as medians with interquartile ranges and compared using the Wilcoxon-Mann-Whitney test.

A multivariable logistic regression model for severe BPD (primary outcome) included the following risk factors from ScR as independent variables: GA, male sex, RDS, hsPDA, HCA, SGA, and bubbly/cystic CXR. A secondary regression model for severe BPD included confirmed risk factors from the primary analysis as the independent variables, adjusted for GA, male sex, antenatal steroid, multiple pregnancy, Apgar score at 5 min <4, and BPD-28d.

To assess whether confirmed risk factors by the primary analysis were independently or incrementally associated with long-term outcomes at 3 years, further multivariable analysis was conducted with adjustments for severe intraventricular hemorrhage, periventricular leukomalacia, sepsis, severe retinopathy of prematurity, and factors used in the primary analysis. Incremental effect analysis used infants without any confirmed risk factors from the primary analysis as a reference. Multicollinearity in the multivariable analysis was evaluated by calculating the variance inflation factor (VIF) for all independent variables.

Overall, 15,834 extremely preterm infants were enrolled between 2003 and 2016. Among them, 1,722 (11%) were classified with severe BPD (NICHD 2001) and analyzed for primary and secondary outcomes. Among 15,599 surviving infants, 8,538 were excluded from the 3-year follow-up due to death (n = 39) or loss to follow-up (n = 8,499) (online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000543810). Finally, 7,061 infants from 130 tertiary centers were evaluated for long-term respiratory and neurodevelopmental outcomes at 3 years (Fig. 1).

To understand the background of eligible infants in the NRNJ database, we analyzed pre- and postnatal characteristics by BPD severity. GA and birth weight were significantly lower with higher BPD severity. Male sex and SGA were more prevalent in the BPD group, especially in moderate-to-severe BPD cases. Premature rupture of membranes, clinical CAM, and HCA were also more prevalent in the BPD group and increased with severity (online suppl. Table 2).

Infants with BPD had more severe respiratory conditions, indicated by a higher prevalence of air leak, bubbly/cystic CXR, steroid treatment for BPD, high-frequency oscillatory ventilation, surfactant therapy, inhaled nitric oxide, tracheostomy, HOT at discharge, and longer durations of mechanical ventilation, oxygen supplementation, and hospital stays. Death prior to discharge was more common in infants with BPD than those without it (online suppl. Table 3).

HCA (adjusted odds ratio [AOR], 1.20; 95% CI, 1.06–1.36), SGA (AOR, 1.73; 95% CI: 1.51–1.98), and bubbly/cystic CXR (AOR, 1.79; 95% CI, 1.60–2.01), extracted from the ScR, were significantly associated with severe BPD. Conversely, GA (AOR, 0.83; 95% CI, 0.79–0.86) and RDS (AOR, 0.79; 95% CI, 0.69–0.91) were negatively associated. HsPDA showed no significant association (AOR, 1.09; 95% CI, 0.97–1.22) (Table 1).

The combination of HCA, SGA, and bubbly/cystic CXR was significantly associated with an incremental increase in severe BPD-36wks prevalence. Notably, combinations, such as bubbly/cystic CXR plus HCA (AOR, 2.54; 95% CI, 2.06–3.14), bubbly/cystic CXR plus SGA (AOR, 3.27; 95% CI, 2.54–4.21), and all three factors (AOR, 3.11; 95% CI, 2.25–4.29) showed significant associations (Fig. 2).

HCA (AOR 1.54; 95% CI: 1.14–2.08), SGA (AOR 1.70; 95% CI: 1.21–2.39), and bubbly/cystic CXR (AOR 2.63; 95% CI: 1.94–3.56) were independently associated with HOT at 3 years (Table 2). These risk factors also showed incremental effects: bubbly/cystic CXR plus HCA (AOR 3.72; 95% CI: 2.11–6.56), bubbly/cystic CXR plus SGA (AOR 4.70; 95% CI: 2.36–9.38), and all three (AOR 5.39; 95% CI: 2.40–12.10) (Fig. 3). For other neurodevelopmental outcomes, only SGA was independently associated with NDIs (AOR 1.55; 95% CI: 1.32–1.83), visual impairment (AOR 1.42; 95% CI: 1.13–1.78), and KSPD DQ <70 (AOR 1.59; 95% CI: 1.33–1.92) (Table 2). SGA, alone (AOR 1.47; 95% CI: 1.15–1.89), with bubbly/cystic CXR (AOR 1.95; 95% CI: 1.41–2.68), or combined with both bubbly/cystic CXR and HCA (AOR 1.58; 95% CI: 1.00–2.48), significantly increased NDIs prevalence (Fig. 4). All multivariable analyses showed VIF values below 5, indicating no multicollinearity concerns.

This retrospective multicenter cohort study has three key findings. First, HCA, SGA, and bubbly/cystic CXR were significant risk factors for severe BPD-36wks even in a large Japanese cohort. Second, they were also associated with an incremental increase in severe BPD prevalence, especially when combined (e.g., bubbly/cystic CXR plus HCA, bubbly/cystic CXR plus SGA, and all three), compared to cases without any of these risk factors. Third, all three risk factors were independently associated with HOT at 3 years of age in infants with BPD-28d and, similarly, showed an incremental increase in prevalence, especially when combined with bubbly/cystic CXR, compared to the reference.

Unlike the US definition of BPD based on respiratory support [3‒5], Japan has employed a unique classification since the 1990s, focusing on BPD-28d diagnosis with three components: RDS, HCA, and bubbly/cystic CXR, which suggest premature lung, intrauterine infection/inflammation, and lung parenchymal/interstitial disease, respectively [6, 7]. However, with the improved survival rates of extremely preterm infants [21], the old Japanese BPD classification requires revisions [9].

Our findings suggest the potential for a novel BPD classification that incorporates two major etiologies, HCA and SGA, together with the phenotype indicated by CXR findings on postnatal day 28. Although Jensen’s classification effectively predicts mortality and severe respiratory morbidity [4], it does not account for etiological factors that distinguish distinct BPD phenotypes. Furthermore, pulmonary phenotype-based classifications [5] often fail to enable timely early interventions. Future BPD treatment strategies may benefit from this classification, which integrates both endotype and phenotype at day 28, enhancing early alerts for disease progression by 36 weeks’ PMA, increasing opportunities for interventions, and improving outcomes compared to assessments at 36 weeks’ PMA [5].

Our classification is robust as it includes both endotype and phenotype aspects with only three universal factors, even as respiratory treatment strategies evolve. First, the routine documentation of bubbly/cystic CXR on postnatal day 28, indicating lung parenchymal/interstitial lesions, is unique to Japan and has been significantly linked to respiratory prognosis in several studies [11‒14]. Radiographic confirmation of parenchymal disease is crucial for BPD diagnosis [3] as early radiologic changes can predict its progression [22]. This approach allows earlier prediction than 36 weeks’ PMA and serves as a prognostic tool for severe BPD and long-term respiratory outcomes in extremely preterm infants.

Second, HCA, a key endotype of prematurity indicating intrauterine infection [5], is strongly linked to BPD risk [23] and pulmonary vascular diseases, such as persistent pulmonary hypertension of the newborn, which may lead to severe BPD [20, 24]. A large US multicenter study confirmed that HCA is associated with increased mortality or severe BPD in extremely preterm infants, emphasizing its importance in BPD classification [25]. Despite diagnostic challenges, HCA is vital for early BPD recognition and intervention.

Third, SGA, indicative of placental dysfunction [26], was the most frequently identified risk factor for moderate/severe BPD in a recent ScR, appearing in all 10 reviewed studies [10]. In-utero hypoxia and malnutrition, common in SGA, hinder alveolar and microvascular development [27]. Our findings also showed a strong association between SGA and NDIs in infants with BPD. Additionally, a systematic review confirmed that SGA adversely affects cognitive development in preterm infants [28]. These findings highlight SGA as a key marker of placental insufficiency and intrauterine growth restriction (IUGR), and a major risk factor for severe BPD, long-term respiratory complications, and abnormal neurodevelopment in extremely preterm infants. Although SGA may not fully capture all aspects of IUGR, it reflects most of its clinical manifestations. Further studies are needed to elucidate the mechanisms linking IUGR and BPD.

However, this study has some limitations. First, classifying BPD based on just three risk factors is controversial due to its multifactorial etiology. The use of CXR findings as a risk factor for severe BPD is debatable due to inconsistent interpretations among radiologists [2]. While commonly used in Japan for BPD evaluation since the 1990s [6, 7], its application varies internationally. Advances in artificial intelligence may help address standardize interpretations. Alternatives such as chest computed tomography (CT) scoring, magnetic resonance imaging (MRI), and lung ultrasound (LUS) are under investigation. LUS, with its high sensitivity, specificity, non-invasiveness, and safety, shows promise for early BPD diagnosis and intervention. However, its reliance on operator expertise remains a limitation compared to CXR. At this stage, it is crucial to acknowledge that each modality – LUS, CT, MRI, and CXR – has unique strengths and limitations. While CT provides detailed imaging, its use is limited by radiation exposure. Similarly, CXR use should be minimized and restricted to essential examinations.

Second, the three components did not cover all BPD-associated phenotypes, such as early vascular lesions, which are linked to late BPD-related pulmonary hypertension [24]. However, these vascular phenotypes are induced by placental insufficiency and intrauterine infection [5]. Thus, HCA and SGA may cover these vascular phenotypes without explicitly including them.

Third, incorporating HCA into BPD classification poses challenges because its diagnosis requires pathologist expertise, which is not available at all centers. While clinical CAM diagnosis varies by obstetricians, HCA provides clear evidence of intrauterine infection, making it a valuable but complex component. Although HCA has a stage-based classification, we excluded stage analysis to prioritize a simplified, practical Japanese BPD classification. Additionally, the global shift from the Blanc to Redline classification [29] suggests stage-based systems like Blanc may lose relevance. The relationship between HCA stages and BPD warrants future research.

Fourth, the low follow-up rate (45.3% at 3 years among surviving in-born infants) may introduce selection bias in evaluating long-term outcomes. However, our large cohort of 15,834 extremely preterm infants provides clinically meaningful insights despite data loss. The cohort size ensures sufficient power to determine the prevalence and correlates of long-term neurodevelopmental outcomes, with careful interpretation. We acknowledge the need to improve follow-up rates and will address this in future studies.

Additionally, our findings, based on a large Japanese cohort, may have limited global generalizability. The retrospective design also hinders evaluation of therapeutic interventions for BPD. Prospective studies exploring BPD phenotypes and serum cytokine profiles alongside these risk factors could refine treatment strategies. Furthermore, applying the concept of precision medicine to BPD phenotyping could be expected to require much more granular typing and mechanistically insightful analysis than these three disparate factors.

In conclusion, HCA, SGA, and bubbly/cystic CXR on postnatal day 28 are significant risk factors for severe BPD and long-term respiratory outcomes. Their combinations, especially with bubbly/cystic CXR, showed a stepwise increase in prevalence. While further research is needed to validate their role in endotype-specific BPD classification, these findings may improve early prognostic assessments and guide timely interventions before 36 weeks’ PMA. Future studies should develop treatments targeting phenotypes linked to these risk factors.

The authors thank all infants and their respective families for their participation in this study. We also thank Dr. Yusuke Okuda, Dr. Tetsuya Isayama, and Dr. Satoshi Kusuda for giving statistical advice. The institutions enrolled in the Neonatal Research Network, Japan, are as follows: Aiiku Hospital, Aizenbashi Hospital, Akita Red Cross Hospital, Almeida Memorial Hospital, Anjo Kosei Hospital, Aomori Prefectural Central Hospital, Asahi General Hospital, Asahikawa Kosei General Hospital, Asahikawa Medical University, Ashikaga Red Cross Hospital, Bell Land General Hospital, Chiba Kaihin Municipal Hospital, Chifune Hospital, Dokkyo Medical University, Ehime Prefectural Central Hospital, Engaru-Kosei General Hospital, Fuji Heavy Industries Health Insurance Society Ota Memorial Hospital, Fujieda Municipal General Hospital, Fukuchiyama City Hospital, Fukui Prefectural Hospital, Fukui University, Fukuoka University, Fukushima Medical University, Fukushima National Hospital, Gifu Prefectural General Medical Center, Gunma Children’s Medical Center, Gunma University, Hamamatsu Medical University, Himeji Red Cross Hospital, Hiroshima City Hiroshima Citizens Hospital, Hiroshima Prefectural Hospital, Hiroshima University, Hyogo College of Medicine Hospital, Hyogo Prefectural Awaji Medical Center, Hyogo Prefectural Kobe Children’s Hospital, Ibaraki Children’s Hospital, Iida Municipal Hospital, Iizuka Hospital, Imakiire General Hospital, Ise Red Cross Hospital, Ishikawa Medical Center for Maternal and Child Health, Iwate Medical University, Iwate Prefectural Kuji Hospital, Iwate Prefectural Ninohe Hospital, Iwate Prefectural Ofunato Hospital, Japan Red Cross Medical Center, Jichi Medical University, Jichi Medical University Saitama Medical Center, Juntendo University, Juntendo University Urayasu Hospital, Kagawa University, Kagoshima City Hospital, Kakogawa West City Hospital, Kameda Medical Center, Kanagawa Children’s Medical Center, Kanazawa Medical Center, Kanazawa Medical University, Kansai Medical University, Katsushika Red Cross Hospital, Kawaguchi Municipal Medical Center, Kawasaki Medical School Hospital, Kiryu Kosei General Hospital, Kitakyushu Municipal Medical Center, Kitano Hospital, Kitasato University, Kobe City Medical Center General Hospital, Kobe University, Kochi Health Science Center, Komaki Municipal Hospital, Konan Kosei Hospital, Koseiren Takaoka Hospital, Kumamoto City Hospital, Kumamoto University, Kurashiki Central Hospital, Kurume University, Kushiro Red Cross Hospital, Kyoto City Hospital, Kyoto Prefectural University of Medicine Hospital, Kyoto Red Cross Daiichi Hospital, Kyoto University, Kyushu University, Matsue Red Cross Hospital, Matsuyama Red Cross Hospital, Mitsubishi Kyoto Hospital, Miyazaki University, NTT Higashinihon Sapporo Hospital, Nagahama Red Cross Hospital, Nagano Children’s Hospital, Nagaoka Red Cross Hospital, Nagoya City West Medical Center, Nagoya Red Cross Daiichi Hospital, Nagoya Red Cross Daini Hospital, Nagoya University, Naha City Hospital, Nakatsu Municipal Hospital, Nara Medical University, National Center for Child Health and Development, National Center for Global Health and Medicine, National Cerebral and Cardiovascular Center, National Hospital Organization Kagawa Children’s Hospital, National Hospital Organization Kokura Medical Center, National Hospital Organization Kure Medical Center, National Hospital Organization Miyakonojo Medical Center, National Hospital Organization Nagara Medical Center, National Hospital Organization Nagasaki Medical Center, National Hospital Organization Okayama Medical Center, National Hospital Organization Saga Hospital, National Hospital Organization Shinshu Ueda Medical Center, National Hospital Organization Yokohama Medical Center, National Kyushu Medical Center, National Maizuru Medical Center, National Mie Central Medical Center, Nayoro City General Hospital, Nihon University Itabashi Hospital, Niigata Municipal Hospital, Niigata Prefectural Central Hospital, Niigata University, Nikko Memorial Hospital, Nippon Medical School Musashi Kosugi Hospital, Nishisaitama-chuo National Hospital, Numazu Municipal Hospital, Obihiro-Kosei General Hospital, Odawara Municipal Hospital, Oita Prefectural Hospital, Okayama Red Cross Hospital, Okazaki Municipal Hospital, Okinawa Prefectural Chubu Hospital, Okinawa Prefectural Nanbu Medical Center & Children’s Medical Center, Okinawa Red Cross Hospital, Osaka City General Hospital, Osaka City Sumiyoshi Hospital, Osaka City University, Osaka General Medical Center, Osaka Medical Center and Research Institute for Maternal and Child Health, Osaka Red Cross Hospital, Osaka University, Otsu Red Cross Hospital, Rinku General Medical Center, Saint Luku’s International Hospital, Saiseikai Hyogoken Hospital, Saiseikai Suita Hospital, Saiseikai Yokohamashi Tobu Hospital, Saitama Children’s Medical Center, Saitama Medical University Saitama Medical Center, Saku General Hospital, Sanikukai Hospital, Sapporo City General Hospital, Sapporo Medical University, Seirei Hamamatsu Hospital, Sendai Red Cross Hospital, Shiga University of Medical Science Hospital, Shimane Prefectural Central Hospital, Shinshu University, Shizuoka Children’s Hospital, Shizuoka Saiseikai Hospital, Showa University, St. Marianna University School of Medicine Hospital, St. Mary’s Hospital, Toyota Memorial Hospital, Takatsuki General Hospital, Takayama Red Cross Hospital, Takeda General Hospital, Teikyo University, Tenshi Hospital, The Japan Baptist Hospital, Toho University, Tokai University, Tokushima Municipal Hospital, Tokushima University, Tokyo Jikei Medical University, Tokyo Medical University, Tokyo Medical and Dental University, Tokyo Metropolitan Bokuto Hospital, Tokyo Metropolitan Children’s Medical Center, Tokyo Metropolitan Otsuka Hospital, Tokyo Women’s Medical University, Tokyo Women’s Medical University Yachiyo Medical Center, Tosei General Hospital, Tottori Prefectural Central Hospital, Tottori University, Toyama Prefectural Central Hospital, Toyama University, Toyohashi Municipal Hospital, Toyonaka Municipal Hospital, Toyooka Public Hospital, Tsuchiura Kyodo Hospital, Tsuchiya General Hospital, Tsukuba University, Tsuruoka Municipal Shonai Hospital, Tsuyama Central Hospital, Uji Tokushukai Hospital, University of Occupational and Environmental Health, Wakayama Medical University, Yaizu City Hospital, Yamagata Prefectural Central Hospital, Yamagata University, Yamaguchi Grand Medical Center, Yamaguchi University, Yamanashi Prefectural Central Hospital, Yao Municipal Hospital, Yodogawa Christian Hospital, Yokkaichi Municipal Hospital, Yokohama City University Medical Center, and Yokohama Rosai Hospital. We also express our gratitude to Editage (www.editage.com) for the English language editing of this manuscript.

The study was approved by the Internal Review Board of Osaka Women’s and Children’s Hospital and the Ethics Committees of all NRNJ-participating institutions, including Kitasato University Hospital (Approval No. 1104), and conducted in line with the principles of the Declaration of Helsinki. Informed consent was obtained in an opt-out manner via the study website due to the retrospective nature of the study. Opt-out informed consent protocol was used for the use of participant data for research purposes. This consent procedure was reviewed and approved by Kitasato University Hospital, Approval No. C20-040, date of decision: May 1, 2020.

The authors have no conflicts of interest relevant to this article to disclose.

This research is supported by the Research Program on Rare and Intractable Diseases, Health, Labor, and Welfare Sciences Research Grants (23FC0104).

Hidehiko Nakanishi: conceptualization, methodology, validation, formal analysis, investigation, data curation, writing – original draft preparation, and visualization. Masahito Ito, Shin Kato, Makoto Saito, Naoyuki Miyahara, and Hirokazu Arai: conceptualization, investigation, and writing – review and editing. Erika Ota: methodology, resources, reviewing, and editing. Fumihiko Namba: conceptualization, writing – review and editing, supervision, and funding acquisition.

The data in this study were obtained from the Neonatal Research Network, Japan, where specific restrictions and licensing agreements may apply. Datasets may be requested from the Neonatal Research Network, Japan, by contacting [email protected].