Abstract
Background
This study aimed to evaluate the associations between severity classified by three bronchopulmonary dysplasia (BPD) diagnostic criteria [2001 National Institute of Child Health and Human Development (NICHD), 2018 NICHD, and 2019 Jensen] and cerebral injury identified on term-equivalent age (TEA) magnetic resonance imaging (MRI) in infants born at <32 weeks’ gestation.
Methods
This single-center retrospective cohort study included 222 preterm infants (mean gestational age 29.3±1.6 weeks) admitted to Shanghai Children’s Hospital [2018–2022]. Brain injury was scored on TEA MRI using the Kidokoro system, yielding white matter (WM), cortical gray matter (GM), deep GM, and cerebellar scores; a global brain abnormality (GBA) score ≥6 defined moderate-to-severe global brain injury (msGBI). We compared brain injury across BPD severity groups for each criterion and analyzed associations between BPD severity and msGBI using Firth’s penalized-likelihood regression, adjusting for gestational age and birth weight.
Results
WM abnormalities were the most frequent injury (70.3%); msGBI was present in 12.6% of infants. Applying the 2019 Jensen criteria revealed a severity-dependent increase in brain injury scores, with significant differences in GBA, WM, and cortical GM scores across severity groups (all P<0.05). Only grade 3 BPD by the 2019 Jensen criteria was independently associated with msGBI (adjusted odds ratio =8.54, 95% confidence interval: 2.04–37.72, P=0.004). No significant associations were found for the other criteria.
Conclusions
The 2019 Jensen criteria uniquely identify a high-risk subgroup (grade 3 BPD) with significantly increased odds of msGBI on TEA MRI, supporting its utility in neurological risk stratification.
Keywords: Bronchopulmonary dysplasia (BPD), brain injury, infants, magnetic resonance imaging (MRI), diagnostic criteria
Highlight box.
Key findings
• Only the 2019 Jensen criteria demonstrated a severity-dependent increase in brain injury scores. Specifically, only grade 3 bronchopulmonary dysplasia (BPD) defined by the 2019 Jensen criteria (invasive mechanical ventilation) was independently associated with moderate-to-severe global brain injury (msGBI) on term-equivalent age (TEA) magnetic resonance imaging (MRI) (adjusted odds ratio =8.54, 95% confidence interval: 2.04–37.72, P=0.004). No significant independent associations were identified for the 2001 National Institute of Child Health and Human Development (NICHD) or 2018 NICHD criteria.
What is known and what is new?
• BPD definitions rely on respiratory support and inconsistently predict neurodevelopmental impairment. TEA MRI objectively detects cerebral injury predictive of long-term outcomes.
• This study found that only the 2019 Jensen criteria stratified infants by cerebral injury burden on MRI. Grade 3 BPD (requiring invasive ventilation) was the sole independent risk factor for msGBI, supporting its use over other definitions for neurological risk stratification.
What is the implication, and what should change now?
• These findings imply that the 2019 Jensen criteria may be clinically valuable in identifying infants at the highest neurological risk. Grade 3 BPD should prompt clinicians to consider early neuroprotective strategies and targeted MRI surveillance for potential brain injury. Our results challenge the interchangeable use of current BPD definitions in neurological risk stratification and suggest that future evaluations of clinical guidelines may consider prioritizing the 2019 Jensen criteria to improve early detection.
Introduction
Bronchopulmonary dysplasia (BPD) is the most common chronic pulmonary disease among surviving extremely or very preterm infants and is closely associated with long-term respiratory sequelae and neurodevelopmental impairments (NDI) (1,2). The establishment of a standardized BPD definition remains a persistent challenge in neonatology. Although diagnostic criteria have undergone iterative modifications, current diagnosis continues to rely predominantly on therapeutic requirements for respiratory support rather than objective measures of pulmonary structure or function, thereby introducing inherent subjectivity into the diagnostic process. Notably, recent translational developments in magnetic resonance imaging (MRI), such as hyperpolarized gas ventilation imaging, have demonstrated the feasibility of directly and noninvasively assessing regional lung ventilation and structure (3,4), highlighting the potential for objective imaging biomarkers to inform future BPD definitions.
Comparative analyses of various BPD definitions for predicting long-term neurodevelopmental outcomes have yielded inconsistent results (5). Some studies indicate that the 2018 National Institute of Child Health and Human Development (NICHD) criteria and the 2019 Jensen criteria perform comparably in predicting the composite outcome of death or NDI at 2 years of corrected age in preterm infants (6-8), with both outperforming the 2001 NICHD criteria (6,7). However, Katz et al. found no statistically significant differences in predictive performance among these three definitions (9). Similarly, Donlon et al. reported no significant difference between the NICHD 2001 and Jensen 2019 criteria in predicting NDI (10). These inconsistent findings likely stem from heterogeneity across studies, including differences in population characteristics, follow-up durations, and the specific definitions of composite long-term outcome measures (5). This heterogeneity underscores the need for an early, objective, and biologically relevant intermediate endpoint to clarify the comparative utility of different BPD criteria for neurological risk.
Cerebral magnetic resonance imaging (cMRI) at term-equivalent age (TEA) has emerged as a promising objective intermediate endpoint for this purpose. While recent advances in fetal brain MRI, including radiomics-based quantitative analysis, hold promise for earlier detection and neurodevelopmental outcome prediction, challenges in prenatal and early neonatal diagnosis persist (11,12). In parallel, emerging early functional assessments such as olfactory testing are being explored for risk stratification (13). At TEA, cMRI accurately identifies structural abnormalities associated with encephalopathy of prematurity, which is primarily characterized by white matter (WM) and gray matter (GM) injury. Accumulated evidence consistently demonstrates that cerebral abnormalities detected by cMRI during the neonatal period serve as important predictors of long-term neurodevelopmental outcomes (14-18). Therefore, cMRI findings at TEA provide an early objective basis that strongly supports the reliability of favorable neurological assessments conducted at 2 years of corrected age (14-16).
A significant knowledge gap exists regarding the relationship between different BPD diagnostic criteria and cerebral injury at TEA in preterm infants. This study therefore aims to systematically evaluate the association between various BPD diagnostic and severity classifications and cerebral injury using TEA cMRI as an objective imaging biomarker. Our findings are expected to establish an empirical basis for developing a more clinically relevant and prognostically significant BPD definition. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0651/rc).
Methods
Study design and participants
This single-center retrospective cohort study included infants who met the following inclusion criteria: born at <32 weeks’ gestation, admitted to the neonatal intensive care unit of Shanghai Children’s Hospital between January 2018 and December 2022, and survived to 36 weeks’ postmenstrual age (PMA). Infants were excluded if they: (I) were admitted to the neonatal intensive care unit >14 days after birth, (II) were readmitted following initial discharge, (III) did not undergo a cMRI scan before discharge, or (IV) had diagnosed or suspected genetic disorders or congenital infections.
This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of Shanghai Children’s Hospital (No. 2024R034-E01). Prior to enrollment, written informed consent was acquired from the parents of all participants, permitting the use of anonymized clinical data for research and publication purposes.
Clinical data collection and definitions
Demographic and clinical data were retrospectively extracted from electronic medical records. BPD severity was assessed at PMA 36 weeks and classified according to the 2001 NICHD (19), 2018 NICHD (20), and 2019 Jensen (21) criteria. Advanced resuscitation referred to the administration of positive pressure ventilation, endotracheal intubation, chest compressions, or epinephrine in the delivery room. Severe retinopathy of prematurity was diagnosed as stage ≥3 or any condition requiring therapy (22). Sepsis refers to life-threatening organ dysfunction resulting from an aberrant host response to infection (23). Necrotizing enterocolitis was defined as stage ≥ IIA (modified Bell’s criteria) (24). Treatment-requiring patent ductus arteriosus indicated a hemodynamically significant patent ductus arteriosus necessitating pharmacologic or surgical intervention (25).
Brain MRI acquisition and analysis
All MRI examinations were performed on a 3-Tesla scanner (Philips Healthcare Ingenia) using a “feed and wrap” technique to minimize motion artifacts. Oral chloral hydrate (20–50 mg/kg) was administered when sedation was necessary. T1‑ and T2‑weighted imaging acquisition parameters are provided in Table S1.
All MRI scans were independently evaluated by an experienced pediatric radiologist (22 years of experience) who was blinded to all clinical information. This study adopted the method proposed by Kidokoro et al. (26) to assess abnormalities in the cerebral WM, cortex, deep GM, and cerebellum. Because images were acquired with a slice thickness of 4 mm, thinning of the corpus callosum was considered unreliable in this study and was therefore excluded from the WM scoring system. Total WM scores were classified as normal/subnormal (0–1) or abnormal (≥2). Cortical GM, deep GM, and cerebellar scores were each defined as normal (0) or abnormal (≥1). The global brain abnormality (GBA) score was computed as the sum of the four regional subscores and dichotomized into normal–mild (0–5) and moderate-to-severe global brain injury (msGBI) (GBA score ≥6) (27).
Statistical analyses
Continuous data were presented as mean ± standard deviation or median with interquartile range (IQR), with comparisons made using the independent samples t-test or Wilcoxon rank-sum test, as appropriate. Categorical and ordinal variables were summarized as frequencies and percentages, and were compared using the Chi-square or Fisher’s exact test. The distributions of brain abnormality scores across BPD severity groups under each diagnostic criterion were compared using the Kruskal-Wallis test, visualized with cumulative distribution curves.
Associations between BPD diagnostic criteria and msGBI were evaluated using Firth’s penalized-likelihood logistic regression to address data sparsity arising from a low outcome event rate and unbalanced distribution of predictor variables. The primary multivariable model was adjusted for two pre-specified core confounders strongly associated with both BPD and brain injury: gestational age (GA) and birth weight (BW) (28,29). To assess the robustness of the association estimates, we performed sensitivity analyses by sequentially adding other clinical covariates potentially associated with brain injury in preterm infants that showed a significant association (P<0.10) in univariate analyses. These covariates included sepsis, use of vasoactive drugs, red blood cell transfusion, and advanced resuscitation. Results were presented as odds ratios (ORs) with 95% confidence intervals (CIs) for each severity category relative to the no-BPD group within each definition. All statistical analyses were performed using SPSS Statistics (version 31.0; IBM Corp., USA) and R (version 4.5.1 for Windows). Statistical significance was defined as a two-sided P<0.05.
Results
Study cohort and baseline characteristics
A total of 222 infants who underwent cMRI at TEA were included in the final analysis (Figure 1). Their baseline demographic and clinical characteristics are summarized in Table 1. According to the 2001 NICHD, 2018 NICHD, and 2019 Jensen diagnostic criteria, the cumulative incidence of BPD was 69.8% (155/222), 35.6% (79/222), and 37.8% (84/222), respectively (Table 1).
Figure 1.
Flowchart of the study population. cMRI, cerebral magnetic resonance imaging; GA, gestational age; PMA, postmenstrual age; TEA, term-equivalent age.
Table 1. Baseline characteristics of the study cohort undergoing cMRI.
| Characteristics | All infants (n=222) |
|---|---|
| Gestational age, weeks | 29.3±1.6 |
| Birth weight, g | 1,295.1±295.9 |
| Male | 127 (57.2) |
| Multiple births | 64 (28.8) |
| Cesarean delivery | 132 (59.5) |
| Assisted reproduction | 57 (25.7) |
| Intrauterine growth retardation | 10 (4.5) |
| Gestational diabetes | 37 (16.7) |
| Pregnancy hypertension | 47 (21.2) |
| Clinical chorioamnionitis | 5 (2.3) |
| Antenatal corticosteroids | 166 (74.8) |
| Antenatal magnesium therapy | 149 (67.1) |
| Advanced resuscitation | 91 (41.0) |
| Apgar score <5 at 5 minutes | 5 (2.3) |
| Surfactant use | 183 (82.4) |
| Postnatal corticosteroid (for BPD) | 20 (9.0) |
| Pneumothorax | 4 (1.8) |
| Pulmonary hemorrhage | 3 (1.4) |
| PDA requiring treatment | 60 (27.0) |
| Sepsis | 59 (26.6) |
| Necrotizing enterocolitis ≥ stage 2 | 15 (6.8) |
| Severe retinopathy of prematurity | 8 (3.6) |
| Days on mechanical ventilator | 1 (0.0–5.3) |
| Days of O2 therapy | 37.5 (20.8–59.0) |
| Duration of caffeine | 40.0 (28.0–53.0) |
| Vasoactive drug | 24 (10.8) |
| Red blood cell transfusion | 143 (64.4) |
| PMA at MRI scan (years) | 37.0 (35.9–38.7) |
| 2001 NICHD definition | |
| Mild BPD | 76 (34.2) |
| Moderate BPD | 29 (13.1) |
| Severe BPD | 50 (22.5) |
| 2018 NICHD definition | |
| Grade I BPD | 30 (13.5) |
| Grade II BPD | 34 (15.3) |
| Grade III BPD | 15 (6.8) |
| 2019 Jensen definition | |
| Grade 1 BPD | 19 (8.6) |
| Grade 2 BPD | 56 (25.2) |
| Grade 3 BPD | 9 (4.1) |
Data are presented as n (%), median (interquartile range) or mean ± standard deviation. BPD, bronchopulmonary dysplasia; cMRI, cerebral magnetic resonance imaging; MRI, magnetic resonance imaging; NICHD, National Institute of Child Health and Human Development; PDA, patent ductus arteriosus; PMA, postmenstrual age.
Owing to the progressive adoption of pre-discharge cMRI as routine clinical practice from 2022 onward, our study cohort was incompletely enrolled. Nevertheless, the group that underwent cMRI comprised a cohort of high clinical relevance, systematically characterized by lower GA, lower BW, and a higher incidence of BPD (all P<0.001; Table S2).
Brain injury scores by BPD severity according to different diagnostic criteria
Analysis based solely on the 2019 Jensen criteria revealed a systematic rightward shift in the cumulative distribution curves for GBA, WM, and cortical GM scores with increasing BPD severity (Figure 2). The grade 3 BPD group exhibited higher median scores for GBA (9; IQR, 1–13.5), WM (6; IQR, 0–9.5), and cortical GM (1; IQR, 0–3.5) compared to the other severity groups. Statistically significant differences were observed across BPD severity groups for GBA score (P=0.04), WM score (P=0.04), and cortical GM score (P=0.02) (Table S3).
Figure 2.
Cumulative distribution curves for brain injury scores according to BPD severity. BPD, bronchopulmonary dysplasia.
Although the distributions of WM, cortical GM, and GBA scores differed significantly among the severity groups defined by the 2001 NICHD criteria (P=0.005, 0.014, and 0.005, respectively), the median values were largely similar across groups, with differences primarily driven by variations in data dispersion (Table S3). In contrast, application of the 2018 NICHD criteria revealed no statistically significant differences in any of these scores across severity groups (all P>0.05; Table S3).
Prevalence of brain injury and its association with BPD severity
At TEA, the most prevalent pattern of injury involved cerebral WM abnormalities (70.3%, 156/222), followed by cortical GM abnormality (27.5%, 61/222) (Table 2). Significant differences in the prevalence of cerebral WM abnormality across BPD severity groups were observed under the 2001 NICHD, 2018 NICHD, and 2019 Jensen criteria (P=0.002, 0.004, and 0.009, respectively) (Table 2; Tables S4,S5). The msGBI was present in 12.6% (28/222) of infants. A significant overall association was identified between BPD severity and the incidence of msGBI only under the 2019 Jensen criteria (P=0.007) (Table 2).
Table 2. Brain injury prevalence per BPD severity (2019 Jensen definition).
| Variable | Total (n=222) | Non (n=138) | Grade 1 (n=19) | Grade 2 (n=56) | Grade 3 (n=9) | P value |
|---|---|---|---|---|---|---|
| Cerebral WM abnormality | 156 (70.3) | 108 (78.3) | 12 (63.2) | 31 (55.4) | 5 (55.6) | 0.009 |
| Focal signal abnormality | 30 (13.5) | 18 (13.0) | 2 (10.5) | 7 (12.5) | 3 (33.3) | 0.36 |
| Cystic lesions | 32 (14.4) | 19 (13.8) | 2 (10.5) | 7 (12.5) | 4 (44.4) | 0.11 |
| Myelination delay | 179 (80.6) | 121 (87.7) | 15 (79.0) | 38 (67.9) | 5 (55.6) | 0.002 |
| Dilated lateral ventricles | 40 (18.0) | 22 (15.9) | 2 (10.5) | 12 (21.4) | 4 (44.4) | 0.13 |
| Volume reduction | 21 (9.5) | 12 (8.7) | 1 (5.3) | 4 (7.1) | 4 (44.4) | 0.02 |
| Cortical GM abnormality | 61 (27.5) | 37 (26.8) | 4 (21.1) | 14 (25.0) | 6 (66.7) | 0.057 |
| Signal abnormality | 9 (4.1) | 4 (2.9) | 1 (5.3) | 1 (1.8) | 3 (33.3) | 0.006 |
| Gyral maturation | 0 | 0 | 0 | 0 | 0 | – |
| Increased extracerebral space | 57 (25.7) | 35 (25.4) | 3 (15.8) | 14 (25) | 5 (55.6) | 0.19 |
| Deep GM abnormality | 29 (13.1) | 16 (11.6) | 2 (10.5) | 8 (14.3) | 3 (33.3) | 0.30 |
| Signal abnormality | 2 (0.9) | 1 (0.7) | 0 | 0 | 1 (11.1) | – |
| Volume reduction | 28 (12.6) | 16 (11.6) | 2 (10.5) | 8 (14.3) | 2 (22.2) | 0.69 |
| Cerebellum abnormality | 13 (5.9) | 9 (6.5) | 1 (5.3) | 1 (1.8) | 2 (22.2) | 0.10 |
| Signal abnormality | 13 (5.9) | 9 (6.5) | 1 (5.3) | 1 (1.8) | 2 (22.2) | 0.10 |
| Volume reduction | 1 (0.5) | 1 (0.7) | 0 | 0 | 0 | – |
| msGBI | 28 (12.6) | 17 (12.3) | 1 (5.3) | 5 (8.9) | 5 (55.6) | 0.007 |
Data are presented as n (%). BPD, bronchopulmonary dysplasia; GM, gray matter; msGBI, moderate to severe global brain injury; WM, white matter.
Risk factors for moderate-to-severe global brain injury
Univariable analysis
Univariable analysis revealed significant associations between msGBI and several neonatal clinical factors (Table 3). When BPD was classified according to the 2019 Jensen criteria, a distinct threshold effect—in contrast to a dose-response relationship—was observed between BPD severity and the risk of msGBI. Specifically, compared with the corresponding no-BPD group, only grade 3 BPD demonstrated a significantly elevated risk of msGBI (OR =8.49, 95% CI: 2.2–34.53, P=0.002), whereas neither grade 1 nor grade 2 BPD showed a significant association. This non-linear pattern was further substantiated by a non-significant result in the overall trend test (P=0.20). In contrast, BPD defined by either the 2018 NICHD or 2001 NICHD criteria showed no significant association with msGBI at any severity grade (Table 3).
Table 3. Risk factors for msGBI at TEA: a Firth-corrected univariate analysis.
| Variables | OR (95% CI) | P value |
|---|---|---|
| Advanced resuscitation | 2.09 (0.95–4.69) | 0.07 |
| Sepsis | 2.73 (1.21–6.70) | 0.02 |
| Vasoactive drug | 2.76 (0.96–7.21) | 0.06 |
| Red blood cell transfusion | 2.09 (0.88–5.68) | 0.10 |
| 2001 NICHD definition | 1.11 (0.78–1.56) | 0.55 |
| Non-BPD | 1.00 | – |
| Mild BPD | 0.59 (0.21–1.59) | 0.30 |
| Moderate BPD | 0.29 (0.03–1.32) | 0.12 |
| Severe BPD | 1.42 (0.55–3.69) | 0.47 |
| 2018 NICHD definition | 1.32 (0.90–1.89) | 0.14 |
| Non-BPD | 1.00 | – |
| Grade Ⅰ BPD | 0.37 (0.04–1.56) | 0.20 |
| Grade Ⅱ BPD | 1.65 (0.58–4.27) | 0.33 |
| Grade Ⅲ BPD | 2.83 (0.78–8.98) | 0.11 |
| 2019 Jensen definition | 1.28 (0.87–1.86) | 0.20 |
| Non-BPD | 1.00 | – |
| Grade 1 BPD | 0.56 (0.06–2.47) | 0.49 |
| Grade 2 BPD | 0.74 (0.25–1.93) | 0.56 |
| Grade 3 BPD | 8.49 (2.2–34.53) | 0.002 |
BPD, bronchopulmonary dysplasia; CI, confidence interval; msGBI, moderate to severe global brain injury; NICHD, National Institute of Child Health and Human Development; OR, odds ratio; TEA, term-equivalent age.
Multivariable analysis
In the multivariable analysis, a Firth-penalized logistic regression model was employed, adjusting for GA and BW (primary model). This analysis confirmed grade 3 BPD (2019 Jensen criteria) as an independent risk factor for msGBI [adjusted OR (aOR) =8.54, 95% CI: 2.04–37.72, P=0.004] (Table 4). To assess the robustness of this association, we performed sensitivity analyses by sequentially incorporating covariates with P<0.10 from the univariable analysis into the primary model. The positive association between grade 3 BPD and msGBI remained stable and statistically significant after separate adjustment for various clinical confounders, including sepsis, vasoactive drug, red blood cell transfusion, and advanced resuscitation. Across all sensitivity models, the aORs ranged from 6.31 to 7.01, with all P<0.05 (Table S6), thereby reinforcing both the robustness and the threshold nature of this association. Application of the same modeling strategy to BPD defined by the 2018 NICHD or 2001 NICHD criteria revealed no significant association with msGBI in any model (Table 4 and Table S6).
Table 4. Independent associations between BPD diagnostic criteria and msGBI at TEA.
| Variables: model 1 (primary) | 2001 NICHD definition | 2018 NICHD definition | 2019 Jensen definition | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
| Gestational age | 0.82 (0.57–1.20) | 0.28 | 0.86 (0.60–1.21) | 0.38 | 0.87 (0.61–1.23) | 0.43 |
| Birth weight | 1.001 (0.999–1.003) | 0.37 | 1.001 (0.999–1.003) | 0.30 | 1.001 (0.999–1.003) | 0.34 |
| Diagnostic criteria for BPD | 1.15 (0.75–1.73) | 0.51 | 1.38 (0.91–2.05) | 0.13 | 1.33 (0.87–2.02) | 0.18 |
| Non-BPD | 1.00 | – | 1.00 | – | 1.00 | – |
| Mild/grade I/1 BPD | 0.55 (0.16–1.76) | 0.31 | 0.40 (0.04–1.79) | 0.26 | 0.60 (0.06–2.75) | 0.55 |
| Moderate/grade II/2 BPD | 0.29 (0.03–1.48) | 0.15 | 1.76 (0.60–4.74) | 0.29 | 0.79 (0.25–2.21) | 0.67 |
| Severe/grade III/3 BPD | 1.34 (0.37–4.78) | 0.66 | 2.91 (0.71–10.78) | 0.13 | 8.54 (2.04–37.72) | 0.004 |
BPD, bronchopulmonary dysplasia; CI, confidence interval; msGBI, moderate to severe global brain injury; NICHD, National Institute of Child Health and Human Development; OR, odds ratio; TEA, term-equivalent age.
Furthermore, the multivariable analysis evaluated the independent effects of other clinical factors on msGBI. Sepsis was consistently identified as significant independent risk factors for msGBI across all models, with aORs ranging from 2.54 to 3.09 (all P<0.05) (Table S6).
Discussion
This study assessed the association of three principal BPD diagnostic criteria with TEA cMRI, an objective neuroimaging biomarker, to identify very preterm infants at high risk for brain injury. Our analysis revealed a critical, definition-dependent disparity: only the 2019 Jensen criteria successfully stratified infants by the burden of cerebral injury, identifying a distinct threshold increase specifically in grade 3 BPD [requiring invasive mechanical ventilation (IMV)], which emerged as an independent risk factor for msGBI. In contrast, neither the 2001 NICHD nor the 2018 NICHD criteria identified a significant independent association with msGBI or a consistent risk gradient. These findings provide crucial neuroimaging evidence that challenges the interchangeability of current BPD definitions for neurodevelopmental risk prediction, and strongly suggest that the 2019 Jensen criteria may offer superior utility in stratifying high-risk infants for targeted surveillance.
The superior discriminative capacity of the 2019 Jensen criteria likely stems from its simplified architecture, using respiratory support at 36 weeks’ PMA as a proxy for disease severity. By classifying grade 3 BPD solely based on IMV need, this criterion selects infants with the most profound and sustained pulmonary dysfunction. We posit that this state of severe respiratory failure creates a convergent pathophysiological milieu conducive to cerebral injury, characterized by persistent hypoxia, hypercapnia, systemic inflammation, and hemodynamic instability. This synergy provides a coherent mechanism for the elevated brain injury risk in the grade 3 subgroup. In contrast, the 2001 NICHD framework lacks granularity in the era of non-invasive support and may dilute risk prediction by including a heterogeneous “mild BPD” population with favorable outcomes (30,31). The 2018 NICHD criteria, though more comprehensive, employs a complex algorithm that may inadvertently obscure the risk associated with the most severe physiological impairment, possibly by grouping infants with less severe compromise into the same severity strata. In essence, the Jensen criteria’s operational simplicity belies its biological precision in targeting the severity of ongoing physiological insult.
Critically, our data reveal that this Jensen-defined severity manifests as a non-linear threshold phenomenon. Injury risk escalates abruptly upon reaching continued IMV dependence at 36 weeks’ PMA, rather than following a graduated dose-response pattern. This threshold likely signifies a critical point of cumulative physiological compromise. Thus, IMV dependence serves as a surrogate for a constellation of severe, interacting insults that synergistically overwhelm the immature brain’s compensatory mechanisms. This neurostructural observation corroborates clinical studies linking only the most severe BPD with elevated NDI risk (6-10), providing an objective neuroanatomical substrate for those clinical correlations.
Furthermore, our study demonstrated that sepsis maintained a consistent and independent significant association with msGBI across all BPD diagnostic frameworks, further corroborating the multifactorial nature of msGBI risk. Previous studies have established that neonatal sepsis is independently associated with periventricular leukomalacia (15.1% vs. 4.4%) (32) and long-term neurological impairments (32,33). These data are highly consistent with our finding of sepsis as an independent risk factor. Clinically, sepsis and severe BPD (2019-Jensen grade 3) frequently coexist in critically ill preterm infants, and both conditions often necessitate similar advanced life-sustaining therapies, including IMV, vasopressor support, and red blood cell transfusion. This suggests that they may converge on shared downstream mechanisms of brain injury, driven by the interplay of systemic inflammation and hemodynamic instability. In this state of high allostatic load, systemic inflammation—exemplified by sepsis—can directly injure oligodendrocyte precursor cells and disrupt myelination, while hemodynamic instability can lead to pressure-passive cerebral circulation, thereby heightening the risk of ischemic WM injury. Concurrently, factors such as repeated transfusions further compound the burdens of oxidative stress and immune activation. These pathways may act synergistically, exacerbating inflammation and oxidative stress while impairing cerebral autoregulation, ultimately culminating in a state of global cerebral vulnerability (34-36). Therefore, our secondary finding supports, from an additional perspective, the inference that grade 3 BPD, as defined by the 2019 Jensen criteria, represents not merely severe lung disease but rather an integrative marker of the high-risk state for brain injury arising from cumulative physiological derangements, captured at 36 weeks’ PMA. The convergent association of sepsis and severe BPD with brain injury risk further strengthens the rationale for using grade 3 BPD as a composite marker for the early identification and stratified management of infants at high risk for neurological injury.
This study has several limitations. First, its single-center design necessitates validation in broader, multicenter cohorts. Second, because cMRI was not uniformly performed but adopted progressively, the included cohort was enriched for infants with lower GA, lower BW, and higher BPD rates; this selection bias may limit the generalizability of our findings. Third, although we employed Firth’s penalized-likelihood regression, the relatively small sample size and limited number of grade 3 BPD cases (n=9) precluded robust stratified analyses by GA subgroups and may affect the precision of the estimates. Although we adjusted for GA as a covariate in our multivariable models, larger cohorts are needed to examine whether the associations differ across specific GA strata. Finally, our analysis focused on structural abnormalities at TEA, and the clinical significance of these MRI findings requires corroboration with long-term neurodevelopmental outcomes.
Conclusions
In summary, this comparative analysis across three BPD diagnostic criteria demonstrates substantial differences in the profile of severe, MRI-defined brain injury depending on the definition used. The 2019 Jensen criteria uniquely identified a threshold phenomenon linking grade 3 BPD to a significantly elevated risks of msGBI—a finding not seen with other criteria. This objective neuroimaging evidence highlights the potential clinical utility of the 2019 Jensen definition for improving risk stratification and targeting neurodevelopmental surveillance in preterm infants.
Supplementary
The article’s supplementary files as
Acknowledgments
The authors are grateful to the patients and their parents for participating in this study and to the clinical staff at Shanghai Children’s Hospital for their assistance.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of Shanghai Children’s Hospital (No. 2024R034-E01). Prior to enrollment, written informed consent was acquired from the parents of all participants, permitting the use of anonymized clinical data for research and publication purposes.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0651/rc
Funding: This work was supported by Shanghai Jiao Tong University (the 2023 “Science and Technology Rejuvenating Mongolia” Action Plan Special Project) (No. 2023XYJG0001-01-09).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0651/coif). The authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0651/dss
References
- 1. Twilhaar ES, Wade RM, de Kieviet JF, et al. Cognitive Outcomes of Children Born Extremely or Very Preterm Since the 1990s and Associated Risk Factors: A Meta-analysis and Meta-regression. JAMA Pediatr 2018;172:361-7.. 10.1001/jamapediatrics.2017.5323 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Horbar JD, Greenberg LT, Buzas JS, et al. Trends in Mortality and Morbidities for Infants Born 24 to 28 Weeks in the US: 1997-2021. Pediatrics 2024;153:e2023064153.. 10.1542/peds.2023-064153 [DOI] [PubMed] [Google Scholar]
- 3. Ariyasingha NM, Oladun C, Samoilenko A, et al. Parahydrogen-Hyperpolarized Propane-d6 Gas Contrast Agent: T1 Relaxation Dynamics and Pilot Millimeter-Scale Ventilation MRI. J Phys Chem A 2025;129:4275-87.. 10.1021/acs.jpca.4c08800 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Chowdhury MRH, Oladun C, Ariyasingha NM, et al. Rapid lung ventilation MRI using parahydrogen-induced polarization of propane gas. Analyst 2024;149:5832-42.. 10.1039/d4an01029a [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Katz TA, de Ridder R, Bancalari E, et al. Prognostic Accuracy of BPD Definitions for Long-Term Outcomes in Preterm Infants: A Systematic Review. Pediatrics 2025;156:e2025070741.. 10.1542/peds.2025-070741 [DOI] [PubMed] [Google Scholar]
- 6. Li R, Wang X, Wu Y, et al. Comparative analysis of three bronchopulmonary dysplasia diagnostic criteria in predicting longterm neurodevelopmental outcomes in preterm infants. Sci Rep 2025;15:29276.. 10.1038/s41598-025-15043-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Jeon GW, Oh M, Lee J, et al. Comparison of definitions of bronchopulmonary dysplasia to reflect the long-term outcomes of extremely preterm infants. Sci Rep 2022;12:18095.. 10.1038/s41598-022-22920-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wang X, Lu YK, Wu YY, et al. Comparison of two novel diagnostic criteria for bronchopulmonary dysplasia in predicting adverse outcomes of preterm infants: a retrospective cohort study. BMC Pulm Med 2023;23:308.. 10.1186/s12890-023-02590-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Katz TA, van Kaam AH, Zuithoff NPA, et al. Association between bronchopulmonary dysplasia severity and its risk factors and long-term outcomes in three definitions: a historical cohort study. Arch Dis Child Fetal Neonatal Ed 2024;110:51-6.. 10.1136/archdischild-2024-326931 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Donlon J, Bhat V, Hunter K, et al. Impact of severity and age with variable definitions of bronchopulmonary dysplasia on neurodevelopmental outcomes. Pediatr Res 2024;96:1243-50.. 10.1038/s41390-024-03304-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Agarwal S, Tarui T, Patel V, et al. Prenatal Neurological Diagnosis: Challenges in Neuroimaging, Prognostic Counseling, and Prediction of Neurodevelopmental Outcomes. Pediatr Neurol 2023;142:60-7.. 10.1016/j.pediatrneurol.2023.02.013 [DOI] [PubMed] [Google Scholar]
- 12. Pacchiano F, Tortora M, Bordin V, et al. Radiomics in fetal brain MRI: a narrative review. Eur Radiol Exp 2026;10:30.. 10.1186/s41747-026-00697-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Perrone S, Beretta V, Tataranno ML, et al. Olfactory testing in infants with perinatal asphyxia: Enhancing encephalopathy risk stratification for future health outcomes. Neurosci Biobehav Rev 2025;169:106029.. 10.1016/j.neubiorev.2025.106029 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Kidokoro H, Anderson PJ, Doyle LW, et al. Brain injury and altered brain growth in preterm infants: predictors and prognosis. Pediatrics 2014;134:e444-53.. 10.1542/peds.2013-2336 [DOI] [PubMed] [Google Scholar]
- 15. Woodward LJ, Anderson PJ, Austin NC, et al. Neonatal MRI to predict neurodevelopmental outcomes in preterm infants. N Engl J Med 2006;355:685-94.. 10.1056/NEJMoa053792 [DOI] [PubMed] [Google Scholar]
- 16. Arulkumaran S, Tusor N, Chew A, et al. MRI Findings at Term-Corrected Age and Neurodevelopmental Outcomes in a Large Cohort of Very Preterm Infants. AJNR Am J Neuroradiol 2020;41:1509-16.. 10.3174/ajnr.A6666 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Mistry KH, Bora S, Pannek K, et al. Diagnostic accuracy of neonatal structural MRI scores to predict 6-year motor outcomes of children born very preterm. Neuroimage Clin 2025;45:103725.. 10.1016/j.nicl.2024.103725 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Cayam-Rand D, Guo T, Grunau RE, et al. Predicting developmental outcomes in preterm infants: A simple white matter injury imaging rule. Neurology 2019;93:e1231-40.. 10.1212/WNL.0000000000008172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Jobe AH, Bancalari E. Bronchopulmonary dysplasia. Am J Respir Crit Care Med 2001;163:1723-9.. 10.1164/ajrccm.163.7.2011060 [DOI] [PubMed] [Google Scholar]
- 20. Higgins RD, Jobe AH, Koso-Thomas M, et al. Bronchopulmonary Dysplasia: Executive Summary of a Workshop. J Pediatr 2018;197:300-8.. 10.1016/j.jpeds.2018.01.043 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Jensen EA, Dysart K, Gantz MG, et al. The Diagnosis of Bronchopulmonary Dysplasia in Very Preterm Infants. An Evidence-based Approach. Am J Respir Crit Care Med 2019;200:751-9.. 10.1164/rccm.201812-2348OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Donaldson RA, Freedman SF, Pittman CR, et al. Revisiting retinopathy of prematurity screening criteria to safely reduce infant stress and provider burden at an academic neonatal intensive care unit: a proof-of-concept study. J AAPOS 2026;30:104808.. 10.1016/j.jaapos.2026.104808 [DOI] [PubMed] [Google Scholar]
- 23. Shane AL, Sánchez PJ, Stoll BJ. Neonatal sepsis. Lancet 2017;390:1770-80.. 10.1016/S0140-6736(17)31002-4 [DOI] [PubMed] [Google Scholar]
- 24. Pérez-Oliver T, Pinilla-Gonzalez A, Gormaz M, et al. Necrotizing enterocolitis: risk factors and predictive modeling in a cohort of preterm infants. A case-control study. Front Pediatr 2026;14:1880723.. 10.3389/fped.2026.1880723 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Shepherd JL, Noori S. What is a hemodynamically significant PDA in preterm infants?. Congenit Heart Dis 2019;14:21-6.. 10.1111/chd.12727 [DOI] [PubMed] [Google Scholar]
- 26. Kidokoro H, Neil JJ, Inder TE. New MR imaging assessment tool to define brain abnormalities in very preterm infants at term. AJNR Am J Neuroradiol 2013;34:2208-14.. 10.3174/ajnr.A3521 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Alison M, Tilea B, Toumazi A, et al. Prophylactic hydrocortisone in extremely preterm infants and brain MRI abnormality. Arch Dis Child Fetal Neonatal Ed 2020;105:520-5.. 10.1136/archdischild-2019-317720 [DOI] [PubMed] [Google Scholar]
- 28. Drommelschmidt K, Mayrhofer T, Hüning B, et al. Incidence of brain injuries in a large cohort of very preterm and extremely preterm infants at term-equivalent age: results of a single tertiary neonatal care center over 10 years. Eur Radiol 2024;34:5239-49.. 10.1007/s00330-024-10592-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Moreira A, Noronha M, Joy J, et al. Rates of bronchopulmonary dysplasia in very low birth weight neonates: a systematic review and meta-analysis. Respir Res 2024;25:219.. 10.1186/s12931-024-02850-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Poindexter BB, Feng R, Schmidt B, et al. Comparisons and Limitations of Current Definitions of Bronchopulmonary Dysplasia for the Prematurity and Respiratory Outcomes Program. Ann Am Thorac Soc 2015;12:1822-30.. 10.1513/AnnalsATS.201504-218OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Ehrenkranz RA, Walsh MC, Vohr BR, et al. Validation of the National Institutes of Health consensus definition of bronchopulmonary dysplasia. Pediatrics 2005;116:1353-60.. 10.1542/peds.2005-0249 [DOI] [PubMed] [Google Scholar]
- 32. Bedetti L, Corso L, Miselli F, et al. Neurodevelopmental Outcome after Culture-Proven or So-Called Culture-Negative Sepsis in Preterm Infants. J Clin Med 2024;13:1140.. 10.3390/jcm13041140 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Alshaikh B, Yusuf K, Sauve R. Neurodevelopmental outcomes of very low birth weight infants with neonatal sepsis: systematic review and meta-analysis. J Perinatol 2013;33:558-64.. 10.1038/jp.2012.167 [DOI] [PubMed] [Google Scholar]
- 34. Khwaja O, Volpe JJ. Pathogenesis of cerebral white matter injury of prematurity. Arch Dis Child Fetal Neonatal Ed 2008;93:F153-61.. 10.1136/adc.2006.108837 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Drommelschmidt K, Mayrhofer T, Müller H, et al. CMRI-detected brain injuries and clinical key risk factors associated with adverse neurodevelopmental outcomes in very preterm infants. Sci Rep 2025;15:18221.. 10.1038/s41598-025-02539-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Boardman JP, Counsell SJ. Invited Review: Factors associated with atypical brain development in preterm infants: insights from magnetic resonance imaging. Neuropathol Appl Neurobiol 2020;46:413-21.. 10.1111/nan.12589 [DOI] [PMC free article] [PubMed] [Google Scholar]