INTRODUCTION
Advances in neonatal intensive care have markedly improved the survival of very preterm infants [1,2]. However, this increased survival has shifted the clinical focus toward long-term outcomes and morbidities associated with prematurity. Among these, respiratory complications remain a primary concern, as the interruption of normal in utero lung development results in structurally immature lungs and surfactant deficiency, predisposing preterm infants to abnormal lung development [3]. Although numerous perinatal risk factors—including lower gestational age, low birth weight, and inflammatory exposure—are well-known contributors to bronchopulmonary dysplasia (BPD), these factors also influence long-term respiratory health beyond the traditional diagnostic framework of BPD at 36 weeks postmenstrual age (PMA) [4,5]. Indeed, recent evidence increasingly supports the concept of ‘prematurity-associated lung disease,’ which recognizes that respiratory impairment in preterm-born individuals extends beyond those diagnosed with BPD and may persist into later childhood and adulthood [6,7].
Assessing pulmonary function at school age using objective measures such as pulmonary function tests (PFTs) is essential for evaluating the true physiological trajectory of lung development in preterm survivors. However, the specific perinatal risk factors that independently predict impaired PFT results in later childhood remain poorly characterized. Moreover, emerging evidence suggests that postnatal growth trajectories—which reflect nutritional adequacy during critical periods of lung development—may independently modulate long-term pulmonary outcomes [8,9].
This study aimed to investigate the perinatal risk factors and postnatal growth trajectories affecting long-term pulmonary function in a cohort of preterm infants. Specifically, we examined the associations of early-life perinatal factors and weight-growth trajectories, both during the initial hospitalization (from birth to 36 weeks PMA) and during the post-discharge period (from 36 weeks PMA to the time of PFT), with pulmonary function at early school age. The objective was to identify high-risk populations requiring more vigilant monitoring and targeted interventions to preserve long-term respiratory health.
MATERIALS AND METHODS
1. Study population
This longitudinal observational cohort study was derived from a previous prospective cohort study conducted at Seoul National University Children’s Hospital in 2021 [10]. The original cohort comprised 100 preterm infants born before 30 weeks of gestation or with a birth weight below 1,250 g who were admitted to the neonatal intensive care unit (NICU) between December 2013 and June 2017. Infants with major congenital anomalies or grade IV intraventricular hemorrhage were excluded from the study.
Among this initial cohort, PFTs were performed at 6 to 7 years of age to evaluate long-term respiratory outcomes. Several infants from the original cohort were ineligible for school-age PFT due to loss to follow-up, refusal to undergo further evaluation, or severe developmental impairment that precluded reliable spirometry. Accordingly, only children who completed valid school-age PFTs were included in the final analysis.
2. Data collection
Maternal and neonatal data were obtained from an established prospective cohort database that included systematically collected clinical and perinatal data. The collected variables included gestational age, birth weight, sex, mode of delivery, and multiple-gestation status, as well as maternal factors such as pregnancy-induced hypertension, histological chorioamnionitis, premature rupture of membranes, oligohydramnios, and antenatal corticosteroid exposure. Perinatal clinical information and major postnatal comorbidities, including respiratory distress syndrome, retinopathy of prematurity, patent ductus arteriosus, and necrotizing enterocolitis, were also recorded. To evaluate growth failure, changes in weight and height z-scores were calculated across three time points: birth, 36 weeks PMA, and the time of PFT [11]. Height and weight z-scores were determined using the Fenton growth charts for birth and 36 weeks PMA and the 2017 Korean National Growth Charts for the assessment at the time of PFT [12,13].
3. Pulmonary function test
Pulmonary function was assessed by spirometry. Spirometric tests were conducted using a Sensormedics VMAX 22 PFT Spirometer (Sensormedics Co.) in accordance with American Thoracic Society guidelines [14]. The measured parameters included forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), and the FEV1/FVC ratio. Z-scores for these parameters were calculated using Global Lung Function Initiative reference equations [15]. A z-score below −1.64 was considered as the lower limit of normal (LLN) [16].
4. Statistical analysis
Continuous variables are expressed as medians with interquartile ranges (IQRs), and categorical variables are expressed as percentages. In this study, the primary outcomes were school-age spirometric measurements. For the multiple linear regression analyses, the continuous outcomes were FEV1 z-score, FVC z-score, and FEV1/FVC z-score. For the multiple logistic regression analyses, the binary outcomes were FEV1 z-score below the LLN, defined as z <−1.64, and FEV1/FVC z-score below the LLN, defined as z <−1.64. The independent variables included in the multivariable models were gestational age, sex, birth weight z-score, antenatal corticosteroid exposure, and Jensen 2019 BPD severity grade as an adjustment covariate, alongside the change in weight z-score from birth to 36 weeks PMA and change in weight z-score from 36 weeks PMA to the time of PFT. Statistical significance was set at P<0.05. All analyses were performed using Python 3.13 using the statsmodels version 0.14.6 (Python Software Foundation), SciPy version 1.17.1, and scikit-learn version 1.8.0 packages.
RESULTS
Among the original cohort, five infants who died, 14 infants whose families declined participation, and nine infants who could not perform PFT due to severe developmental delay were excluded. A total of 72 children were included in this study (Figure 1). No statistically significant demographic differences between the excluded patients and the study population were observed (Supplementary Table 1). The median gestational age at birth was 27.9 weeks (IQR, 27.4 to 28.6), and the median birth weight was 940 g (IQR, 816.5 to 1,264.3) (Table 1). Chorioamnionitis and preterm premature rupture of membranes were present in 42.6% and 20.8% of infants, respectively. Small-for-gestational-age status was observed in 31.9% of patients, and 43.1% were from multiple gestations. According to the 2019 Jensen classification, 19.4% of infants had grade I BPD, 20.8% had grade II BPD, and 2.8% had grade III BPD. PFTs were conducted at a median age of 7.0 years (IQR, 6.0 to 7.1).
At the time of PFT, the median body weight and height were 22.0 kg (IQR, 18.9 to 26.4) and 119.5 cm (IQR, 116.0 to 122.0), respectively. The median weight and height z-scores were −0.5 (IQR, −1.7 to 0.6) and −0.57 (IQR, −1.1 to 0.1), respectively. The median FEV₁ z-score was −0.93 (IQR, −1.76 to −0.16), and 23 children (31.9%) had an FEV₁ z-score below −1.64. The median FVC z-score was −0.29 (IQR, −1.59 to 0.76), and the median FEV1/FVC z-score was −1.06 (IQR, −1.92 to 0.28), with 28 children (38.9%) having an FEV1/FVC z-score below −1.64 (Table 2). Three patients with missing weight z-score data at 36 weeks PMA were excluded from the regression analyses, resulting in an analytical cohort of 69 patients.
Univariate regression analysis identified several factors associated with FEV1 z-score (Table 3). In simple linear regression, birth weight z-score was positively associated with FEV₁ z-score (B=0.241; 95% confidence interval [CI], 0.076 to 0.405; P=0.005), whereas the change in weight z-score from birth to 36 weeks PMA was negatively associated with FEV₁ z-score (B=−0.340; 95% CI, −0.613 to −0.066; P=0.016). Male sex (B=−0.772; 95% CI, −1.370 to −0.173; P=0.012) and cesarean delivery (B=−0.764; 95% CI, −1.396 to −0.132; P=0.018) were also significantly associated with lower FEV₁ z-scores. For FEV₁/FVC z-score, no variable reached statistical significance in univariate linear regression.
In a multiple linear regression analysis with FEV₁ z-score as the dependent variable (n=69), lower gestational age (β=0.18, P=0.026), male sex (β=−0.57, P=0.038), lower birth weight z-score (β=0.48, P<0.001), and smaller change in weight z-score from 36 weeks PMA to PFT (β=0.19, P=0.048) were significantly associated with lower FEV1 z-score. Jensen BPD grade, antenatal corticosteroid exposure, and change in weight z-score from birth to 36 weeks PMA were not significant. The model explained 31.8% of the variance observed in FEV1 z-score (adjusted R2=0.318, F=5.45, P<0.001) (Table 4).
In multiple linear regression analysis with FVC z-score as the dependent variable (n=69), lower gestational age (β=0.24, P=0.031), lower birth weight z-score (β=0.89, P<0.001), and smaller post-discharge weight gain (β=0.37, P=0.007) were significantly associated with lower FVC z-score. The change in weight z-score from birth to 36 weeks PMA showed a borderline positive association (β=0.42, P=0.076). Male sex, antenatal corticosteroid exposure, and Jensen BPD grade were not significant. The model explained 30.1% of the variance in FVC z-score (adjusted R2=0.301, F=5.19, P<0.001).
In logistic regression with FEV₁ z-score <−1.64 as the outcome (n=69), lower gestational age (odds ratio [OR], 0.52; 95% CI, 0.32 to 0.84; P=0.007), male sex (OR, 4.74; 95% CI, 1.22 to 18.38; P=0.024), and lower birth weight z-score (OR, 0.57; 95% CI, 0.33 to 0.98; P=0.044) were significantly associated with increased odds of impaired FEV₁. The change in weight z-score from birth to 36 weeks PMA was not a significant predictor (OR, 1.66; 95% CI, 0.70 to 3.93; P=0.251). Neither Jensen BPD grade (OR, 0.93; 95% CI, 0.40 to 2.15; P=0.860) nor antenatal corticosteroid exposure (OR, 1.00; 95% CI, 0.12 to 8.03; P=0.998) was significantly associated with impaired FEV₁.
In contrast, different factors were associated with FEV1/FVC z-score. In the linear model, Jensen BPD grade was significantly associated with lower FEV₁/FVC z-score (β=−0.52, P=0.036), along with birth weight z-score (β=−0.47, P=0.005) and change in weight z-score from birth to 36 weeks PMA (β=−0.63, P=0.009). Gestational age, sex, and antenatal corticosteroid use were not statistically significant. The model had an adjusted R2 of 0.066 (Table 4). In logistic regression with FEV₁/FVC z-score <−1.64 as the outcome, Jensen BPD grade was significantly associated with increased odds of obstructive impairment (OR, 2.17; 95% CI, 1.02 to 4.63; P=0.044), as were birth weight z-score (OR, 2.39; 95% CI, 1.27 to 4.52; P=0.007) and change in weight z-score from birth to 36 weeks PMA (OR, 3.97; 95% CI, 1.64 to 9.57; P=0.002) (Table 5).
To address the possibility that longitudinal height-growth trajectories might also influence school-age pulmonary function, we conducted parallel multiple linear and multiple logistic regression analyses using height z-scores in place of weight z-scores. In multiple linear regression, birth height z-score was independently associated with FEV1 z-score (β=0.28; 95% CI, 0.03 to 0.53; P=0.030), whereas no other height-related variable was significantly associated with either FEV₁ or FEV1/FVC (Supplementary Tables 2, 3).
DISCUSSION
In this single-center longitudinal cohort study of preterm infants born before 30 weeks of gestation or with a birth weight below 1,250 g, lower gestational age, male sex, and lower birth weight z-score were independently associated with lower FEV1 z-score at school age. These findings are consistent with established perinatal risk factors for long-term respiratory impairment in preterm populations [5-7]. Furthermore, smaller postdischarge weight gain from 36 weeks PMA to the time of PFT was independently associated with lower FEV1 z-score, suggesting that postnatal growth during early childhood continues to influence school-age pulmonary function. For the FEV1/FVC ratio, Jensen BPD grade and the extent of in-hospital weight z-score change emerged as independent correlates. Interpreted alongside the FVC analysis described below, these findings are consistent with a dysanapsis-like mechanism.
The finding that approximately one-third (31.9%) of preterm-born children exhibited an FEV1 z-score below the LLN is consistent with those of previous studies reporting significant airflow limitation in this vulnerable population [17,18]. This high prevalence underscores the substantial and persistent burden of respiratory impairment that remains clinically relevant well beyond the neonatal period, suggesting that a considerable proportion of preterm survivors may enter adulthood with reduced ventilatory reserve.
Birth weight z-score was consistently and positively associated with FEV1 across the linear and logistic regression analyses, supporting the view that intrauterine growth plays a fundamental role in determining the overall scale of school-age pulmonary function, independent of gestational age. Intrauterine growth restriction may impair lung growth through nutritional deficiency, an altered hormonal milieu, and reduced airway caliber at birth [8,19]. The predominant association of gestational age, birth weight, and male sex with FEV₁, but not with the FEV₁/FVC ratio, likely reflects their primary role in establishing overall lung size during the canalicular and early saccular stages of fetal development [7]. FEV₁ is influenced by both airway function and overall lung volume, whereas FEV1/FVC captures airflow relative to lung volume and is therefore more sensitive to mismatches between airway caliber and parenchymal size [20,21].
Comparison of the independent effects of each growth variable on FEV1, FVC, and FEV₁/FVC in the multivariable models (Table 4) suggests a pattern of disproportionate growth between lung volume and airflow indices. Higher birth weight z-score was associated with larger FVC (β=0.89) and, to a smaller extent, larger FEV1 (β=0.48), producing a net decrease in the FEV1/FVC ratio (β=−0.47). The change in weight z-score from birth to 36 weeks PMA was positively, although not statistically significantly, associated with FVC (β=0.42, P=0.076), while exhibiting essentially no effect on FEV₁ (β=−0.04) and resulting in a corresponding decrease in the FEV1/FVC ratio (β=−0.63). Post-discharge weight gain was positively and significantly associated with both FEV1 (β=0.19) and FVC (β=0.37), with only a non-significant trend toward lower FEV1/FVC ratio (β=−0.21). The common feature across these three growth windows is that each unit of somatic growth was accompanied by a numerically larger increase in FVC than in FEV1. Such a pattern, in which lung volume expansion appears to exceed the proportional increase in airflow, is compatible with the concept of dysanapsis, a relative mismatch between parenchymal or somatic growth and airway growth that has been associated with reduced FEV₁/FVC in preterm-born and other populations [22,23]. Because airway branching is essentially complete by 16 to 17 weeks of gestation, whereas alveolar multiplication and parenchymal expansion continue into postnatal life [7,24], accelerated somatic growth during windows after airway branching is complete could, in principle, produce a relative excess of lung volume over airway caliber.
Jensen BPD grade was independently associated with lower FEV1/FVC ratio (β=−0.52, P=0.036), without a significant independent effect on either FEV1 or FVC after adjustment for gestational age, birth weight, and growth variables. In univariate analysis (Table 3), BPD was associated with a numerical reduction across all three spirometric outcomes, although none reached statistical significance. The residual association with FEV1/FVC that persisted in the multivariable model likely reflects BPD-specific features that are not fully captured by gestational age, birth weight, or growth trajectories. Therefore, both growth variables and BPD grade contribute to the observed reductions in FEV1/FVC; however, their patterns across FEV1 and FVC are not identical, and partly distinct biological pathways may be involved.
Several caveats should be considered when interpreting these findings. The adjusted R² of the FEV1/FVC model was modest, indicating that the included variables accounted for only a small proportion of the variance in this outcome. The logistic regression models were fitted with 22 to 28 events for seven covariates, below conventional events-per-variable thresholds; therefore, the larger ORs should be interpreted with caution. Post-discharge weight gain in our cohort showed only a non-significant trend with FEV1/FVC, whereas the dysanapsis phenotype described by Cousins et al. [22] was most strongly linked to weight gain across childhood, from 2 to 11 years of age. However, our findings should not be used to suggest that postnatal growth in the NICU should be restricted. The lower FEV1/FVC ratio associated with greater early weight gain may be attributable to a proportionally larger increase in FVC rather than to a reduction in FEV1 itself. It is therefore uncertain whether the NICU-period and childhood growth windows act through a shared mechanism or reflect distinct phenomena; this cannot be resolved using the present data. In addition, the analyses performed herein are limited to spirometric indices; confirmation of the suggested dysanapsis pattern would ideally incorporate direct measurements of airway caliber or imaging-based assessments.
Post-discharge weight gain was independently associated with FEV1 (β=0.19, P=0.048), suggesting that nutrition and somatic growth during early childhood may continue to contribute to the overall scale of lung development [25,26]. The modest effect size may explain why the association was detectable across the continuous distribution of FEV₁ z-scores but insufficient to predict LLN threshold crossing in logistic regression. Weight gain in preterm infants reflects not only nutritional intake but also the cumulative burden of neonatal and early-childhood morbidities, including BPD severity, feeding difficulties, and other comorbidities [8,9]. Post-discharge weight gain should therefore be interpreted as a composite marker of overall biological status rather than as a direct proxy for nutritional adequacy alone.
Several strengths of this study support the validity of these findings, including the use of a well-defined cohort, comprehensive neonatal data, and standardized PFTs performed in school-aged children; however, several limitations persist. The sample size was modest, and some infants were excluded due to developmental disability or loss to follow-up, which may have introduced selection bias. The evaluation was confined to objective pulmonary function measures; therefore, other aspects of respiratory outcomes, such as respiratory medication use and symptoms, were not assessed. In addition, the cohort was drawn from a single tertiary center, potentially limiting generalizability.
In summary, in a cohort of preterm survivors, lower gestational age, male sex, lower birth weight z-score, and smaller post-discharge weight gain were independently associated with lower school-age FEV1. Higher Jensen BPD grade and a smaller decline in weight z-score during hospitalization were independently associated with lower FEV1/FVC ratio. Growth variables were generally associated with larger increases in FVC than with those in FEV1, a pattern compatible with a dysanapsis-like process. Although the modest variance explained in the FEV1/FVC model and the limited agreement with prior literature on the timing of postnatal weight gain indicate that these findings should be regarded as hypothesis-generating. Because growth in preterm infants reflects both nutritional adequacy and cumulative disease burden, sustained monitoring of growth trajectories from fetal life through early childhood may nonetheless help identify children at greater risk of long-term respiratory impairment. Consequently, confirmation of these observations in larger cohorts with direct airway assessments is warranted.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
