School-age lung function after preterm birth: early-life determinants in a paediatric pulmonology cohort with comparison with NHANES

Abstract

Introduction:

Lung-function outcomes among preterm-born children referred for pulmonology care are highly heterogeneous, and early determinants remain incompletely defined. We aimed to identify early-life factors associated with school-age lung function after preterm birth and to examine whether school-age spirometry patterns relate to subsequent lung-function trajectories.

Methods:

Early-life analyses (n=511) examined predictors of FEV₁ and FVC z-scores at ages 6-8 years using multivariable regression. Spirometry-based phenotypes included prematurity-associated preserved-ratio impaired spirometry (pPRISm), prematurity-associated obstructive lung disease (POLD), and prematurity-associated dysanapsis. Longitudinal mixed-effects models assessed lung-function trajectories from ages 8-21 years. Parallel analyses were conducted in NHANES.

Results:

Lower weight-for-height at 4 years was non-linearly associated with both FEV₁ and FVC at school age. Invasive ventilation in the first year of life was associated with lower FEV₁ (β −0.62, 95% CI −0.96 to −0.28) and FVC (β −0.54, 95% CI −0.92 to −0.18). In exploratory analyses, pPRISm was inversely associated with neighborhood income (RR 0.63 per SD increase, 95% CI 0.46-0.88), and POLD was associated with invasive ventilation (RR 8.07, 95% CI 3.33-19.5). Similar subtypes and associations were observed in NHANES. School-age pPRISm was associated with progressive FEV₁ z-score decline (age x pPRISm β −0.08 SD/year, 95% CI −0.15 to −0.02), while POLD was associated with improving FVC z-score (age x POLD β 0.13 SD/year, 95% CI 0.07-0.20).

Discussion:

Early-childhood growth, neonatal respiratory exposures, and school-age spirometry patterns help stratify long-term pulmonary risk among preterm-born children referred for pulmonary care.

Introduction

Preterm birth disrupts coordinated airway, alveolar, and vascular development and remains a major determinant of lifelong respiratory morbidity. [1, 2] Population-based studies demonstrate lower mean forced expiratory volume in one second (FEV₁) and forced vital capacity (FVC) among individuals born preterm compared with term-born peers, with deficits evident by school age and often persisting into adulthood. [1, 3–9] Despite this, the early-life factors that shape distinct spirometric patterns and lung-function trajectories remain incompletely defined. [8–11]

Within the framework of prematurity-associated lung disease (PLD), spirometry-based phenotypes such as prematurity-associated obstructive lung disease (POLD), preserved-ratio impaired spirometry (pPRISm), and prematurity-associated dysanapsis have been described and may reflect distinct developmental pathways. [12] These phenotypes have been reported in selected paediatric cohorts and parallel adult spirometric patterns, but their determinants and prognostic value among preterm-born children requiring subspecialty pulmonary care remain areas of active investigation. [2, 12–16]

We conducted a retrospective cohort study of children born preterm and followed at a tertiary paediatric pulmonology center. Our primary objective was to identify early-life clinical, demographic, and growth-related factors associated with lung function at school age. Our secondary objective was to characterize school-age spirometric subtypes and examine their association with longitudinal lung-function trajectories through adolescence. To contextualize these findings beyond a clinically referred cohort, we also examined National Health and Nutrition Examination Survey (NHANES) data to assess whether similar spirometric patterns are present in the general paediatric population. [17] The goal of this work is to inform risk stratification within paediatric pulmonology practice rather than to estimate population-average effects of prematurity.

Methods

a. Population

We conducted a retrospective cohort study at the Children’s Hospital of Philadelphia (CHOP) including children evaluated by the Division of Pulmonary and Sleep Medicine from January 1, 2010, onward with documented gestational age <37 weeks and at least one spirometry test between ages 6 and 21 years. Demographics, birth anthropometrics, growth measures, and problem-list diagnoses were extracted, together with indicators of invasive ventilation in the first year of life and systemic corticosteroid bursts from ages 0-4 years (14-day lockout between prescriptions). Children with conditions expected to independently alter lung function (e.g., cystic fibrosis, genetic syndromes, interstitial lung disease, neuromuscular disorders, oncologic diagnoses, prior transplant) were excluded. Spirometry results were converted to Global Lung Function Initiative (GLI) race-neutral z-scores. [18] Analyses were restricted to spirometry performed at age ≥6 years; implausible values (z < −5 or >5) were excluded. All spirometry values analysed were pre-bronchodilator measurements.

Spirometric subtypes were defined as pPRISm (FEV₁ < lower limit of normal [LLN; z-score −1.645] with preserved FEV₁/FVC), POLD (FEV₁ and FEV₁/FVC < LLN), and pDysanapsis (FEV₁/FVC < LLN with preserved FEV₁), all remaining tests were assigned to the reference (Normal) category. BPD status was defined by clinician-documented diagnosis in the problem list. ZIP codes were linked to median household income. [19] Ethical approval was not required as the study was deemed exempt under category 4 by the CHOP IRB (protocol 24-022076).

b. Determinants of school-age lung function

The primary analysis examined determinants of FEV₁ and FVC z-scores at ages 6-8 years. When multiple spirometry tests were available within this window, the first test was used to avoid preferential selection of higher values. Because alveolar development extends through early childhood, World Health Organization weight-for-height at approximately 4 years (±185 days; wfh_4y) served as a marker of nutritional status during this period. [20] Children were included if birth-weight data, growth measures enabling wfh_4y calculation, and school-age spirometry were available.

c. Determinants of lung function across the paediatric range

To examine whether early spirometric phenotypes were associated with subsequent lung-function trajectories, we identified children with at least one spirometry assessment between ages 6 and 8 years and at least one additional assessment after age 8 years. For longitudinal analyses, BMI-for-age z-scores were calculated using CDC growth references. [21]

d. Correlation to nationally representative cohort

To assess whether similar spirometry-based patterns occur beyond a clinically referred preterm cohort, we analysed NHANES 2007-2012 data among children aged 6-8 years with acceptable spirometry quality control and complete covariates. [17] Spirometric subtypes were constructed using GLI z-scores. Birth weight, race/ethnicity, and poverty-income ratio were available. Gestational age was not consistently available and was not used.

e. Statistical analysis

Analyses were performed in R (version 4.4.1). [22] For the CHOP cohort, analyses were performed separately for three aims, determinants of school-age FEV₁ and FVC at ages 6-8 years, determinants of school-age spirometric subtype at ages 6-8 years, and longitudinal lung-function trajectories from ages 8-21 years.

The first FEV1 and FVC z-scores measured between ages 6-8 years were treated as co-primary endpoints; a Bonferroni-corrected two-sided significance threshold of α=0.025 was applied.. All remaining analyses, including spirometric subtype and longitudinal trajectory analyses, were considered secondary or exploratory and were interpreted primarily on the basis of effect estimates, confidence intervals, and consistency of patterns across related analyses.

For primary analyses, we fit multivariable linear regression models for FEV₁ and FVC z-scores. Covariates were selected a priori based on clinical relevance and included perinatal factors (birth weight, gestational age, BPD, invasive ventilation), early-childhood morbidity and growth (systemic corticosteroid bursts ages 0-4 years; wfh_4y), demographic variables (recorded race, ethnicity, neighborhood median household income), and birth year. [1, 2, 4–6, 8, 9] Birth weight and wfh_4y were modeled using natural cubic splines (3 degrees of freedom) to allow non-linear associations; inference was based on joint Wald tests and visualisation of predicted values across the observed range.

Given multiple mutually exclusive nominal categories, determinants of spirometric subtypes at ages 6-8 years were examined using multinomial logistic regression with Normal spirometry as the reference category and the same covariates.

Longitudinal trajectories were evaluated using linear mixed-effects models with random intercepts for participants. In these models, the time variable was age at spirometry in years. Age was modeled linearly so that subtype-by-age interaction terms could be interpreted as differences in annual change in lung-function z-score. Fixed effects included school-age spirometric subtype, age, the subtype-by-age interaction term, BMI-for-age z-score at the time of spirometry, BPD, invasive ventilation in the first year of life, systemic corticosteroid exposure from ages 0-4 years, recorded race, ethnicity, neighbourhood median household income, and birth year.

Missing data were handled using complete-case analysis for each model; analytic sample size therefore differed across analyses according to the availability of required covariates. Because complete-case restriction may introduce selection bias, we report characteristics of the analytic subgroup relative to excluded children.

Model diagnostics included residual plots, influential observation checks, variance inflation factors for collinearity, and heteroscedasticity-robust (HC3) standard errors. For mixed-effects and multinomial models we evaluated residuals, fitted trajectories, and estimate stability. Because gestational age and birth weight were correlated, we performed sensitivity analyses excluding gestational age from the primary models. Additional sensitivity analyses included adjustment for an EHR-derived atopy proxy and restriction to children born at <32 weeks’ gestation.

In NHANES, analyses incorporated survey weights scaled for pooled cycles and accounted for strata and primary sampling units. [23–27] Associations of birth weight, race/ethnicity, and poverty-income ratio with spirometric subtype were evaluated using survey-weighted multinomial regression.

Results

a. Population

We identified 1,959 children born preterm who completed at least one spirometry assessment between 6 and 21 years of age (Table 1). Slightly more than half were male (55.5%). Recorded race was “Black/African American” for 30.6% and “White” for 51.6% of participants and 95.6% were classified as non-Hispanic. Mean gestational age was 32.0 weeks (±4.1) and mean birth weight was 1,868 g (±866). Neighbourhood median household income averaged $67,402 (±28,816).

Table 1.

Population overview.

Number of patients	1,959
sex = Male (%)	1,087 (55.5)
Recorded race (%)
White	1,011 (51.6)
Black or African American	600 (30.6)
Other or NA	348 (17.8)
Ethnicity = Hispanic or Latino (%)	87 (4.4)
Recorded birth weight - grams (mean (SD))	1868 (866)
Gestational age - weeks (mean (SD))	32 (4.1)
Median household Income - USD (mean (SD))	$67,402 (28,816)
Number of spirometry measurements per patient (mean (SD))	3.84 (4.65)
First spirometry subtype between 6 and 8 years of age (%)
Normal	435 (56.6)
pdysanapsis	74 (9.6)
pPRISm	147 (19.1)
POLD	112 (14.6)

Markers of early respiratory morbidity included documented invasive ventilation in 127 children (6.5%). Among those, the mean duration of invasive ventilation in the first year of life was 53.3 (±65.2) days. Overall, 8.6% had BPD and 77.8% had asthma documented on the problem list. The mean number of systemic steroid courses in the first four years of life was 1.48 ± 2.41.

Among children with school-age spirometry (6-8 years), the first test was Normal in 56.6%, while pPRISm, pDysanapsis, and POLD patterns were present in 19.1%, 9.6%, and 14.6%, respectively (43.3% abnormal overall). More severe prematurity and lower birth weight were enriched among children with pPRISm and POLD (Table 2).

Table 2. Population by school age spirometry pattern.

Baseline demographic and perinatal characteristics are shown for children classified into Normal, pDysanapsis, pPRISM, or POLD spirometry patterns. P-values reported are between different pollution categories and were calculated using a chi-square test for categorical variables (with continuity correction) and ANOVA for continous variables. P-values below 0.05 are bolded.

	Normal	pDysanapsis	pPRISM	POLD	p-value
Number of patients	435	74	147	112
Gestational age (Bands based on weeks)					<0.001
<28	63 (17.0)	21 (31.8)	59 (44.7)	39 (37.5)
28-32	88 (23.8)	15 (22.7)	25 (18.9)	28 (26.9)
33-34	69 (18.6)	11 (16.7)	21 (15.9)	13 (12.5)
35-36	150 (40.5)	19 (28.8)	27 (20.5)	24 (23.1)
Recorded birth weight (Bands based on grams)					<0.001
<1000	51 (14.7)	17 (27.4)	59 (43.7)	37 (38.1)
1000-1499	43 (12.4)	11 (17.7)	20 (14.8)	18 (18.6)
1500-2499	151 (43.5)	19 (30.6)	38 (28.1)	26 (26.8)
>2500	102 (29.4)	15 (24.2)	18 (13.3)	16 (16.5)
Recorded race (%)					<0.001
White	261 (60.0)	43 (58.1)	48 (32.7)	40 (35.7)
Black or African American	86 (19.8)	15 (20.3)	69 (46.9)	49 (43.8)
Other or NA	88 (20.2)	16 (21.6)	30 (20.4)	23 (20.5)
Ethnicity = Hispanic or Latino (%)	21 (4.8)	2 (2.7)	7 (4.8)	1 (0.9)	0.251
Sex = Male (%)	254 (58.4)	38 (51.4)	90 (61.2)	64 (57.1)	0.566
Median household Income - USD (mean (SD))	69,597 (28,544)	71,149 (27,224)	54,765 (23,269)	66,140 (29,474)	<0.001
Diagnosis of BPD (%)	28 (6.4)	7 (9.5)	29 (19.7)	22 (19.6)	<0.001
Diagnosis of Asthma (%)	354 (81.4)	65 (87.8)	123 (83.7)	96 (85.7)	0.44

For early-life determinants of lung function, we analysed a sub-cohort of 511 children with complete birth and anthropometric data and spirometry at 6-8 years (Supplementary table 1). Compared with excluded children, this sub-cohort was born earlier, had lower birth weight, higher prevalence of BPD and asthma, and a greater proportion of abnormal spirometric patterns, consistent with greater respiratory morbidity.

b. Early life spirometry determinants - Primary analysis

In multivariable models, wfh_4y was non-linearly associated with both FEV₁ and FVC z-scores at first spirometry between 6-8 years (overall p < 0.001 for both; Fig. 1). Children who received invasive ventilation in the first year of life had 0.62 z-score lower FEV₁ (p < 0.001) and 0.54 z-score lower FVC (p = 0.004) compared with those who were never invasively ventilated. Children with recorded race “Black/African American” had markedly lower lung function compared with those with recorded “White” race, with effect sizes of −0.86 z-scores for FEV₁ (p < 0.001) and −0.85 z-scores for FVC (p < 0.001). Children in the “Other/Unknown” race category also showed significantly lower FEV₁ (β = −0.44, p = 0.005). Systemic steroid courses from ages 0-4 years were associated with a modest reduction in FEV₁ (β = −0.07, p = 0.01), although the association with FVC (β = −0.06, p = 0.025) did not meet the Bonferroni-corrected significance threshold. Neither BPD nor neighbourhood median household income reached statistical significance.

Figure 1. Association between weight-for-height at four years of age and school-age lung function among children born preterm.

(A) Estimated association between weight-for-height (WFH) z-score at four years and FEV₁ z-score at the first spirometry (ages 6-8 years), adjusted for birth weight (spline), gestational age, BPD, early-life steroid bursts, invasive ventilation, race, ethnicity, neighbourhood median household income, and birth year. (B) Corresponding adjusted effect estimates for FVC z-score. Curves represent fitted values from multivariable linear regression models using natural cubic splines for WFH; shaded bands indicate 95% confidence intervals. Marks on the x-axis denote individual measurements.

Birth weight showed a borderline association with FEV₁ and a significant non-linear association with FVC, whereas gestational age was not associated with either outcome. Given collinearity, subsequent analyses excluded gestational age(Supplementary Figs. 1-2). Additionally, adjustment for atopy or restriction to children born at <32 weeks’ gestation did not materially alter the results (Supplementary Results).

c. Early life determinants of spirometry patterns

We evaluated early-life predictors of spirometric subtype at ages 6-8 years using multinomial models (Fig. 2). pPRISm was associated with socioeconomic and growth-related factors. Each standard deviation increase in neighbourhood median household income was associated with a lower relative risk of pPRISm versus Normal (relative risk ratio [RR] 0.63, p = 0.006). Birth weight and weight-for-height at 4 years were both non-linearly associated with pPRISm, with predicted probabilities indicating higher risk at lower birth weight and a U-shaped association across the distribution of early-childhood weight-for-height. Children with recorded “Black or African American” race had higher relative risk of pPRISm versus Normal compared with those with recorded “White” race (RR 1.99, p = 0.041). Other early-life characteristics, including BPD, ethnicity, invasive ventilation, and systemic steroid exposure in the first four years, were not independently associated with pPRISm in adjusted models.

Figure 2. Predicted probability of pPRISm and POLD spirometry subtype by weight-for-height at age 4.

(A) Predicted probability of pPRISm according to weight-for-height z-score at approximately 4 years of age (wfh_4y), estimated from the multinomial logistic regression model with Normal spirometry as the reference category and adjusted for neighbourhood median household income, birth weight, race, ethnicity, BPD, systemic steroid exposure in the first four years, invasive ventilation, and birth year. The curve demonstrates a U-shaped association, with higher predicted probability of pPRISm at both low and high wfh_4y compared with mid-range values. (B) Predicted probability of POLD spirometry pattern as a function of wfh_4y from the same adjusted multinomial model. The POLD curve shows decreasing predicted probability with higher wfh_4y, indicating greater risk of POLD among children with lower early-childhood weight-for-height z-scores. For both panels, all other covariates are held at representative values (single “typical” child), and wfh_4y is displayed across its observed range.

In contrast, POLD was most strongly associated with markers of early respiratory morbidity. Lower birth weight was associated with higher POLD risk, and weight-for-height at 4 years showed a positive, non-linear association, with higher predicted probability of POLD at lower early-childhood weight-for-height values (Fig. 2). Recorded “Black or African American” race was also associated with increased POLD risk compared with recorded “White” race (RR 4.87, p < 0.001).

Few early-life determinants were clearly associated with pDysanapsis. Lower birth weight was associated with higher risk, but other estimates were imprecise.

d. Early Spirometry subtypes and lung trajectory

Among 349 children with spirometry at ages 6-8 and at least one later test (1,375 assessments, Supplementary table 2), spirometric subtype at school age was associated with both subsequent lung-function level and trajectory through adolescence.

As expected, baseline FEV₁ and FVC differed across subtypes (Fig. 3). Relative to children with normal spirometry, those with early POLD had substantially lower FEV₁ and FVC (β −1.87, p<0.001 and β −2.03, p<0.001 respectively), while those with pPRISm had lower FVC and borderline lower FEV₁ (β −1.11, p=0.001 and β −0.70, p=0.06 respectively). Children with pDysanapsis had higher FVC with preserved FEV₁ (β 0.77, p=0.027 and β 0.16, p=0.68 respectively).

Figure 3. Predicted FEV₁ z-score trajectories from school-age through adolescence by early spirometry subtype among preterm-born children.

Lines show model-based mean FEV₁ (A) and FVC (B) z-scores from a linear mixed-effects model including age (linear), early spirometry subtype (normal, pdysanapsis, pPRISm, POLD), and their interaction, adjusted for BMI-for-age z-score, race, ethnicity, bronchopulmonary dysplasia, systemic corticosteroid bursts by age four, history of invasive ventilation in the first year of life, neighbourhood median household income, and birth year. Shaded bands represent 95% confidence intervals.

Subtype-by-age interaction terms demonstrated divergent longitudinal trajectories. Children with early pPRISm showed a progressive decline in FEV₁ relative to peers with normal spirometry (age x pPRISm β −0.08 SD per year, p=0.015), whereas the FEV₁ slope for children with early POLD did not differ significantly from the normal group. For FVC, early POLD was associated with a steeper increase over time (age x POLD β 0.13 SD per year, p<0.001), while trajectories for pPRISm and pDysanapsis did not differ significantly from normal.

Although both pPRISm and POLD were associated with declining FEV₁/FVC over time, these patterns reflected distinct component trajectories. Namely, progressive FEV₁ decline in pPRISm and disproportionate FVC catch-up in POLD (Supplementary Results, Supplementary Fig. 3).

a. Correlation to nationally representative cohort

To assess whether similar spirometric patterns are present beyond a tertiary pulmonology cohort, we analysed NHANES 2007-2012 data among children aged 6-8 years with acceptable spirometry quality control (unweighted n = 1,394). Using GLI-based definitions, the same spirometry patterns were observed, with weighted prevalence of 87.2% Normal, 2.7% pPRISm-like, 2.1% POLD-like, and 8.1% pDysanapsis-like patterns.

Among participants with available birth weight data (n = 714), higher birth weight was associated with lower relative risk of POLD-like patterns (RR per kg 0.30, p < 0.001) and pDysanapsis-like patterns (RR 0.69, p = 0.036). The association between birth weight and pPRISm-like patterns did not reach statistical significance (RR 0.68, p = 0.16).

Children recorded as non-Hispanic Black had higher relative risk of both pPRISm-like (RR 10.1, p = 0.017) and POLD-like patterns (RR 6.95, p = 0.014) compared with non-Hispanic White peers. Poverty-income ratio was not associated with spirometric subtype risk after multivariable adjustment (all p > 0.52).

Overall, the presence of these spirometric patterns and the direction of key associations for birth weight and race/ethnicity paralleled findings in the preterm-born pulmonology cohort.

Discussion

In this clinically referred cohort of children born preterm, we found that early-childhood growth and neonatal respiratory exposures were associated with lung function at school age, and that spirometric phenotypes identified at ages 6-8 years were associated with distinct longitudinal trajectories through adolescence. These findings highlight substantial heterogeneity in pulmonary outcomes among preterm-born children referred for pulmonology care and suggest that early-life characteristics and school-age spirometry provide complementary information for risk stratification within subspecialty care.

Wfh_4y demonstrated a robust, non-linear association with later lung function. Although causal inference is not possible, and lower weight-for-height may reflect comorbid conditions, this observation is consistent with the concept that nutritional status during a critical period of airway and alveolar development influences attained lung volume. [1, 28] Importantly, such deficits may be underestimated when relying solely on height-corrected spirometry z-scores. Nevertheless, the presence of pPRISm, pDysanapsis, and POLD phenotypes in our cohort aligns closely with the spirometric subtypes described by Cousins et al., reinforcing the notion that these patterns represent robust, recurring features of prematurity-associated lung disease. [12]

School-age spirometric subtypes were also clinically informative. Children exhibiting pPRISm had marked deficits in FEV₁ and FVC at baseline and showed progressive widening of their FEV₁ deficit over time, consistent with impaired lung-growth trajectories and conceptually aligned with adult PRISm phenotypes. [2, 13, 29] The U-shaped association between wfh_4 and pPRISm risk suggests heterogeneity within this phenotype. Lower wfh_4 may reflect impaired somatic and pulmonary growth in early life, whereas higher values may reflect disproportionate weight gain or early adiposity influencing lung mechanics or dysanaptic growth. Because wfh_4 is an indirect nutritional proxy that does not capture body composition, these findings should be interpreted as hypothesis-generating.

The risk-factor profile we observed aligns with adult data linking lower BMI, nutritional deficits, socioeconomic adversity, and environmental exposures to PRISm. [14, 29] In contrast, those with early POLD had the lowest absolute spirometry values but demonstrated partial catch-up in FVC over time. This pattern is consistent with infant and early-childhood data from Friedrich et al., who reported normal or improving lung volumes but persistently reduced forced expiratory flows in preterm-born children, consistent with altered airway development in the setting of preserved parenchymal growth. [30] Together, these findings suggest at least two developmental pathways after preterm birth including a low-growth pathway associated with pPRISm, and an early-injury airway-predominant pathway associated with POLD. In contrast, pDysanapsis showed weaker associations with early-life exposures, suggesting etiologic heterogeneity and possible overlap with milder obstructive disease.

In complementary NHANES analyses, similar spirometric patterns were observed at lower prevalence, and associations with birth weight and race/ethnicity were directionally concordant. These findings indicate that spirometry-based phenotypes and their correlates are detectable beyond a tertiary pulmonology setting, although NHANES cannot distinguish prematurity-associated disease.

Observed differences in lung function by recorded race should be interpreted as reflecting social and environmental, rather than biological, determinants. Children from minority racial groups experience disproportionately higher exposure to ambient air pollution and other adverse environmental conditions, which likely contribute to the lung function deficits seen at the population level. [31–38] While we adjusted for recorded race and neighbourhood income, these remain imperfect proxies for structural and environmental factors shaping respiratory health.

Several limitations warrant consideration. First, this was a retrospective, single-center study conducted in a tertiary paediatric pulmonology clinic using clinically obtained spirometry and EMR-derived exposures. Residual confounding by unmeasured socioeconomic, environmental, neonatal, and treatment-related factors is likely, and causal inference is not possible. The cohort represents a convenience sample of children with ongoing pulmonary needs and does not include a term-born control group, therefore, results should not be interpreted as estimating population-average effects of prematurity. Second, although our cohort is large for a preterm-born population with longitudinal spirometry, some analyses were constrained by modest sample sizes and yielded imprecise estimates, particularly for less common spirometric subtypes and subgroups. The relatively low proportion of children with documented BPD may have limited power to detect independent effects of BPD after adjustment for birth weight, invasive ventilation, and other factors. Third, weight-for-height at 4 years is an imperfect proxy for early-life nutritional status and cannot distinguish nutritional insufficiency from chronic illness, healthcare utilization patterns, or genetic determinants of growth. In addition, enteral feeding dependence and aspiration-related morbidity are clinically important determinants of long-term pulmonary outcomes after prematurity but could not be reliably captured from structured EHR data in this cohort and were not included as covariates. Similarly, neighbourhood median household income and recorded race/ethnicity may not capture more proximal determinants such as household stress, food insecurity, caregiver education, or specific environmental exposures. Fourth, spirometry subtypes assigned at ages 6-8 years may be influenced by technique variability or transient illness, although the strong links to longitudinal trajectories suggest that they capture stable physiologic differences for many children. Residual measurement error due to variable effort may contribute to misclassification of spirometric subtypes, which would be expected to bias associations toward the null. In addition, our spirometry-based definition of pDysanapsis may reflect true airway-parenchymal disproportion in some children but may also capture mild or evolving obstructive disease. Definitive discrimination of true pDysanapsis from early POLD will likely require imaging-based airway phenotyping, which was not available in this dataset. Similarly, spirometry is simply not possible for some of the sickest patients followed in subspecialty care. Finally, follow-up extends only through adolescence, and it remains unknown whether the observed spirometric trajectories in this subspecialty cohort converge with adult PRISm or COPD pathways or whether early intervention could modify these trajectories.

Despite these limitations, our findings have implications for subspecialty providers caring for children with a history of preterm birth. First, early-childhood growth, as reflected by weight-for-height, appears to be a clinically accessible marker that stratifies long-term pulmonary risk in preterm-born children referred for pulmonology care. This suggests that optimizing nutritional status and linear growth during the critical window of airway and alveolar development may represent a promising target for intervention. Second, the resemblance between school-age pPRISm in our cohort and the low-growth adult PRISm phenotype supports the use of early spirometry, particularly identification of pPRISm and POLD, to flag children at highest risk for persistent deficits and to prioritize closer pulmonary surveillance. Third, the POLD subtype, strongly linked to low birth weight and invasive ventilation, appears to reflect early airway injury with partial volume catch-up, suggesting that preventive strategies for this group may need to focus on minimizing neonatal and perinatal lung injury and optimizing early respiratory care, in addition to addressing ongoing exposures.

Future studies should include prospective multicenter cohorts with detailed environmental, nutritional, imaging, and genomic phenotyping to clarify causal pathways and identify modifiable targets influencing lung-function trajectories.

. For subspecialty clinicians, integrating early clinical markers with longitudinal trajectory modeling may enable a shift from reactive treatment of established lung disease to proactive modification of developmental pathways. Ultimately, this approach could help tailor surveillance and intervention strategies for preterm-born children at highest risk, with the goal of improving lifelong respiratory outcomes in this vulnerable, clinically referred population. From a life-course perspective, early identification of spirometric trajectories resembling adult PRISm or obstructive patterns may also inform transition planning to adult pulmonary care, where these phenotypes are associated with persistent impairment and increased morbidity. Recognizing high-risk trajectories before transfer may support continuity of surveillance and earlier preventive strategies.

What is already known on this topic:

On average, children born preterm have lower lung function than term-born peers, but respiratory outcomes are highly heterogeneous, and the early-life factors that shape long-term lung-function trajectories in clinically referred populations remain unclear.

What this study adds:

Early-childhood growth and neonatal respiratory exposures were strongly associated with school-age lung function, and spirometric subtypes at ages 6 to 8 years, particularly prematurity-associated preserved-ratio impaired spirometry and prematurity-associated obstructive patterns, were associated with lung-function trajectories. These subtypes and associations were also observed in NHANES, supporting their broader relevance.

How this study might affect research, practice or policy:

Routine growth monitoring and early spirometry may help stratify pulmonary risk after preterm birth and guide targeted surveillance and intervention for those most at risk.

Data availability:

Data are available from the corresponding author upon reasonable request.