Introduction

Mortality after preterm birth has significantly decreased over the past decades, but many of the surviving children suffer the long-term consequences to various organ systems [1]. This is particularly true for the lungs, which are underdeveloped at birth, which often leads to a prolonged need for respiratory support [2]. Supportive therapies may induce secondary lung injury leading to an arrest in lung development, a condition often referred to as bronchopulmonary dysplasia (BPD) [3].

Preterm-born children, especially those with BPD, are at risk of an impaired lung function leading to reduced physical activity and sports participation in childhood and adolescence [4], which in turn may contribute to other health risks such as obesity, vascular disease and osteoporosis [5–8]. The respiratory symptoms associated with preterm birth and BPD are often falsely attributed to asthma [9–11]. However, the pathophysiology is considerably different from asthma, and presumably requires a different treatment strategy [12]. Understanding and quantifying the relation between preterm birth and long-term pulmonary function is essential for (early) identification of children at high risk, better recognition of symptoms, and for developing therapeutic interventions specifically tailored to this population to optimise adult outcome.

Several cohort studies have investigated the relation between preterm birth and long-term pulmonary outcome, primarily focusing on the forced expiratory volume in 1 s (FEV1). To improve the robustness of the evidence, the results of these studies, comparing FEV1 in preterm and term-born children, have been captured in several systematic reviews [13–16]. However, these systematic reviews have important limitations. First, most of them included studies that were conducted before the introduction of exogenous surfactant and antenatal corticosteroids [13, 15–17]. These interventions are nowadays widely implemented in clinical care and both are known for their beneficial effect on neonatal respiratory outcome [18]. Therefore, these studies no longer reflect current clinical practice and including these studies in a systematic review may compromise the applicability of the results for children born in the current era. Second, previous systematic reviews have also included studies with selected populations (for example infants with BPD or asthmatic complaints), which have limited value for effect estimation in the general preterm population. Finally, previous systematic reviews have not explored effect modifiers other than BPD for a compromised lung function after preterm birth at a meta-analytic level.

The aim of the current study was to conduct a systematic review and meta-analysis on FEV1 as main pulmonary outcome in preterm children, born since the introduction of surfactant and antenatal corticosteroids. Only studies with samples representative for the general preterm population were included. Furthermore, meta-regression analysis was performed to investigate the modifying effect of perinatal and demographic variables on pulmonary outcome in preterm-born children.

Methods

This systematic review and meta-analysis was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [19]. The review protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO, identifier #CRD42020165900) [20].

Study selection

Studies were included if 1) the sample consisted of preterm-born children (<37 weeks’ gestational age); 2) children were born in 1990 or later (i.e. after the introduction of antenatal corticosteroids and surfactant); 3) age of lung function measurement was ≥5 years (considered the minimal age from which spirometry can reliably be performed); 4) outcome was assessed through spirometry and the study reported on FEV1 outcome; 5) a selected term-born control group was included, or the outcome was compared to a normative population (i.e. use of z-score); and 6) the study was published in a peer-reviewed, English-language journal. Studies were excluded if additional criteria besides gestational age and/or birthweight were used to select the sample (for example BPD, asthmatic complaints). In case of unclear age at assessment or birth year of the cohort, the authors were contacted to provide additional information.

A systematic search was set up with help from a qualified clinical librarian and performed in MEDLINE, Embase, Cochrane Library, Web of Science and Scopus using combinations of simple and hierarchical terms for “preterm children” AND “pulmonary outcome” (last search 13 September 2023; full search strategy available in the supplementary material). In case of overlapping cohorts, selection was based on the following criteria (in order of importance): 1) the study with the largest sample size; 2) the study with the longest follow-up interval (i.e. oldest age at assessment); 3) the study reporting on z-score rather than non-normalised FEV1 scores (i.e. % predicted or L); and 4) the study reporting the highest number of moderator variables. Finally, the reference lists of included articles were screened for articles meeting the inclusion criteria. Two authors (M.R. van Boven and R. Richardson) independently performed the study selection. Any disagreements were resolved by discussion or by consulting a third author (A.H. van Kaam or G.J. Hutten).

Outcomes and covariates

The following data were extracted from the included articles: 1) sample size, mean standardised or raw scores and accompanying standard deviations of lung function parameters (i.e. FEV1, forced vital capacity (FVC), FEV1/FVC ratio and forced expiratory flow at 25–75% of FVC (FEF25–75%)) for all preterm cohorts and, if applicable, control groups; 2) study characteristics, including sex, birth year and age at assessment; 3) potential (clinical) moderating variables of pulmonary outcome as listed in supplementary table e1. Data extraction was performed by one author (R. Richardson or M.R. van Boven) and carefully checked by a second author (M.R. van Boven or R. Richardson). If only the median and (interquartile) range were reported, normality of the data was checked using the method described by Shi et al. [21] before estimating the sample mean±sd by the method of Luo et al. [22] and Wan et al. [23]. Authors of studies were contacted if recalculation of the primary outcome measurement (FEV1) was not possible with the data provided in the study. Moderating variables were extracted in means, percentages or medians where appropriate.

Study quality

Study quality of the included studies was assessed by two authors (M.R. van Boven and R. Richardson) independently, using the Newcastle–Ottawa Scale for cohort studies [24]. The scale was adapted to fit the goal of this study (supplementary material). Each study was rated on a seven-point rating scale, with higher scores representing better study quality. Any disagreements were resolved through discussion or by consulting a third author (A.H. van Kaam or G.J. Hutten).

Statistical analysis

Analyses of the data were performed using Comprehensive Meta-Analysis software (version 4.0; Biostat). The standardised mean difference (Cohen's d) between pulmonary outcomes in preterm and controls or normative data was used as effect size. In case no control group was included in the study, we used normative data for spirometry (i.e. z-score with mean=0, sd=1) assuming the same sample size as the preterm sample. Effect sizes were labelled as small (d<0.5), medium (d=0.5–0.8) or large (d>0.8), according to Cohen [25], and the accompanying relative risk for adverse pulmonary outcome was used to interpret the results (scoring below lower limit of normal (LLN; z= −1.64 sd)). If studies reported data for independent subgroups, a combined effect across subgroups was computed using the built-in option in the software. Random-effects meta-analysis was performed to calculate summary estimates. The effect size of each study was weighted by the inverse of its variance to account for sample size and measurement error. Estimates and 95% confidence intervals were graphically presented using forest plots. Heterogeneity in effect sizes was quantified using I [2]. Publication bias was assessed by visually inspecting funnel plots for asymmetry and performing Egger's one-tailed test of the intercept. Random-effects meta-regression analyses were performed on each moderator variable with >10 observations to quantify their association with the studies’ effect sizes for FEV1. As sensitivity analyses, study quality, age at assessment and birth year were added in the meta-regression analysis. Furthermore, subgroup analysis was done comparing studies with and without a control group to rule out bias through using reference values. Additional sensitivity analysis utilising the one-study-removed method assessed the impact of each individual study on the overall effect.

Certainty of evidence

Certainty of evidence was assessed using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) approach as outlined in the GRADE handbook [26]. Two authors (M.R. van Boven and R. Richardson) independently assessed the certainty of evidence for each of the outcomes, starting with a “high” initial certainty regardless of study design as the risk of confounding and selection bias was incorporated in the risk of bias assessment using the Newcastle–Ottawa Scale for cohort studies [27]. Any disagreements were resolved through discussion or by consulting a third author (A.H. van Kaam or G.J. Hutten).

Results

Full-text examination revealed k=42 studies for inclusion in this meta-analysis, published between 2007 and 2023 and comprising 4743 preterm-born children. The flowchart illustrating the process of study selection is shown in supplementary figure e1. 34 studies contained a control group, representing 9843 term-born children. Median gestational age and birthweight of the mean values in the studies were 28.0 weeks (range 25.0–35.2 weeks, weighted mean 29.1 weeks) and 1050 g (range 750–2603 g, weighted mean 1318 g), respectively. Birth year of the cohorts ranged from 1990 to 2015 and mean age at assessment from 5.6 to 16.7 years (supplementary table e2 presents details of all included studies).

Study quality

The overall study quality in the included studies ranged from four to seven on a seven-point rating scale. Most studies (71%) scored a high risk of bias on adequacy of follow-up, as in these studies the number of children lost to follow-up was either not reported or high (>20%), with important potentially confounding differences between the children included and not included in the follow-up. Furthermore, some of the studies had risk of bias due to selection of the nonexposed cohort (31%), limited comparability of the cohorts (29%) or limited ascertainment of the exposure (5%). The other criteria were scored as low risk of bias for all studies. The overall risk-of-bias score is illustrated in figure 1. Details on the study quality at study level can be found in supplementary table e3.

FIGURE 1

Overall risk of bias.

Main pulmonary outcome

42 studies measured FEV1 in preterm-born children compared with controls or normative data [28–69]. The results reveal a medium-sized aggregated effect of d −0.58 (95% CI −0.69– −0.47, p<0.001; 42 studies, 4743 children, high-certainty of evidence; figure 2, table 1), implying a decrease of 0.58 sd in FEV1 in preterm-born children, leading to a three-fold increased risk of abnormal FEV1 (relative risk 2.9, 95% CI 2.4–3.4). There was strong heterogeneity in the effect sizes of the individual studies (I2=81%); however, this was largely explained by the moderator variables in the meta-regression (table 2). No relationship was found between FEV1 and study quality, age at assessment or birth year. In a sensitivity analysis with the one-study-removed method, the meta-analytic effect remained significant after iterative exclusion of every single study. The meta-analytic effect remained unchanged in the subgroup analysis using studies with a term-born control group only (d= −0.61, medium-sized effect). We found no evidence for publication bias in the visual inspection of the funnel plot (supplementary figure e2) and Egger's regression intercept (p=0.09). Certainty of evidence was upgraded due to a large magnitude of effect (relative risk 2.9) and evidence indicating a strong dose–response gradient. For details on GRADE assessment see supplementary tables e4 and e5.

FIGURE 2

Forest plot of meta-analysis of forced expiratory volume in 1 s.

TABLE 1

Summary of findings: pulmonary outcomes in preterm-born children compared to full-term-born children since the 1990s

TABLE 2

Results of univariate meta-regression analyses of the association between moderator variables and FEV1 z-scores

Meta-regression analysis

Results of all univariate meta-regression analyses are presented in table 2. Gestational age, birthweight, BPD and invasive mechanical ventilation (IMV) significantly explained heterogeneity in effect size across studies. Figure 3 displays the relationship between the four significant moderator variables and FEV1. For gestational age and birthweight, the correlation corresponds to a decrease of FEV1 of 0.10 sd for every week of lower gestation, and a decrease of 0.43 sd for every kilogram decline in birthweight. For BPD and IMV, each percentage increase in BPD or IMV rate across studies was associated with a decrease of FEV1 of 0.01 sd and 0.006 sd, respectively. Multivariate analysis was deemed not feasible, as only eight studies reported on all four significant moderator variables.

FIGURE 3

Scatterplots of univariate meta-regression analysis. BPD: bronchopulmonary dysplasia.

Secondary outcomes

All secondary outcomes were significantly lower for preterm-born children compared with controls or normative data with a small-sized effect for FVC (d= −0.33, 95% CI −0.42– −0.23, relative risk 1.9, 95% CI 1.6–2.2; 40 studies; 4629 children, moderate certainty of evidence; table 1) [29–54, 56–69]; small-sized effect for FEV1/FVC (d= −0.48, 95% CI −0.60– −0.36, relative risk 2.4, 95% CI 2.0–3.0; 32 studies, 4148 children, very low certainty of evidence) [28–39, 41–47, 50, 51, 53, 54, 59–64, 66, 68, 69]; and a medium-sized effect for FEF25–75% (d= −0.73, 95% CI −0.90– −0.56, relative risk 2.6, 95% CI 2.8–4.6; 27 studies, 3860 children, high certainty of evidence) [30, 31, 33–39, 41–47, 49–51, 54, 59, 61–63, 65, 68, 69]. The high heterogeneity (I2=76–91%) was largely explained by the moderator variables for FVC and FEF25–75% and only partly for FEV1/FVC (supplementary table e4). Evidence of publication bias was found for FEV1/FVC (supplementary figure e2). For details on GRADE assessment see the supplementary material (supplementary tables e4 and e5).

Discussion

This is the first study to report pulmonary outcome for the general preterm population born since the adoption of surfactant and antenatal corticosteroids in clinical practice. Our study shows that preterm-born children have impaired long-term pulmonary outcomes, as demonstrated by a 0.58 sd lower FEV1 score, compared with controls. In addition, meta-regression identified lower gestational age, lower birthweight, the presence of BPD, and IMV as risk factors for poorer pulmonary outcomes.

The effect size of −0.58 sd on FEV1 following preterm birth suggests that ∼72% of the preterm-born children have an FEV1 below the average of the control group, and that 15% of the preterm-born children perform below the LLN (−1.64 sd), resulting in a three-fold increase in the prevalence of an abnormal lung function. There are two large previous systematic reviews assessing the relation between prematurity and long-term lung function. Doyle et al. [17] assessed the effect of prematurity in young adolescents born in the pre-surfactant and antenatal corticosteroid era and found a stronger effect of −0.78 sd on FEV1. However, comparison with our findings is difficult, as only infants with a gestational age <32 weeks were included in that systematic review. A systematic review by Kotecha et al. [15] also included studies from both the pre- and post-surfactant era and also found a slightly stronger effect of prematurity on FEV1. Furthermore, they found an association between birth year and FEV1, showing an improvement of FEV1 in studies with later cohorts. In our study, including only cohorts over a range of 25 years after the introduction of surfactant and antenatal corticosteroids, we found no impact of birth year on FEV1, while the mean gestational age of the different cohorts remained stable. These findings suggest that the introduction of surfactant and antenatal steroids resulted in an improvement of FEV1 in preterm-born infants, but after that no significant improvement has been made over time.

The importance of the findings in our study is underscored by longitudinal studies showing that poorer lung function in preterm children persists over time, thereby never achieving their full airway growth potential [17, 70, 71]. Moreover, abnormal FEV1 in preterm children is associated with reduced physical activity [49, 72], which also appears to persist in adolescence and adulthood [73]. This may contribute to other health risks in later life, such as obesity, vascular disease and osteoporosis [5–8]. This underscores the importance of early identification of children at risk for abnormal lung function, with physical activity also serving as a potential target for intervention to reduce these later health risks, whose safety and positive effects have already been demonstrated in several other chronic lung diseases [74, 75].

Meta-regression was used to study a broad range of demographic and clinical moderator variables for pulmonary outcome. The results showed that lower gestational age and birthweight and the presence of BPD and IMV were associated with worse pulmonary outcome in terms of FEV1 (R2=36–96%). Gestational age and birthweight are highly interrelated, and they both reflect the more premature and vulnerable infant, facing a higher degree of immaturity of the pulmonary system. Our results indicate that children born at 25 weeks gestational age will face an average decrease of almost 1 sd in FEV1. It is plausible that this will be even lower in preterm children born at 22 or 23 weeks gestational age, who are increasingly being offered active treatment and show a steady increase in survival [76]. The strong relationship of BPD and gestational age with FEV1 is in accordance with previous research [13–15]. Thus far, the impact of IMV on pulmonary outcome in preterm-born children has not been investigated, and might provide new opportunities for improvement, for example by promoting a strict noninvasive respiratory support strategy in the treatment of preterm children. Unfortunately, we could not study the impact of duration of IMV as this moderator was missing in most studies. The same is true for many other moderator variables such as days on CPAP or oxygen, family asthma or socioeconomic status. Furthermore, it is unclear if the associations of significant moderator variables will hold in multivariate analysis, as they are all highly interrelated. This emphasises the need for international standards on data collection and reporting, for example according to the approach of the Chronic Airway Diseases Early Stratification collaboration [77].

In addition to FEV1, we also found impairments in other spirometry parameters, with the most profound effect on FEF25–75% with a difference of −0.73 sd between preterm and term-born children. This finding is supported by previous meta-analyses [13, 17], which also reported stronger effects on FEF25–75% compared with FEV1 in preterm populations. It has been suggested that this flow measurement, also known as the maximal mid-expiratory flow rate, reflects flow resistance in the smaller conducting airways [78]. This finding might therefore indicate that preterm children also face abnormal development of the smaller airways. However, others have questioned the value of FEF25–75% in assessing pulmonary outcome [79]. Furthermore, FEF25–75% is highly shunned due to its high variability between and within patients, and the fact that it does not add substantial information to the diagnosis of airflow obstruction based on FEV1, FVC and FEV1/FVC [79–81]. It therefore remains difficult to draw firm conclusions on what these differences in FEF25–75% actual mean in the preterm population.

One limitation of the present meta-analysis is the risk of bias in the studies, which was qualified as serious, mainly due to the high loss to follow-up in the preterm cohorts. Most of the studies lacked information on representativeness of the follow-up cohort compared to the original cohort. However, it is likely that mortality before follow-up and inability to perform spirometry, for example due to neurological comorbidity, were important reasons for this loss to follow-up. This limitation may have impacted our results, although it remains unclear to what extent. Nevertheless, GRADE evaluation of the evidence certified the evidence resulting from this meta-analysis as high for both FEV1 and FEF25–75%. A limitation of the meta-regression analyses is the number of missing details for demographic and perinatal moderator variables. This is a potential source of bias, as univariate meta-regression analyses could only be performed on a subsample of the available studies. Additionally, it also prevented us to perform a multivariate meta-regression analysis which would allow us to determine the relative contribution of the presumably interrelated individual risk factors. Strengths of our study include the careful selection of studies, resulting in a representative reflection of the general preterm population in the current surfactant and antenatal corticosteroids era. Furthermore, we expressed the result of our meta-analysis in mean standardised difference, which is superior to % predicted used in some of the previous systematic reviews. The meaning of deviation in % predicted is harder to interpret as its between-subject variability depends on the setting (sex, age, height and ethnicity). Hence, z-scores from the Global Lung Function Initiative are currently preferred over % predicted [82]. Moreover, this approach enabled the inclusion of nine additional studies that reported z-scores, but did not include a control group, by benchmarking them to the reference value.

As the limit of viability continues to decline, the survival of extreme preterm infants increases, and no significant improvement of pulmonary outcome has been made since the 1990s, the need for preventative or therapeutic interventions becomes more important. Therefore, it is worth considering implementing a structural pulmonary follow-up for all children at highest risk of pulmonary impairment (e.g. <28 weeks or a diagnosis of BPD) to identify those with abnormal lung function and possible development of chronic lung disease in adult life. This also aids future research on possible interventions, thereby increasing chances for improvement of both pulmonary and other health outcomes. Furthermore, the limitations of this study emphasise the need for international standards on follow-up of preterm infants, and data collection and reporting, aiming for increased accessibility to aggregate data (e.g. in meta-analysis and meta-regression) and thereby increase power to detect risk factors and possibilities for improved treatment strategies.

Conclusion

In this systematic review and meta-analysis, robust evidence was found for an increased risk of impaired lung function in preterm children born after the introduction of surfactant and antenatal corticosteroids. Gestational age, birthweight, BPD and IMV were identified as significant risk factors. Our findings underscore the importance of identifying strategies to further improve pulmonary outcome in preterm children.

Points for clinical practice and questions for future research

Preterm infants born in the current surfactant and antenatal corticosteroid era are at risk of impaired lung function.

Gestational age, birthweight, bronchopulmonary dysplasia and invasive mechanical ventilation are risk factors for impaired FEV1.

There is need for improvement in follow-up standards and data-collection and reporting to increase power to detect risk factors and possibilities for improved treatment strategies.