Bronchopulmonary dysplasia (BPD) is a common complication in preterm infants. This complication can lead to asthma, wheezing, limitation of activity tolerance, and recurrent low respiratory tract infections (1, 2). Infants with lower birth weight (BW) and gestational age (GA) are at risk for developing BPD. In addition, maternal smoking, chorioamnionitis, male sex, oxygen therapy, duration of mechanical ventilation, and co-morbidities, including sepsis, necrotizing enterocolitis, and pulmonary hemorrhage have been found to be associated with BPD (3–5). Epidemiological studies have reported that approximately 80% of preterm infants born between 22 and 24 weeks of gestation will develop BPD to a certain degree (6). As a result of an increased survival rate of extremely preterm infants, the incidence of BPD has increased (7, 8); however, no specific or preventive treatment has developed.

Red blood cell transfusions (RBCT) are commonly used in neonatal intensive care units. A previous survey reported that the vast majority of extremely low BW infants and 58% of preterm infants (<32 weeks of GA) received RBCT in the neonatal period (9–11). However, the effects of RBCT on short and long-term outcomes remain unclear. Recent studies have reported an association between RBCT and intra-ventricular hemorrhage, retinopathy of prematurity, and necrotizing enterocolitis (12–14). Few observational studies have found that patients with BPD receive higher RBCT volume and the number of RBCT is associated with BPD development (15–17). Vecchio et al. instituted new guidelines for administering RBCT in their neonatal intensive care unit, and the transfusion rate was concordant with a lower incidence of BPD (18). However, other studies found no relationship between BPD and RBCT (19, 20). Therefore, we aimed to conduct a systematic review and meta-analysis of the available literature to investigate the association between RBCT and BPD.

A total of 1,011 articles were retrieved from the databases using systematic and manual searches. After removing the duplicates, and performing screening of abstracts and full-text articles, 21 studies (5, 15, 16, 19, 20, 26–41) were included. These studies included a total of 6,567 control infants and 1,476 patients with BPD, and were eligible for the meta-analysis. (Supplementary Material Figure S1). The characteristics of eligible studies are presented in Table 1.

All included participants had a GA of less than 32 weeks at birth or BW of less than 1,500 g. Studies included populations from different countries, with being China and the USA being the most common. Although all studies reported the association between BPD and RBCT, their objectives varied. In half of the included studies the primary objective was to investigate risk factors of BPD including RBCT. Two studies collected data based on medical records (16, 26), and others defined BPD as neonates requiring oxygen support at 28 days of life or at 36 weeks of corrected GA. Study design included cross-sectional (n = 10), cohort (n = 9), and case–control (n = 2).

The NOS scores of the included studies ranged from 4 to 8, and detailed information is provided in the supplemental files. In all included studies, exposed and non-exposed controls were drawn from the same community. However, common weaknesses were lack of adequate confounder adjustments such as neonatal sepsis, maternal diseases during pregnancy, and outcome assessment based on medical records without a description of blind.

The pooled ORs obtained using a random-effect model showed a significant association between RBCT and BPD in both crude analysis (OR 4.01, 95% CI: 2.31–6.97) and adjusted analysis (OR 5.11, 95% CI: 3.11–8.40). There was a substantial heterogeneity between studies (I²= 56%) (Figure 1). In addition, we conducted subgroup-analyses of the number of transfusions, and the association remained significant (Figure 2).

A pooled analysis of RBCT including 227 cases and 465 controls (Supplementary Material Figure S2) found that patients with BPD were more likely to have more RBCT than those without BPD (standard mean difference 2.62, 95% CI 0.99–4.24, I² = 89%, p < 0.01).

Three studies including 255 cases and 500 controls were pooled for the total transfusion volumes. The standard mean difference of total RBCT volumes was 55.48 (95% CI: 3.82–107.4, I² = 95%, p < 0.01) between patients with BPD and controls. (Supplementary Material Figure S2).

Except for the analysis of the total transfusion volume, there were no apparent changes in the pooled results when any single study was removed. Visual examination of the funnel plots of the studies included in this meta-analysis showed no apparent publication bias.

The aim of this systematic review was to evaluate the association between RBCT and BPD based on the available evidence to provide suggestions for future investigations. The meta-analysis of the unadjusted data showed a significant association between RBCT and BPD. The odds of BPD among infants having RBCT were 4.01 times higher than the odds of those without transfusion with statistically significant adjusted estimate (OR 5.11, 95% CI, 3.11–8.40). These estimate association between RBCT and BPD for different numbers of transfusions did not differ between groups. Our study provides data on the total volume of transfusions, and the number of transfusions between BPD and non-BPD groups.

The relationship between RBCT and BPD may be explained by several mechanisms. When immature lungs are exposed to a hyperoxic environment, this could lead to a disruption of the alveolar-capillary barrier, overexpression of inflammatory mediators, vascular leakage, and cell death (42). RBCT increases adult hemoglobin in neonates, and shifts the oxygen disassociation curve to the right, increasing lung susceptibility to hyperoxia. Additionally, iron levels and other inflammatory mediators in stored blood products can promote free radical generation, infection, and fibrosis (43). Patel et al. (33) reported that very low BW infants undergoing multiple RBCT had excessive iron stores compared to non-transfused infants, and those who received more transfusions were more prone to developing BPD. This finding supports the hypothesis that free radicals and iron overload play key roles in the pathogenesis of BPD. Evidence also suggests that RBCT can potentially impact the lungs via an acute inflammatory response (28). Some neonates that received RBCT required mechanical ventilation, and the ventilator parameters could not be reduced in these patients (28). The hypothesis of transfusion related acute lung injury may be explained by this fact. The production of inflammatory cytokines and immunoactivation of the endothelium observed after the transfusion of RBCT in preterm infants may be a direct causal link between transfusion and major neonatal morbidity (44).

Investigations exploring the association between RBCT and BPD should consider several other factors based on our study. First, the number or the volume of transfusions showed substantial heterogeneity across studies, ranging from one to more than seven. We grouped the studies according to the number of transfusions performed. Although the ORs of BPD did not differ within the groups, heterogeneity was significantly reduced in each subgroup. As hypothesized, a higher number of transfusions may increase the risk of BPD. This might be explained by the small number of studies and variables that were adjusted differently between studies. The second factor that could have influenced the results on the association between RBCT and BPD may be the timing of transfusion. Duan et al. (27) investigated the risk factors for BPD and showed that early but not late anemia increased the risk of BPD. Early exposure to free radicals and iron load during this critical period would influence lung development. However, none of the included studies that described RBCT provided information about the timing of transfusion.

RBCT in neonates could be a lifesaving intervention in certain emergencies, but it is also known to cause adverse events, such as dissemination of infectious disease and inappropriate immune responses (45). Our findings raise the awareness of the risks associated with RBCT. This is the first study to systematically review pooled evidence for an association between BPD and RBCT. We used rigorous methods to search for potential studies, select studies, extract data, and pool the results to identify potential confounders. However, our study has a few limitations. The GA and BW from the included studies varied, and there was limited data to perform subgroup analysis with these parameters. Additionally, the definition of BPD differed among included studies, and misclassification of BPD could lead to possible bias. Primary outcomes in the majority of included studies did not investigate the association between BPD and RBCT, therefore, they may not have adequately controlled for potential confounders such as neonatal co-morbidities. Because more ill neonates often require closer monitoring and have more complications of prematurity and longer hospital stays, more RBCT and higher rates of BPD will be observed in these patients. Furthermore, although we performed a subgroup analysis of crude and adjusted analyses, the adjusted variables were not consistent between studies. There were insufficient data to perform meta-regression to identify potential effectors. The number of studies was lower than needed to interpret the results of subgroup analysis.

This is the most comprehensive review and analysis of the evidence on the relationship between RBCT and BPD. However, substantial heterogeneity across the body of studies was clearly identified. Many aspects, such as lack of or inadequate confounder adjustments, severity of BPD, and the number of transfusions significantly varied. The association between RBCT and BPD needs more studies to confirmit in the future. This review provides empirical evidence that demonstrating the importance of accounting for potential confounders when interpreting data in this area.