Our study showed that the survival rate of neonates with respiratory failure after receiving VV ECMO was 88%, higher than that (73%) of neonates with respiratory failure treated by ECMO according to ELSO registry report in January 2019 . The reason might be that the data of ELSO come from the mixed population of VA ECMO and VV ECMO, and most patients who receive VA ECMO have hemodynamic instability and need cardiac support, thus reduce the survival rate. According to the ELSO database, the survival rate of VA ECMO for neonatal respiratory failure between 2012 and 2017 was 70%, while that of VV ECMO was 80% .
Our results also showed that mortality rate of neonates in the Kugdman et al.’s study was lowest , while that in the Chevalier et al.’s study was highest . According to the ELSO database, neonates with MAS have the highest survival rate, followed by neonates with PPHN and CDH . On one hand, neonates with MAS enrolled in the Kugdman et al.’s study might have more stable respiratory status, plus new treatment modalities (NO, HFV, PS) were used and the ECMO team was more experienced at that time, thus improve the survival rate. On the other hand, in the Chevalier et al.’s study, cannula applied on neonates was small, indicating that this group of neonates were small, and ECMO equipment was not advanced at the early time, all these factors might result in the relatively high mortality of this study.
An overall survival rate of 88% was seen in the 347 neonates, higher than that of other age groups by VV perfusion according to the ELSO database. Actually, different age groups have different disease spectrum. For neonatal ECMO, the most common diagnoses are CDH, MAS, and PPHN, accounting for almost 75% of all neonatal respiratory ECMO cases . Whilst for pediatric ECMO and adult ECMO, the most common diagnoses are pneumonia and acute respiratory distress syndrome (ARDS) . Prognosis of neonates with MAS, RDS and PPHN is promising due to good response to supplemental therapies such PS and iNO. In contrast, no studies have shown the beneficial effects of surfactant for adult and pediatric ARDS, which may explain the lower survival rate of pediatric and adult ECMO for respiratory failure caused by ARDS and pneumonia. In 2017, the international ARDS collaborative group provided the first consensus definition for neonatal ARDS . However, the above studies of neonatal ECMO were performed in the pre-ARDS era, in which ARDS was usually considered as neonatal RDS. Actually, ARDS and RDS are two significant different diseases with different reactions to surfactant, and they should be diagnosed and treated independently. Besides, mortality rate is also associated with other factors such as annual hospital ECMO volume for neonates and adults, but not for pediatric cases .
In our study, complications including mechanical complications, bleeding, hypertension, seizure, and renal failure occurred during hospitalization. According to the ELSO, the most common complication of neonatal ECMO for respiratory failure is mechanical complication, such as clots in the ECMO circuit , which is consistent with our study results. Bleeding and clots complications are multifactorial. Even though an ideal test of anticoagulation for patients is lacking, continuous unfractionated heparin and close monitoring of anticoagulation are required to reduce the risk of thrombosis and hemorrhage . In our study, the rates of neurologic complications such as intracranial hemorrhage (ICH)/infarction and seizure are high as well, with 6.6% and 14.9%, respectively. When analyzing the ELSO registry report in 2016, neonates using ECMO have the highest rate of neurologic complications, with an ICH incidence of around 7.6% . Various pre-existing factors like low birth weight, acidosis, hypoxia, hypotension, and organ failure have been found to be associated with neurologic injury. Besides, some ECMO factors such as modality of ECMO, hemorrhage, seizures, and development of new organ failure increase the risk of central neural system injuries further . Therefore, understanding of risk factors associated with neonates and knowing how to deal with them are important to reduce complications. With the evolving indications for ECMO and the dramatically changed monitoring technology and supportive therapies over these years, the outcomes of patients have been improved greatly. Further attempts, such as by improving the equipment of ECMO, are needed to determine whether such events can be reduced.
Since a double-lumen catheter was designed in 1989, VV ECMO has been increasingly used in neonatal respiratory failure [27-28]. VV ECMO has a few advantages over VA ECMO. During VV ECMO, ligation of the carotid arteries is avoided, pulmonary circulation and coronary artery perfusion are maintained well, thus reduce the left ventricular afterload. Studies have showed that VV ECMO compared favorably to VA ECMO for cardiovascular support [29-30]. Some previous studies have also shown that VV ECMO was associated with lower rates of neurologic complications as compared with VA ECMO [27,31-32].
In this study, to minimize potential bias of observational study, we established inclusion and exclusion criteria strictly to provide accurate prevalence and incidence estimation, and we limited the minimum sample size of each study to 50 to reduce publication bias. Moreover, we excluded the studies published in the ELSO database to avoid data duplication and reduce selection bias, because only the selected medical centers have the chance to register in the ELSO database, which will increase selection bias. Therefore, detailed VV ECMO data of other medical centers outside the ELSO database was collected in this study.
There are some limitations in our study. Firstly, all the studies were non-RCT studies, which increased the risk of bias. Statistic quality of systematic review and meta-analysis is best assessed by RCTs. However, a pure randomized study is rare, whereas accurate studies are relatively common and provide most of the available evidence . Secondly, only studies written in English were included, which might cause language bias. Thirdly, less than 10 studies were included, and publication bias and meta regression analysis were not performed, which might pose a potential risk of publication bias. Fourthly, the number of included studies was small and there was moderate heterogeneity among the studies. Fifthly, Some data in the original study could not be obtained, such as pump type and membrane type, and the baseline standards of each study might be inconsistent, many potential factors might play a role in our analysis. Lastly, the inclusion criteria might also result in the omission of potentially important studies, such as case reports and small sample studies. However, small sample studies might be affected by publication bias, historical bias, selective reporting, and other methodological deficiencies, which increase the risk of bias.