Extracorporeal membrane oxygenation (ECMO) was introduced in the UK in 1989 for the treatment of neonates with acute hypoxic respiratory failure (AHRF). Between 1993 and 1995 the UK Collaborative ECMO Trial was conducted,1 showing a marked survival, and survival without disability, benefit of ECMO over conventional treatment, which has been maintained to 7 years’ follow up.1 2 As a result of the UK ECMO Trial, the Department of Health established a national service for neonatal ECMO with four centres: Glasgow, Newcastle, Leicester and London.

Over the past 15 years, advanced respiratory therapies for neonates with AHRF have evolved, in particular high-frequency oscillatory ventilation (HFOV), inhaled nitric oxide (iNO) and surfactant.3^–6 Related to this, there has been worldwide reduction in the use of ECMO for neonatal ARHF, reported by the ELSO registry.7^–10

This paper aims to describe and evaluate the use of ECMO for neonatal AHRF in the UK since the service was set up in 1993. Furthermore, we set out to establish the risk factors for death in these patients.

METHODS

We included all neonates treated with ECMO for AHRF in the UK between January 1993 and December 2005. Neonates with congenital heart disease were excluded.

The following data were collected for each neonate:

• Pre-ECMO characteristics: diagnosis, sex, age (0–28 days), gestational age, maximum oxygenation index (OI), use of iNO, surfactant and HFOV. Diagnosis was categorised as one of:

  • meconium aspiration syndrome (MAS);

  • persistent pulmonary hypertension of the newborn (PPHN);

  • sepsis (bacterial or viral);

  • respiratory distress syndrome (RDS, previously known as hyaline membrane disease);

  • congenital diaphragmatic hernia (CDH);

  • “others” (including congenital lung dysplasias such as surfactant protein B deficiency, alveolar-capillary dysplasia and congenital lymphangiectasia).

• ECMO characteristics: type of ECMO used: veno-venous (VV) or veno-arterial (VA), and length of ECMO run.

• Outcome measure: survival to hospital discharge.

Statistical analysis

Summaries are expressed as median values with interquartile ranges (IQRs). For investigation of variation within the cohort, the Mann–Whitney test was used to compare skewed data between groups, and the Kruskal–Wallis test was used where there were more than two groups. For frequency data, a χ² test was used to test for association between two factors and a χ² test for trend was performed where one of the factors was ordered, for example the year of treatment with ECMO. Spearman rank correlation coefficient was used to test for a trend over time for continuous factors. Investigation of risk factors for death was performed using logistic regression analysis and fractional polynomials were used to obtain the best fitting model for total length of time on ECMO. Since the CDH group differed from the rest in terms of ECMO benefit in the UK ECMO trial1 and CDH patients are distinguishable at the time of ECMO referral, risk factors for death were analysed separately in the CDH and non-CDH patients.

RESULTS

Cohort characteristics

In total, 718 neonates (399 boys, 319 girls) were treated with ECMO for AHRF between 1993 and 2005. The median birth weight was 3.3 kg (IQR 2.9, 3.7), gestational age 40 weeks (IQR 38, 41) and patient age when ECMO was initiated was 24 h (IQR 20, 50).

The most common diagnosis leading to ECMO was MAS (n = 345; 48.0% patients), followed by CDH (n = 114; 15.9%), sepsis (n = 114; 15.9%), PPHN (n = 68; 9.5%) RDS (n = 40; 6.6%) and others (n = 37; 5.1%). The spectrum of diagnoses remained fairly constant over the study period except for RDS, where the relative proportion decreased from 9.4% in 1996 to 5.9% in 2005 (p<0.01). Overall survival to discharge was 79.7%, with significant differences in survival rates for different diagnoses (see table 1).

The degree of respiratory failure was determined by highest pre-ECMO OI: the median highest OI was 48 (IQR 35, 65); the highest OI was noted to decrease over the study period (p<0.001). There was a difference in the pre-ECMO OI of the different diagnostic groups, with CDH patients having a median OI of 53 (IQR 39, 68), MAS patients having a median OI of 48 (IQR 35, 65) and RDS having a median OI of 42 (IQR 35, 52) (p = 0.03).

With respect to ECMO support, in 413 (57.6%) patients managed on VV ECMO the survival rate was 362/413 (87.7%) compared with a survival rate of 193/263 (73.4%) among the 263 (36.8%) patients who were managed on VA ECMO. Forty (5.6%) patients needed conversion from VV to VA ECMO; the survival rate for these was only 17/40 (42.5%). The proportion of VV ECMO was the highest in MAS patients (75.4% cases) and the lowest in sepsis (38.6%) and CDH (36.3%) patients (table 1). Notably, of those needing conversion from VV to VA, 18% were patients from the “others” diagnostic group. The median length of the ECMO run was 119 h (IQR 84, 180). The length of ECMO run related to both diagnosis and type of ECMO (VV or VA) support: VV runs were significantly shorter (109 h (IQR 77, 156)) than VA runs (136 h (IQR 94, 203)) (p<0.001); the differences between the diagnoses are shown in table 1.

Risk factors for mortality

Pre-ECMO risk factors

In the non-CDH group of patients (MAS, PPHN, sepsis, RDS and others), lower birth weight, lower gestational age, older age at ECMO initiation and higher OI were all associated with increased risk of death. In the CDH patients, only lower birth weight and younger age at ECMO initiation were identified as risk factors for death. The age at which ECMO was initiated proved to be a strong risk factor for death in both non-CDH and CDH patients, but of note, in opposite directions (table 2).

ECMO-related risk factors

Both the type of ECMO and the length of ECMO run were significantly associated with outcome. In the non-CDH group, patients managed on VV ECMO had a better survival rate (90.9%) compared with those that had VA ECMO (77.8%) (p<0.001). Unsurprisingly, neonates treated on VA ECMO were sicker, for example in MAS patients on VA ECMO the OI was higher (57 (IQR 44, 80)) than in the VV group (45 (IQR 32, 61)) (p<0.001). In the CDH patients the survival rate was similar regardless of type of ECMO: survival was 58.5% in the VV ECMO group and 60.9% in the VA ECMO group (p = 0.81). Conversion from VV to VA ECMO carried a significant mortality risk for both CDH and non-CDH patients (p<0.001). The length of ECMO run had significant association with outcome in all groups of patients, showing bimodal distribution: there was a higher mortality risk with very short runs and with runs of extended length. This reflects early deaths in neonates with severe end organ injury and late deaths in those who failed to recover despite a long ECMO run (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Probability of death plotted against the duration of ECMO run in hours.

Changes in ECMO practice over time

Whilst the ECMO trial was in progress (1993–5), approximately 30 neonates with AHRF were treated with ECMO per year: this low number reflects the fact that only half were randomised to ECMO. Following the trial, the number of runs per year stabilised from 1996. During the last 2 years of the study, 2004–5, the numbers decreased slightly to 53 and 51 ECMO cases per annum, respectively (fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Number of ECMO runs by calendar year: 1989–2005.

With respect to pre-ECMO management, the use of advanced respiratory therapies (iNO, HFOV and surfactant) significantly increased over the study period. In contrast to the beginning of the study period (1993–5), when the use of iNO, HFOV and surfactant pre-ECMO was minimal, in 2005, iNO use rose to 92.2%, HFOV to 60.8% and surfactant to 62.8% (p<0.001). When the potential effects of this were reviewed, it was noted that the highest pre-ECMO OI decreased over the study period, from a median OI of 57 (49, 86) in 1993 to a median OI of 45 (IQR 36, 62) in 2005 (p<0.001). The age when ECMO was initiated did not alter over the study period (p = 0.85).

When ECMO practice and outcome were evaluated, it was noted that ECMO runs became significantly shorter over the study period (p = 0.004) and the use of VV ECMO increased significantly (p<0.001). While VV ECMO was used in only 36.7% of patients during the ECMO trial from 1993 to 1995, VV ECMO use increased to 74% between 2003 and 2005 (p<0.001). Despite this increase in the use of VV ECMO, the conversion rate from VV to VA ECMO did not change over the study period (p = 0.59). There was a trend towards increased mortality over the study period (p = 0.06) and notable annual fluctuations: while the survival rate was 82–88% between 1993 and 1998, it fell to a minimum of 71.2% in 1999, and then improved to 78.8% and 80.4% in 2004 and 2005, respectively. We analysed data from 1999 looking for reasons for the worse outcome that year and failed to identify any risk factor.

DISCUSSION

As far as we are aware this is the first detailed review of a national ECMO service. Our data show that outcome for UK patients considered by diagnosis is overall better than that reported by the international ELSO Registry apart from sepsis (ELSO 75% vs UK 62.3%) and the “others” group (64% vs 24.3%).7 The ELSO Registry reports sepsis and pneumonia (survival 59%) separately, whereas these were combined in our dataset. The narrow case definition for the “others” group in our cohort meant that this group contained only neonates with irreversible congenital lung dysplasias that have a very poor outcome.11

Several US studies have reported the widespread use of advanced respiratory therapies prior to ECMO initiation and an associated reduction in neonatal ECMO cases.8^–10 Our study confirms that these treatments are commonly used in UK ECMO candidates, although we have not seen a clear downturn in ECMO cases. There has been concern that the more widespread use of iNO and HFOV in the UK might lead to a delay in initiation of ECMO support. Given the observed relationship between older age at ECMO initiation in non-CDH neonates and higher mortality risk, which was also reported by Gill et al12 from the US, this would have been an important drawback. We were pleased to note that the increased use of advanced respiratory therapies prior to ECMO in the UK was associated with improved oxygenation (OI), but no trend towards later initiation of ECMO over time. This is in keeping with studies in term non-CDH neonates that indicate advanced respiratory therapies, particularly iNO, are associated with an improvement in OI and a reduction in need for ECMO, but have no effect on mortality.5 13

Given the significant changes in pre-ECMO management over the study period, the population treated during the UK ECMO trial (1994–5) is likely to have differed to the population treated with ECMO in 2005, and this must be considered when interpreting our data. For instance, the observed reduction in OI over time could be perceived as a lowering of the threshold for ECMO and inclusion of “less sick patients”. We consider it more likely that the reduction in OI reflects the increased use of iNO and HFOV, and that the babies needing ECMO support in the current era, were in fact sicker than those needing ECMO 10 years ago. Indeed some authors have suggested that the improvement in oxygenation observed with iNO should be used as a window for safer ECMO cannulation and that neonates treated with HFOV and iNO should be considered for ECMO at a lower OI than those treated conventionally.9 14

A trend towards increased mortality in neonatal ECMO has been observed by ELSO10 and was also suggested by our data. In the ELSO registry, this trend in mortality has been attributed to an increasing proportion of CDH babies treated with ECMO, but in the UK this was not the case since the diagnostic case mix was stable. We therefore speculate that the mortality trend in the UK is another reflection of the fact that the neonates who required ECMO despite the current aggressive pre-ECMO management strategies had a greater severity of illness.

With respect to the support approach, in the UK there has been a switch to VV ECMO in recent years. The proportion of VV ECMO in the neonatal population differs from the worldwide experience and the US experience reported by the ELSO registry, which indicate that VA ECMO still remains the predominant mode of support.7 15 Although VV ECMO has some theoretical advantages over VA (such as avoidance of instrumentation of a carotid artery) and is associated with an improved survival and reduced neurological morbidity compared with VA, the improved survival for VV ECMO is biased by the selection of less sick neonates for VV compared with VA ECMO.16 In the UK, VA ECMO is essentially reserved for infants who cannot be cannulated for VV ECMO for technical reasons or who have severe myocardial dysfunction and cardiovascular instability. It is worth noting that the patients who require VV to VA conversion have the worst outcome of all groups and it is important to note that the rate of conversion has not changed over the study period, indicating that selection of candidates for VV ECMO over VA ECMO in the UK is fairly robust.

Our data regarding risk factors for outcome is informative to clinicians considering inclusion – exclusion criteria for neonatal ECMO. For example, the higher mortality in CDH neonates needing very early ECMO initiation likely reflects deaths in neonates with severe pulmonary hypoplasia and no tolerance for mechanical ventilation from birth. These data may help the bedside clinician to recognise the CDH neonates with irreversible forms of lung hypoplasia and therefore avoid futile ECMO runs. Following the UK ECMO trial, an OI of 40 has been used as a key threshold for ECMO referral.1 The relationship between higher OI and mortality in non-CDH neonates reinforces the continued relevance of this parameter in the current era: in our study every 5-point increase in OI raised the risk of death by 5%. Furthermore, older age at ECMO initiation proved to be strongly associated with poor outcome in non-CDH patients. Similar findings have been noted in other studies particularly in patients ventilated for more than a week before ECMO.12 17 18 This underlines the importance of timely ECMO referral in the non-CDH patients.

The length of ECMO run had important association with outcome in both CDH and non-CDH patients, showing bimodal distribution. We observed a similar relationship between ECMO run duration and outcome in cardiac ECMO cases, where early deaths were due to severe end organ damage and later deaths were predominantly due to failure to recover.19 In this neonatal cohort, most of the early deaths within 24 h of ECMO initiation were due to withdrawal of ECMO support in patients in whom severe brain injury became apparent. Longer ECMO runs were also associated with lower survival rates, for example, the estimated survival rate after 2 weeks on ECMO dropped to 57% for non-CDH and 42% for CDH patients. In our experience the failure to wean from ECMO after several weeks of support usually signifies irreversible lung pathology. While the benefit of continuing ECMO support has to be carefully judged on individual basis, in those patients who fail to come off ECMO in an expected time frame every attempt should be made to reach the correct diagnosis and aggressively treat the underlying pathology.

CONCLUSION

The UK neonatal ECMO service achieves good outcomes when compared with international results, with the overall survival to discharge rate reaching 80%. Advanced respiratory therapies are used widely in UK ECMO patients. Identification of higher OI and older age at ECMO initiation as two risk factors in the non-CDH patients reinforces the importance of timely referral for ECMO.