Background

ECMO is a potentially life-saving therapy for neonates suffering from severe respiratory failure. Randomized studies, especially the UK ECMO Trial, showed improved survival for neonates treated with ECMO in comparison with conventional treatment [1, 2]. First used by Bartlett et al. [3, 4] in 1975, currently more than 20,000 neonates treated with ECMO have been reported to the Extracorporeal Life Organization (ELSO), founded in 1989 [5].

Typical indications for use of ECMO are congenital diaphragmatic hernia (CDH), meconium aspiration syndrome (MAS), sepsis, pneumonia, and persistent pulmonary hypertension of the newborn (PPHN). Generally, diseases treated should be of potentially reversible character. Contraindications include gestational age <34 weeks or weight <2,000 g, coagulopathy or uncontrolled bleeding, intracranial hemorrhage exceeding grade 2, uncorrectable cardiac malformations, lethal anomalies, and irreversible brain injury [6]. Common entry criteria for neonatal ECMO according to the ELSO guidelines are listed in Table 1.

Table 1 Neonatal ECMO criteria
Full size table

During the last 10–15 years, implementation of therapies such as surfactant, high-frequency oscillatory ventilation (HFOV), and inhaled nitric oxide (iNO) have widened therapeutic options in the treatment of critically ill neonates. As a consequence, the need for ECMO has decreased for most diagnoses, excepting CDH [7, 8]. However, improvement with HFOV and iNO is frequently only temporary, and ECMO cannot be thereby avoided. Thus, administration of those therapies may lead to delay in ECMO initiation [9, 10].

This retrospective study reports on 20 years of experience in ECMO in Mannheim, particularly on changing demographics and diagnoses as well as on the impact of additional therapies. Furthermore, it states the importance of timely transfer of children born outside an ECMO center.

Methods

Data on newborns suffering from respiratory failure who received ECMO during a 20- year period (1987–2006, divided into four 5-year periods) were collected. Four diagnostic categories were constructed: CDH, MAS, sepsis and/or pneumonia, and a mixed group (“others”). The following variables were registered: inborn/outborn, age at referral, age at ECMO initiation, use of HFOV and iNO, PaO2, PaCO2, AaDO2, oxygenation index (OI) (latest values prior to ECMO) and maximum peak inspiratory pressure (PIP) prior to ECMO, ECMO duration, days on ventilator, length of hospital stay, and survival to discharge or transfer. Referral within the first 24 h of life was classified as “early referral” and afterwards as “late referral.” Age at ECMO initiation and survival of children receiving HFOV and/or iNO prior to ECMO were compared with children who had no trial of HFOV and/or iNO. HFOV was available in Mannheim since 1981, and during the first period all transferred patients were standardized to be put on HFOV as a trial prior to ECMO. With the rising number of patients with CDH (whereby we saw that nearly all with rescue trial of HFOV were put on ECMO) this standard trial was withdrawn. iNO was available since 1992 and was applied in nearly all patients prior to ECMO since 1993.

To identify factors associated with improved survival, logistic regression analysis was performed using the following variables: diagnosis, early versus late referral, and pretreatment with HFOV and/or iNO.

Continuous variables were analyzed by using the Mann–Whitney U-test (two independent samples) and the Kruskal–Wallis test (three or more independent samples). For discrete variables, the χ 2 test was adopted, and Fisher’s exact test in absence of normal distribution. Significance level was fixed at 5%. All statistical calculations were carried out by using Statistical Analysis System (SAS) software version 8.02.

Results

During the study period, 321 newborns with respiratory failure were treated with ECMO at our institution. Mean gestational age was 38.73 weeks, and mean birth weight was 3,163 g, without relevant changes over the different periods. The number of patients increased with each 5-year period (Fig. 1). CDH was the most common diagnosis (53%), followed by MAS (21%), sepsis and/or pneumonia (13%), and “others” (13%). The distribution of diagnoses changed significantly over time. In the early periods of our study, children with CDH only represented 28% of all ECMO patients, with the percentage increasing up to 75% in period 4. Until 1999, MAS had been the most common diagnosis. In the last 5 years, also absolute numbers for MAS decreased.

Fig. 1
figure 1

Changing demographics at the ECMO center Mannheim

Full size image

Duration of ECMO increased over the study period for all diagnoses. Children with CDH had significantly longer ECMO runs (193.1 ± 70.95 h) than other children (MAS: 149.75 ± 94.94 h; sepsis/pneumonia: 153.48 ± 105.91 h; “others”: 131.97 ± 83.80 h) (p ≤ 0.0016). They also needed more days on ventilator after termination of ECMO (CDH: 25.99 ± 14.74 days; MAS: 9.21 ± 6.64 days; sepsis/pneumonia: 8.32 ± 4.63 days; “others”: 12.0 ± 10.25 days) (p < 0.0001). Time on ventilator tended to increase over the study period for all diagnoses. Length of hospital stay remained relatively stable (mean 73.8 days, range 18–445 days) (Table 2).

Table 2 Blood gas and ventilatory support prior to ECMO; clinical outcome
Full size table

Values for PaO2, AaDO2, and OI prior to ECMO improved (Table 2). CDH patients showed the strongest decline in maximum PIP, while PaCO2 levels increased.

The overall survival rate was 67%, whereas best rates were seen for MAS (94%), followed by sepsis and/or pneumonia (69%), CDH (62%), and other diagnoses (43%). Both overall and diagnoses-specific mortality rates did not change significantly (Table 2).

The use of HFOV was 100% in period 1 and decreased to 61% in period 4, varying by diagnosis (CDH: 55%; MAS: 82%; sepsis/pneumonia: 90%; “others”: 57%). Children pretreated with HFOV had initiation of ECMO delayed by 34.56 h (53.52 ± 71.52 h versus 18.96 ± 21.12 h; p < 0.0001), but no difference concerning survival rate was seen (Table 3). Inhaled NO was introduced at our center in 1992, so from 1993 onwards all patients had inhaled NO trial prior to ECMO. In only 12% of these cases put on ECMO was there a temporarily response to treatment (defined as increase of paO2 levels by 15 mmHg or to above 50 mmHg). Initiation of ECMO was delayed for 37.92 h in comparison with nonresponders (75.84 ± 69.84 h versus 37.92 ± 58.56 h) (p = 0.0023), though mortality rate showed no difference (Table 3).

Table 3 Survival and pretreatment with HFOV and iNO before ECMO
Full size table

A total of 224 patients (70%) were transferred from other hospitals. In 137 cases, referral was performed during the first 24 h of life (early referral), and in 76 cases beyond that period (late referral). Eleven children were transported on ECMO (six of them during the first 24 h of life); they were added to neither of the two groups. The diagnostic mix between inborn and outborn children differed significantly: 89% of inborn neonates had CDH as compared with 37% of the outborn children. The percentage of inborn children increased from 10% to 47% over the last decade. A comparison between inborn and outborn children showed significantly higher PaO2 levels and lower OI for inborn children. The survival rate of inborn children did not differ from that of outborn children (Table 4).

Table 4 Comparison of PaO2, OI, and survival rate for inborn versus outborn and for early versus late referral
Full size table

The distribution of diagnoses in case of early referral was 42% for CDH, 38% for MAS, 13% for sepsis/pneumonia, and 7% for “other.” In contrast, the diagnoses in case of late referral were 22% for CDH, 18% MAS, 28% sepsis/pneumonia, and 32% “other” (p < 0.0001). Early referral was associated with a better survival rate: 77% of early-referred children survived, as compared with only 54% of the late-referred children (p = 0.0004). The difference could mostly be attributed to the subgroup of patients with the diagnosis CDH: the survival rate in case of early referral was 67% (39 of 58) compared with only 35% (6 of 17) in case of late referral (p = 0.02). Looking at survival rates and time of referral in other diagnoses we found a trend of better survival in early referral for sepsis/pneumonia, but this did not reach statistical significance (Table 5).

Table 5 Subgroup analysis of survival by diagnoses and time of referral
Full size table

To identify factors associated with improved survival, a logistic regression model was applied. Only MAS (compared with diagnostic category “others”) showed an association with survival (odds ratio 42; p < 0.0001). The two variables of diagnostic category CDH and early/late referral only just failed to reach statistical significance, whereas pretreatment with HFOV or iNO clearly did (Table 6).

Table 6 Factors associated with survival using logistic regression analysis
Full size table

Discussion

In contrast to the international development towards decreasing need for ECMO reported by the ELSO [5, 11, 12], our center showed increasing numbers of patients due to our high degree of specialization in ECMO and focusing on CDH patients since the year 2000. During the last 5 years, over 200 CDH patients reached our center, and we put nearly 40% of these patients on ECMO.

These changes in diagnoses, except for CDH, resulted from growing use of therapies such as surfactant, HFOV, and iNO [13]. With the availability of exogenous surfactant, ECMO treatment became dispensable in many cases, especially for RDS (last treatment in our center was in 1994) and MAS [7, 14, 15]. Use of HFOV also might be responsible for decreased need for ECMO, although this has not yet been proven in a randomized controlled trial [16]. Various studies showed reduced need for ECMO contributing to iNO, except for CDH patients [8, 17–19, 20]. This might be explained by the pathophysiology of CDH: in addition to hypoplasia of the lung, structural alteration of pulmonary vessels may play an important role. The decrease in pulmonary vascular bed and medial muscular hypertrophy lead to persistent pulmonary hypertension [21–25], which is difficult to treat. According to several studies, neither surfactant therapy nor iNO were able to reduce need for ECMO in infants with CDH [18, 21, 26]. Nevertheless, HFOV and iNO may help to stabilize patients during cannulation or transfer for ECMO.

Our data show worsening severity of illness, indicated by prolonged ECMO duration and more days on ventilator. This development was seen for all diagnoses. Neonates with CDH had significantly longer ECMO runs compared with non-CDH patients, underlining their high severity of illness. Improved values for PaO2, AaDO2, and OI do not necessarily indicate healthier children, because they probably resulted from alternative treatments (surfactant, HFOV, iNO). Respiratory values may not be suitable to express severity of underlying disease. Roy et al. [12] used the same argument to explain an increase in mean PaO2/FiO2 ratio which was found by analyzing data of the ELSO registry. Changes in maximum PIP and PaCO2 can be explained by the implementation of “gentle ventilation,” which constitutes prevention of chronic lung disease by lowering maximum PIP while accepting higher PaCO2 levels [23, 27]. Lowering mean airway pressure also contributed to increased values of OI.

In spite of a tendency towards a higher percentage of children with CDH as well as higher degree of severity of illness in general, a stable survival rate was achieved. A comparison of overall survival rate between the ELSO registry and our center is limited by differences in the diagnostic mix [5]. Our center showed lower overall survival rate (67% versus 76%) but higher percentage of patients with CDH (53% versus 24%). The survival rate for CDH patients was 62% in our center in comparison with 52% reported by the ELSO [5].

Even if additional therapies contributed to fewer transfer patients and decreasing need for ECMO, especially in MAS, negative effects in cases of poor response have to be considered. Gill and colleagues hypothesized a delay of ECMO initiation caused by use of HFOV and iNO. They reported increased mortality and morbidity for neonates with MAS if ECMO initiation was postponed to more than 96 h of age [28]. Our results show a delay in ECMO initiation for almost 35 (trial of HFOV) and 38 h (iNO responder), respectively. Survival rate, however, did not differ significantly.

Analyzing the data of the ELSO, Fliman et al. [9] found a minimally increased age at cannulation when comparing the periods 1996–1999 (2.5 days) and 2000–2003 (2.7 days). Increasing age at cannulation was also reported by Hui et al. [29], who compared three periods between 1989 and 2001 in a population of 499 neonates. In accordance with our results in patients with MAS, sepsis/pneumonia, and “other,” both studies did not find an adverse effect of this late initiation on mortality rate. Two other ECMO centers did not even note increased age at cannulation [30, 31].

Awareness of the limitations of HFOV and iNO in case of severe respiratory failure is indispensable for not missing the most appropriate moment for transfer or cannulation.

The percentage of inborn children has increased over the last decade due to improvement of prenatal diagnostic skills, thus allowing early diagnosis of CDH. This opened up the possibility for a well-prepared birth in our center, and best medical care could be offered to mother and child from the very beginning. Concerning measurements of fetal lung volume by magnetic resonance imaging, Buesing et al. [32] recently reported negative selection of severely affected CDH patients in Mannheim. It is known that centralization is a factor in evidence for better survival rates in patients with CDH [33]. The ELSO recordings show a smaller increase in percentage of patients with CDH among those treated with ECMO [2, 13]. The fact that we increased our capacity from two ECMO beds in the first time period to four beds in the last period contributes to the higher number of patients treated at our center. Despite low evidence for advantages of ECMO treatment in patients with CDH [33] and an ongoing discussion about entry criteria for ECMO treatment, we have the experience that applying the classic criteria of acute hypoxemia with PaO2 below 40 mmHg can be used for fast entry in case of preductal measurement. A high amount of ductal shunting can be interpreted as a sign of severe pulmonary hypertension, especially in combination with circulatory failure. Although the mortality rate did not differ between inborn and outborn children, the higher PaO2 levels and lower OI found in inborn neonates might indicate more stable clinical condition. Based on our data, patients with CDH who qualify for ECMO treatment should be transferred to an ECMO center within the first 24 h of life. Mortality caused by late referral does not seem to be such an issue in MAS. However, also in patients with sepsis/pneumonia, the data seem to show an advantage of early transfer. A higher risk of mortality in late transfer may occur in patients with diseases such as CDH or sepsis that are more often accompanied by additional circulatory failure and resulting multi-organ dysfunction.

We suggest that the ECMO center should be contacted when best PaO2 levels only reach 50 mmHg despite use of adequate inotropic support, intensive conventional ventilation (peak pressure 30–35 mbar, oxygen 100%), or if HFOV and iNO are unable to improve pulmonary condition within 12 h (oxygenation index higher than 30).

Boedy and colleagues analyzed 158 cases of neonates referred to their ECMO program and found a hidden mortality rate of 11.3% [34]. We speculate that early referral also contributes to a lower hidden mortality rate. The management of timely but also safe transport requires specialized transport programs. Therefore, our findings should also motivate further elaboration of such programs.

The focus on neonates with CDH is a special feature of our center. In recent years, additional treatments (HFOV and iNO) have more often been used by the referring hospitals, leading to delay in ECMO initiation. Due to this development, transportation of critically ill patients on HFOV and iNO will be necessary. Early referral (<24 h) was superior to late referral in patients with CDH, and possibly in other diseases with poor response to additional treatments. It will be a future challenge to identify these patients by registering, e.g., hemodynamic parameters, as we have discussed that circulatory failure may be responsible for the higher mortality in patients transferred at a late stage.