Keywords
diaphragmatic hernia - mortality - prognostic indices - best oxygenation index
Introduction
Congenital diaphragmatic hernia (CDH) is a complex condition with an incidence of
1 to 5 per 10,000 live births in Western Europe.[1]
[2] Survival rates have slowly improved from approximately 50 to perhaps 70 to 80% in
most institutions although the proportion remains lower in those with a high proportion
of antenatally referred patients.[3]
[4]
[5]
[6]
[7] This is despite significant advances in neonatal intensive care and the widespread
availability of extracorporeal membrane oxygenation (ECMO), inhaled nitric oxide (NO),
and high frequency oscillatory ventilation (HFOV).[7]
[8]
[9] More recently, fetal intervention has been proposed to try and prevent or even reverse
lung hypoplasia. In particular, fetal endoluminal tracheal occlusion (FETO) has been
shown to ameliorate and even reverse impaired lung growth by preventing the egress
of liquid from the lung, and it is currently in use in some European centers.[10]
[11]
[12]
Prediction of survival is important in CDH to identify the infants with an impaired
prognosis, target resources, and aid parental counseling. A plethora of prenatal and
postnatal indices have been described at one time or another and include fairly nonspecific
indices of infant maturity at birth, such as gestational age and birth weight,[13]
[14]
[15] presence of other congenital anomalies,[16] and the side of the defect.[17] More sophisticated fetal ultrasound-derived features, such as presence or absence
of liver herniation,[18] thoracic position of the stomach, and the measured lung-to-head ratio (LHR),[19]
[20] have all attempted to quantify the effects of herniation on the ipsilateral lung.
More recently, direct measurements of lung volume have also been reported by fetal
magnetic resonance imaging.[21]
[22]
As soon as the infant is born, the effects of lung hypoplasia become immediately evident
and this crucial element can be quantified. Thus, Apgar scores[14]
[23] and achieved arterial blood gases (best Po2, best Pco2) on day 1 have long been evaluated individually,[24] or in combination (e.g., Wilford Hall/Santa Rosa index, CDH study group index).[25]
[26] The oxygenation index (OI) belongs to the latter group and we have reported that
the best achieved OI on day 1 of life (BOI-d1) proved to be an effective independent
predictor in single-center series of infants with antenatally detected CDH.[27] The value of BOI, either as a standalone parameter or within a set of variables,
has also been confirmed in other series to predict survival and plan CDH surgical
repair,[28]
[29] or to predict the need for ECMO in neonates with respiratory failure.[30]
[31]
The aim of this study was to evaluate the predictive value of BOI-d1 in CDH patients
recruited from four disparate European centers.
Materials and Methods
A retrospective analysis of outcome was performed in all infants with CDH born between
2000 and 2009 in four European tertiary pediatric surgery centers following ethical
approval from their institutional review boards. The centers included Hospital for
Children and Adolescents, Helsinki, Finland (no. 1), Hospital Universitario La Paz,
Madrid, Spain (no. 2); King's College Hospital, London, United Kingdom (no. 3), and
Universitaetsklinikum Graz, Austria (no. 4).
Center no. 1 and no. 4 offer ECMO for selected infants. Including criteria for ECMO
are birth weight > 2 kg, gestational age > 38 weeks, no intracranial hemorrhage, and
failed maximal conventional ventilation treatment in center no. 1, and the EURO CDH
consortium guidelines for ECMO referral in center no. 4.[8] Center no. 3 offered FETO for poor prognosis fetuses selected on the basis of LHR
(≤ 1.0) and “liver-up” position.[27] Outcomes and some elements of their morbidity (e.g., tracheomegaly) from the FETO
group in center no. 3 have previously also been reported.[15]
[32]
Prenatal ultrasound characteristics (second trimester LHR, hernial contents) were
available from centers no. 1, 2, and 3. Perinatal variables included gestational age,
birth weight, defect side, and need for prosthetic patch in those coming to surgery.
BOI-d1 was calculated from the best blood gas on day 1 of life as follows:
% Fractional inspired O2 × mean airway pressure (in cm H2O)/Pao2 (in kPa).[27]
Survival was defined as overall survival and survival to 28 days. Comparison was made
between survivors and nonsurvivors using univariate χ2 analysis. Receiver operating characteristic (ROC) curves were used to determine cutoff
values for continuous variables. Nonparametric statistical tests (e.g., Kruskal–Wallis
test, Mann–Whitney U test, Spearman rank correlation) were used to determine differences
between groups and any degree of association. Data are quoted as median (interquartile
range [IQR]), p < 0.05 was considered significant.
Results
There were 235 infants (center no. 1, n = 29; center no. 2, n = 64; center no. 3, n = 113; and center no. 4, n = 29) within the study period. Of these, 136 (58%) patients were male. An infant
from center no. 4 had a 5-day period of ECMO while 66 infants from center no. 3 had
undergone a FETO procedure. A total of 174 (74%) infants underwent operative repair
of CDH with 65 (37%) requiring a prosthetic patch. There was a significant difference
in the need for prosthetic patch across the centers (no. 1, 44%; no. 2, 14%; no. 3,
58%; no. 4, 27%; p < 0.001). Median gestational age at delivery was 38 (35–39) weeks. This varied across
the centers (p < 0.0001) with center no. 3 having a significantly lower median gestation age (36
[34–38] weeks) than all others.
Best Oxygenation Index (Day 1)
Median (IQR) BOI-d1 was 32 (18–91) and values were available in 189 infants. BOI-d1
was significantly lower in non-FETO than in FETO infants (29 [18–77] vs. 52 [21–153];
p = 0.023) and in those who underwent primary repair compared with those requiring
a prosthetic patch repair (21 [13–29] vs. 46 [22–81]; p < 0.0001). There was a low degree of inverse correlation between BOI-d1 and both
birth weight (rS = −0.12; p = 0.05) and gestational age (rS = −0.16; p = 0.01). There was a significant difference in BOI-d1 between the four centers (no.
1, 68 [29–92]; no. 2, 24 [12–47]; no. 3, 42 [22–156], and no. 4, 22 [18–41]; analysis
of variance [ANOVA] p < 0.0001) ([Fig. 1]).
Fig. 1 Best oxygenation index on day 1 of life by center. *p < 0.05, **p < 0.01.
Prenatal values for LHR were available in 83 (36%) infants and typically from the
single center no. 3. There was a significant inverse correlation between first measured
LHR and BOI-d1 (rS = −0.33; p < 0.003). These could be separated into those who then underwent intervention and
those who followed the natural course (rS = 0.22; p = 0.11 and (rS = −0.42; p = 0.03, respectively). Indeed there was no significant difference between FETO and
non-FETO infants (restricted to center no. 3 [52 {21–153} vs. 34 {22–199}; p = 0.88]).
Outcome
There were 76 (32.3%) deaths (< 28 days [n = 65], ≥28 days [n = 11]) in the series and of those, 23 (30.3%) had undergone operative repair of the
diaphragm. The overall survival (discharge from hospital) was 67.6% (n = 159) and the 28-day survival was 72.3% (n = 170). This was chosen as the preferred outcome measure and varied by center from
60 (no. 3) to 90% (no. 1) (ANOVA p < 0.0001) ([Fig. 2]).
Fig. 2 Outcome of congenital diaphragmatic hernia by center.
Prognostic Factors
[Table 1] illustrates univariate χ2 analysis for the outcome—survival to 28 days using categorical data and included
“liver-down” position at fetal ultrasound (p < 0.0001) and need for primary repair (p = 0.02). There was no difference in outcome depending on the side of the diaphragm
defect (p = 0.83).
Table 1
Categorical variables in congenital diaphragmatic hernia
|
Variable
|
n
|
28-day survival (%)
|
p-Value
|
|
Fetal variables
|
|
Liver position (down/up)
|
89/100[a]
|
88 vs. 62%
|
< 0.0001
|
|
Side (left/right)
|
198/31[b]
|
73 vs. 71%
|
0.83
|
|
Postnatal variables
|
|
Surgery (primary/patch repair)
|
109/65
|
96 vs. 88%
|
0.03
|
a Total excludes missing data; p < 0.05.
b Total excludes bilateral congenital diaphragmatic hernia or missing data; p < 0.01.
[Table 2] illustrates the values for AUROC, sensitivities, and specificities for the continuous
variables of gestational age and birth weight for overall survival and BOI-d1 both
for overall and for survival to 28 days. The most useful predictive variable (for
28-day survival) was the BOI-d1 (AUROC, 0.91) ([Fig. 3]). Using a cutoff of ≤82, the sensitivity was 91% (95% confidence interval [CI],
85–95%) and specificity was 80% (95% CI, 66–90%). Using a cutoff of < 40, the sensitivity
fell to 73% (95% CI, 65–81%) and specificity rose to 92% (95% CI, 81–98%).
Fig. 3 Receiver operating characteristic curve of best oxygenation index on day 1 of life
across four centers (n = 189). Prediction of 28-day survival.
Table 2
Continuous variables in congenital diaphragmatic hernia
|
Variable
|
Predicted survival
|
AUROC
(95% CI)
|
Cutoff
|
Sensitivity (%)
(95% CI)
|
Specificity (%)
(95% CI)
|
|
Gestational age (wk)
|
Overall
|
0.70
(0.62–0.78)
|
34
|
42
(30–55)
|
89
(84–94)
|
|
Birth weight (kg)
|
Overall
|
0.68
(0.60–0.76)
|
2.5
|
51
(38–64)
|
75
(68–82)
|
|
BOI-d1
|
Overall
|
0.88
(0.82–0.94)
|
82
|
94
(88–97)
|
73
(60–84)
|
|
40
|
75
(66–82)
|
85
(73–93)
|
|
28 d
|
0.91
(85–96)
|
82
|
91
(85–95)
|
80
(66–90)
|
|
40
|
73
(65–81)
|
92
(81–98)
|
Abbreviations: AUROC, area under the receiver operating characteristic; BOI-d1, best
oxygenation index on day 1 of life; CI, confidence interval.
Comparison of Centers
To examine whether BOI-d1 was also applicable outside of the originating center (no.
3), we repeated the prediction of 28-day survival using data from the other three
centers (no. 1, 2, and 4). This showed entirely comparable results with an AUROC of
0.94 (95% CI, 90–98%). The relevant sensitivity and specificity for a cutoff of 82
was 86 (95% CI, 71–95%) and 89 (80–94%); and a cutoff of 40 was 97 (86–99.9%) and
67 (56–77%).
To try and correct for the effect of FETO in center no. 3 when looking at BOI-d1,
center no. 3 was divided according to whether patients had undergone FETO. This showed
that the lowest BOI-d1 was in center no. 4 and the highest in center no. 1 (overall
p < 0.0001) ([Fig. 1]).
Discussion
We have previously shown in a single-center retrospective series that BOI-d1 is a
consistent, independent predictor of survival in CDH.[27] Further data from that center were combined with three other European centers and
have shown a similar robust relationship between this and the relevant early outcome
of 28-day mortality. We chose this as a preferable index and a more appropriate outcome
measure with something measured only on day 1 of life.[33] Nevertheless, the high AUROC (∼0.9) for BOI-d1 was seen with either index or outcome
and indeed was consistent even when we excluded the original center (no. 3).
There were differences, both in outcome and BOI-d1 across the European centers. The
largest center (no. 3) had the worst outcome. We attribute this in part to it having
the largest referral base for antenatally diagnosed fetuses with CDH for consideration
for FETO. Similar survival rates have been reported for other institutions with similarly
active fetal medicine programs.[34] An indirect measure of this is the need for prosthetic patches. Infants with large
diaphragmatic defects have a poorer outcome[35] and are those typically who form the referral group for fetal intervention. By contrast,
an overall survival of > 90% was achieved in one center (no. 1), and it also had one
of the highest median BOI-d1 and showed an increased rate of prosthetic patch use.
This is certainly comparable with the best reported outcomes both from Europe and
North America.[3]
[4]
[33]
Only one center offered therapeutic intervention (FETO) for selected poor prognosis
fetuses over this period beginning in January 2002. Since then, there have been increasing
referrals of poor prognosis fetuses from all over Europe and a decrease in overall
survival within the center, as previously documented by Sinha et al.[27] We believe that this selection bias was manifest in several ways. Thus, the patient
cohort of center no. 3 had a significantly lower median gestational age and significantly
higher rates of patch requirements—all due to the effects of an increasing proportion
of poor-prognosis infants surviving to be born. At 60% overall survival, center no.
3 also had the lowest survival rate and the highest number of patients that had died
more than 28 days postsurgery. Severe lung hypoplasia was even the main cause of death
in 4 out of the 10 late deaths.
Correlation between LHR, as the most standardized measurement for lung hypoplasia
and, therefore, “poor prognosis,” and BOI-d1, as predictor for ultimate outcome, was
tested and showed that there is a statistical relationship between LHR and BOI-d1.
Although not a direct aim of the study, we used the BOI-d1 to try and reflect the
change in their natural history. Thus, the BOI-d1 in center no. 3 was now similar
in the two groups (FETO and non-FETO), and broadly comparable with the other centers
despite FETO candidates starting off with lower LHR and being born at a younger gestational
age.[15] These results are consistent with our previous report and we obviously infer that
FETO improves lung function.[27]
Tan et al modified our relatively straightforward concept of “best” OI in their study
on 24 antenatally diagnosed infants from a United Kingdom center.[28] They measured OI four times in the first 24 hours and then derived an area under
the curve from it—suggesting increased predictive precision. Nonetheless, we did not
think this would be repeatable outside of a single center. They also alluded to its
use as a serial tool when measured daily. Obviously, OI accuracy as a predictive tool
for survival increases with the age of the child as those who have already died are
excluded. Nevertheless, calculation of daily BOI summarizes the improvement in gas
exchange and stability of the infant presurgery and one could envisage its use in
protocols designed to standardize the timing of surgical intervention.
Attempted prediction of survival in CDH has a long history and many indices have been,
at one time, put forward. Just showing a difference between survivors and nonsurvivors,
however, is not useful and a more valid appreciation of the value can be made by comparing
AUROC values. Thus, the congenital diaphragmatic hernia study group prediction formula,
based upon birth weight and Apgar score, has an AUROC of 0.74,[25] the Wilford Hall/Santa Rosa formula (based on best Pao2 and Paco2) has a reported AUROC of 0.87,[14] while a Korean arterial blood gas formula (Pao2 and Paco2) is slightly lower at 0.8.[25]
[36] Our BOI-d1 across the centers is consistently around 0.9. As with any clinical test,
determining the appropriate cutoff point has to be balanced by the inverse relationship
between sensitivity and specificity and we illustrate our analysis using two cutoff
points: 40 and 82.
In conclusion, this series has confirmed the validity of BOI-d1 when used in a multicenter
setting. BOI-d1 is relatively straightforward to calculate at bedside and may have
use in indicating the need for more intensive medical (e.g., HFOV) and surgical care
(e.g., prosthetic patch need) and formalizing the timing of surgery. On the other
hand, BOI-d1 can also be used to characterize the severity of a group of infants with
CDH to enable more appropriate comparison between centers.
