- Email: [email protected]
Cardiac arrest before repair or extracorporeal membrane oxygenation cannulation does not increase the mortality rate associated with congenital diaphragmatic hernia
Cardiac Arrest Before Repair or Extracorporeal Membrane Oxygenation Cannulation Does Not Increase the Mortality Rate Associated With Congenital Diaphragmatic Hernia By Anita
t? Courcoulas,
Kimberly K. Reblock, Marc I. Rowe, and Henri Pittsburgh, Pennsylvania
l Despite recent advances in the management of high-risk congenital diaphragmatic hernia KDH), mortality remains high. Deaths occur later because infants with inadequate pulmonary parenchyma are treated aggressively but eventually succumb to respiratory failure. In an attempt to identify absolute predictors of mortality the authors examined retrospectively their experience with CDH to determine if cardiac arrest before repair or initiation of extracorporeal membrane oxygenation (ECMO) invariably increased mortality. The authors reviewed the charts of 119 infants who had high-risk CDH treated between 1981 and 1994. They were divided into two groups: those that suffered cardiopulmonary arrest (CA, n = 21) before CDH repair or ECMO cannulation; and those that did not (NCA, N = 98). The authors compared mortality rate, ventilatory parameters, duration of, and complications on ECMO, as well as length of hospitalization between groups. Twenty-one infants suffered CA before initiation of ECMO support or CDH repair. Three infants (14%) suffered CA before arrival at our institution; seven (33%) after, and 11 (53%) both before and after arrival. There was no difference between the CA and NCA groups in terms of birth weight, gestational age, race and gender mix, or pregnancy and delivery complications. Five-minute Apgar scores were significantly lower in the CA group compared with the NCA group (4.6 Y 5.7, P = ,041. The CA group also had significantly worse “best postductal” blood gas and ventilatory parameters. There was no significant difference in length of hospitalization, time from admission to ECMO cannulation or CDH repair, or incidence of complications while on ECMO between the two groups. CA cases were more likely to require ECMO support (76% v 48%, P = .02) and to stay on ECMO for a more prolonged period than NCA cases (5.8 ~3.8 days, P = NSI. However, there was no significant difference in overall survival between CA and NCA cases (43% v 51%, P = NS). Cardiopulmonary arrest before repair of CDH or ECMO cannulation is not a univariate independent predictor of mortality and therefore should not preclude these highrisk infants from maximum intensive care therapy, including ECMO cannulation. Copyright Q 1997 by W.B. Saunders Company
INDEX WORDS: Congenital diaphragmatic cannulation, predictors of mortality, cardiac tory failure.
hernia, arrest,
ECMO respira-
T
HE MORTALITY RATE for infants presenting at birth with congenital diaphragmatic hernia (CDH) remains high. Series from many institutions, including ours, quote overall mortality rates of approximately 50%.1.2 Respiratory distress develops at birth in the most seriously affected infants. In the past. many of these infants died shortly after birth. However, with the inception of newer modalities,3-7 such as extracorporeal membrane oxygenation (ECMO), many critically ill infants Journal
of Pediatrrc
Surgery,
Vol 32, No 7 (July),
1997:
pp 953-957
R. Ford
who have CDH initially survive, only to succumb later to progressive respiratory failure or refractory pulmonary hypertension. Recent efforts have focused on the development of specific criteria to identify those infants who have critical pulmonary hypoplasia that is incompatible with life. Before the ECMO era, Bohn et al* used ventilatory parameters to predict outcome by plotting Pacoz versus ventilation index (VI). He predicted near 100% mortality in infants with Paco2 levels greater than 30 and VI values greater than 1,000.9 Since the advent of ECMO therapy, predicting absolute mortality has become more difficult. In some infants, ECMO is able to reverse clinical deterioration and provide salvage therapy. In others ECMO may only delay the inevitable mortality. Wilson et allo have attempted to identify predictors of severity that may be useful to determine outcome in the ECMO era. These investigators reported that patients not responding to maximum conventional ventilation with best postductal oxygen saturation less than 100% and with ventilation indices greater than 1,000 were not able to be saved and did not benefit from ECMO cannulation.rO Although one would assume that cardiac arrest before CDH repair or ECMO cannulation is uniformly fatal, to date, no study has examined the effect of such an event on ultimate outcome in patients who have CDH. This study was undertaken to determine if cardiac arrest before CDH repair or ECMO cannulation invariably increases mortality. If an arrest event could be shown to be an absolute predictor of mortality, the need for aggressive theiapy, including ECMO cannulation, would be obviated in these high-risk infants. MATERIALS
AND
METHODS
The medical records of 135 Infants who had CDH treated from 198 1 to 1994 at the Children’s Hospital of Pittsburgh (CHP) were reviewed retrospectively. The following variables were analyzed: demographics,
From the Depnrhrmt of Surgery, University ofPittsburgh School of Medicine, and the Deparfment of Pediafric Surgery, Chrldren k Hospital of Pittsburgh, Prttsburgh, PA. Presented at the 1996 Annual Meetmg of the Section on Surgery of the American Academy of Pedzatncs. Boston, Massachusetts, October 26-30, 1996. Address reprrnt requests to Anita P Courcoulas, MD, 300 Kaufiann Building, 3471 Fi@ Ave, Pzftsburgh, PA 15213. Copyright o 1997 by WB. Saunders Company
0022-3468/97/3207-0003$03.00/O 953
954
COURCOULAS
length of stay, gestabonal age, birth weight, pregnancy-related comphcations, delivery mode, Apgar scores, and associated congenital anomalies. In addition, the site of the diaphragmatic hernia (left v right). need for immediate endotracheal intubation, cardiopulmonary arrest before transfer to CHP or before to CDH repair or ECMO therapy. and laboratory parameters before CDH repair (best postductal arterial blood gas, mode of ventilation, mean arterial pressure [MAP], alveolararterial oxygen tension difference, oxygenation Index, and VI) were also entered in the database. Only the high-risk infants who required endotracheal intubation, a subset of 119 of the 135 patients. formed the basis of this study. Patients were divided into two groups: those who sustamed a cardiopulmonary arrest (CA) before repair of CDH or ECMO cannulation and those that did not (NCA). All patients were candidates for ECMO once this therapy became available at CHP m 1985. Cardiopulmonary arrest events were defined as any event that resulted in cardiopulmonary collapse requiring treatment including, but not limited to, chest compression. Patients were also categorized into seven predefined repair groups depending on the timing of repair and need for ECMO therapy: (A) repair of CDH and no need for ECMO. (B) repair after ECMO therapy. (C) repair before ECMO, (D) repair during ECMO, (E) repair before transfer to CHP with and without ECMO, and (F) no repair with some ECMO support. To better define absolute predictors of mortality m our patients, we examined outcome retrospectively for CDH patients in each group in the context of a Bohn plot,* or the method proposed by Wilson (Pal v VI).t” Outcome variables collected included: length of hospital stay, need for, length of, and complications on ECMO therapy: and mortality. The data were abstracted into a database and analyzed using StatView for the Macintosh by Abacus Concepts (Berkeley, CA). Continuous variables were compared using t tests and categorical variables were compared using either x2 or Fisher’s Exact tests. The CA group was compared with the NCA group in a univariate fashion. All results are presented as percentages or as mean i SD.
RESULTS
Twenty-one of the 119 high-risk infants (17.6%) suffered a cardiopulmonary arrest (CA group) event before CDH repair or initiation of ECMO therapy. Ninety-eight (82.4%) of the 119 had no history of an arrest (NCA group). Among the 21 infants in the CA group, 3 (14%) experienced CA before arrival at our institution, 7 (33%) after arrival, and 11 (53%) experienced arrest both before and after arrival. Table 1 shows baseline variables. The only significant difference between groups was 5-minute Apgar scores, Table
I. Baseline
Variables
Pregnancy (% with
complications none)
Delivery mode vaginal) I-min Apat 5-min Apgar
NCA in = 98)
P Value
NS
57143
2886 t 659 38.1 i 2.5 7.2/90.8/2.0 52148
65
60
NS
71 3.7 i 2.4 4 6 -t 2.8 60.0
67 4 1 2 2.3 5.7 i- 2.1
NS NS .04
3089
2771
39.0 2 2.1 0/100/0
NS NS NS
(% spontaneous
Outside city referral Left-sided CDH (%)
t%)
85.7
and Ventilatory
PO2 Pco, >50 FIO, Mean
59.2 816
NS NS
airway
VI Respiratory
pressure rate
PH
Arterial
Blood
Gas
Parameters
CA (n = 21)
NCA (n = 98)
P Value
51 i 33
122 + 127
.03
60% 1.0 f 0.01
28%
.03
16.6 r 5.3
0.87 + 0.21 12.8 t 4.3
.04 .Ol
1,047 + 559 58 k 16
666 + 456 49 + 18
.03
7.23 I! 0.30
7.33 f 0.23
NS NS
which were significantly worse (4.6 v 5.7) for the CA group. There was no difference in the incidence of congenital anomalies or the need for patch repair between groups. As shown in Table 2, the CA group had significantly worse “best postductal” arterial blood gas values and ventilatory parameters than the NCA group. To define the effect of the timing of CDH repair or ECMO cannulation on ultimate outcome, we classified the patients into various subgroups (A through F, Table 3). CA patients were relatively evenly distributed among the various subgroups. In contrast, nearly half the NCA patients underwent CDH repair without requiring ECMO support (Group A). Overall, the timing of CDH repair or ECMO had no significant effect on outcome. However, CA patients who underwent CDH repair before ECMO support (Group C) had a 100% survival rate compared with NCA patients in the same group who only had a 35% survival rate. Table 4 shows a comparison of the outcome variables. CA patients required ECMO support more frequently than NCA patients (P = .02). There was a trend toward longer duration of ECMO support in the CA group. Length of hospitalization and time from admission to therapy (either CDH repair or initiation of ECMO) were comparable for both groups. NCA patients experienced fewer complications than CA patients while on ECMO, but this difference was not statistically significant (P = .08). Similarly, the survival rate was comparable for both groups. Interestingly, 8 of 16 CA patients who Table 3. Timing (Repair
by Group
CA(n = 211 Birth weight(g) Gestational age (wk) Race, black/white/other (%I Gender male/female (%)
Table 2. Best Postductal
ET AL
of Repair and Need for ECMO Group) and Survival
CAla=
NCA tn = 981
n(%) Cases
n(%) Alive
nl%l Cases
nl%) Alive
A B
4 (19.1) 2 (9.5)
1 (25) 1150)
48 (49) 4 (4.1)
33 (69) 4(100)
C D E
5 (23.7) 2 (9.5) 4 (19.1)
5 1100) 0 10) 2 (50)
23 (23.4) 9 (9.2) 8 (8.2)
8 (35) 0 (0) 5 (63)
F
4 (19.1)
0 (0)
6 (6.1)
0 (0)
*A, repair of CDH and no need for ECMO; B, reparr after ECMO therapy; C, repair before ECMO, D, repair during ECMO; E, reparr before transfer to CHP wrth and without ECMO; F, and no reparr with some ECMO support.
CARDIAC
ARREST
AND
MORTALITY
Table 4. Outcome
IN CDH
Variables CAM
955
by Group
Table 6. Cause
= 211
NCA (n = 98)
P Value
Length of hosprtal stay id) Time from admit to ECMO
21.3 + 24 4
30.8 + 44.6
NS
None DNR
or repair(h) Need for ECMO
22.6 f 30.5
20.5 i 30.1
(%)
Sepsis Arrest
(d)
48 3.8 2 5 2
NS .02
Time on ECMO
76 5.8 2 7 1 42.9 43
29.6 51
NS NS
Bleeding Pulmonaryfailure
NS
Other*
Complrcations Survival (%)
on ECMO
(%)
Cause of Death
required ECMO survived compared with 16 of 48 NCA patients (P = NS). Of note, of the five CA patients who did not receive ECMO support, only one survived. Table 5 outlines specific complications seen during ECMO. There was no significant difference in the rate or type of complications between the two groups. Causes of death in both groups are shown in Table 6. Cardiopulmonary arrest was the terminal event that accounted for 5 of 12 deaths in the CAgroup and 12 of 36 deaths in the NCA group. Overall, there was no significant difference between the two groups in death rate or cause of death. According to Bohn’s predictions, patients in the right upper quadrant (with a Pace, value greater than 40 and a VI value greater than 1,000) should have 100% mortality. Three of 21 CA patients fell in that quadrant, and they all died (Fig 1). However, one patient in the NCA group who also fell in the right upper quadrant survived. Fig 2 shows the correlation of patient outcome with best postductal Pao2 and VI, as described by Wilson et al.1° Patients who fall in the right lower quadrant (Pao, value less than 100 and VI value greater than 1,000) should have a 100% mortality rate. However, our data show survivors from both groups in this area. DISCUSSION
Because a number of CDH patients suffer cardiopulmonary arrest shortly after birth, we hypothesized that these infants may represent a subset of “extremely high-risk” CDH patients with critical pulmonary hypoplasia that may be incompatible with life and who should not receive aggressive therapy. We conducted a retrospective analysis of 119 high-risk CDH patients treated at CHP over a 13-year period. Twenty-one of these infants suffered
= 21)
NCA (n = 98)
9 0
Abbrevratron: *One
of Death
CA(n
0
50 1 7
5 3
12 7
4 0
20 1
DNR, do not resuscitate.
persrstent
fetal crrculation
as “other”
cause
of death
cardiopulmonary arrest before CDH repair or initiation of ECMO therapy. Our data show that cardiopulmonary arrest before CDH repair or ECMO cannulation is not a univariate independent predictor of mortality. Baseline variables were comparable between CA and NCA patients. Fiveminute Apgar scores were significantly lower in the CA group, consistent with the fact that these infants probably experienced earlier and more severe respiratory distress than their NCA counterparts. This is further supported by the significantly worse best postductal arterial blood gases (Pao, and Pace,) in the CA compared with the NCA group (Table 2). Although these differences would suggest that CA patients experienced critical pulmonary hypoplasia, the overall mortality rate did not vary appreciably between CA and NCA patients (43% v 51%). Although it is not readily apparent why the mortality rates did not differ between the two groups. it is likely that initiation of ECMO support may have played a role in the CA group because 76% of these patients underwent ECMO therapy. The survival rate for these patients was no different than their NCA counterparts who also required ECMO. However, the survival rate for CA patients who received ECMO support was 50% compared with 20% for patients who did not undergo ECMO cannulation. The frequency and pattern of complications 1
’
’
’
1
’
’
’
’
’
’
i
X
110 I
Table 5. Complications Compllcatlon
CA (n = 21)
NCA (n = 981
12
69 3 10
None Sepsrs Bleedrng lntraventricular Other* More
than
0 0 2
hemorrhage
7 0
onet
*Six anuria/renal
farlure,
ity, one ischemic bowel. tSepsis and intraventncular
five pulmonary, hemorrhage.
two hemodynamic
lOi
8 7 1
0 instabil-
Fig 1. Outcome VI > 1,000 predict alive, CA; 0, alive,
500
1000
for each patient 100% mortality). NCA.
1500 VI
2000
2500
by Bohn plot (Pace, > 40 and x, dead, CA; +, dead, NCA; A,
COURCOULAS
loo-
m 0 %33Q%$hx+ I I 0 500
b I
I
1000
I
I
1500 VI
n I
I
.&,? I
2000
+ I
2500
Fig 2. Outcome for each patient by Wilson’s criteria (Pao, < 100 and VI > 1,000 predict 100% mortality). x, dead, CA; +, dead, NCA; A, alive, CA; 0, alive, NCA.
encountered during ECMO was similar in both groups. There was no increased predisposition to bleeding complications in the CA group despite a trend toward a longer ECMO run. Bohn’s criteria accurately predicted 100% mortality in 3 of 2 1 CA patients who fell in the right upper quadrant of the Bohn plot (Fig 1). However, the same criteria would have excluded a survivor from the NCA group, thus
ET AL
suggesting that Bohn’s criteria may not be optimal for predicting mortality. Similarly, best postductal Pao, in combination with the VI failed to accurately predict the likelihood of 100% mortality. Our data suggest that CA before CDH repair or ECMO cannulation is not a clinical marker for inadequate pulmonary parenchyma that may be incompatible with life. Infants who sustain cardiopulmonary arrest and who meet Bohn’s criteria may define an “extremely high-risk” group that is likely to have a poor outcome. Presently, however, our numbers are too small to draw such a conclusion. One potential problem in our database is that the length of each arrest event and arrest events that occurred during ECMO cannulation or CDH repair were not recorded because they were not easily obtainable from the charts. Also, we did not include long-term follow-up of CA patients to determine if these infants sustained more long-term neurological impairment than the NCA group. Our findings suggest that many CDH patients who sustained cardiopulmonary arrest may still have adequate pulmonary parenchyma to survive and therefore should be treated aggressively. A larger series of patients, or a meta analysis encompassing experiences from multiple centers will be required to develop a multivariate model of predictors of mortality.
REFERENCES 1. Wiener ES: Congenital posterolateral diaphragmatic hernia: New &mensions in management. Surgery 92:670-681, 1982 2. Lessm MS, Thompson IM, Deprez MF, et al: Congenital diaphragmatic hernia with or without extra corporeal membrane oxygenation: Are we making progress? JAm Co11 Surg 181:65-71, 1995 3. Heiss K, Manning P, Oldham KT, et al: Reversal of mortality for congenital diaphragmatic hernia with ECMO. Ann Surg 209:225-230, 1989 4. Wilson JM, Lund DP, Lillehei CW, et al: Delayed repair and preoperative ECMO does not improve survival in high-risk congenital diaphragmatlc hernia. J Pediatric Surg 27:368-375, 1992 5. West KW, Bengston K. Rescorla FJ, et al: Delayed surgical repair and ECMO improves survival in congenital diaphragmatic hernia. Ann Surg 216:454-462, 1992
6. Puri P: Congenital diaphragmatic hernia. Gun Probl Surg 16:790846, 1994 7. Slgalet DL, Tiemey A. Adolph V, et al: Timing of repair of congenital diaphragmatic hernia requiring extra corporeal membrane oxygenation support. J Pediatric Surg 30:1183-1187, 1995 8. Bohn DJ. Tamura M, Perein D, et al: Ventilatory predictors of pulmonary hypoplasia in congenital diaphragmatic hernia confirmed by morphometry J Pediatr 111:423-431, 1987 9. Bohn D: Ventilatory and blood gas parameters m predicting survival in congenital diaphragmatic hernia. Pediatr Surg Int 2:336-340, 1987 10. Wilson JM, Lund DP, Lillehei CW, et al: Congemtal diaphragmatic hernia: Predictors of severity in the ECMO era. J Pediatric Surg 26:1028-1034, 1991
Discussion J. Wilson (Boston, MA): The authors reviewed 119 infants with congenital diaphragmatic hernia (CDH) and compared outcomes between 21 patients who suffered pre-ECMO or preoperative cardiac arrest with 98 who did not. They concluded that the cardiac arrest group appeared to be a sicker population based on lower Apgar scores, worse postductal gas values, requirements for ECMO, and duration of ECMO. However, the survival was not significantly affected. They concluded, therefore, that pre-ECMO or preoperative cardiac arrest did not adversely affect survival in such patients and these patients should warrant an attempt to save.
Did any patient undergo cannulation during arrest itself versus being resuscitated from the arrest and then undergoing recannulation? In our patients there is a difference in survival in those groups. Did you categorize the cause of the arrest as true hemodynamic collapse, which is probably the case in some patients, versus a plugged endotracheal tube during transport or other mechanical problem in others who were expected to have a better survival? Under the causes of death there were 12 patients in the no-arrest group that had his cause of death listed as arrest. Did these patients die of arrest after ECMO or after
CARDIAC
ARREST
AND
MORTALITY
IN CDH
repair? Was this really an unexpected sudden arrest or was this the final common pathway of death in a patient who was suffering from pulmonary failure? The patients who had no need for ECMO had a 65% survival rate. Could you discuss the causes of death in these patients? Lastly, if you ventilate patients harshly enough to get into the box that predicts 100% mortality, you are not predicting mortality, you are assuring it. If you ventilate patients that harshly, you will destroy their lungs. Dr Bohn and I have long conversations about Wilson boxes and Bohn boxes, and harsh ventilation will put patients in the boxes that predict a poor outcome. Ad? Cow-codas (response): To answer your first question, in this study we looked only at patients with arrest events that occurred either before cannulation or before repair, so that was not one of our variables. To answer your question, regarding categorization of cause of arrest, just like cause of death, typically in a retrospective epidemiological study both cause of death
957
and cause of arrest are not accurately recorded in the charts. We did not have adequate information to obtain this, although this would be an interesting variable to study prospectively. You asked the most relevant question, were these mechanical problems or pulmonary parenchymal problems? We had no technical complications related to intubation problems as a cause of arrest. We had 12 patients in the no-arrest group whose ultimate common pathway or cause of death was arrest, and, as you point out, this is probably the final common pathway. This was not one of our variables. Again, we defined arrest as a pre-ECMO or prerepair variable. Finally, you asked about patients who were underwent repair and did not need ECMO. The survival rate for this group was 65%. Many of these cases occurred between the years of 1985 and 1988 when ECMO was fairly new and some parents decided not to go ahead with cannulation.
t? Courcoulas,
Kimberly K. Reblock, Marc I. Rowe, and Henri Pittsburgh, Pennsylvania
l Despite recent advances in the management of high-risk congenital diaphragmatic hernia KDH), mortality remains high. Deaths occur later because infants with inadequate pulmonary parenchyma are treated aggressively but eventually succumb to respiratory failure. In an attempt to identify absolute predictors of mortality the authors examined retrospectively their experience with CDH to determine if cardiac arrest before repair or initiation of extracorporeal membrane oxygenation (ECMO) invariably increased mortality. The authors reviewed the charts of 119 infants who had high-risk CDH treated between 1981 and 1994. They were divided into two groups: those that suffered cardiopulmonary arrest (CA, n = 21) before CDH repair or ECMO cannulation; and those that did not (NCA, N = 98). The authors compared mortality rate, ventilatory parameters, duration of, and complications on ECMO, as well as length of hospitalization between groups. Twenty-one infants suffered CA before initiation of ECMO support or CDH repair. Three infants (14%) suffered CA before arrival at our institution; seven (33%) after, and 11 (53%) both before and after arrival. There was no difference between the CA and NCA groups in terms of birth weight, gestational age, race and gender mix, or pregnancy and delivery complications. Five-minute Apgar scores were significantly lower in the CA group compared with the NCA group (4.6 Y 5.7, P = ,041. The CA group also had significantly worse “best postductal” blood gas and ventilatory parameters. There was no significant difference in length of hospitalization, time from admission to ECMO cannulation or CDH repair, or incidence of complications while on ECMO between the two groups. CA cases were more likely to require ECMO support (76% v 48%, P = .02) and to stay on ECMO for a more prolonged period than NCA cases (5.8 ~3.8 days, P = NSI. However, there was no significant difference in overall survival between CA and NCA cases (43% v 51%, P = NS). Cardiopulmonary arrest before repair of CDH or ECMO cannulation is not a univariate independent predictor of mortality and therefore should not preclude these highrisk infants from maximum intensive care therapy, including ECMO cannulation. Copyright Q 1997 by W.B. Saunders Company
INDEX WORDS: Congenital diaphragmatic cannulation, predictors of mortality, cardiac tory failure.
hernia, arrest,
ECMO respira-
T
HE MORTALITY RATE for infants presenting at birth with congenital diaphragmatic hernia (CDH) remains high. Series from many institutions, including ours, quote overall mortality rates of approximately 50%.1.2 Respiratory distress develops at birth in the most seriously affected infants. In the past. many of these infants died shortly after birth. However, with the inception of newer modalities,3-7 such as extracorporeal membrane oxygenation (ECMO), many critically ill infants Journal
of Pediatrrc
Surgery,
Vol 32, No 7 (July),
1997:
pp 953-957
R. Ford
who have CDH initially survive, only to succumb later to progressive respiratory failure or refractory pulmonary hypertension. Recent efforts have focused on the development of specific criteria to identify those infants who have critical pulmonary hypoplasia that is incompatible with life. Before the ECMO era, Bohn et al* used ventilatory parameters to predict outcome by plotting Pacoz versus ventilation index (VI). He predicted near 100% mortality in infants with Paco2 levels greater than 30 and VI values greater than 1,000.9 Since the advent of ECMO therapy, predicting absolute mortality has become more difficult. In some infants, ECMO is able to reverse clinical deterioration and provide salvage therapy. In others ECMO may only delay the inevitable mortality. Wilson et allo have attempted to identify predictors of severity that may be useful to determine outcome in the ECMO era. These investigators reported that patients not responding to maximum conventional ventilation with best postductal oxygen saturation less than 100% and with ventilation indices greater than 1,000 were not able to be saved and did not benefit from ECMO cannulation.rO Although one would assume that cardiac arrest before CDH repair or ECMO cannulation is uniformly fatal, to date, no study has examined the effect of such an event on ultimate outcome in patients who have CDH. This study was undertaken to determine if cardiac arrest before CDH repair or ECMO cannulation invariably increases mortality. If an arrest event could be shown to be an absolute predictor of mortality, the need for aggressive theiapy, including ECMO cannulation, would be obviated in these high-risk infants. MATERIALS
AND
METHODS
The medical records of 135 Infants who had CDH treated from 198 1 to 1994 at the Children’s Hospital of Pittsburgh (CHP) were reviewed retrospectively. The following variables were analyzed: demographics,
From the Depnrhrmt of Surgery, University ofPittsburgh School of Medicine, and the Deparfment of Pediafric Surgery, Chrldren k Hospital of Pittsburgh, Prttsburgh, PA. Presented at the 1996 Annual Meetmg of the Section on Surgery of the American Academy of Pedzatncs. Boston, Massachusetts, October 26-30, 1996. Address reprrnt requests to Anita P Courcoulas, MD, 300 Kaufiann Building, 3471 Fi@ Ave, Pzftsburgh, PA 15213. Copyright o 1997 by WB. Saunders Company
0022-3468/97/3207-0003$03.00/O 953
954
COURCOULAS
length of stay, gestabonal age, birth weight, pregnancy-related comphcations, delivery mode, Apgar scores, and associated congenital anomalies. In addition, the site of the diaphragmatic hernia (left v right). need for immediate endotracheal intubation, cardiopulmonary arrest before transfer to CHP or before to CDH repair or ECMO therapy. and laboratory parameters before CDH repair (best postductal arterial blood gas, mode of ventilation, mean arterial pressure [MAP], alveolararterial oxygen tension difference, oxygenation Index, and VI) were also entered in the database. Only the high-risk infants who required endotracheal intubation, a subset of 119 of the 135 patients. formed the basis of this study. Patients were divided into two groups: those who sustamed a cardiopulmonary arrest (CA) before repair of CDH or ECMO cannulation and those that did not (NCA). All patients were candidates for ECMO once this therapy became available at CHP m 1985. Cardiopulmonary arrest events were defined as any event that resulted in cardiopulmonary collapse requiring treatment including, but not limited to, chest compression. Patients were also categorized into seven predefined repair groups depending on the timing of repair and need for ECMO therapy: (A) repair of CDH and no need for ECMO. (B) repair after ECMO therapy. (C) repair before ECMO, (D) repair during ECMO, (E) repair before transfer to CHP with and without ECMO, and (F) no repair with some ECMO support. To better define absolute predictors of mortality m our patients, we examined outcome retrospectively for CDH patients in each group in the context of a Bohn plot,* or the method proposed by Wilson (Pal v VI).t” Outcome variables collected included: length of hospital stay, need for, length of, and complications on ECMO therapy: and mortality. The data were abstracted into a database and analyzed using StatView for the Macintosh by Abacus Concepts (Berkeley, CA). Continuous variables were compared using t tests and categorical variables were compared using either x2 or Fisher’s Exact tests. The CA group was compared with the NCA group in a univariate fashion. All results are presented as percentages or as mean i SD.
RESULTS
Twenty-one of the 119 high-risk infants (17.6%) suffered a cardiopulmonary arrest (CA group) event before CDH repair or initiation of ECMO therapy. Ninety-eight (82.4%) of the 119 had no history of an arrest (NCA group). Among the 21 infants in the CA group, 3 (14%) experienced CA before arrival at our institution, 7 (33%) after arrival, and 11 (53%) experienced arrest both before and after arrival. Table 1 shows baseline variables. The only significant difference between groups was 5-minute Apgar scores, Table
I. Baseline
Variables
Pregnancy (% with
complications none)
Delivery mode vaginal) I-min Apat 5-min Apgar
NCA in = 98)
P Value
NS
57143
2886 t 659 38.1 i 2.5 7.2/90.8/2.0 52148
65
60
NS
71 3.7 i 2.4 4 6 -t 2.8 60.0
67 4 1 2 2.3 5.7 i- 2.1
NS NS .04
3089
2771
39.0 2 2.1 0/100/0
NS NS NS
(% spontaneous
Outside city referral Left-sided CDH (%)
t%)
85.7
and Ventilatory
PO2 Pco, >50 FIO, Mean
59.2 816
NS NS
airway
VI Respiratory
pressure rate
PH
Arterial
Blood
Gas
Parameters
CA (n = 21)
NCA (n = 98)
P Value
51 i 33
122 + 127
.03
60% 1.0 f 0.01
28%
.03
16.6 r 5.3
0.87 + 0.21 12.8 t 4.3
.04 .Ol
1,047 + 559 58 k 16
666 + 456 49 + 18
.03
7.23 I! 0.30
7.33 f 0.23
NS NS
which were significantly worse (4.6 v 5.7) for the CA group. There was no difference in the incidence of congenital anomalies or the need for patch repair between groups. As shown in Table 2, the CA group had significantly worse “best postductal” arterial blood gas values and ventilatory parameters than the NCA group. To define the effect of the timing of CDH repair or ECMO cannulation on ultimate outcome, we classified the patients into various subgroups (A through F, Table 3). CA patients were relatively evenly distributed among the various subgroups. In contrast, nearly half the NCA patients underwent CDH repair without requiring ECMO support (Group A). Overall, the timing of CDH repair or ECMO had no significant effect on outcome. However, CA patients who underwent CDH repair before ECMO support (Group C) had a 100% survival rate compared with NCA patients in the same group who only had a 35% survival rate. Table 4 shows a comparison of the outcome variables. CA patients required ECMO support more frequently than NCA patients (P = .02). There was a trend toward longer duration of ECMO support in the CA group. Length of hospitalization and time from admission to therapy (either CDH repair or initiation of ECMO) were comparable for both groups. NCA patients experienced fewer complications than CA patients while on ECMO, but this difference was not statistically significant (P = .08). Similarly, the survival rate was comparable for both groups. Interestingly, 8 of 16 CA patients who Table 3. Timing (Repair
by Group
CA(n = 211 Birth weight(g) Gestational age (wk) Race, black/white/other (%I Gender male/female (%)
Table 2. Best Postductal
ET AL
of Repair and Need for ECMO Group) and Survival
CAla=
NCA tn = 981
n(%) Cases
n(%) Alive
nl%l Cases
nl%) Alive
A B
4 (19.1) 2 (9.5)
1 (25) 1150)
48 (49) 4 (4.1)
33 (69) 4(100)
C D E
5 (23.7) 2 (9.5) 4 (19.1)
5 1100) 0 10) 2 (50)
23 (23.4) 9 (9.2) 8 (8.2)
8 (35) 0 (0) 5 (63)
F
4 (19.1)
0 (0)
6 (6.1)
0 (0)
*A, repair of CDH and no need for ECMO; B, reparr after ECMO therapy; C, repair before ECMO, D, repair during ECMO; E, reparr before transfer to CHP wrth and without ECMO; F, and no reparr with some ECMO support.
CARDIAC
ARREST
AND
MORTALITY
Table 4. Outcome
IN CDH
Variables CAM
955
by Group
Table 6. Cause
= 211
NCA (n = 98)
P Value
Length of hosprtal stay id) Time from admit to ECMO
21.3 + 24 4
30.8 + 44.6
NS
None DNR
or repair(h) Need for ECMO
22.6 f 30.5
20.5 i 30.1
(%)
Sepsis Arrest
(d)
48 3.8 2 5 2
NS .02
Time on ECMO
76 5.8 2 7 1 42.9 43
29.6 51
NS NS
Bleeding Pulmonaryfailure
NS
Other*
Complrcations Survival (%)
on ECMO
(%)
Cause of Death
required ECMO survived compared with 16 of 48 NCA patients (P = NS). Of note, of the five CA patients who did not receive ECMO support, only one survived. Table 5 outlines specific complications seen during ECMO. There was no significant difference in the rate or type of complications between the two groups. Causes of death in both groups are shown in Table 6. Cardiopulmonary arrest was the terminal event that accounted for 5 of 12 deaths in the CAgroup and 12 of 36 deaths in the NCA group. Overall, there was no significant difference between the two groups in death rate or cause of death. According to Bohn’s predictions, patients in the right upper quadrant (with a Pace, value greater than 40 and a VI value greater than 1,000) should have 100% mortality. Three of 21 CA patients fell in that quadrant, and they all died (Fig 1). However, one patient in the NCA group who also fell in the right upper quadrant survived. Fig 2 shows the correlation of patient outcome with best postductal Pao2 and VI, as described by Wilson et al.1° Patients who fall in the right lower quadrant (Pao, value less than 100 and VI value greater than 1,000) should have a 100% mortality rate. However, our data show survivors from both groups in this area. DISCUSSION
Because a number of CDH patients suffer cardiopulmonary arrest shortly after birth, we hypothesized that these infants may represent a subset of “extremely high-risk” CDH patients with critical pulmonary hypoplasia that may be incompatible with life and who should not receive aggressive therapy. We conducted a retrospective analysis of 119 high-risk CDH patients treated at CHP over a 13-year period. Twenty-one of these infants suffered
= 21)
NCA (n = 98)
9 0
Abbrevratron: *One
of Death
CA(n
0
50 1 7
5 3
12 7
4 0
20 1
DNR, do not resuscitate.
persrstent
fetal crrculation
as “other”
cause
of death
cardiopulmonary arrest before CDH repair or initiation of ECMO therapy. Our data show that cardiopulmonary arrest before CDH repair or ECMO cannulation is not a univariate independent predictor of mortality. Baseline variables were comparable between CA and NCA patients. Fiveminute Apgar scores were significantly lower in the CA group, consistent with the fact that these infants probably experienced earlier and more severe respiratory distress than their NCA counterparts. This is further supported by the significantly worse best postductal arterial blood gases (Pao, and Pace,) in the CA compared with the NCA group (Table 2). Although these differences would suggest that CA patients experienced critical pulmonary hypoplasia, the overall mortality rate did not vary appreciably between CA and NCA patients (43% v 51%). Although it is not readily apparent why the mortality rates did not differ between the two groups. it is likely that initiation of ECMO support may have played a role in the CA group because 76% of these patients underwent ECMO therapy. The survival rate for these patients was no different than their NCA counterparts who also required ECMO. However, the survival rate for CA patients who received ECMO support was 50% compared with 20% for patients who did not undergo ECMO cannulation. The frequency and pattern of complications 1
’
’
’
1
’
’
’
’
’
’
i
X
110 I
Table 5. Complications Compllcatlon
CA (n = 21)
NCA (n = 981
12
69 3 10
None Sepsrs Bleedrng lntraventricular Other* More
than
0 0 2
hemorrhage
7 0
onet
*Six anuria/renal
farlure,
ity, one ischemic bowel. tSepsis and intraventncular
five pulmonary, hemorrhage.
two hemodynamic
lOi
8 7 1
0 instabil-
Fig 1. Outcome VI > 1,000 predict alive, CA; 0, alive,
500
1000
for each patient 100% mortality). NCA.
1500 VI
2000
2500
by Bohn plot (Pace, > 40 and x, dead, CA; +, dead, NCA; A,
COURCOULAS
loo-
m 0 %33Q%$hx+ I I 0 500
b I
I
1000
I
I
1500 VI
n I
I
.&,? I
2000
+ I
2500
Fig 2. Outcome for each patient by Wilson’s criteria (Pao, < 100 and VI > 1,000 predict 100% mortality). x, dead, CA; +, dead, NCA; A, alive, CA; 0, alive, NCA.
encountered during ECMO was similar in both groups. There was no increased predisposition to bleeding complications in the CA group despite a trend toward a longer ECMO run. Bohn’s criteria accurately predicted 100% mortality in 3 of 2 1 CA patients who fell in the right upper quadrant of the Bohn plot (Fig 1). However, the same criteria would have excluded a survivor from the NCA group, thus
ET AL
suggesting that Bohn’s criteria may not be optimal for predicting mortality. Similarly, best postductal Pao, in combination with the VI failed to accurately predict the likelihood of 100% mortality. Our data suggest that CA before CDH repair or ECMO cannulation is not a clinical marker for inadequate pulmonary parenchyma that may be incompatible with life. Infants who sustain cardiopulmonary arrest and who meet Bohn’s criteria may define an “extremely high-risk” group that is likely to have a poor outcome. Presently, however, our numbers are too small to draw such a conclusion. One potential problem in our database is that the length of each arrest event and arrest events that occurred during ECMO cannulation or CDH repair were not recorded because they were not easily obtainable from the charts. Also, we did not include long-term follow-up of CA patients to determine if these infants sustained more long-term neurological impairment than the NCA group. Our findings suggest that many CDH patients who sustained cardiopulmonary arrest may still have adequate pulmonary parenchyma to survive and therefore should be treated aggressively. A larger series of patients, or a meta analysis encompassing experiences from multiple centers will be required to develop a multivariate model of predictors of mortality.
REFERENCES 1. Wiener ES: Congenital posterolateral diaphragmatic hernia: New &mensions in management. Surgery 92:670-681, 1982 2. Lessm MS, Thompson IM, Deprez MF, et al: Congenital diaphragmatic hernia with or without extra corporeal membrane oxygenation: Are we making progress? JAm Co11 Surg 181:65-71, 1995 3. Heiss K, Manning P, Oldham KT, et al: Reversal of mortality for congenital diaphragmatic hernia with ECMO. Ann Surg 209:225-230, 1989 4. Wilson JM, Lund DP, Lillehei CW, et al: Delayed repair and preoperative ECMO does not improve survival in high-risk congenital diaphragmatlc hernia. J Pediatric Surg 27:368-375, 1992 5. West KW, Bengston K. Rescorla FJ, et al: Delayed surgical repair and ECMO improves survival in congenital diaphragmatic hernia. Ann Surg 216:454-462, 1992
6. Puri P: Congenital diaphragmatic hernia. Gun Probl Surg 16:790846, 1994 7. Slgalet DL, Tiemey A. Adolph V, et al: Timing of repair of congenital diaphragmatic hernia requiring extra corporeal membrane oxygenation support. J Pediatric Surg 30:1183-1187, 1995 8. Bohn DJ. Tamura M, Perein D, et al: Ventilatory predictors of pulmonary hypoplasia in congenital diaphragmatic hernia confirmed by morphometry J Pediatr 111:423-431, 1987 9. Bohn D: Ventilatory and blood gas parameters m predicting survival in congenital diaphragmatic hernia. Pediatr Surg Int 2:336-340, 1987 10. Wilson JM, Lund DP, Lillehei CW, et al: Congemtal diaphragmatic hernia: Predictors of severity in the ECMO era. J Pediatric Surg 26:1028-1034, 1991
Discussion J. Wilson (Boston, MA): The authors reviewed 119 infants with congenital diaphragmatic hernia (CDH) and compared outcomes between 21 patients who suffered pre-ECMO or preoperative cardiac arrest with 98 who did not. They concluded that the cardiac arrest group appeared to be a sicker population based on lower Apgar scores, worse postductal gas values, requirements for ECMO, and duration of ECMO. However, the survival was not significantly affected. They concluded, therefore, that pre-ECMO or preoperative cardiac arrest did not adversely affect survival in such patients and these patients should warrant an attempt to save.
Did any patient undergo cannulation during arrest itself versus being resuscitated from the arrest and then undergoing recannulation? In our patients there is a difference in survival in those groups. Did you categorize the cause of the arrest as true hemodynamic collapse, which is probably the case in some patients, versus a plugged endotracheal tube during transport or other mechanical problem in others who were expected to have a better survival? Under the causes of death there were 12 patients in the no-arrest group that had his cause of death listed as arrest. Did these patients die of arrest after ECMO or after
CARDIAC
ARREST
AND
MORTALITY
IN CDH
repair? Was this really an unexpected sudden arrest or was this the final common pathway of death in a patient who was suffering from pulmonary failure? The patients who had no need for ECMO had a 65% survival rate. Could you discuss the causes of death in these patients? Lastly, if you ventilate patients harshly enough to get into the box that predicts 100% mortality, you are not predicting mortality, you are assuring it. If you ventilate patients that harshly, you will destroy their lungs. Dr Bohn and I have long conversations about Wilson boxes and Bohn boxes, and harsh ventilation will put patients in the boxes that predict a poor outcome. Ad? Cow-codas (response): To answer your first question, in this study we looked only at patients with arrest events that occurred either before cannulation or before repair, so that was not one of our variables. To answer your question, regarding categorization of cause of arrest, just like cause of death, typically in a retrospective epidemiological study both cause of death
957
and cause of arrest are not accurately recorded in the charts. We did not have adequate information to obtain this, although this would be an interesting variable to study prospectively. You asked the most relevant question, were these mechanical problems or pulmonary parenchymal problems? We had no technical complications related to intubation problems as a cause of arrest. We had 12 patients in the no-arrest group whose ultimate common pathway or cause of death was arrest, and, as you point out, this is probably the final common pathway. This was not one of our variables. Again, we defined arrest as a pre-ECMO or prerepair variable. Finally, you asked about patients who were underwent repair and did not need ECMO. The survival rate for this group was 65%. Many of these cases occurred between the years of 1985 and 1988 when ECMO was fairly new and some parents decided not to go ahead with cannulation.