Adenovirus is a common cause of respiratory tract infection, particularly in children and immunocompromised hosts.1–6 It occurs worldwide in the form of epidemic, endemic, or sporadic disease, causing approximately 5–11% of acute viral pneumonia and bronchiolitis in infants and children. Occasionally, certain adenovirus serotypes can cause outbreaks of severe respiratory infections in healthy adults.7–11
Extracorporeal membrane oxygenation (ECMO) support for severe adenoviral infection has been reported since the early days of ECMO use.12–16 However, all these studies have either been case reports,13–16 represent an earlier cohort till up to 1994,12 or have not provided detailed analysis.12–16 Despite prior studies investigating risk factors on ECMO for acute respiratory failure,17–21 to date there are no studies that have investigated risk factors associated with mortality for patients with adenoviral infection supported with ECMO. The objective of this retrospective study was to review data obtained from the Extracorporeal Life Support Organization (ELSO) registry for pediatric patients with adenovirus infection. Our aim was to define for this patient cohort, the 1) clinical characteristics, 2) survival to hospital discharge, and 3) factors associated with death before hospital discharge.
The ELSO registry was founded in 1982 and collects patient data from member institutions on extracorporeal support in children and adults. Data are collected and sent from the contributing centers with a standardized data sheet containing patient demographics, diagnosis and procedure information, ECMO technique, complications during ECMO, and patient outcomes. Currently, 200 centers contribute data to the registry. Each individual member institution approves data reporting through its local institutional review board.
This study was approved by the Institutional Review Board of the University of Arkansas for Medical Sciences. Informed consent was exempted for this analysis of retrospective deidentified registry data.
The ELSO registry was queried for all patients between the ages of 0 and 18 with documented adenoviral infection from 1998 to 2009.
The ELSO registry was founded in 1982 and collects patient data from member institutions on extracorporeal support in children and adults. Data are collected and sent from the contributing centers with a standardized data sheet containing patient demographics, diagnosis and procedure information, ECMO technique, complications during ECMO, and patient outcomes. Currently, 200 centers contribute data to the registry. Each individual member institution approves data reporting through its local institutional review board. This database was queried using International Classification of Disease (ICD)-9 diagnostic code (079.0) 480.0 also and the organism code used by the ELSO registry for adenovirus (code number 49) to identify the study cohort. Each patient’s diagnosis was assessed for accurate identification independently by two investigators (P.P., R.T.F.), and any discrepancy was resolved independently by a third investigator (A.T.B.).
The primary outcome variable was death before hospital discharge. Patient demographics, pre-ECMO comorbidities/secondary diagnoses, pre-ECMO oxygenation, ventilation, respiratory support, and data on the type and duration of ECMO support were collected. The ELSO registry provides data on respiratory parameters as the worst values within 6 hours of ECMO initiation. Data were also collected on a number of predefined complications which occurred while on ECMO support.16
Categorical variables were summarized using frequencies and percentages, and continuous variables using median and ranges. Either Fisher’s exact or Pearson’s chi-square test was used to test for associations of categorical variables with in-hospital mortality. The Wilcoxon–Mann–Whitney test was used to compare continuous variables by in-hospital mortality.
Logistic regression models were used to evaluate association of patient characteristics and pre-ECMO risk factors with in-hospital mortality in patients with adenovirus infections. A separate model was used to evaluate the association of ECMO complications with in-hospital mortality. Variables associated with hospital survival in the univariate analysis with a p value of 0.10 or less were considered for inclusion in a stepwise multiple logistic regression analysis followed by a refit model using backwards selection with 0.1 as selection probability to form different models based on 1) pre-ECMO variables (models 1s, 2), 2) on-ECMO variables (models 3, 4). Those variables with more than 20% missing values were not included in the final multivariate models. All variables were assessed for multicollinearity. When the final model was identified, factors that had not been retained in the model were re-evaluated for inclusion to achieve a parsimonious model. In the model for complications, several variables of potential interest had zero cells (e.g., no patients who developed cardiac tamponade experienced death in hospital). To accommodate potential predictors with zero cells, Firth’s method was used to fit the logistic regression analysis.22,23 Cumulative incidence plots with Gray’s test were used to compare the shared hazard of death in-hospital adjusting for patients discharged alive. The time intervals used for this analysis were a composite of time of admission to time on ECMO support, time on ECMO support, and time off ECMO support to death.
As the frequency of complications on ECMO support could potentially increase with increased duration of ECMO, hours of ECMO was forced into the multivariate models as depicted in model 3. As the risk of death is very nonlinear for time on ECMO, ECMO duration was parameterized using restricted cubic splines with four knots at: 69.4, 206.2, 370.8, and 859.8 hours (approximately at the 5th, 35th, 65th, and 95th percentiles). Odds ratio (OR) calculations are for changes from the median (290) versus the first quartile (161) and the third quartile (480) versus the median (290) time on ECMO. A p value less than 0.05 was considered significant. All statistical calculations were performed using STATA version 11.1 (College Station, TX).
For the years 1998–2009, 163 cases of adenoviral infection among patients supported with ECMO were identified from the ELSO registry. During the same time period, the total number of cases of ECMO support among neonates and children for respiratory indication reported to the ELSO registry was 13,404 (ELSO International Summary; January 2010). Overall, survival at hospital discharge occurred in 38% of cases (62/163). Among neonates (<31 days of age), the survival at hospital discharge was only 11% (7/55). There were 38% (n = 39/101) of survivors in 2003–2008 which was unchanged compared with 38% (n = 23/61) of survivors in 1998–2002.
Table 1 lists demographic and other data for survivors and nonsurvivors. Neonatal presentation, pre-ECMO use of high-frequency oscillatory ventilation (HFOV), lower pH, PaO2, SaO2, higher PaCO2, and sepsis were all associated with in-hospital mortality. Acidosis was associated with mortality, but PaO2/FiO2 ratio and oxygenation index were not associated with mortality.
Table 2 lists complications while supported on ECMO between survivors and nonsurvivors. Pneumothorax, pH less than 7.2, serum creatinine greater than 3 mg/dl, and receipt of dialysis during ECMO were significantly different between survivors and nonsurvivors.
Table 3 shows the results of three logistic regression models for variables associated with death before hospital discharge in children with adenovirus infection supported on ECMO. Models 1 and 2: includes pre-ECMO variables only; models 3 and 4: on-ECMO complications with and without central nervous system (CNS) hemorrhage included in the model. In multivariable logistic regression analysis (model 1), neonatal age (OR, 4.3; 95% confidence interval [CI], 1.62–10.87), a decrease of 0.1 unit in pre-ECMO pH (OR, 1.77; 95% CI, 1.3–2.42), the presence of sepsis (OR, 4.55; 95% CI, 1.47–14.15), and increased PIP (OR, 1.04; 95% CI, 1.01–1.08) were all independently associated with in-hospital mortality.
Multivariate analysis for on-ECMO complications (model 2) demonstrates that presence of pneumothorax (OR, 3.6; 95% CI, 1.2–10.7), pH less than 7.2 (OR, 5.9; 95% CI, 1.0–34.1), and CNS hemorrhage (OR, 25.4; 95% CI, 1.5–436.7) were independently associated with in-hospital mortality. However, the frequency of complications on ECMO can increase as a function of duration of ECMO. Thus in model 3, in which ECMO duration as a variable was forced into the models, only CNS hemorrhage on ECMO (OR, 36; 95% CI, 2–655.8) and hours of ECMO (p < 0.001) were independently associated with in-hospital mortality. Figure 1 depicts graphically the probability of death by hours on ECMO by presence or absence of CNS hemorrhage.
Table 4 shows the comparison of neonatal and non-neonatal patient population. Neonates had higher mortality, and were more likely to be supported by HFOV, receive surfactant, inhaled nitric oxide pre-ECMO compared with older patients. In contrast, older patients were more likely to suffer pre-ECMO cardiac arrest, have longer duration of pre-ECMO ventilation, have higher PaO2/FiO2 ratios, and have higher mean arterial pressures.
Although time to death was not of primary interest, we used survival methods to visually compare rates of death by time in hospital. Figure 2 depicts the cumulative incidence of in-hospital mortality, adjusted for the competing risk of hospital discharge, stratified by age (neonate versus non-neonates). The hazard of death is significantly higher in the neonates based on Gray’s test (p < 0.001).
This investigation offers the first systematic evaluation of clinical characteristics (demographic features, comorbidities, and complications) on hospital survival for children with adenovirus infection supported on ECMO. In this review of 163 cases from the ELSO registry, neonatal presentation, degree of acidosis immediately before ECMO support, the presence of sepsis, and increased PIP were identified as factors that independently associated with in-hospital mortality existing before initiation of ECMO in multivariate analysis. Brain hemorrhage and duration of ECMO support also were identified as independent risk factors for death.
Adenoviral infection severe enough to require ECMO support is an infrequent occurrence, as evidenced by the relatively small number of cases reported to an international registry during a 12 year period. The strength of this study is the use of the ELSO registry to report multicenter data on the largest cohort of children with adenovirus infection supported with ECMO and to describe risk factors associated with in-hospital mortality in these patients. Meyer et al.12 reported on the collective ELSO registry experience of children with viral infection supported on ECMO between 1988 and 1994, showing that adenovirus was the third most common etiology (12/127 cases) for ECMO support and had the poorest overall survival (25%) as compared with other viruses. Recently, Allibhai et al.13 reported the successful use of prolonged (>30 days) ECMO support for two children with overwhelming adenoviral pneumonia. These investigators also mention in their case report discussion that they queried the ELSO registry (era 1998–2005) for cases of adenoviral pneumonia supported with ECMO and reported 58 cases with a 52% survival.13 However, the authors provided no details on how potential cases from the ELSO registry were identified, the clinical characteristics of the identified patients, and unlike the current study, they did not assess the risk factors associated with in-hospital mortality in these patients.
Neonates with adenovirus infection requiring ECMO support have significantly higher in-hospital mortality rates (89%) compared with older infants and children (48%). These results are similar to reports with other viral infections (respiratory syncytial virus and influenza) where mortality is higher in the younger age groups.24–26 Shay et al.,26 during a 19 year study period, noted that among 1,806 bronchiolitis-associated deaths 79% occurred among infants less than 1 year of age. These studies, however, do not provide subgroup analysis specifically for neonates. A surprising finding was that most neonates in our study cohort did not have secondary diagnoses noted and seemed to be term infants without significant comorbidity. We speculate that neonates with immature organ systems and immune mechanisms may be more prone to disseminated adenoviral disease compared with older children. There may be other epidemiological risk factors involved, but the ELSO registry database is limited in its ability to assess these variables.
It is likely that many patients in this cohort had disseminated adenovirus disease (DAD), an entity used to describe systemic adenoviral infection associated with involvement of two or more organs. Munoz et al.27 reported that DAD with multiorgan involvement occurred in 11 of 440 (2.5%) adenovirus-infected patients: six of 11 (54%) were immunocompromised and five were immunocompetent. DAD was more common in immunocompromised hosts (12.5%) compared with 1.5% in immunocompetent hosts. Mortality was higher among the immunodeficient host (83%) compared with the immunocompetent host (60%).
We do note in this study that CNS hemorrhage was significantly associated with mortality (Table 3: models 2 and 3). However, despite the statistical significance and large OR, the 95% CIs are extremely large suggesting lesser confidence in this variable despite its statistical significance. This may also be reflective of the small event rates for CNS hemorrhage in the study cohort to preclude meaningful conclusions. There may be a role for neurosurgical intervention in CNS bleeding and the risk versus benefit has to be considered on a case-by-case basis.28
Our study has certain limitations. As a retrospective study of registry data, it is subject to considerable bias, both in reporting and in selection. Registry data provide no information regarding how the diagnosis of adenoviral disease was made (e.g., viral respiratory culture, rapid antigen testing, polymerase chain reaction testing), and patients were identified for the study via an ICD-9 code for adenoviral disease. The accuracy of the diagnosis neither can be assessed from the available data, nor can one know with certainty whether an ICD-9 code for adenovirus represents the patient’s primary pathophysiologic reason for requiring ECMO or a comorbidity. In addition, little standardization exists in the coding of pre-ECMO comorbidities in the ELSO registry. For instance, “sepsis” was found to be a significant risk factor for mortality in this cohort. In this study, patients with “sepsis” were identified by coding in the ELSO registry, rather than by application of accepted criteria for the definition of sepsis. One might surmise that most infants with disseminated adenoviral disease and severe adenoviral pneumonia would have an inflammatory response of such a magnitude as to meet internationally accepted criteria for sepsis. However, use of such strict criteria is not required by the ELSO registry, so one cannot know whether the diagnosis of sepsis in this cohort was under-reported or over-reported. In addition, the data from the ELSO registry do not allow one to determine whether the primary pathophysiology requiring ECMO support was overwhelming shock or refractory hypoxemic respiratory failure.
However, ELSO has maintained an international registry since the 1980s with great success, and the registry has been used to help answer a number of research questions. No other mechanism currently exists to collect information regarding such a large number of patients suffering from what is essentially a rare but critical clinical problem. The ELSO data are limited as they do not have information related to severity of illness scores (e.g., Pediatric Index of Mortality, Pediatric Risk of Mortality, Pediatric Logistic Organ Dysfunction) during hospitalization. It is possible that the indication and timing of ECMO deployment may vary among the various centers submitting data to the registry. There are no indicators to assess practice variation among the various centers in the ELSO Registry. It is difficult to determine absolute criteria used by different institutions for deciding the type and indication for ECMO support in this population. This investigation is not designed either to evaluate the impact of novel therapies on ECMO outcomes or to investigate residual morbidity among survivors.
A strength of this study is its use of more than one model for logistic regression analysis, particularly with respect to controlling for the increasing likelihood of ECMO complications as the duration of ECMO increases. The association of increasing ECMO duration with mortality may be reflective of the severity of the underlying disease, an increase in ECMO-related complications, or both. It is not surprising that CNS hemorrhage was independently associated with mortality in all models. Such hemorrhage can be neurologically devastating and is usually an indication for cessation of ECMO support, because of the risk of hemorrhage extension with continued anticoagulation. The presence of pneumothorax while on ECMO as a risk factor for in-hospital mortality most likely reflects severity of underlying pneumonia, and in particular the possibility of necrotizing pneumonia.
In this largest cohort study of pediatric patients supported with ECMO for adenoviral infection, survival to hospital discharge occurred in approximately 38% of cases. Neonatal presentation, degree of acidosis, sepsis, pneumothorax, and brain hemorrhage were independently associated with mortality. Neonates (<31 days of age) with overwhelming adenovirus infection requiring ECMO support have a particularly grim prognosis (11% survival). Knowledge of these factors may be of use to clinicians in patient selection for ECMO support and in counseling families regarding prognosis.
1. Ahmad NM, Ahmad KM, Younus F. Severe adenovirus pneumonia (AVP) following infliximab infusion for the treatment of Crohn’s disease. J Infect. 2007;54:e29–e32
2. Castro-Rodriguez JA, Daszenies C, Garcia M, Meyer R, Gonzales R. Adenovirus pneumonia in infants and factors for developing bronchiolitis obliterans: A 5-year follow-up. Pediatr Pulmonol. 2006;41:947–953
3. Doan ML, Mallory GB, Kaplan SL, et al. Treatment of adenovirus pneumonia with cidofovir in pediatric lung transplant recipients. J Heart Lung Transplant. 2007;26:883–889
4. Ou ZY, Zeng QY, Wang FH, et al. Retrospective study of adenovirus in autopsied pulmonary tissue of pediatric fatal pneumonia in South China. BMC Infect Dis. 2008;8:122
5. Refaat M, McNamara D, Teuteberg J, et al. Successful cidofovir treatment in an adult heart transplant recipient with severe adenovirus pneumonia. J Heart Lung Transplant. 2008;27:699–700
6. Gupta P, Tobias JD, Goyal S, et al. Prolonged mechanical support in children
with severe adenoviral infections: A case series and review of the literature. J Intensive Care Med. 2011;26:267–272
7. Yamasaki S, Heike Y, Mori S, et al. Infectious complications in chronic graft-versus
-host disease: A retrospective study of 145 recipients of allogeneic hematopoietic stem cell transplantation with reduced- and conventional-intensity conditioning regimens. Transpl Infect Dis. 2008;10:252–259
8. Brosch L, Tchandja J, Marconi V, et al. Adenovirus serotype 14 pneumonia at a basic military training site in the United States, spring 2007: A case series. Mil Med. 2009;174:1295–1299
9. Cunha BA. Severe adenovirus community-acquired pneumonia mimicking Legionella. Eur J Clin Microbiol Infect Dis. 2009;28:313–315
10. Louie JK, Kajon AE, Holodniy M, et al. Severe pneumonia due to adenovirus serotype 14: A new respiratory threat? Clin Infect Dis. 2008;46:421–425
11. Hakim FA, Tleyjeh IM. Severe adenovirus pneumonia in immunocompetent adults: A case report and review of the literature. Eur J Clin Microbiol Infect Dis. 2008;27:153–158
12. Meyer TA, Warner BW. Extracorporeal life support for the treatment of viral pneumonia: Collective experience from the ELSO registry. Extracorporeal Life Support Organization. J Pediatr Surg. 1997;32:232–236
13. Allibhai TF, Spinella PC, Meyer MT, Hall BH, Kofos D, DiGeronimo RJ. Survival after prolonged pediatric extracorporeal membrane oxygenation
support for adenoviral pneumonia. J Pediatr Surg. 2008;43:e9–e11
14. Splaingard ML, Frazier OH, Jefferson LS, Stein F, Harrison GM. Extracorporeal membrane oxygenation
: Its role in the survival of a child with adenoviral pneumonia and myocarditis. South Med J. 1983;76:1171–1173
15. Aebi C, Headrick CL, McCracken GH, Lindsay CA. Intravenous ribavirin therapy in a neonate with disseminated adenovirus infection undergoing extracorporeal membrane oxygenation
: Pharmacokinetics and clearance by hemofiltration. J Pediatr. 1997;130:612–615
16. Kinney JS, Hierholzer JC, Thibeault DW. Neonatal pulmonary insufficiency caused by adenovirus infection successfully treated with extracorporeal membrane oxygenation
. J Pediatr. 1994;125:110–112
17. Weber TR, Connors RH, Tracy TF Jr, Bailey PV, Stephens C, Keenan W. Prognostic determinants in extracorporeal membrane oxygenation
for respiratory failure in newborns. Ann Thorac Surg. 1990;50:720–723
18. Domico MB, Ridout DA, Bronicki R, et al. The impact of mechanical ventilation time before initiation of extracorporeal life support on survival in pediatric respiratory failure: A review of the Extracorporeal Life Support Registry. Pediatr Crit Care Med. 2012;13:16–21
19. Stolar CJ, Snedecor SM, Bartlett RH. Extracorporeal membrane oxygenation
and neonatal respiratory failure: Experience from the extracorporeal life support organization. J Pediatr Surg. 1991;26:563–571
20. Moler FW, Custer JR, Bartlett RH, et al. Extracorporeal life support for severe pediatric respiratory failure: An updated experience 1991–1993. J Pediatr. 1994;124:875–880
21. Moler FW, Palmisano J, Custer JR. Extracorporeal life support for pediatric respiratory failure: Predictors of survival from 220 patients. Crit Care Med. 1993;21:1604–1611
22. Heinze G, Schemper M. A solution to the problem of monotone likelihood in Cox regression. Biometrics. 2001;57:114–119
23. Heinze G. A comparative investigation of methods for logistic regression with separated or nearly separated data. Stat Med. 2006;25:4216–4226
24. Bhat N, Wright JG, Broder KR, et al.Influenza Special Investigations Team. Influenza-associated deaths among children
in the United States, 2003–2004. N Engl J Med. 2005;353:2559–2567
25. Poehling KA, Edwards KM, Weinberg GA, et al.New Vaccine Surveillance Network. The underrecognized burden of influenza in young children
. N Engl J Med. 2006;355:31–40
26. Shay DK, Holman RC, Roosevelt GE, Clarke MJ, Anderson LJ. Bronchiolitis-associated mortality
and estimates of respiratory syncytial virus-associated deaths among US children
, 1979–1997. J Infect Dis. 2001;183:16–22
27. Munoz FM, Piedra PA, Demmler GJ. Disseminated adenovirus disease in immunocompromised and immunocompetent children
. Clin Infect Dis. 1998;27:1194–1200
28. Hervey-Jumper SL, Annich GM, Yancon AR, Garton HJ, Muraszko KM, Maher CO. Neurological complications of extracorporeal membrane oxygenation
. J Neurosurg Pediatr. 2011;7:338–344